id	sid	tid	token	lemma	pos
work_cxyxl5vs6zd6xlanudcx46qoqy	1	1	Joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	1	2	Semantic	Semantic	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	3	Synthesis	Synthesis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	1	5	Morphological	Morphological	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	6	Analysis	Analysis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	1	9	Derived	Derived	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	10	Word	Word	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	11	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	12	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	13	Department	Department	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	15	Computer	Computer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	16	Science	Science	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	17	Johns	Johns	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	18	Hopkins	Hopkins	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	19	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	20	ryan.cotterell@jhu.edu	ryan.cotterell@jhu.edu	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	1	21	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	22	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	23	CIS	CIS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	24	LMU	LMU	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	25	Munich	Munich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	26	inquiries@cislmu.org	inquiries@cislmu.org	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	1	27	Abstract	Abstract	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	28	Much	much	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	1	29	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	30	sentences	sentence	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	1	31	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	32	composed	compose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	33	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	34	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	1	35	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	1	36	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	1	37	themselves	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	38	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	1	39	composed	compose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	40	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	1	41	smaller	small	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	1	42	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	1	43	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	2	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	2	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	2	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	2	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	2	5	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	2	6	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	2	7	questionably	questionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	2	8	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	2	9	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	2	10	analyzed	analyze	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	2	11	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	2	12	question+able+ly	question+able+ly	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	2	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	3	1	However	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	3	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	3	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	3	4	structural	structural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	3	5	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	3	8	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	9	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	3	10	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	3	11	directly	directly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	3	12	give	give	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	3	13	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	3	14	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	3	15	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	3	16	representation	representation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	18	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	3	19	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	20	’s	’s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	3	21	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	3	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	4	1	Since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	2	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	3	obeys	obey	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	4	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	4	5	principle	principle	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	7	compositionality	compositionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	4	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	4	10	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	4	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	4	13	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	14	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	4	15	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	4	16	systematically	systematically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	4	17	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	18	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	4	20	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	21	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	4	22	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	4	23	parts	part	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	4	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	5	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	2	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	3	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	5	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	5	6	propose	propose	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	5	7	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	8	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	5	9	probabilistic	probabilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	5	10	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	12	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	13	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	14	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	15	captures	capture	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	5	16	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	18	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	19	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	20	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	21	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	22	w	w	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	23	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	24	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	5	25	constituent	constituent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	26	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	5	27	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	5	28	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	29	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	30	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	32	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	33	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	34	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	5	35	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	36	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	37	mean-	mean-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	5	38	ings	ing	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	5	39	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	5	40	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	5	41	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	5	42	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	6	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	6	2	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	6	3	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	6	4	learns	learn	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	6	5	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	6	6	segment	segment	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	6	7	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	6	8	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	6	9	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	6	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	6	11	compose	compose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	6	12	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	6	13	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	6	14	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	6	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	6	16	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	6	17	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	6	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	7	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	2	experiment	experiment	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	3	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	7	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	7	5	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	7	6	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	7	7	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	8	CELEX	CELEX	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	9	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	7	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	7	11	German	german	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	7	12	DErivBase	derivbase	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	7	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	7	14	Zeller	Zeller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	15	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	7	16	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	17	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	7	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	7	19	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	7	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	7	21	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	7	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	8	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	8	2	show	show	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	8	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	4	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	8	5	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	8	6	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	8	7	increases	increase	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	8	8	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	8	9	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	10	accuracy	accuracy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	8	12	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	8	13	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	8	14	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	15	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	16	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	8	17	%	%	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	18	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	8	19	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	8	20	%	%	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	8	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	9	1	Additionally	additionally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	9	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	9	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	9	4	investigate	investigate	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	9	5	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	9	6	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	9	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	9	8	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	9	9	compo-	compo-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	9	10	sition	sition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	9	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	9	12	showing	show	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	9	13	that	that	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	9	14	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	9	15	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	9	16	networks	network	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	9	17	yield	yield	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	9	18	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	9	19	improvement	improvement	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	9	20	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	9	21	simple	simple	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	9	22	additive	additive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	9	23	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	9	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	10	1	Finally	finally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	10	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	10	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	10	4	study	study	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	10	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	10	6	degree	degree	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	10	8	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	10	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	10	10	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	10	11	correspond	correspond	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	10	12	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	10	14	linguist	linguist	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	15	’s	’s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	16	notion	notion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	18	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	10	19	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	10	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	11	1	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	11	2	Introduction	introduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	11	3	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	11	4	most	most	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	11	5	languages	language	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	11	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	11	7	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	11	8	decompose	decompose	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	11	9	further	far	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	11	10	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	11	11	smaller	small	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	11	12	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	11	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	11	14	termed	term	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	11	15	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	11	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	12	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	12	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	12	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	12	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	12	5	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	12	6	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	12	7	questionably	questionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	12	8	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	12	9	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	12	10	analyzed	analyze	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	12	11	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	12	12	question+able+ly	question+able+ly	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	12	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	13	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	2	structural	structural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	13	3	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	6	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	13	8	however	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	13	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	13	10	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	11	itself	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	13	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	13	13	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	13	14	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	15	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	13	16	rep-	rep-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	17	resentation	resentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	18	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	20	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	21	’s	’s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	13	22	meaning;1	meaning;1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	13	23	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	13	24	further	further	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	13	25	re-	re-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	13	26	quire	quire	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	13	27	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	28	account	account	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	29	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	30	how	how	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	13	31	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	13	32	synthesize	synthesize	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	13	33	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	34	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	35	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	36	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	13	37	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	13	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	14	1	Fortunately	fortunately	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	14	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	14	3	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	14	4	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	14	5	just	just	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	14	6	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	7	phrases	phrase	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	14	8	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	14	9	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	11	large	large	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	12	extent	extent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	13	obey	obey	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	15	principle	principle	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	16	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	17	compositionality	compositionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	18	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	14	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	20	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	14	21	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	22	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	23	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	24	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	14	25	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	14	26	systematically	systematically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	14	27	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	28	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	30	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	31	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	32	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	14	33	parts.2	parts.2	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	14	34	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	35	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	36	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	37	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	14	38	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	14	39	propose	propose	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	14	40	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	41	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	42	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	43	prob-	prob-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	44	abilistic	abilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	45	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	46	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	47	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	48	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	49	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	50	captures	capture	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	51	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	52	structural	structural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	14	53	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	54	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	55	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	56	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	57	w	w	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	58	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	59	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	14	60	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	61	stituent	stituent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	62	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	14	63	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	14	64	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	65	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	66	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	67	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	14	68	’s	’s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	14	69	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	70	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	71	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	72	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	73	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	14	74	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	14	75	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	14	76	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	15	1	Morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	15	2	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	15	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	15	5	structured	structured	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	15	6	pre-	pre-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	7	diction	diction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	8	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	9	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	15	10	seeks	seek	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	15	11	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	15	12	break	break	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	15	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	15	14	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	15	up	up	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	15	16	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	17	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	15	18	constituent	constituent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	15	19	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	15	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	16	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	16	2	output	output	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	3	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	4	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	16	5	been	be	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	6	shown	show	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	16	8	aid	aid	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	16	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	16	10	diverse	diverse	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	16	11	set	set	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	13	applications	application	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	16	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	15	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	16	16	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	17	automatic	automatic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	16	18	speech	speech	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	19	recognition	recognition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	20	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	21	Afify	Afify	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	22	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	16	23	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	24	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	26	2006	2006	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	16	27	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	29	keyword	keyword	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	30	spotting	spot	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	16	31	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	32	Narasimhan	Narasimhan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	33	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	16	34	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	35	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	36	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	37	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	16	38	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	39	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	40	machine	machine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	41	translation	translation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	16	42	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	43	Clifton	Clifton	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	44	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	16	45	Sarkar	Sarkar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	46	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	47	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	16	48	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	49	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	16	50	parsing	parse	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	16	51	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	52	Seeker	Seeker	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	53	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	16	54	Çetinoğlu	Çetinoğlu	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	16	55	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	16	56	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	16	57	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	16	58	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	17	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	2	contrast	contrast	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	3	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	4	much	much	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	17	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	6	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	17	7	prior	prior	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	17	8	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	17	10	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	17	11	focus	focus	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	17	12	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	13	supervised	supervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	17	14	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	17	16	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	17	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	17	18	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	17	19	provide	provide	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	17	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	17	21	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	22	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	23	gold	gold	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	17	24	segmentations	segmentation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	17	25	during	during	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	26	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	27	time	time	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	17	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	18	1	Instead	instead	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	18	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	3	sur-	sur-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	4	1There	1there	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	18	5	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	18	6	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	18	7	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	18	8	linguistic	linguistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	18	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	18	10	computational	computational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	18	11	theo-	theo-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	12	ries	rie	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	18	13	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	14	interpreting	interpret	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	18	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	18	16	structural	structural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	18	17	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	18	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	19	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	18	20	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	18	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	19	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	19	4	un-	un-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	19	5	often	often	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	19	6	signifies	signify	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	19	7	negation	negation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	19	9	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	19	10	effect	effect	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	11	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	12	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	19	13	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	19	14	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	19	15	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	19	16	modeled	model	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	17	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	18	theories	theory	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	19	19	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	20	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	21	logic	logic	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	19	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	20	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	20	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	3	ad-	ad-	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	20	4	dresses	dress	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	20	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	20	6	question	question	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	8	structural	structural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	20	9	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	20	11	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	20	12	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	20	15	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	20	16	framework	framework	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	20	18	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	20	19	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	20	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	21	1	2Morphological	2morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	2	research	research	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	4	theoretical	theoretical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	21	6	computational	computational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	7	linguistics	linguistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	21	8	often	often	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	21	9	focuses	focus	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	10	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	11	noncompositional	noncompositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	12	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	21	13	less	less	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	21	14	com-	com-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	15	positional	positional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	16	phenomena	phenomenon	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	21	17	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	21	18	simply	simply	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	21	19	because	because	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	20	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	21	deriva-	deriva-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	22	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	23	poses	pose	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	24	fewer	fewer	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	21	25	interesting	interesting	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	21	26	research	research	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	21	27	problems	problem	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	21	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	22	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	22	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	3	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	4	true	true	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	5	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	22	6	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	22	7	just	just	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	8	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	22	9	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	10	frequent	frequent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	11	multiword	multiword	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	12	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	22	13	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	22	14	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	15	completely	completely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	16	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	17	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	22	18	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	19	frequent	frequent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	20	derivations	derivation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	22	21	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	22	22	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	22	24	refusal	refusal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	22	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	22	26	fit-	fit-	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	22	27	ness	ness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	22	28	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	22	29	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	22	30	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	31	completely	completely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	22	32	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	22	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	23	1	An	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	23	2	indication	indication	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	4	non-	non-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	23	5	lexicalized	lexicalize	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	23	6	derivations	derivation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	23	7	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	23	8	usually	usually	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	23	9	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	23	10	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	23	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	23	12	fact	fact	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	13	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	14	standard	standard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	23	15	dictionaries	dictionary	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	23	16	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	17	OUP	OUP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	23	18	editors	editor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	23	19	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	23	20	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	23	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	23	22	list	list	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	23	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	23	24	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	23	25	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	26	their	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	23	27	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	23	28	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	23	30	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	31	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	23	32	hedge	hedge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	33	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	34	they	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	23	35	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	23	36	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	23	37	occur	occur	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	23	38	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	39	part	part	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	40	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	23	41	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	23	42	partially	partially	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	23	43	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	23	44	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	23	45	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	24	1	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	24	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	24	3	Haspelmath	Haspelmath	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	24	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	24	5	Sims	Sims	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	24	6	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	24	7	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	24	8	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	24	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	24	10	§	§	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	24	11	5.3.6	5.3.6	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	24	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	25	1	33	33	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	25	2	Transactions	transaction	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	25	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	25	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	25	5	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	25	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	25	7	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	25	8	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	25	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	25	10	vol	vol	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	25	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	26	1	6	6	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	26	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	26	3	pp	pp	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	26	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	27	1	33–48	33–48	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	27	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	27	3	2018	2018	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	27	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	28	1	Action	Action	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	28	2	Editor	Editor	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	28	3	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	28	4	Regina	Regina	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	28	5	Barzilay	Barzilay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	28	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	29	1	Submission	submission	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	29	2	batch	batch	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	29	3	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	29	4	5/2016	5/2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	29	5	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	29	6	Revision	revision	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	29	7	batch	batch	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	29	8	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	29	9	10/2016	10/2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	29	10	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	29	11	Published	publish	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	29	12	1/2018	1/2018	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	29	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	30	1	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	30	2	©	©	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	30	3	2018	2018	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	30	4	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	30	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	30	6	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	30	7	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	30	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	31	1	Distributed	distribute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	31	2	under	under	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	31	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	31	4	CC	cc	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	31	5	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	31	6	BY	BY	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	31	7	4.0	4.0	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	31	8	license	license	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	31	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	32	1	face	face	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	2	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	4	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	32	5	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	6	performs	perform	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	32	7	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	32	8	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	9	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	32	10	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	11	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	32	12	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	13	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	15	2016a	2016a	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	32	16	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	32	17	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	18	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	19	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	20	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	22	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	32	23	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	32	24	Kann	Kann	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	25	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	26	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	27	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	29	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	32	30	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	32	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	32	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	32	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	34	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	35	allows	allow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	32	36	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	32	37	induc-	induc-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	32	38	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	39	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	40	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	32	41	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	32	42	together	together	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	32	43	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	44	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	32	45	seg-	seg-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	46	mentation	mentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	32	47	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	32	48	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	32	49	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	32	50	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	32	51	typical	typical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	32	52	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	33	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	33	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	33	3	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	33	4	questionably	questionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	33	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	33	6	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	33	7	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	33	8	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	33	9	restore	restore	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	33	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	33	11	deleted	delete	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	33	12	char-	char-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	33	13	acters	acter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	33	14	le	le	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	33	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	33	16	yielding	yield	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	33	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	33	18	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	33	19	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	33	20	question	question	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	33	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	33	22	able	able	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	33	23	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	33	24	ly	ly	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	33	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	34	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	2	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	34	3	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	34	5	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	34	6	primary	primary	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	7	contribution	contribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	8	lies	lie	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	9	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	34	11	integration	integration	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	13	continuous	continuous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	14	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	15	vec-	vec-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	16	tors	tor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	34	17	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	18	supervised	supervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	19	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	20	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	21	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	34	22	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	34	23	present	present	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	34	24	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	34	25	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	26	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	27	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	28	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	29	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	30	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	34	31	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	34	32	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	33	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	34	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	34	35	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	36	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	34	37	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	34	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	35	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	35	2	experimentally	experimentally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	35	3	investigate	investigate	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	35	4	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	35	5	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	35	6	aspects	aspect	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	35	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	35	8	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	35	9	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	35	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	36	1	•	•	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	36	2	First	first	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	36	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	36	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	36	5	show	show	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	36	6	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	36	7	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	36	8	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	36	9	continu-	continu-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	36	10	ous	ous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	36	11	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	36	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	36	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	36	14	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	36	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	36	16	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	36	17	phemes	pheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	36	18	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	36	19	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	36	20	allows	allow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	36	21	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	36	22	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	36	23	improve	improve	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	36	24	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	36	25	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	36	26	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	36	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	37	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	37	3	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	37	4	portion	portion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	6	CELEX	CELEX	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	37	8	Baayen	Baayen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	9	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	10	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	11	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	37	13	1993	1993	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	37	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	37	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	37	16	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	17	achieve	achieve	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	18	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	37	19	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	37	20	point	point	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	21	improvement	improvement	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	22	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	23	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	24	accuracy	accuracy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	25	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	37	26	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	37	27	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	37	28	point	point	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	29	improvement	improvement	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	30	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	37	31	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	32	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	37	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	38	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	38	3	German	german	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	38	4	DErivBase	derivbase	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	5	dataset	dataset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	6	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	38	7	achieve	achieve	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	38	8	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	38	9	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	38	10	point	point	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	11	improvement	improvement	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	12	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	13	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	14	accu-	accu-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	15	racy	racy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	16	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	38	17	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	38	18	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	38	19	point	point	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	20	improvement	improvement	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	21	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	38	22	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	38	23	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	38	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	39	1	•	•	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	39	2	Second	second	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	39	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	39	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	39	5	explore	explore	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	39	6	improved	improved	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	39	7	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	39	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	39	9	vec-	vec-	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	39	10	tor	tor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	39	11	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	39	12	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	39	13	synthesizing	synthesize	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	39	14	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	39	15	mean-	mean-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	39	16	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	39	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	40	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	40	2	find	find	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	40	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	40	4	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	40	5	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	40	6	network	network	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	40	7	im-	im-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	40	8	proves	prove	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	40	9	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	40	10	previously	previously	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	40	11	proposed	propose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	40	12	additive	additive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	40	13	mod-	mod-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	40	14	els	els	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	40	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	41	1	Moreover	moreover	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	41	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	41	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	41	4	find	find	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	41	5	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	6	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	41	7	syntactically	syntactically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	41	8	oriented	orient	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	9	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	41	10	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	41	11	Levy	Levy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	41	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	41	13	Goldberg	Goldberg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	41	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	41	15	2014a	2014a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	41	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	41	17	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	41	18	better	well	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	41	19	suited	suited	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	41	20	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	21	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	22	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	23	bag	bag	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	24	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	41	25	of-	of-	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	41	26	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	41	27	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	41	28	BOW	BOW	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	41	29	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	41	30	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	41	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	42	1	•	•	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	42	2	Finally	finally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	42	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	42	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	42	5	explore	explore	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	42	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	42	7	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	42	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	42	9	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	42	10	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	42	11	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	42	12	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	42	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	42	14	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	42	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	42	16	distribu-	distribu-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	42	17	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	42	18	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	42	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	43	1	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	43	2	Derivational	Derivational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	43	3	Morphology	Morphology	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	43	4	Two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	43	5	important	important	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	43	6	goals	goal	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	43	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	8	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	43	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	43	11	linguistic	linguistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	43	12	study	study	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	43	15	internal	internal	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	43	16	structure	structure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	18	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	43	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	43	20	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	43	21	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	43	22	de-	de-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	43	23	scribe	scribe	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	43	24	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	43	25	relation	relation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	26	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	27	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	43	28	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	43	29	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	30	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	43	31	lexicon	lexicon	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	32	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	43	33	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	43	34	decompose	decompose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	43	35	them	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	43	36	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	37	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	43	38	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	43	39	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	43	40	smallest	small	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	43	41	linguistic	linguistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	43	42	unit	unit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	43	bearing	bear	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	43	44	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	43	45	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	44	1	Morphol-	morphol-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	44	2	ogy	ogy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	44	3	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	44	4	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	44	5	divided	divide	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	44	6	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	44	7	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	44	8	types	type	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	44	9	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	44	10	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	44	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	44	12	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	44	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	45	1	Inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	45	2	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	45	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	45	5	set	set	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	7	processes	process	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	45	8	through	through	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	9	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	45	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	45	11	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	12	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	13	outwardly	outwardly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	45	14	displays	display	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	45	15	syntactic	syntactic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	45	16	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	45	18	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	45	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	45	20	verb	verb	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	45	21	tense	tense	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	45	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	46	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	46	2	follows	follow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	46	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	46	4	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	46	5	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	46	6	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	46	7	typically	typically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	46	8	neither	neither	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	46	9	changes	change	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	46	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	46	11	part	part	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	46	12	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	46	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	46	14	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	46	15	speech	speech	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	46	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	46	17	POS	POS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	46	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	46	19	nor	nor	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	46	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	46	21	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	46	22	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	46	23	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	46	24	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	46	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	47	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	47	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	47	5	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	47	6	verb	verb	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	47	8	run	run	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	47	9	takes	take	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	47	10	various	various	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	47	11	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	47	12	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	47	13	run	run	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	47	15	runs	run	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	47	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	47	17	ran	run	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	47	18	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	47	19	running	run	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	47	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	47	21	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	47	22	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	23	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	47	24	convey	convey	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	47	25	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	47	26	moving	move	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	47	27	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	28	foot	foot	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	29	quickly	quickly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	47	30	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	47	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	47	32	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	47	33	ap-	ap-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	47	34	pear	pear	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	35	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	36	complementary	complementary	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	47	37	syntactic	syntactic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	47	38	contexts	contexts	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	47	39	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	48	1	Derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	2	deals	deal	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	48	3	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	48	5	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	7	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	48	8	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	48	9	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	48	10	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	48	11	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	48	12	shifts	shift	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	48	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	14	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	48	16	often	often	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	48	17	includ-	includ-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	48	18	ing	e	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	48	19	POS	POS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	48	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	48	21	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	48	22	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	48	23	tightly	tightly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	48	24	intertwined	intertwine	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	25	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	26	lexical	lexical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	48	27	se-	se-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	28	mantics	mantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	48	29	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	48	30	Light	light	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	48	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	48	32	1996	1996	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	48	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	48	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	49	1	Consider	consider	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	49	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	49	3	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	49	6	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	49	7	noun	noun	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	8	discontentedness	discontentedness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	49	10	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	49	11	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	49	12	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	13	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	49	15	adjective	adjective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	16	discontented	discontent	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	49	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	50	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	50	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	3	true	true	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	4	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	5	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	50	6	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	50	7	share	share	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	50	8	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	50	9	close	close	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	10	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	11	relationship	relationship	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	50	13	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	50	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	50	15	transformation	transformation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	16	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	17	clearly	clearly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	50	18	more	more	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	50	19	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	20	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	50	21	simple	simple	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	22	inflec-	inflec-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	23	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	50	24	marking	marking	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	25	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	26	syntax	syntax	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	50	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	51	1	Indeed	indeed	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	51	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	51	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	51	4	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	51	5	go	go	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	51	6	one	one	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	51	7	step	step	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	51	8	further	far	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	51	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	51	10	define	define	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	51	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	51	12	chain	chain	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	51	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	51	14	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	51	15	content	content	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	51	16	7→	7→	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	51	17	contented	contented	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	51	18	7→	7→	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	51	19	discontented	discontent	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	51	20	7→	7→	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	51	21	discontentedness	discontentedness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	51	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	52	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	52	3	computational	computational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	52	4	literature	literature	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	52	6	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	52	7	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	8	phology	phology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	9	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	52	10	received	receive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	11	less	less	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	52	12	attention	attention	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	13	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	52	14	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	52	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	53	1	There	there	EX
work_cxyxl5vs6zd6xlanudcx46qoqy	53	2	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	53	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	53	4	however	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	53	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	53	6	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	53	7	bodies	body	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	53	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	53	9	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	53	10	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	53	11	deriva-	deriva-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	53	12	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	53	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	53	14	computational	computational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	53	15	linguistics	linguistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	53	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	54	1	First	first	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	54	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	54	3	there	there	EX
work_cxyxl5vs6zd6xlanudcx46qoqy	54	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	54	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	54	6	series	series	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	54	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	54	8	papers	paper	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	54	9	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	54	10	explore	explore	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	54	12	relation	relation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	54	13	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	54	14	lexical	lexical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	54	15	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	54	16	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	54	17	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	54	18	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	54	19	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	20	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	54	21	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	22	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	54	24	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	54	25	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	54	26	Zeller	Zeller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	27	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	54	28	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	29	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	54	31	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	54	32	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	54	33	Padó	Padó	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	34	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	35	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	36	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	37	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	54	38	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	54	39	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	54	40	Kisse-	Kisse-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	41	lew	lew	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	42	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	43	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	44	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	54	45	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	54	46	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	54	47	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	54	48	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	55	1	All	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	55	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	3	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	55	4	assume	assume	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	55	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	55	6	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	7	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	8	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	55	9	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	55	11	primarily	primarily	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	55	12	focus	focus	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	55	13	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	55	15	ef-	ef-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	55	16	fect	fect	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	18	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	19	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	55	20	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	55	21	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	55	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	56	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	2	second	second	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	3	body	body	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	5	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	7	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	10	unsupervised	unsupervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	11	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	12	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	13	segmenter	segmenter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	14	MORFESSOR	MORFESSOR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	56	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	56	16	Creutz	Creutz	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	56	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	56	18	La-	la-	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	56	19	gus	gus	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	21	2007	2007	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	56	22	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	56	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	24	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	25	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	26	deal	deal	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	27	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	28	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	56	29	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	56	30	makes	make	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	31	no	no	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	32	distinction	distinction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	33	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	34	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	35	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	56	36	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	37	morphology.3	morphology.3	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	38	Even	even	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	39	though	though	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	40	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	41	boundary	boundary	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	42	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	43	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	44	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	56	45	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	46	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	47	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	48	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	49	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	50	tinuum	tinuum	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	51	rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	52	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	53	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	54	rigid	rigid	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	55	divide	divide	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	56	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	56	57	Haspelmath	Haspelmath	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	56	58	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	56	59	Sims	Sims	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	56	60	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	61	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	56	62	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	56	63	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	56	64	there	there	EX
work_cxyxl5vs6zd6xlanudcx46qoqy	56	65	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	66	still	still	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	67	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	56	68	clear	clear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	69	distinction	distinction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	70	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	71	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	72	changes	change	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	73	meaning	mean	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	56	74	whereas	whereas	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	75	inflection	inflection	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	56	76	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	56	77	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	56	78	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	57	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	57	2	goal	goal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	4	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	57	5	paper	paper	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	57	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	57	8	develop	develop	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	57	9	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	57	10	account	account	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	12	how	how	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	57	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	57	14	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	16	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	57	17	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	18	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	19	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	57	20	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	57	21	computed	compute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	22	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	57	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	57	24	combining	combine	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	57	25	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	57	26	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	57	27	lines	line	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	57	28	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	29	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	57	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	58	1	Productivity	Productivity	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	58	2	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	58	3	Semantic	Semantic	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	58	4	Coherence	Coherence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	58	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	59	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	59	2	highlight	highlight	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	59	3	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	59	4	related	related	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	59	5	issues	issue	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	59	6	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	59	7	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	59	8	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	59	9	moti-	moti-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	59	10	vated	vate	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	59	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	59	12	development	development	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	59	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	59	14	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	59	15	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	59	16	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	59	17	productivity	productivity	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	59	18	3Narasimhan	3Narasimhan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	59	19	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	59	20	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	59	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	60	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	60	2	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	60	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	60	4	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	60	5	make	make	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	60	6	no	no	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	60	7	distinction	distinction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	8	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	9	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	60	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	60	11	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	60	12	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	60	14	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	60	15	their	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	60	16	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	17	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	60	18	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	60	19	exception	exception	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	20	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	21	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	22	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	60	23	includes	include	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	60	24	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	25	similarity	similarity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	26	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	60	28	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	60	29	fea-	fea-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	30	ture	ture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	60	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	61	1	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	61	2	§	§	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	61	3	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	61	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	61	5	discussion	discussion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	61	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	62	1	34	34	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	62	2	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	62	3	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	62	4	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	62	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	63	1	Roughly	roughly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	63	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	63	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	63	4	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	63	5	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	63	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	63	7	one	one	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	63	8	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	63	9	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	63	10	still	still	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	63	11	actively	actively	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	63	12	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	63	13	employed	employ	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	63	14	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	63	15	form	form	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	63	16	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	63	17	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	63	18	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	63	19	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	63	20	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	63	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	64	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	64	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	64	5	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	64	6	nominalizing	nominalizing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	7	affix	affix	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	64	8	ness	ness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	9	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	64	10	red7→red+ness	red7→red+ness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	64	12	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	64	13	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	64	14	attached	attach	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	15	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	16	just	just	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	64	17	about	about	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	64	18	any	any	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	64	19	adjective	adjective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	64	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	64	21	including	include	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	64	22	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	64	23	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	64	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	65	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	65	2	contrast	contrast	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	65	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	65	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	65	5	archaic	archaic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	65	6	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	65	7	nominal-	nominal-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	65	8	izing	ize	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	65	9	affix	affix	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	65	10	th	th	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	65	11	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	65	12	dear7→dear+th	dear7→dear+th	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	65	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	65	14	heal7→heal+th	heal7→heal+th	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	65	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	65	16	steal7→steal+th	steal7→steal+th	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	65	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	65	18	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	65	19	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	65	20	allow	allow	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	65	21	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	65	22	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	65	23	form	form	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	65	24	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	65	25	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	65	26	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	65	27	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	65	28	cheapth	cheapth	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	65	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	66	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	66	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	66	4	crucial	crucial	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	5	issue	issue	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	6	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	7	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	8	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	9	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	10	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	66	11	would	would	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	66	12	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	14	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	15	want	want	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	66	16	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	66	17	analyze	analyze	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	18	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	19	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	66	20	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	21	having	have	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	66	22	been	be	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	23	formed	form	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	24	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	25	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	26	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	27	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	66	28	endings	ending	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	66	29	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	66	30	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	66	32	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	66	33	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	66	34	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	35	want	want	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	36	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	66	37	analyze	analyze	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	38	hearth	hearth	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	66	39	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	40	hear+th	hear+th	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	66	41	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	66	42	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	66	43	wugth	wugth	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	44	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	66	45	wug+th	wug+th	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	66	46	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	66	47	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	67	1	Relations	relation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	67	2	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	67	3	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	4	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	67	5	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	6	heal	heal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	67	8	health	health	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	9	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	67	10	lexicalized	lexicalize	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	11	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	12	they	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	67	13	no	no	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	67	14	longer	longer	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	67	15	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	67	16	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	67	17	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	18	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	67	19	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	67	20	processes	process	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	67	21	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	67	22	Bauer	Bauer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	67	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	67	24	1983	1983	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	67	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	67	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	68	1	Under	under	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	2	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	68	3	generative	generative	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	68	4	treatment	treatment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	5	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	68	6	Chomsky	Chomsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	68	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	68	8	1965	1965	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	68	9	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	68	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	11	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	68	13	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	14	becomes	become	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	68	15	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	68	16	central	central	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	68	17	no-	no-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	68	18	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	19	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	20	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	68	21	grammar	grammar	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	22	needs	need	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	68	23	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	68	24	account	account	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	68	25	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	26	active	active	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	68	27	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	28	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	29	processes	process	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	68	30	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	68	32	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	68	33	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	68	34	Aronoff	Aronoff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	68	35	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	68	36	1976	1976	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	68	37	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	68	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	69	1	Defining	define	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	69	2	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	69	3	precisely	precisely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	69	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	69	5	however	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	69	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	69	7	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	69	8	tricky	tricky	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	69	9	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	69	10	Aronoff	Aronoff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	69	11	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	69	12	1976	1976	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	69	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	69	14	writes	write	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	69	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	69	16	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	69	17	one	one	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	69	18	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	69	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	69	20	central	central	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	69	21	mysteries	mystery	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	69	22	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	69	23	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	69	24	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	69	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	70	1	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	71	1	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	72	1	[	[	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	72	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	72	3	that	that	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	72	4	]	]	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	72	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	73	1	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	74	1	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	75	1	though	though	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	2	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	3	things	thing	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	75	4	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	75	5	possible	possible	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	6	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	7	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	75	9	some	some	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	75	10	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	75	11	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	75	12	possible	possible	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	13	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	14	others	other	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	75	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	75	16	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	75	17	Nevertheless	nevertheless	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	75	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	75	19	speakers	speaker	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	75	20	often	often	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	75	21	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	75	22	clear	clear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	23	intuitions	intuition	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	75	24	about	about	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	25	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	75	26	af-	af-	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	75	27	fixes	fix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	75	28	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	75	30	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	31	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	75	32	productive.4	productive.4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	75	33	Related	related	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	34	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	35	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	36	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	37	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	75	38	notion	notion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	39	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	40	seman-	seman-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	75	41	tic	tic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	75	42	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	75	43	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	76	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	76	2	principle	principle	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	4	compositionality	compositionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	5	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	76	6	Frege	Frege	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	76	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	76	8	1892	1892	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	76	9	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	76	10	Heim	Heim	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	76	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	76	12	Kratzer	Kratzer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	76	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	76	14	1998	1998	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	76	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	76	16	applies	apply	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	76	17	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	18	interpretation	interpretation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	19	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	20	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	76	21	just	just	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	76	22	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	23	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	76	24	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	76	25	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	76	26	phrases	phrase	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	76	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	77	1	In-	in-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	77	2	deed	deed	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	77	4	compositionality	compositionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	77	6	often	often	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	77	7	taken	take	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	8	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	77	9	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	77	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	77	11	sig-	sig-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	12	nal	nal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	13	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	14	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	77	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	77	16	Aronoff	Aronoff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	77	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	77	18	1976	1976	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	77	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	77	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	78	1	When	when	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	78	2	de-	de-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	78	3	ciding	cide	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	78	4	whether	whether	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	78	5	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	78	6	further	further	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	78	7	decompose	decompose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	78	8	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	78	9	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	78	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	78	11	ask-	ask-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	78	12	ing	e	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	78	13	whether	whether	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	78	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	78	15	parts	part	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	78	16	sum	sum	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	78	17	up	up	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	78	18	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	78	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	78	20	whole	whole	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	78	21	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	78	22	of-	of-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	78	23	ten	ten	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	78	24	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	78	25	good	good	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	78	26	indicator	indicator	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	78	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	79	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	3	case	case	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	5	questionably	questionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	79	6	7→	7→	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	79	7	question+able+ly	question+able+ly	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	79	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	79	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	10	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	79	11	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	79	13	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	79	14	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	15	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	16	manner	manner	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	17	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	18	could	could	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	79	19	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	79	20	questioned	question	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	21	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	79	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	79	23	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	24	corresponds	correspond	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	79	25	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	26	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	27	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	28	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	79	30	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	79	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	80	1	Contrast	contrast	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	80	2	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	80	3	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	80	5	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	6	unquiet	unquiet	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	80	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	80	8	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	80	9	means	mean	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	80	10	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	80	11	restless	restless	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	80	12	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	80	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	80	14	rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	80	15	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	16	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	80	17	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	80	18	quiet	quiet	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	80	19	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	80	20	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	80	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	80	22	compound	compound	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	23	blackmail	blackmail	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	80	25	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	80	26	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	80	27	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	80	28	refer	refer	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	80	29	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	30	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	80	31	letter	letter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	32	written	write	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	33	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	34	black	black	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	80	35	ink	ink	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	80	36	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	81	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	81	2	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	81	4	will	will	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	81	5	describe	describe	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	81	6	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	7	§	§	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	81	8	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	81	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	81	11	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	12	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	14	both	both	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	81	15	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	16	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	81	18	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	19	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	81	20	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	21	4It	4it	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	81	22	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	23	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	81	24	important	important	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	25	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	81	26	distinguish	distinguish	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	81	27	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	28	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	29	creativity	creativity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	30	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	81	31	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	81	32	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	33	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	81	34	rule	rule	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	35	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	81	36	governed	govern	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	37	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	38	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	39	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	40	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	81	41	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	81	42	Lyons	Lyons	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	81	43	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	81	44	1977	1977	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	81	45	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	81	46	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	82	1	As	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	2	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	82	3	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	5	creativity	creativity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	82	7	consider	consider	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	82	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	82	9	cre-	cre-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	10	ation	ation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	12	portmanteaux	portmanteaux	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	82	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	82	14	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	82	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	82	16	dramedy	dramedy	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	82	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	82	18	soundscape	soundscape	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	82	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	83	1	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	83	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	83	3	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	83	4	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	83	6	judged	judge	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	7	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	83	8	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	83	9	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	10	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	11	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	83	12	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	13	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	83	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	83	15	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	83	16	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	83	17	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	18	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	83	19	degree	degree	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	20	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	83	21	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	83	22	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	83	23	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	83	24	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	83	25	semantically	semantically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	83	26	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	83	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	84	1	Implicit	implicit	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	84	2	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	3	such	such	PDT
work_cxyxl5vs6zd6xlanudcx46qoqy	84	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	84	5	treatment	treatment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	84	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	84	8	desire	desire	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	9	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	84	10	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	84	11	segment	segment	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	84	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	84	13	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	14	if	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	84	16	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	17	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	84	18	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	19	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	20	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	84	21	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	84	22	process	process	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	84	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	85	1	While	while	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	2	most	most	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	85	3	prior	prior	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	4	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	5	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	6	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	7	segmen-	segmen-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	8	tation	tation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	9	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	10	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	85	11	explicitly	explicitly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	85	12	modeled	model	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	13	productivity,5	productivity,5	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	14	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	85	15	believe	believe	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	85	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	85	17	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	18	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	85	19	computational	computational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	20	modeling	modeling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	21	perspective	perspective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	85	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	85	23	segmenting	segment	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	85	24	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	85	25	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	26	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	85	27	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	28	preferable	preferable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	85	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	86	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	86	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	86	3	analogous	analogous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	86	4	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	86	6	modeling	modeling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	8	phrase	phrase	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	9	compo-	compo-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	10	sitionality	sitionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	11	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	12	embedding	embed	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	86	13	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	86	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	86	15	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	16	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	86	17	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	86	18	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	19	bet-	bet-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	86	20	ter	ter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	21	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	22	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	23	further	further	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	24	decompose	decompose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	25	noncompositional	noncompositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	86	26	mul-	mul-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	86	27	tiword	tiword	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	28	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	86	29	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	30	named	name	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	86	31	entities	entity	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	86	32	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	86	33	idiomatic	idiomatic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	86	34	ex-	ex-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	86	35	pressions	pression	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	86	36	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	86	37	see	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	38	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	86	39	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	86	40	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	86	41	Mikolov	Mikolov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	86	42	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	86	43	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	86	44	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	87	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	87	2	2013b	2013b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	87	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	87	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	87	5	Wang	Wang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	87	6	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	87	7	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	87	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	88	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	2	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	88	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	88	5	Yin	Yin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	88	7	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	9	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	88	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	88	12	Yaghoobzadeh	Yaghoobzadeh	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	88	14	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	16	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	88	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	88	19	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	88	20	Hashimoto	Hashimoto	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	21	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	88	22	Tsuruoka	Tsuruoka	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	23	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	88	24	2016).6	2016).6	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	88	25	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	26	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	88	27	paper	paper	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	88	29	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	30	refer	refer	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	88	31	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	32	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	88	33	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	88	34	aspect	aspect	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	35	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	36	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	88	37	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	38	either	either	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	88	39	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	40	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	88	41	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	42	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	88	43	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	44	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	88	45	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	89	1	These	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	89	2	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	89	3	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	89	4	ways	way	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	89	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	89	6	looking	look	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	89	7	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	89	8	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	89	9	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	89	10	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	89	11	related	relate	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	89	12	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	89	13	follows	follow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	89	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	90	1	If	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	3	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	4	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	90	5	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	90	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	90	7	composi-	composi-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	90	8	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	9	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	90	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	12	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	15	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	16	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	17	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	18	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	19	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	20	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	21	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	90	22	parts	part	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	90	23	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	90	24	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	25	regular	regular	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	90	26	application	application	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	27	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	28	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	29	linguistic	linguistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	90	30	rules	rule	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	90	31	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	32	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	90	34	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	90	35	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	90	36	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	37	so	so	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	90	38	constructed	construct	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	90	39	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	90	40	coherent	coherent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	90	41	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	91	1	These	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	91	2	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	91	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	91	4	cases	case	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	91	5	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	91	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	91	7	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	91	8	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	91	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	91	10	expected	expect	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	91	11	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	91	12	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	91	13	beneficial	beneficial	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	91	14	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	91	15	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	91	16	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	91	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	91	18	interpretation	interpretation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	91	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	92	1	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	92	2	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	92	3	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	4	Model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	5	From	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	6	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	92	7	NLP	NLP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	8	perspective	perspective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	92	10	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	92	11	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	92	13	Naradowsky	Naradowsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	92	15	Goldwater	Goldwater	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	92	17	2009	2009	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	92	18	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	92	19	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	20	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	21	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	22	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	92	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	92	24	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	92	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	92	26	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	92	27	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	92	28	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	29	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	92	30	seeks	seek	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	92	31	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	92	32	algorithmically	algorithmically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	92	33	de-	de-	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	92	34	compose	compose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	92	35	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	92	36	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	37	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	38	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	92	39	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	92	40	sequence	sequence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	41	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	42	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	92	43	phemes	pheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	92	44	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	93	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	93	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	93	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	93	4	version	version	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	6	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	93	7	segmenta-	segmenta-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	93	8	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	9	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	93	10	requires	require	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	93	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	93	12	learner	learner	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	13	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	93	14	handle	handle	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	93	15	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	93	16	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	93	17	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	93	18	take	take	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	93	19	place	place	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	20	during	during	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	21	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	22	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	93	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	94	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	94	2	believe	believe	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	94	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	94	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	94	6	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	94	7	natural	natural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	8	formulation	formulation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	10	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	11	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	12	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	13	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	94	14	especially	especially	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	94	15	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	94	17	processing	processing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	18	5Note	5note	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	94	19	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	20	segmenters	segmenter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	94	21	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	22	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	23	MORFESSOR	MORFESSOR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	94	24	utilize	utilize	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	94	25	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	94	26	prin-	prin-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	27	ciple	ciple	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	28	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	29	minimum	minimum	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	30	description	description	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	31	length	length	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	94	33	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	94	34	implicitly	implicitly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	94	35	encodes	encode	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	94	36	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	37	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	94	38	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	39	order	order	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	40	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	94	41	guide	guide	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	94	42	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	94	43	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	95	1	6As	6as	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	95	2	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	95	3	reviewer	reviewer	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	4	points	point	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	95	5	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	95	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	95	7	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	9	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	95	10	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	95	12	se-	se-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	13	mantic	mantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	95	14	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	95	17	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	95	18	formed	form	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	19	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	95	20	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	95	21	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	95	22	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	95	23	perfectly	perfectly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	95	24	aligned	aligned	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	95	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	96	1	Nonproductive	nonproductive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	96	2	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	96	3	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	96	4	produce	produce	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	96	5	semantically	semantically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	96	6	coher-	coher-	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	96	7	ent	ent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	96	8	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	96	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	96	10	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	96	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	96	12	warm	warm	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	96	13	7→warm+th	7→warm+th	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	96	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	97	1	Productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	97	2	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	97	3	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	97	4	pro-	pro-	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	97	5	duce	duce	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	97	6	semantically	semantically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	97	7	incoherent	incoherent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	97	8	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	97	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	97	10	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	97	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	97	12	canny	canny	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	97	13	7→un+canny	7→un+canny	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	97	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	98	1	Again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	98	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	98	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	98	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	98	5	analogous	analogous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	98	6	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	98	7	multiword	multiword	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	98	8	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	98	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	99	1	However	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	99	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	99	3	there	there	EX
work_cxyxl5vs6zd6xlanudcx46qoqy	99	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	99	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	99	6	strong	strong	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	99	7	correlation	correlation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	99	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	99	9	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	99	10	experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	99	11	show	show	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	99	12	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	99	13	relying	rely	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	99	14	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	99	15	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	99	16	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	99	17	good	good	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	99	18	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	99	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	100	1	35	35	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	100	2	un	un	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	3	question	question	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	4	able	able	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	100	5	ly	ly	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	100	6	su	su	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	7	ffi	ffi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	8	x	x	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	9	su	su	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	10	ffi	ffi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	11	x	x	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	12	st	st	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	13	em	-PRON-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	14	pr	pr	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	15	ef	ef	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	16	ix	ix	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	17	unquestionably	unquestionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	100	18	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	100	19	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	100	20	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	100	21	unquestionablely	unquestionablely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	100	22	⇡	⇡	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	100	23	surface	surface	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	24	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	25	underlying	underlie	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	100	26	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	27	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	28	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	29	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	30	ta	ta	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	31	rg	rg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	32	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	33	Figure	Figure	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	100	34	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	100	35	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	100	36	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	100	37	depiction	depiction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	38	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	39	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	100	40	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	100	41	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	42	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	100	43	makes	make	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	100	44	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	100	45	relation	relation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	46	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	47	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	100	48	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	100	49	factors	factor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	100	50	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	100	51	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	100	52	observed	observed	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	100	53	sur-	sur-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	54	face	face	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	55	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	100	56	explicit	explicit	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	100	57	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	101	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	101	2	show	show	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	101	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	101	4	simple	simple	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	101	5	additive	additive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	101	6	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	8	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	9	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	10	ease	ease	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	12	explication	explication	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	101	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	102	1	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	102	2	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	102	3	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	102	4	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	102	5	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	102	6	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	102	7	draws	draw	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	102	8	heavily	heavily	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	102	9	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	102	10	linguistic	linguistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	102	11	notions	notion	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	102	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	102	13	see	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	102	14	§	§	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	102	15	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	102	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	102	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	103	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	103	2	main	main	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	103	3	innovation	innovation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	103	5	present	present	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	103	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	103	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	103	8	augmen-	augmen-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	103	9	tation	tation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	11	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	103	12	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	13	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	103	14	take	take	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	103	15	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	16	ac-	ac-	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	103	17	count	count	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	103	18	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	103	19	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	20	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	103	21	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	103	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	104	1	Con-	Con-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	104	2	sider	sider	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	104	4	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	5	hypercuriosity	hypercuriosity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	104	7	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	104	8	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	104	9	seg-	seg-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	10	mentation	mentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	11	hyper+curious+ity	hyper+curious+ity	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	104	12	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	104	13	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	104	14	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	104	15	seg-	seg-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	16	mentation	mentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	17	seeks	seek	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	104	18	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	104	19	decompose	decompose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	104	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	104	21	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	22	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	23	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	104	24	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	25	stituent	stituent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	26	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	104	27	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	104	28	account	account	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	104	29	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	104	30	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	104	31	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	104	32	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	105	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	105	2	amounts	amount	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	105	3	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	105	5	structural	structural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	105	6	decomposi-	decomposi-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	7	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	105	10	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	105	12	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	105	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	105	14	how	how	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	105	15	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	105	16	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	105	17	break	break	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	105	18	up	up	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	105	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	105	20	string	string	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	21	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	22	characters	character	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	105	23	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	105	24	chunks	chunk	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	105	25	?	?	.
work_cxyxl5vs6zd6xlanudcx46qoqy	106	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	106	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	106	3	similar	similar	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	106	4	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	106	6	de-	de-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	106	7	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	106	10	sentence	sentence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	11	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	106	13	parse	parse	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	14	tree	tree	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	106	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	107	1	How-	How-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	107	2	ever	ever	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	107	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	107	4	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	107	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	107	6	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	107	7	natural	natural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	107	8	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	107	9	consider	consider	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	107	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	107	11	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	107	12	com-	com-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	13	positionality	positionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	15	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	107	16	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	107	18	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	107	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	107	20	how	how	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	107	21	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	107	22	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	107	23	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	24	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	25	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	107	26	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	27	synthesized	synthesize	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	28	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	107	30	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	31	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	107	32	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	107	33	indi-	indi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	107	34	vidual	vidual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	107	35	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	107	36	?	?	.
work_cxyxl5vs6zd6xlanudcx46qoqy	108	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	108	2	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	108	3	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	108	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	108	5	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	108	6	questions	question	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	108	7	together	together	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	108	8	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	108	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	108	10	single	single	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	108	11	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	108	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	108	13	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	108	14	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	108	15	would	would	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	108	16	like	like	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	108	17	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	108	18	place	place	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	108	19	high	high	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	108	20	probability	probability	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	108	21	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	108	22	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	108	23	segmentations	segmentation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	108	24	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	108	25	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	108	26	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	108	27	semantically	semantically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	108	28	coherent	coherent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	108	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	109	1	Returning	return	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	109	2	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	3	hy-	hy-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	109	4	percuriosity	percuriosity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	109	6	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	109	7	could	could	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	109	8	further	further	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	109	9	decompose	decompose	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	109	10	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	109	11	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	12	hyper+cure+ous+ity	hyper+cure+ous+ity	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	109	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	14	analogy	analogy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	15	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	109	17	say	say	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	109	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	109	19	vice	vice	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	109	20	7→	7→	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	109	21	vicious	vicious	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	109	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	110	1	Nothing	nothing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	2	about	about	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	110	4	surface	surface	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	5	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	7	curi-	curi-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	8	ous	ous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	110	9	alone	alone	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	110	10	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	110	11	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	110	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	110	13	strong	strong	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	110	14	cue	cue	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	15	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	110	16	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	110	17	should	should	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	110	18	rule	rule	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	110	19	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	110	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	110	21	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	110	22	cure+ous	cure+ous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	110	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	111	1	Turning	turn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	111	2	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	3	distri-	distri-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	111	4	butional	butional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	111	5	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	111	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	111	7	however	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	111	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	111	9	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	111	10	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	111	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	111	12	case	case	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	13	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	111	15	contexts	contexts	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	16	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	17	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	111	18	curious	curious	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	111	19	occurs	occur	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	111	20	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	111	21	quite	quite	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	111	22	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	111	23	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	24	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	111	25	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	26	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	111	27	cure	cure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	111	28	occurs	occur	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	111	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	112	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	112	2	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	112	3	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	112	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	112	5	strong	strong	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	112	6	cue	cue	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	112	7	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	112	8	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	112	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	112	10	correct	correct	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	112	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	113	1	Formally	formally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	113	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	3	given	give	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	5	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	6	string	string	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	7	w	w	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	8	∈	∈	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	9	Σ∗	Σ∗	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	113	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	11	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	113	12	Σ	Σ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	13	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	113	14	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	15	discrete	discrete	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	113	16	alphabet	alphabet	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	18	characters	character	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	113	19	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	113	20	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	21	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	22	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	23	could	could	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	24	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	113	25	as	as	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	113	26	simple	simple	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	113	27	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	28	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	29	26	26	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	30	letter	letter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	31	lowercase	lowercase	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	32	alpha-	alpha-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	33	bet	bet	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	34	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	113	35	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	36	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	113	37	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	38	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	39	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	40	v	v	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	41	∈	∈	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	42	V	V	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	43	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	44	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	113	45	V	v	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	46	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	113	47	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	48	set	set	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	49	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	50	low	low	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	113	51	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	113	52	dimensional	dimensional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	113	53	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	54	embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	113	55	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	56	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	57	define	define	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	58	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	113	59	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	60	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	61	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	113	62	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	63	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	64	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	65	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	66	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	67	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	68	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	69	|w	|w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	70	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	113	71	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	72	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	73	Zθ(w	Zθ(w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	74	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	113	75	exp	exp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	76	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	113	77	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	78	2σ2	2σ2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	79	||v	||v	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	80	−Cβ(s	−cβ(s	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	113	81	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	82	l)||22	l)||22	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	83	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	113	84	f(s	f(s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	85	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	86	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	87	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	88	u)>η	u)>η	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	113	89	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	113	90	g(u	g(u	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	113	91	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	113	92	w)>ω	w)>ω	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	113	93	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	113	94	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	114	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	2	1	1	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	114	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	4	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	5	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	7	composed	compose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	9	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	114	10	factors	factor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	114	11	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	114	12	composi-	composi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	13	tion	tion	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	14	factor	factor	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	16	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	114	17	2σ2	2σ2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	114	18	||v−Cβ(s	||v−Cβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	20	l)||22	l)||22	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	23	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	24	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	25	f	f	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	26	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	114	27	transduction	transduction	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	28	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	29	g.	g.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	114	30	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	31	parameters	parameter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	114	32	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	33	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	34	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	35	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	36	θ	θ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	37	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	114	38	{	{	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	39	β	β	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	40	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	41	η	η	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	42	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	43	ω	ω	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	44	}	}	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	45	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	46	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	47	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	48	Cβ	Cβ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	49	composes	compose	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	50	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	51	vectors	vectors	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	52	together	together	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	114	53	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	54	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	55	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	56	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	57	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	58	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	59	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	60	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	61	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	62	labeling	labeling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	63	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	64	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	65	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	114	66	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	114	67	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	68	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	69	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	70	underlying	underlying	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	71	representation	representation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	72	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	114	73	Zθ(w	Zθ(w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	114	74	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	114	75	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	114	76	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	114	77	partition	partition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	78	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	114	79	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	115	1	Note	note	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	115	2	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	115	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	115	4	conditional	conditional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	115	5	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	115	6	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	7	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	8	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	115	10	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	115	12	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	115	14	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	115	16	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	115	17	Gaussian	Gaussian	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	115	18	distributed	distribute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	115	19	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	115	20	construction	construction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	115	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	116	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	116	2	visualiza-	visualiza-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	116	3	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	116	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	116	5	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	116	6	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	116	7	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	116	8	found	find	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	116	9	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	116	10	Figure	Figure	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	116	11	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	116	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	117	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	117	2	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	117	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	117	5	conditional	conditional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	6	random	random	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	7	field	field	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	117	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	117	9	CRF	CRF	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	117	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	117	11	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	117	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	13	mixed	mixed	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	117	15	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	117	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	117	17	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	117	18	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	19	defined	define	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	117	20	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	117	21	both	both	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	117	22	discrete	discrete	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	23	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	117	24	continuous	continuous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	25	random	random	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	117	26	variables	variable	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	117	27	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	117	28	Koller	Koller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	117	29	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	117	30	Friedman	Friedman	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	117	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	117	32	2009	2009	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	117	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	117	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	118	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	118	2	restrict	restrict	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	118	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	118	4	range	range	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	118	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	118	6	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	118	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	118	8	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	118	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	118	10	subset	subset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	118	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	118	12	Σ|w|+k	Σ|w|+k	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	118	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	118	14	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	118	15	k	k	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	118	16	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	118	17	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	118	18	insertion	insertion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	118	19	limit	limit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	118	20	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	118	21	Dreyer	Dreyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	118	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	118	23	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	118	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	118	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	119	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	119	2	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	119	3	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	119	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	119	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	119	6	take	take	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	119	7	k	k	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	119	8	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	119	9	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	119	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	120	1	Explicitly	explicitly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	120	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	120	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	120	4	partition	partition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	5	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	7	defined	define	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	8	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	9	Zθ(w	Zθ(w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	11	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	12	∫	∫	:
work_cxyxl5vs6zd6xlanudcx46qoqy	120	13	∑	∑	.
work_cxyxl5vs6zd6xlanudcx46qoqy	120	14	l′,s′,u′	l′,s′,u′	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	15	exp	exp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	17	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	120	18	2σ2	2σ2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	120	19	||v′	||v′	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	120	20	−Cβ(s′	−Cβ(s′	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	120	22	l′)||22	l′)||22	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	23	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	120	24	f(s′	f(s′	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	120	26	l′,u′)>η	l′,u′)>η	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	27	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	120	28	g(u′,w)>ω	g(u′,w)>ω	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	29	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	30	dv′	dv′	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	120	32	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	33	2	2	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	120	34	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	35	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	120	36	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	37	guaranteed	guarantee	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	38	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	120	39	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	120	40	finite.7	finite.7	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	120	41	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	120	42	CRF	CRF	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	43	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	44	simply	simply	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	120	45	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	120	46	globally	globally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	120	47	renormalized	renormalize	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	48	prod-	prod-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	49	uct	uct	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	50	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	120	51	several	several	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	52	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	53	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	54	negative	negative	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	120	55	factors	factor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	120	56	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	57	Sutton	Sutton	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	58	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	120	59	McCallum	McCallum	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	120	60	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	120	61	2006	2006	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	120	62	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	120	63	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	121	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	121	2	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	121	4	composed	compose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	6	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	121	7	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	121	8	transduction	transduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	121	10	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	121	12	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	13	factors	factor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	121	14	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	121	15	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	121	16	describe	describe	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	121	17	each	each	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	121	18	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	19	turn	turn	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	121	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	122	1	3.1	3.1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	122	2	Transduction	Transduction	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	3	Factor	Factor	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	4	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	5	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	6	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	7	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	8	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	11	transduction	transduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	12	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	13	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	122	14	exp	exp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	122	16	g(u	g(u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	122	18	w)>ω	w)>ω	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	122	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	122	21	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	22	scores	score	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	23	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	24	surface	surface	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	25	7Since	7Since	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	26	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	27	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	28	capped	cap	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	30	insertion	insertion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	31	limit	limit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	122	33	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	34	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	35	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	36	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	37	number	number	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	38	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	39	values	value	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	122	40	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	41	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	42	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	122	43	take	take	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	122	44	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	45	any	any	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	46	w.	w.	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	47	Thus	thus	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	122	48	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	122	49	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	50	follows	follow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	51	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	52	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	53	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	54	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	55	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	56	number	number	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	57	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	58	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	59	segmentations	segmentation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	122	60	s.	s.	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	61	Hence	hence	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	122	62	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	63	take	take	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	122	64	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	122	65	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	66	number	number	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	67	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	122	68	Gaussian	gaussian	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	122	69	integrals	integral	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	122	70	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	123	1	These	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	123	2	integrals	integral	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	123	3	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	123	4	converge	converge	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	123	5	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	123	6	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	123	7	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	123	8	fixed	fix	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	123	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	123	10	covariance	covariance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	123	11	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	123	12	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	123	13	σ2I	σ2I	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	123	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	123	15	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	123	16	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	123	17	positive	positive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	123	18	definite	definite	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	123	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	124	1	36	36	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	124	2	representation	representation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	3	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	124	4	SR	SR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	124	5	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	124	6	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	124	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	124	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	124	9	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	10	string	string	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	11	ob-	ob-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	124	12	served	serve	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	124	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	14	raw	raw	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	124	15	text	text	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	124	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	124	18	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	124	19	underlying	underlying	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	124	20	represen-	represen-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	124	21	tation	tation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	22	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	124	23	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	124	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	124	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	124	26	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	124	27	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	28	string	string	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	29	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	30	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	124	31	processes	process	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	124	32	reversed	reverse	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	124	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	125	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	125	2	aim	aim	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	4	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	125	5	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	125	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	125	8	place	place	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	9	high	high	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	125	10	weight	weight	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	11	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	12	good	good	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	125	13	pairs	pair	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	125	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	125	15	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	125	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	125	18	pair	pair	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	19	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	125	20	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	125	21	=	=	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	125	22	questionably	questionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	125	24	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	125	25	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	125	26	questionablely	questionablely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	27	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	125	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	125	29	so	so	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	30	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	125	31	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	125	32	ac-	ac-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	33	curately	curately	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	34	restore	restore	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	125	35	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	36	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	125	37	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	125	38	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	125	39	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	126	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	126	2	encode	encode	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	126	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	126	4	portion	portion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	126	7	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	8	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	126	10	weighted	weighted	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	126	11	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	126	12	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	126	13	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	14	machine	machine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	15	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	16	ease	ease	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	18	computation	computation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	126	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	127	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	127	2	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	3	generalizes	generalize	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	127	4	probabilistic	probabilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	127	5	edit	edit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	6	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	127	8	Ristad	Ristad	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	127	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	127	10	Yianilos	Yianilos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	127	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	127	12	1998	1998	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	127	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	127	14	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	15	looking	look	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	127	16	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	17	additional	additional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	127	18	input	input	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	19	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	127	20	output	output	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	21	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	127	22	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	127	23	see	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	127	24	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	127	25	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	127	26	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	127	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	128	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	128	2	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	128	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	128	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	128	5	de-	de-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	128	6	tails	tail	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	128	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	129	1	As	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	129	2	mentioned	mention	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	129	3	above	above	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	129	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	129	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	129	6	contrast	contrast	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	129	7	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	129	8	Cot-	Cot-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	129	9	terell	terell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	129	10	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	129	11	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	129	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	130	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	130	2	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	130	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	130	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	130	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	130	6	bound	bind	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	130	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	130	8	insertion	insertion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	9	limit	limit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	130	12	edit	edit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	13	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	14	model.8	model.8	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	130	15	Computing	compute	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	130	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	130	17	score	score	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	18	be-	be-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	130	19	tween	tween	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	20	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	130	21	strings	string	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	130	22	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	130	23	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	130	24	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	130	25	requires	require	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	130	26	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	130	27	dynamic	dynamic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	130	28	pro-	pro-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	29	gram	gram	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	30	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	130	31	runs	run	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	130	32	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	130	33	O(|u|·|w|	O(|u|·|w|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	130	34	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	130	35	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	131	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	131	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	131	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	131	4	generalization	generalization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	131	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	131	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	131	7	forward	forward	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	131	8	algorithm	algorithm	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	131	9	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	131	10	Hidden	Hidden	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	131	11	Markov	Markov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	131	12	Models	Models	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	131	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	131	14	HMMs	HMMs	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	131	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	131	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	131	17	Rabiner	Rabiner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	131	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	131	19	1989	1989	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	131	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	131	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	132	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	2	employ	employ	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	3	standard	standard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	132	4	feature	feature	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	5	templates	template	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	132	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	132	8	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	9	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	132	10	look	look	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	11	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	12	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	132	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	14	edit	edit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	15	operations	operation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	132	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	132	17	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	132	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	132	19	substi-	substi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	20	tute	tute	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	21	i	i	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	22	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	23	y	y	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	132	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	132	25	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	26	varying	vary	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	132	27	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	132	28	granularities	granularity	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	132	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	133	1	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	133	2	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	133	3	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	133	4	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	133	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	134	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	134	2	2016b	2016b	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	134	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	134	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	134	5	details	detail	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	134	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	135	1	Recent	recent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	135	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	3	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	135	4	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	135	5	explored	explore	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	6	weighting	weighting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	8	WFST	WFST	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	9	arcs	arc	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	135	10	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	11	scores	score	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	135	12	computed	compute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	13	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	14	LSTMs	LSTMs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	135	16	Hochreiter	Hochreiter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	135	18	Schmidhuber	Schmidhuber	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	135	20	1997	1997	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	135	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	135	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	135	23	obviating	obviate	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	135	24	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	135	25	need	need	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	26	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	27	human	human	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	135	28	selection	selection	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	29	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	30	feature	feature	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	135	31	templates	template	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	135	32	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	135	33	Rastogi	Rastogi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	34	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	35	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	36	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	135	37	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	135	38	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	135	39	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	135	40	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	136	1	3.2	3.2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	136	2	Segmentation	Segmentation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	136	3	Factor	Factor	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	136	4	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	136	5	second	second	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	136	6	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	136	7	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	136	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	136	9	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	136	10	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	136	11	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	136	12	exp	exp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	136	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	136	14	f(s	f(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	136	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	136	16	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	136	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	136	18	u)>η	u)>η	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	136	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	136	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	137	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	137	2	goal	goal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	137	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	137	4	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	137	5	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	137	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	137	7	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	137	8	score	score	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	137	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	137	10	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	137	11	s	s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	137	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	137	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	137	14	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	137	15	u.	u.	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	138	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	138	2	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	138	3	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	138	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	138	5	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	138	6	scores	score	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	138	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	138	8	input	input	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	138	9	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	138	10	output	output	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	138	11	pair	pair	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	138	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	138	13	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	138	14	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	138	15	questionablely	questionablely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	138	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	138	17	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	138	18	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	138	19	question+able+ly	question+able+ly	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	138	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	138	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	139	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	139	2	additionally	additionally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	139	3	scores	score	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	139	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	139	5	labeling	labeling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	139	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	139	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	139	8	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	139	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	140	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	140	2	label	label	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	3	set	set	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	140	4	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	5	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	140	6	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	7	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	140	8	L	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	9	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	140	10	{	{	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	140	11	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	140	13	prefix	prefix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	140	15	suffix	suffix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	140	16	}	}	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	140	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	141	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	141	2	proper	proper	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	141	3	labeling	labeling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	141	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	141	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	141	6	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	141	7	above	above	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	141	8	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	141	9	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	141	10	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	141	11	question	question	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	141	12	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	141	13	stem+able	stem+able	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	141	14	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	141	15	suffix+ly	suffix+ly	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	141	16	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	141	17	suffix	suffix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	141	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	142	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	2	label-	label-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	3	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	142	5	critical	critical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	142	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	7	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	142	8	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	9	functions	function	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	142	10	Cβ	Cβ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	11	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	142	12	Cot-	Cot-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	13	terell	terell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	14	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	142	15	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	16	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	142	18	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	142	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	142	20	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	142	21	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	22	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	142	23	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	24	used	use	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	25	depends	depend	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	142	26	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	27	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	28	label	label	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	29	given	give	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	30	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	32	segment	segment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	33	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	142	34	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	142	35	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	142	36	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	37	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	142	38	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	39	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	40	prefix	prefix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	41	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	142	42	post	post	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	43	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	142	44	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	142	45	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	142	46	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	47	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	142	48	post	post	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	142	49	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	142	50	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	142	51	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	142	52	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	143	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	143	2	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	143	3	view	view	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	143	4	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	143	5	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	6	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	7	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	143	8	unnormalized	unnormalized	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	143	9	first-	first-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	10	8As	8as	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	143	11	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	143	12	transduction	transduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	13	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	14	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	143	15	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	143	16	unnormalized	unnormalized	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	143	17	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	18	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	19	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	143	20	CRF	crf	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	143	22	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	143	23	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	143	24	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	143	25	require	require	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	143	26	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	143	27	local	local	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	143	28	normalization	normalization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	29	discussed	discuss	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	30	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	143	31	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	143	32	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	143	33	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	143	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	144	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	144	2	2014)—a	2014)—a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	144	3	weight	weight	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	144	4	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	144	5	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	144	6	edge	edge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	144	7	may	may	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	144	8	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	144	9	any	any	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	144	10	non-	non-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	144	11	negative	negative	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	144	12	real	real	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	144	13	number	number	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	144	14	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	144	15	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	144	16	will	will	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	144	17	renormalize	renormalize	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	144	18	later	later	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	144	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	145	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	145	2	un-	un-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	145	3	derlying	derlying	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	145	4	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	145	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	145	6	however	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	145	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	145	8	remains	remain	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	145	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	145	10	same	same	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	145	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	146	1	model	model	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	2	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	3	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	4	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	5	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	6	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	7	∑N	∑N	.
work_cxyxl5vs6zd6xlanudcx46qoqy	146	8	i=1	i=1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	9	1li	1li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	10	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	11	stemm	stemm	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	12	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	13	si	si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	14	mult	mult	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	15	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	16	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	17	⊙N	⊙N	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	18	i=1	i=1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	19	m	m	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	20	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	21	si	si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	22	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	146	23	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	24	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	25	∑N	∑n	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	26	i=1	i=1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	27	m	m	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	28	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	29	si	si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	30	wadd	wadd	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	31	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	32	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	33	∑N	∑N	.
work_cxyxl5vs6zd6xlanudcx46qoqy	146	34	i=1	i=1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	35	αim	αim	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	36	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	37	si	si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	38	fulladd	fulladd	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	39	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	40	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	41	∑N	∑N	.
work_cxyxl5vs6zd6xlanudcx46qoqy	146	42	i=1	i=1	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	146	43	Uim	uim	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	146	44	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	45	si	si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	46	LDS	LDS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	47	hi	hi	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	146	48	=	=	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	146	49	Xhi−1	Xhi−1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	50	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	51	Umlisi	Umlisi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	52	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	53	hi	hi	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	146	54	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	146	55	tanh(Xhi−1	tanh(xhi−1	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	56	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	146	57	Umlisi	Umlisi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	58	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	146	59	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	60	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	146	61	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	146	62	Composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	63	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	146	64	Cβ(s	Cβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	146	65	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	146	66	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	67	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	146	68	used	use	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	69	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	70	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	146	71	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	146	72	prior	prior	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	146	73	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	146	74	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	147	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	147	2	representation	representation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	147	5	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	147	7	hN	hN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	147	8	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	147	10	dynamic	dynamic	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	147	12	c	c	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	13	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	147	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	147	15	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	147	16	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	147	17	dynamic	dynamic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	147	18	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	147	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	148	1	Note	note	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	148	2	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	148	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	148	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	148	5	dynamic	dynamic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	148	6	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	148	7	h0	h0	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	148	8	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	148	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	148	10	learned	learned	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	148	11	parameter	parameter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	148	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	149	1	order	order	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	149	2	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	149	3	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	149	4	CRF	crf	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	149	5	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	149	6	Sarawagi	Sarawagi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	149	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	149	8	Cohen	Cohen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	149	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	149	10	2005	2005	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	149	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	149	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	150	1	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	150	2	putation	putation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	150	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	150	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	150	5	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	150	6	again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	150	7	requires	require	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	150	8	dynamic	dynamic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	150	9	pro-	pro-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	150	10	gramming	gramming	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	150	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	151	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	151	2	algorithm	algorithm	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	151	5	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	6	generaliza-	generaliza-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	7	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	151	10	forward	forward	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	11	algorithm	algorithm	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	12	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	13	HMMs	hmm	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	151	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	151	15	one	one	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	16	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	151	17	extends	extend	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	18	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	151	19	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	151	21	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	22	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	23	Markov	markov	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	151	24	case	case	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	151	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	152	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	152	2	algorithm	algorithm	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	152	3	runs	run	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	152	4	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	152	5	O(|u|2	o(|u|2	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	152	6	·	·	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	152	7	|L|2	|L|2	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	152	8	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	152	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	153	1	Features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	153	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	154	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	154	2	again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	154	3	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	154	4	standard	standard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	154	5	feature	feature	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	154	6	templates	template	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	154	7	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	154	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	154	9	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	154	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	155	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	155	2	create	create	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	155	3	atomic	atomic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	155	4	indicator	indicator	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	155	5	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	155	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	155	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	155	8	individual	individual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	155	9	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	155	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	156	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	156	2	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	156	3	conjoin	conjoin	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	156	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	156	5	atomic	atomic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	156	6	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	156	7	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	156	8	left	left	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	156	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	156	10	right	right	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	156	11	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	156	12	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	156	13	as	as	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	156	14	well	well	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	156	15	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	156	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	156	17	label	label	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	156	18	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	156	19	create	create	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	156	20	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	156	21	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	156	22	feature	feature	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	156	23	templates	template	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	156	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	157	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	157	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	157	3	include	include	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	157	4	transition	transition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	157	5	features	feature	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	157	6	that	that	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	157	7	fire	fire	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	157	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	157	9	pairs	pair	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	157	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	157	11	sequential	sequential	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	157	12	labels	label	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	157	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	158	1	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	158	2	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	158	3	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	158	4	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	158	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	159	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	159	2	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	159	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	159	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	159	5	details	detail	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	159	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	160	1	Recent	recent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	160	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	3	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	160	4	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	160	5	showed	show	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	6	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	7	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	160	8	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	160	9	parameterization	parameterization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	10	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	160	11	remove	remove	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	160	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	160	13	need	need	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	14	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	15	manual	manual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	160	16	feature	feature	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	17	design	design	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	160	18	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	160	19	Kong	Kong	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	160	20	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	160	21	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	160	22	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	160	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	160	24	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	160	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	160	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	161	1	3.3	3.3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	161	2	Composition	Composition	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	161	3	Factor	Factor	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	161	4	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	5	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	6	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	7	takes	take	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	9	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	11	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	12	unnormalized	unnormalized	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	13	multivariate	multivariate	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	14	Gaussian	gaussian	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	15	density	density	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	16	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	161	17	exp	exp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	18	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	161	19	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	161	20	2σ2	2σ2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	161	21	||v	||v	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	161	22	−Cβ(s	−cβ(s	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	161	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	161	24	l)||22	l)||22	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	161	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	161	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	161	27	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	161	28	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	29	mean	mean	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	30	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	31	computed	compute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	32	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	33	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	34	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	161	35	potentially	potentially	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	161	36	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	37	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	38	linear	linear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	39	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	161	40	compo-	compo-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	41	sition	sition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	42	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	43	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	161	44	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	161	45	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	46	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	161	47	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	161	48	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	161	49	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	50	covariance	covariance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	51	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	52	σ2I	σ2I	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	161	53	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	54	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	161	55	diagonal	diagonal	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	161	56	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	161	57	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	162	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	162	2	goal	goal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	162	5	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	6	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	7	Cβ(s	Cβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	162	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	162	9	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	162	11	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	162	12	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	162	13	stitch	stitch	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	162	14	together	together	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	162	15	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	162	16	embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	162	17	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	162	18	approximate	approximate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	162	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	162	20	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	21	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	22	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	162	23	entire	entire	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	162	24	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	162	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	163	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	163	2	simplest	simple	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	163	3	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	163	6	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	7	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	8	Cβ(s	Cβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	163	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	163	10	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	163	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	163	13	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	163	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	163	15	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	163	16	additive	additive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	163	17	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	18	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	163	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	163	20	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	163	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	164	1	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	164	2	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	164	3	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	164	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	164	5	each	each	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	164	6	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	164	7	mlisi	mlisi	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	164	8	refers	refer	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	164	9	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	164	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	164	11	morpheme-	morpheme-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	164	12	37	37	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	164	13	specific	specific	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	164	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	164	15	label	label	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	164	16	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	164	17	dependent	dependent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	164	18	embedding	embedding	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	164	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	165	1	If	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	165	2	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	165	3	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	165	4	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	165	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	165	6	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	165	7	si	si	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	165	8	represents	represent	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	165	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	165	10	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	165	11	morpheme	morpheme	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	165	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	166	1	Given	give	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	2	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	3	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	166	4	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	166	6	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	166	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	166	8	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	166	9	si	si	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	10	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	166	11	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	166	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	166	13	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	14	gen-	gen-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	15	erally	erally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	166	16	itself	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	166	17	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	166	18	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	166	19	entry	entry	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	20	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	166	22	lexicon	lexicon	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	23	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	166	24	v(si	v(si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	166	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	166	26	∈	∈	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	166	27	V	v	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	166	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	167	1	If	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	2	v(si	v(si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	167	4	6∈	6∈	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	167	5	V	v	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	167	7	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	167	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	9	set	set	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	10	v(si	v(si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	167	12	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	13	0.9	0.9	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	167	14	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	15	optimize	optimize	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	16	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	17	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	167	18	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	19	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	20	∈	∈	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	21	{	{	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	167	22	prefix	prefix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	167	24	suffix	suffix	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	25	}	}	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	167	26	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	27	they	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	28	corre-	corre-	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	29	spond	spond	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	30	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	167	31	bound	bound	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	32	morphemes	morphemes	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	167	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	168	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	168	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	168	3	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	168	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	168	5	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	168	6	expressive	expressive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	168	7	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	168	8	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	168	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	168	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	168	11	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	168	12	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	168	13	network	network	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	168	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	168	15	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	168	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	168	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	169	1	Let	let	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	169	2	N	n	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	169	3	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	169	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	169	5	number	number	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	169	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	169	7	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	169	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	170	1	Then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	170	2	Cβ(s	Cβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	170	4	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	5	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	170	6	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	7	hN	hN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	8	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	170	9	hi	hi	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	170	10	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	170	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	170	12	hidden	hidden	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	170	13	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	170	15	defined	define	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	16	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	170	18	re-	re-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	170	19	cursion:10	cursion:10	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	20	hi	hi	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	170	21	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	22	tanh	tanh	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	23	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	170	24	Xhi−1	Xhi−1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	25	+	+	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	26	Um	um	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	170	27	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	28	si	si	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	170	29	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	170	30	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	170	31	Elman	Elman	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	170	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	170	33	1990	1990	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	170	34	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	170	35	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	171	1	Again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	171	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	171	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	4	optimize	optimize	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	171	6	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	7	embed-	embed-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	171	8	dings	ding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	171	9	mlisi	mlisi	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	171	10	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	171	11	when	when	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	171	12	li	li	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	13	6=	6=	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	171	14	stem	stem	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	15	along	along	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	171	16	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	171	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	171	18	other	other	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	171	19	parameters	parameter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	171	20	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	171	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	171	22	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	171	24	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	171	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	171	26	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	171	27	matrices	matrix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	171	28	U	U	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	171	29	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	171	30	X.	X.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	172	1	4	4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	172	2	Inference	Inference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	172	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	172	4	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	172	5	Exact	exact	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	172	6	inference	inference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	7	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	172	8	intractable	intractable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	172	9	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	10	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	172	11	allow	allow	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	172	12	ar-	ar-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	13	bitrary	bitrary	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	172	14	segment	segment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	15	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	172	16	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	17	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	172	18	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	172	20	canonicalized	canonicalize	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	21	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	172	22	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	172	23	u.	u.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	173	1	Since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	3	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	4	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	5	CRF	crf	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	6	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	7	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	8	fea-	fea-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	9	tures	ture	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	173	10	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	11	fire	fire	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	12	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	13	substrings	substring	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	173	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	173	15	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	173	16	would	would	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	173	17	need	need	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	173	18	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	19	dy-	dy-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	173	20	namic	namic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	21	programming	programming	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	22	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	23	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	24	each	each	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	25	substring	substring	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	26	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	27	each	each	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	28	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	30	exponentially	exponentially	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	173	31	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	32	settings	setting	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	173	33	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	34	u	u	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	173	35	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	173	36	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	37	breaks	break	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	38	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	173	39	dynamic	dynamic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	173	40	program	program	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	173	41	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	174	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	174	2	thus	thus	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	174	3	turn	turn	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	174	4	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	174	5	approximate	approximate	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	174	6	inference	inference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	174	7	through	through	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	174	8	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	174	9	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	174	10	sampling	sampling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	174	11	routine	routine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	174	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	174	13	Rubinstein	Rubinstein	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	174	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	174	15	Kroese	Kroese	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	174	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	174	17	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	174	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	174	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	175	1	4.1	4.1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	175	2	Inference	inference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	3	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	4	Importance	Importance	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	175	5	Sampling	Sampling	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	175	6	Rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	175	7	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	8	considering	consider	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	175	9	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	175	10	underlying	underlie	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	175	11	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	175	12	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	175	13	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	175	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	175	15	segmentations	segmentations	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	175	16	s	s	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	175	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	175	18	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	175	19	sample	sample	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	175	20	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	21	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	175	22	tractable	tractable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	175	23	proposal	proposal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	24	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	25	q	q	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	175	26	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	175	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	175	28	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	29	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	175	30	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	175	31	segmentations	segmentation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	175	32	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	176	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	176	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	176	3	following	following	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	176	4	equations	equation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	176	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	6	omit	omit	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	176	8	dependence	dependence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	176	9	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	176	10	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	176	12	notational	notational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	176	13	brevity	brevity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	176	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	176	15	define	define	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	176	16	h(l	h(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	176	18	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	176	20	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	176	22	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	176	23	f(s	f(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	176	25	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	176	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	176	27	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	28	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	176	29	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	176	30	g(u	g(u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	176	32	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	176	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	176	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	177	1	Cru-	Cru-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	2	cially	cially	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	177	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	177	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	177	5	partition	partition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	6	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	7	Zθ(w	Zθ(w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	8	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	177	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	177	10	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	177	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	177	12	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	14	parameter	parameter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	15	subvector	subvector	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	16	β	β	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	177	18	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	177	19	gradient	gradient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	20	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	21	re-	re-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	177	22	9This	9This	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	23	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	177	24	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	177	25	changed	change	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	26	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	27	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	177	29	so	so	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	30	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	177	31	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	177	32	v(si	v(si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	177	34	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	177	35	0	0	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	177	36	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	37	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	177	38	final	final	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	177	39	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	177	40	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	178	1	Clearly	clearly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	178	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	178	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	178	4	could	could	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	178	5	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	178	6	improved	improve	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	8	future	future	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	178	9	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	10	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	178	12	reviewer	reviewer	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	13	points	point	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	178	14	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	178	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	178	16	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	178	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	178	18	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	19	setting	set	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	178	20	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	178	21	v(si	v(si	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	178	22	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	178	23	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	24	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	178	25	average	average	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	26	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	178	28	suitable	suitable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	178	29	chosen	choose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	30	set	set	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	31	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	32	known	know	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	33	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	178	34	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	178	35	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	179	1	10We	10We	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	179	2	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	3	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	179	4	explore	explore	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	179	5	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	179	6	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	179	7	RNNs	rnn	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	179	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	179	9	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	179	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	179	11	LSTMs	LSTMs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	179	13	Hochreiter	Hochreiter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	179	15	Schmidhuber	Schmidhuber	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	179	17	1997	1997	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	179	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	179	19	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	179	20	GRUs	gru	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	179	21	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	179	22	Cho	Cho	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	23	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	24	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	25	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	179	27	2014a	2014a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	179	28	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	179	29	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	179	30	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	179	31	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	179	32	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	179	33	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	179	34	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	35	≤7	≤7	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	179	36	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	179	37	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	180	1	These	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	180	2	archi-	archi-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	3	tectures	tecture	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	180	4	make	make	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	180	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	180	6	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	8	long	long	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	180	9	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	10	dependencies	dependency	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	180	11	eas-	eas-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	180	12	ier	ier	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	180	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	180	14	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	180	15	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	180	16	no	no	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	180	17	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	180	18	powerful	powerful	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	180	19	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	20	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	180	21	Elman	elman	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	180	22	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	180	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	180	24	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	25	least	least	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	180	26	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	27	theory	theory	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	180	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	181	1	Note	note	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	181	2	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	3	perhaps	perhaps	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	181	4	if	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	5	applied	apply	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	6	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	7	languages	language	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	181	8	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	9	richer	rich	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	181	10	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	181	11	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	12	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	13	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	181	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	181	15	considering	consider	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	181	16	more	more	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	181	17	com-	com-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	18	plex	plex	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	19	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	181	20	architectures	architecture	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	181	21	would	would	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	181	22	make	make	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	181	23	sense	sense	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	181	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	182	1	spect	spect	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	182	2	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	3	β	β	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	182	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	182	5	0.11	0.11	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	182	6	Recall	Recall	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	182	7	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	8	computing	compute	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	182	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	182	10	gradi-	gradi-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	11	ent	ent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	182	14	log	log	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	15	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	182	16	partition	partition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	17	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	18	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	182	19	equivalent	equivalent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	182	20	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	182	22	problem	problem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	23	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	24	marginal	marginal	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	182	25	inference	inference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	182	26	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	182	27	Wainwright	Wainwright	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	182	28	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	182	29	Jor-	Jor-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	182	30	dan	dan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	182	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	182	32	2008	2008	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	182	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	182	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	183	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	2	derive	derive	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	3	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	183	4	estimator	estimator	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	5	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	6	follows	follow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	183	7	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	183	8	∇θ	∇θ	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	9	log	log	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	10	Z	z	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	11	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	183	12	E	e	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	14	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	16	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	18	u)∼p	u)∼p	.
work_cxyxl5vs6zd6xlanudcx46qoqy	183	19	[	[	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	20	h(l	h(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	22	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	24	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	26	]	]	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	27	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	28	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	183	29	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	30	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	31	∑	∑	.
work_cxyxl5vs6zd6xlanudcx46qoqy	183	32	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	34	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	35	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	36	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	37	p(l	p(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	38	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	39	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	40	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	41	u)h(l	u)h(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	42	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	43	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	44	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	45	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	46	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	47	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	48	4	4	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	183	49	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	50	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	51	∑	∑	.
work_cxyxl5vs6zd6xlanudcx46qoqy	183	52	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	53	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	54	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	55	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	56	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	57	q(l	q(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	58	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	59	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	60	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	61	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	62	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	63	q(l	q(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	64	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	65	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	66	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	67	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	68	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	69	p(l	p(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	70	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	71	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	72	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	73	u)h(l	u)h(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	74	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	75	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	76	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	77	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	78	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	79	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	80	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	183	81	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	82	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	83	E	e	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	84	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	85	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	86	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	87	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	88	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	89	u)∼q	u)∼q	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	90	[	[	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	91	p(l	p(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	92	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	93	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	94	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	95	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	96	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	97	q(l	q(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	98	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	99	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	100	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	101	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	102	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	103	h(l	h(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	104	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	105	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	106	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	107	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	108	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	109	]	]	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	110	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	183	111	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	112	6	6	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	183	113	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	114	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	183	115	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	116	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	117	omitted	omit	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	118	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	183	119	dependence	dependence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	120	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	121	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	122	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	123	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	183	124	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	125	condition	condition	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	126	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	183	127	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	128	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	183	129	v	v	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	183	130	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	131	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	183	132	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	133	marginalize	marginalize	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	134	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	183	135	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	183	136	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	184	1	So	so	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	184	2	long	long	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	184	3	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	184	4	q	q	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	5	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	184	6	support	support	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	184	7	everywhere	everywhere	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	184	8	p	p	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	184	9	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	184	10	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	184	11	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	184	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	184	13	p(l	p(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	184	15	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	184	17	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	184	19	>	>	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	184	20	0	0	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	184	21	⇒	⇒	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	184	22	q(l	q(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	184	24	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	184	26	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	27	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	184	28	>	>	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	184	29	0	0	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	184	30	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	184	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	184	32	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	184	33	estimate	estimate	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	184	34	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	184	35	un-	un-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	184	36	biased	biased	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	184	37	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	185	1	Unfortunately	unfortunately	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	185	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	4	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	185	5	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	185	6	efficiently	efficiently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	185	7	com-	com-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	185	8	pute	pute	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	185	9	p(l	p(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	11	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	13	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	185	15	up	up	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	185	16	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	185	17	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	185	18	constant	constant	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	185	19	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	185	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	21	p(l	p(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	23	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	25	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	26	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	185	27	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	28	p̄(l	p̄(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	30	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	185	32	u)/Z′θ(w	u)/Z′θ(w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	185	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	185	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	186	1	Thus	thus	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	186	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	186	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	4	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	186	6	indirect	indirect	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	7	impor-	impor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	8	tance	tance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	9	sampling	sample	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	186	10	estimator	estimator	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	186	12	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	186	13	∑M	∑M	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	14	i=1	i=1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	15	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	17	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	19	M∑	M∑	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	20	i=1	i=1	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	21	w(i)h(l(i),s(i),u(i	w(i)h(l(i),s(i),u(i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	186	22	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	23	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	186	25	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	26	7	7	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	186	27	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	28	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	186	29	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	30	l(1),s(1),u(1	l(1),s(1),u(1	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	186	31	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	32	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	186	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	187	1	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	188	1	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	189	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	2	l(M),s(M),u(M	l(M),s(M),u(M	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	4	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	5	i.i.d.∼	i.i.d.∼	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	6	q	q	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	189	8	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	189	9	weights	weight	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	189	10	w(i	w(i	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	12	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	13	defined	define	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	189	14	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	189	15	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	189	16	w(i	w(i	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	18	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	19	p̄(l(i),s(i),u(i	p̄(l(i),s(i),u(i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	189	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	22	q(l(i),s(i),u(i	q(l(i),s(i),u(i	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	189	23	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	189	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	190	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	190	2	8)	8)	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	190	3	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	190	4	indirect	indirect	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	190	5	estimator	estimator	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	190	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	190	7	biased	bias	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	190	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	190	9	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	190	10	consistent.12	consistent.12	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	190	11	Proposal	Proposal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	190	12	Distribution	Distribution	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	190	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	191	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	191	2	success	success	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	4	impor-	impor-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	5	tance	tance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	6	sampling	sampling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	7	depends	depend	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	191	10	choice	choice	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	191	13	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	191	14	good	good	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	15	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	191	16	proposal	proposal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	17	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	191	19	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	191	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	191	21	one	one	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	191	22	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	191	23	ideally	ideally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	191	24	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	25	close	close	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	26	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	27	p.	p.	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	28	Since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	29	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	191	30	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	191	31	fully	fully	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	191	32	supervised	supervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	33	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	34	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	35	time	time	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	36	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	191	37	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	191	38	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	191	39	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	191	40	option	option	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	41	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	42	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	43	locally	locally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	191	44	normalized	normalized	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	45	distributions	distribution	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	191	46	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	191	47	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	191	48	individual	individual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	191	49	components	component	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	191	50	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	192	1	Con-	Con-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	2	cretely	cretely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	192	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	192	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	5	train	train	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	6	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	192	7	proposal	proposal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	8	distributions	distribution	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	192	9	q1(u	q1(u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	10	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	11	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	12	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	192	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	192	14	q2(l	q2(l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	192	16	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	17	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	18	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	192	20	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	21	take	take	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	22	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	23	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	24	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	25	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	26	WFST	WFST	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	27	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	192	28	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	29	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	30	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	31	CRF	crf	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	192	33	respectively	respectively	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	192	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	192	35	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	192	36	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	192	37	identical	identical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	38	11The	11the	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	192	39	subvector	subvector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	40	β	β	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	41	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	42	responsible	responsible	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	43	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	44	computing	compute	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	192	45	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	192	46	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	47	mean	mean	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	48	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	49	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	50	Gaussian	gaussian	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	51	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	52	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	192	53	thus	thus	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	192	54	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	55	no	no	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	192	56	impact	impact	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	57	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	58	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	192	59	nor-	nor-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	192	60	malization	malization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	61	coefficient	coefficient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	192	62	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	192	63	Murphy	Murphy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	192	64	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	192	65	2012	2012	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	192	66	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	192	67	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	193	1	12Informally	12informally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	193	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	193	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	193	4	indirect	indirect	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	193	5	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	6	sampling	sample	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	193	7	estimate	estimate	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	8	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	9	verges	verge	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	193	10	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	193	12	true	true	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	193	13	expectation	expectation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	14	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	15	M	M	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	193	16	→	→	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	193	17	∞	∞	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	193	18	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	193	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	193	20	definition	definition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	21	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	22	statistical	statistical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	193	23	consistency	consistency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	193	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	193	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	194	1	38	38	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	194	2	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	194	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	194	4	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	194	5	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	194	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	195	1	Each	each	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	195	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	195	3	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	195	4	distributions	distribution	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	195	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	195	6	tractable	tractable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	195	7	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	195	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	195	9	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	195	10	compute	compute	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	195	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	195	12	marginals	marginal	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	195	13	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	195	14	dy-	dy-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	195	15	namic	namic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	195	16	programming	programming	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	195	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	195	18	thus	thus	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	195	19	sample	sample	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	195	20	efficiently	efficiently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	195	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	196	1	To	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	196	2	draw	draw	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	196	3	samples	sample	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	196	4	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	196	5	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	196	7	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	196	9	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	196	11	∼	∼	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	12	q	q	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	196	14	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	15	sample	sample	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	16	sequentially	sequentially	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	196	17	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	196	18	q1	q1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	19	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	196	20	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	196	21	q2	q2	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	196	23	conditioned	condition	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	196	24	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	196	25	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	196	26	output	output	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	196	27	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	196	28	q1	q1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	196	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	197	1	4.2	4.2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	197	2	Learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	197	3	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	4	optimize	optimize	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	197	6	log	log	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	7	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	197	8	likelihood	likelihood	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	197	11	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	12	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	197	13	ADAGRAD	ADAGRAD	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	197	15	Duchi	Duchi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	16	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	17	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	18	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	197	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	197	20	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	197	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	197	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	197	23	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	197	24	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	197	25	SGD	sgd	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	26	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	197	28	special	special	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	197	29	per	per	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	30	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	197	31	parameter	parameter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	32	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	33	rate	rate	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	197	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	198	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	198	2	full	full	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	198	3	gra-	gra-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	198	4	dient	dient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	198	7	objective	objective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	8	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	9	one	one	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	10	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	11	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	198	13	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	198	14	∇θ	∇θ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	15	log	log	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	16	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	18	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	20	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	22	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	23	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	24	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	26	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	27	f(s	f(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	29	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	31	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	32	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	33	>	>	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	198	34	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	198	35	g(u	g(u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	36	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	37	w	w	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	38	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	39	>	>	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	40	−	−	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	41	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	42	σ2	σ2	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	43	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	44	v	v	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	45	−Cβ(s	−cβ(s	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	46	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	47	l))∇θCβ(s	l))∇θCβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	48	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	49	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	50	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	51	−∇θ	−∇θ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	52	log	log	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	53	Zθ(w	Zθ(w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	54	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	55	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	56	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	57	9	9	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	58	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	59	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	198	60	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	61	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	62	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	198	63	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	64	sampling	sampling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	65	algorithm	algorithm	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	66	described	describe	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	67	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	68	§	§	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	69	4.1	4.1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	70	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	198	71	approximate	approximate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	198	72	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	198	73	gradient	gradient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	74	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	75	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	198	76	log	log	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	77	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	198	78	partition	partition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	79	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	198	80	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	198	81	following	follow	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	198	82	Bengio	Bengio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	83	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	198	84	Senecal	Senecal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	198	85	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	86	2003	2003	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	198	87	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	198	88	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	199	1	Note	note	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	199	2	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	199	3	∇θCβ(s	∇θCβ(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	199	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	199	5	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	199	6	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	199	7	depends	depend	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	199	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	199	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	199	10	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	199	11	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	199	12	used	use	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	199	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	200	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	200	3	most	most	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	200	4	com-	com-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	5	plicated	plicate	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	200	6	case	case	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	7	when	when	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	200	8	Cβ	Cβ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	200	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	200	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	200	11	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	200	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	200	13	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	200	14	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	200	15	com-	com-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	200	16	pute	pute	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	17	∇βCβ(s	∇βcβ(s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	200	19	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	200	21	efficiently	efficiently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	200	22	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	23	backpropagation	backpropagation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	24	through	through	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	25	time	time	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	200	26	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	200	27	Werbos	Werbos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	200	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	200	29	1990	1990	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	200	30	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	200	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	201	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	201	2	take	take	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	201	3	M	M	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	201	4	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	201	5	10	10	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	201	6	im-	im-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	7	portance	portance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	8	samples	sample	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	201	9	;	;	,
work_cxyxl5vs6zd6xlanudcx46qoqy	201	10	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	201	11	so	so	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	201	12	few	few	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	201	13	samples	sample	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	201	14	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	201	15	lead	lead	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	201	16	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	17	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	201	18	poor	poor	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	201	19	estimate	estimate	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	20	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	201	22	gradient	gradient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	201	24	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	201	25	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	26	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	201	27	application	application	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	201	28	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	201	29	suffices	suffice	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	201	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	202	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	202	2	employ	employ	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	202	3	L2	l2	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	202	4	regularization	regularization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	202	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	203	1	4.3	4.3	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	203	2	Decoding	decode	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	203	3	Decoding	decode	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	203	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	203	5	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	203	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	203	7	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	203	8	intractable	intractable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	203	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	204	1	To	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	204	2	approxi-	approxi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	204	3	mate	mate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	204	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	204	5	solution	solution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	204	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	204	7	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	204	8	again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	204	9	employ	employ	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	204	10	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	204	11	sam-	sam-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	204	12	pling	pling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	204	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	205	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	205	2	take	take	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	205	3	M	M	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	205	4	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	205	5	10,000	10,000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	205	6	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	205	7	samples	sample	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	205	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	205	9	select	select	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	205	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	205	11	highest	high	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	205	12	weighted	weighted	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	205	13	sample	sample	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	205	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	206	1	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	206	2	Related	related	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	206	3	Work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	4	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	206	5	idea	idea	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	6	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	7	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	8	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	206	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	206	10	useful	useful	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	206	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	12	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	13	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	206	14	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	206	15	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	206	16	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	206	17	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	206	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	207	1	Count	Count	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	2	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	207	3	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	207	4	Salton	Salton	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	207	6	1971	1971	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	207	7	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	207	8	Turney	Turney	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	207	10	Pantel	Pantel	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	207	12	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	207	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	207	14	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	15	been	be	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	16	shown	show	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	17	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	207	18	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	207	19	beneficial	beneficial	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	207	20	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	207	22	unsupervised	unsupervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	23	induction	induction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	24	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	25	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	207	26	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	207	27	Schone	Schone	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	28	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	207	29	Jurafsky	Jurafsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	207	31	2000	2000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	207	32	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	207	33	Schone	Schone	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	34	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	207	35	Jurafsky	Jurafsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	207	36	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	207	37	2001	2001	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	207	38	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	207	39	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	208	1	Embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	208	2	were	be	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	208	3	shown	show	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	208	4	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	208	5	act	act	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	208	6	similarly	similarly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	208	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	208	8	Soricut	Soricut	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	208	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	208	10	Och	Och	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	208	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	208	12	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	208	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	208	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	209	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	209	2	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	209	3	differs	differ	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	209	4	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	209	5	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	209	6	line	line	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	209	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	209	8	research	research	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	209	9	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	209	10	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	209	11	key	key	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	209	12	ways	way	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	209	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	210	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	210	2	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	210	4	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	210	5	present	present	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	210	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	210	7	probabilistic	probabilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	210	8	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	210	11	pro-	pro-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	12	cess	cess	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	14	synthesizing	synthesize	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	210	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	210	16	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	17	’s	’s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	210	18	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	19	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	210	21	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	22	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	210	23	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	210	24	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	210	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	211	1	Prior	prior	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	211	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	211	3	was	be	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	211	4	ei-	ei-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	211	5	ther	th	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	211	6	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	211	7	probabilistic	probabilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	211	8	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	211	9	did	do	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	211	10	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	211	11	explicitly	explicitly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	211	12	model	model	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	211	13	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	211	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	212	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	212	2	ii	ii	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	212	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	212	4	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	212	5	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	212	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	212	7	supervised	supervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	212	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	212	9	fo-	fo-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	212	10	cuses	cuse	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	212	11	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	212	12	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	212	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	213	1	Schone	Schone	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	2	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	213	3	Jurafsky	Jurafsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	4	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	213	5	2000	2000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	213	6	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	213	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	213	8	Soricut	Soricut	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	213	10	Och	Och	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	11	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	213	12	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	213	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	213	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	213	15	being	be	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	213	16	fully	fully	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	213	17	unsupervised	unsupervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	213	19	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	20	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	213	21	distinguish	distinguish	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	213	22	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	23	inflection	inflection	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	24	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	213	25	deriva-	deriva-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	26	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	27	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	213	28	Schone	Schone	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	29	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	213	30	Jurafsky	Jurafsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	213	31	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	213	32	2001	2001	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	213	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	213	34	focus	focus	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	35	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	36	in-	in-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	213	37	flection	flection	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	213	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	214	1	More	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	214	2	recently	recently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	214	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	214	4	Narasimhan	Narasimhan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	214	5	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	214	6	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	214	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	215	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	215	2	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	215	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	215	4	look	look	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	215	5	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	215	7	unsupervised	unsupervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	8	induction	induction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	10	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	215	11	morpholog-	morpholog-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	215	12	ical	ical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	215	13	chains	chain	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	215	14	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	215	15	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	16	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	215	17	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	215	18	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	19	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	215	20	crucial	crucial	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	215	21	fea-	fea-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	22	ture	ture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	215	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	216	1	Their	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	216	2	goal	goal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	216	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	216	4	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	216	5	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	216	6	figure	figure	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	216	7	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	216	8	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	216	9	ordering	ordering	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	216	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	216	11	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	216	12	formation	formation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	216	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	216	14	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	216	15	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	216	16	segmenta-	segmenta-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	216	17	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	216	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	216	19	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	216	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	216	21	play7→playful7→playfulness	play7→playful7→playfulness	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	216	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	217	1	While	while	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	2	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	217	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	217	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	217	5	rich	rich	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	217	6	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	7	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	8	ours	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	217	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	217	10	theirs	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	217	11	differs	differ	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	217	12	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	13	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	14	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	217	15	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	217	16	un-	un-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	217	17	supervised	supervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	217	18	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	217	19	uses	use	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	217	20	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	217	21	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	22	features	feature	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	217	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	217	24	rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	217	25	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	26	explicitly	explicitly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	217	27	treating	treat	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	217	28	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	29	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	217	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	218	1	All	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	218	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	218	4	above	above	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	218	5	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	6	focuses	focus	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	218	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	8	surface	surface	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	9	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	218	11	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	218	12	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	218	13	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	218	15	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	218	16	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	218	17	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	218	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	219	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	219	2	related	related	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	219	3	line	line	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	219	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	219	5	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	219	6	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	219	7	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	219	8	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	219	9	goals	goal	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	219	10	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	219	11	cerns	cern	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	219	12	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	219	13	generation	generation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	219	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	220	1	Two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	220	2	recent	recent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	220	3	papers	paper	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	220	4	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	220	5	address	address	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	220	6	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	220	7	problem	problem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	220	8	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	220	9	deep	deep	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	220	10	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	220	11	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	220	12	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	220	13	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	220	14	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	220	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	221	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	221	2	2016a	2016a	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	221	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	221	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	221	5	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	221	6	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	221	7	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	221	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	222	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	222	2	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	222	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	222	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	223	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	2	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	223	3	older	old	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	223	4	line	line	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	6	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	223	8	Yarowsky	Yarowsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	223	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	223	10	Wicen-	Wicen-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	223	11	towski	towski	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	223	13	2000	2000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	223	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	223	15	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	223	16	Wicentowski	Wicentowski	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	223	17	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	223	18	2002	2002	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	223	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	223	20	exploit	exploit	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	21	log	log	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	223	22	frequency	frequency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	23	ratios	ratio	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	223	24	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	25	inflectionally	inflectionally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	26	related	relate	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	27	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	223	28	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	223	29	tease	tease	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	30	apart	apart	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	31	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	223	33	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	223	35	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	223	36	past	past	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	223	37	tense	tense	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	38	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	39	sing	sing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	40	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	223	41	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	42	singed	singe	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	223	43	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	223	44	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	223	45	instead	instead	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	223	46	sang	sing	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	223	47	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	224	1	Related	related	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	3	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	4	Dreyer	Dreyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	224	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	224	6	Eisner	Eisner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	224	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	224	8	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	224	9	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	224	10	uses	use	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	224	12	Dirichlet	Dirichlet	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	224	13	process	process	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	14	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	224	15	model	model	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	224	16	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	224	17	corpus	corpus	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	18	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	19	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	224	20	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	224	21	mixture	mixture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	22	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	23	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	224	24	paradigm	paradigm	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	25	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	224	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	224	27	allowing	allow	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	224	28	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	224	30	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	31	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	32	supervised	supervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	33	incorporation	incorporation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	34	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	35	distribu-	distribu-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	36	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	37	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	224	38	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	39	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	224	40	structured	structured	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	41	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	42	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	43	inflec-	inflec-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	224	44	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	224	45	paradigm	paradigm	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	224	46	completion	completion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	224	47	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	225	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	225	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	225	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	225	4	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	225	5	related	relate	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	225	6	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	225	7	recent	recent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	225	8	attempts	attempt	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	225	9	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	225	10	in-	in-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	225	11	tegrate	tegrate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	225	12	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	225	13	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	225	14	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	225	15	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	225	16	em-	em-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	225	17	bedding	bed	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	225	18	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	225	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	226	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	226	4	Botha	Botha	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	226	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	226	6	Blun-	Blun-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	226	7	som	som	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	226	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	226	9	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	226	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	226	11	train	train	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	226	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	226	13	log	log	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	14	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	226	15	bilinear	bilinear	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	16	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	17	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	18	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	226	19	models	model	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	226	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	226	21	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	22	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	23	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	226	24	structure	structure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	226	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	227	1	Likewise	likewise	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	227	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	227	3	Luong	Luong	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	227	4	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	227	5	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	227	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	228	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	228	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	228	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	228	4	train	train	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	228	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	228	6	recursive	recursive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	228	7	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	228	8	network	network	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	9	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	228	10	Goller	Goller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	228	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	228	12	Küchler	Küchler	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	228	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	228	14	1996	1996	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	228	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	228	16	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	17	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	228	18	heuristi-	heuristi-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	19	cally	cally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	228	20	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	21	tree	tree	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	22	structure	structure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	23	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	228	24	learn	learn	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	228	25	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	228	26	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	27	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	228	28	continuous	continuous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	228	29	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	228	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	229	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	229	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	229	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	229	4	different	different	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	229	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	229	6	that	that	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	229	7	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	229	8	learn	learn	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	229	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	229	10	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	229	11	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	229	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	229	13	segmen-	segmen-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	229	14	tation	tation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	229	15	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	229	16	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	229	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	230	1	Moreover	moreover	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	230	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	230	3	supervised	supervise	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	230	4	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	230	5	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	230	6	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	230	7	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	230	8	drastically	drastically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	230	9	outperform	outperform	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	230	10	unsu-	unsu-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	230	11	pervised	pervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	230	12	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	230	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	230	14	Ruokolainen	Ruokolainen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	230	15	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	230	16	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	230	17	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	230	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	230	19	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	230	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	230	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	231	1	Early	early	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	3	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	4	Kay	Kay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	231	5	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	231	6	1977	1977	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	231	7	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	231	8	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	231	9	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	231	10	interpreted	interpret	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	11	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	12	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	13	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	231	14	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	15	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	16	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	231	18	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	231	19	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	231	20	neither	neither	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	231	21	ad-	ad-	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	231	22	dresses	dress	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	231	23	nor	nor	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	231	24	experimentally	experimentally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	231	25	evaluates	evaluate	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	26	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	231	27	question	question	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	28	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	29	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	30	modeling	modeling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	31	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	32	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	33	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	34	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	231	35	se-	se-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	36	mantic	mantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	231	37	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	231	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	232	1	Moreover	moreover	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	232	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	232	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	4	may	may	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	5	view	view	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	232	6	canoni-	canoni-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	7	39	39	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	8	dev	dev	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	9	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	10	Model	Model	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	11	Acc	Acc	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	12	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	13	Edit	Edit	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	14	Acc	Acc	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	15	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	16	Edit	Edit	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	17	E	E	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	18	N	N	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	19	Semi	Semi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	20	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	232	21	CRF	CRF	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	22	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	23	Baseline	Baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	25	0.55	0.55	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	26	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	27	.018	.018	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	28	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	29	0.75	0.75	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	30	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	31	.014	.014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	32	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	33	0.80	0.80	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	34	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	35	.043	.043	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	36	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	37	0.54	0.54	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	38	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	39	.018	.018	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	40	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	41	0.75	0.75	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	42	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	43	.014	.014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	44	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	45	0.78	0.78	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	46	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	47	.034	.034	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	48	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	49	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	50	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	51	Baseline	Baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	52	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	53	0.77	0.77	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	54	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	55	.011	.011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	56	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	57	0.87	0.87	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	58	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	59	.007	.007	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	60	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	61	0.41	0.41	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	62	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	63	.029	.029	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	64	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	65	0.77	0.77	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	66	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	67	.013	.013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	68	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	69	0.87	0.87	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	70	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	71	.007	.007	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	72	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	73	0.43	0.43	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	74	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	75	.029	.029	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	76	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	77	Joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	78	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	232	79	Vec	vec	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	80	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	81	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	232	82	Work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	83	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	84	0.83	0.83	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	85	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	86	.014	.014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	87	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	88	0.91	0.91	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	89	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	90	.008	.008	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	91	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	92	0.31	0.31	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	93	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	94	.019	.019	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	95	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	96	0.82	0.82	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	97	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	98	.020	.020	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	99	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	100	0.90	0.90	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	101	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	102	.011	.011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	103	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	104	0.32	0.32	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	105	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	106	.038	.038	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	107	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	108	Joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	109	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	232	110	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	111	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	112	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	113	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	114	0.94	0.94	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	115	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	116	.015	.015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	117	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	118	0.96	0.96	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	119	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	120	.009	.009	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	121	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	122	0.07	0.07	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	123	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	124	.016	.016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	125	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	126	0.94	0.94	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	127	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	128	.011	.011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	129	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	130	0.96	0.96	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	131	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	132	.007	.007	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	133	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	134	0.07	0.07	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	135	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	136	.011	.011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	137	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	138	Joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	139	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	232	140	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	141	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	232	142	Vec	vec	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	143	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	144	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	145	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	146	0.95	0.95	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	147	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	148	.011	.011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	149	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	150	0.97	0.97	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	151	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	152	.007	.007	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	153	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	154	0.05	0.05	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	155	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	156	.013	.013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	157	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	158	0.95	0.95	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	159	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	160	.023	.023	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	161	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	162	0.97	0.97	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	163	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	164	.006	.006	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	165	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	166	0.05	0.05	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	167	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	168	.025	.025	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	169	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	170	D	d	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	171	E	E	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	172	Semi	Semi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	173	-	-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	174	CRF	CRF	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	175	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	176	Baseline	Baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	177	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	178	0.39	0.39	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	179	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	180	.062	.062	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	181	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	182	0.68	0.68	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	183	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	184	.039	.039	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	185	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	186	1.15	1.15	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	187	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	188	.230	.230	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	189	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	190	0.39	0.39	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	191	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	192	.058	.058	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	193	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	194	0.68	0.68	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	195	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	196	.042	.042	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	197	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	198	1.14	1.14	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	199	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	200	.240	.240	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	201	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	202	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	203	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	204	Baseline	Baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	205	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	206	0.79	0.79	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	207	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	208	.107	.107	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	209	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	210	0.88	0.88	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	211	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	212	.069	.069	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	213	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	214	0.40	0.40	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	215	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	216	.313	.313	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	217	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	218	0.79	0.79	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	219	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	220	.099	.099	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	221	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	222	0.87	0.87	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	223	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	224	.063	.063	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	225	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	226	0.41	0.41	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	227	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	228	.282	.282	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	229	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	230	Joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	231	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	232	232	Vec	vec	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	233	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	234	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	232	235	Work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	236	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	237	0.82	0.82	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	238	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	239	.102	.102	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	240	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	241	0.90	0.90	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	242	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	243	.067	.067	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	244	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	245	0.33	0.33	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	246	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	247	.312	.312	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	248	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	249	0.82	0.82	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	250	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	251	.096	.096	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	252	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	253	0.90	0.90	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	254	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	255	.061	.061	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	256	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	257	0.33	0.33	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	258	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	259	.282	.282	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	260	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	261	Joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	262	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	232	263	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	264	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	265	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	266	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	267	0.86	0.86	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	268	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	269	.108	.108	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	270	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	271	0.90	0.90	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	272	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	273	.070	.070	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	274	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	275	0.25	0.25	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	276	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	277	.288	.288	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	278	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	279	0.86	0.86	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	280	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	281	.100	.100	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	282	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	283	0.90	0.90	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	284	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	285	.064	.064	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	286	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	287	0.25	0.25	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	288	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	289	.268	.268	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	290	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	291	Joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	292	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	232	293	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	294	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	232	295	Vec	vec	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	296	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	297	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	298	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	299	0.87	0.87	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	300	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	301	.106	.106	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	302	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	303	0.92	0.92	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	304	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	305	.069	.069	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	306	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	307	0.20	0.20	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	308	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	309	.285	.285	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	310	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	311	0.88	0.88	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	312	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	313	.096	.096	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	314	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	315	0.93	0.93	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	316	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	317	.062	.062	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	318	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	319	0.19	0.19	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	320	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	321	.263	.263	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	322	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	232	323	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	324	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	232	325	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	232	326	Results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	232	327	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	328	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	232	329	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	232	330	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	232	331	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	332	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	333	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	232	334	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	335	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	232	336	German	German	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	232	337	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	233	1	Standard	standard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	233	2	deviation	deviation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	233	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	233	4	given	give	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	233	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	233	6	parentheses	parenthesis	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	233	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	234	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	2	compare	compare	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	3	against	against	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	4	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	234	5	baselines	baseline	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	234	6	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	234	7	do	do	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	8	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	234	9	make	make	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	234	10	use	use	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	12	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	234	13	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	234	14	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	234	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	16	i	i	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	18	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	234	19	Semi	Semi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	20	-	-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	21	CRF	CRF	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	22	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	23	baseline	baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	25	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	234	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	234	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	234	28	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	234	29	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	234	30	CRF	crf	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	31	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	234	32	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	234	33	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	234	34	account	account	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	234	35	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	36	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	234	37	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	234	38	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	234	39	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	40	ii	ii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	41	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	42	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	234	43	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	44	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	45	Baseline	Baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	234	46	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	234	47	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	234	48	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	234	49	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	234	50	version	version	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	51	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	52	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	234	53	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	234	54	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	55	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	234	56	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	234	57	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	235	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	235	3	compare	compare	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	4	against	against	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	5	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	235	6	oracle	oracle	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	235	7	version	version	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	8	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	9	access	access	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	10	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	11	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	12	URs	ur	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	235	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	235	14	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	235	15	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	16	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	235	17	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	18	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	235	19	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	235	21	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	235	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	235	23	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	235	24	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	25	+	+	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	235	26	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	235	27	+	+	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	235	28	Vec	vec	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	29	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	235	30	Oracle	oracle	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	31	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	235	32	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	235	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	235	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	235	35	revealing	reveal	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	235	36	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	37	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	235	38	toughest	tough	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	235	39	part	part	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	40	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	41	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	235	42	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	235	43	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	44	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	235	45	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	235	46	reversing	reverse	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	235	47	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	235	48	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	235	49	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	235	50	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	236	1	calization	calization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	236	2	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	236	3	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	236	4	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	236	5	analogue	analogue	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	236	6	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	236	7	phonology	phonology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	236	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	237	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	2	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	237	3	interpretation	interpretation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	237	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	237	6	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	237	7	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	237	8	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	9	systems	system	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	237	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	11	Ka-	Ka-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	237	12	plan	plan	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	237	14	Kay	Kay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	237	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	237	16	1994	1994	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	237	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	237	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	237	19	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	237	20	computationally	computationally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	237	21	apply	apply	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	237	22	SPE	SPE	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	237	23	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	237	24	style	style	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	25	phonological	phonological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	237	26	rules	rule	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	237	27	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	237	28	Chomsky	Chomsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	237	29	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	237	30	Halle	Halle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	237	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	237	32	1968	1968	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	237	33	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	237	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	237	35	may	may	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	237	36	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	237	37	run	run	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	237	38	backwards	backwards	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	237	39	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	237	40	get	get	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	237	41	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	237	42	un-	un-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	237	43	derlying	derlye	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	237	44	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	237	45	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	238	1	6	6	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	238	2	Experiments	Experiments	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	238	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	238	4	Results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	238	5	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	238	6	conduct	conduct	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	238	7	experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	238	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	238	9	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	238	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	238	11	German	german	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	238	12	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	238	13	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	238	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	239	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	239	2	analyze	analyze	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	239	3	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	239	4	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	239	5	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	6	’s	’s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	239	7	ability	ability	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	8	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	239	9	segment	segment	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	239	10	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	239	11	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	12	their	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	239	13	canoni-	canoni-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	14	cal	cal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	15	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	239	16	as	as	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	239	17	well	well	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	239	18	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	19	its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	239	20	ability	ability	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	21	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	22	composition-	composition-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	239	23	ally	ally	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	239	24	derive	derive	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	239	25	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	239	26	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	239	27	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	239	28	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	239	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	240	1	Finally	finally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	240	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	240	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	240	4	ex-	ex-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	240	5	plore	plore	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	240	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	240	7	relationship	relationship	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	240	8	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	240	9	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	240	10	seman-	seman-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	240	11	tics	tic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	240	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	240	13	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	240	14	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	240	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	241	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	241	2	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	241	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	241	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	241	5	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	241	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	241	7	pretrained	pretrained	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	241	8	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	241	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	241	10	Levy	Levy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	241	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	241	12	Goldberg	Goldberg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	241	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	241	14	2014a	2014a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	241	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	241	16	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	241	17	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	241	18	experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	241	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	242	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	242	2	German	German	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	242	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	242	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	242	5	train	train	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	242	6	word2vec	word2vec	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	242	7	skip	skip	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	242	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	242	9	gram	gram	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	242	10	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	242	11	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	242	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	242	13	German	german	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	242	14	Wikipedia	Wikipedia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	242	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	243	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	2	first	first	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	243	3	describe	describe	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	4	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	243	5	En-	en-	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	243	6	glish	glish	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	243	7	dataset	dataset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	243	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	243	10	subset	subset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	243	13	English	english	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	243	14	portion	portion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	243	17	CELEX	CELEX	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	18	lexical	lexical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	243	19	database	database	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	20	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	243	21	Baayen	Baayen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	22	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	23	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	24	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	243	26	1993	1993	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	243	27	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	243	28	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	243	29	was	be	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	243	30	selected	select	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	31	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	243	32	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	33	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	34	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	243	35	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	244	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	244	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	244	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	244	4	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	244	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	244	6	dataset	dataset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	244	7	contains	contain	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	244	8	10,000	10,000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	244	9	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	244	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	245	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	245	2	allows	allow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	245	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	245	4	com-	com-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	245	5	parison	parison	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	245	6	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	245	7	previously	previously	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	245	8	proposed	propose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	245	9	methods	method	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	245	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	246	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	246	2	make	make	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	246	3	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	246	4	modifications	modification	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	246	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	247	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	247	2	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	247	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	247	4	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	247	5	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	247	6	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	247	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	248	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	248	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	248	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	248	4	make	make	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	248	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	248	6	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	248	7	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	248	8	morpheme	morpheme	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	248	9	assumption	assumption	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	248	10	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	248	11	every	every	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	248	12	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	248	13	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	248	14	composed	compose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	248	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	248	16	exactly	exactly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	248	17	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	248	18	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	248	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	249	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	249	2	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	249	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	249	4	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	249	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	249	6	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	249	7	true	true	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	249	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	249	9	so	so	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	249	10	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	249	11	further	further	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	249	12	segment	segment	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	249	13	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	249	14	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	249	15	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	249	16	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	249	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	249	18	corpus	corpus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	249	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	250	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	250	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	250	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	250	4	friendless+ness	friendless+ness	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	250	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	250	6	further	further	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	250	7	segmented	segment	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	250	8	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	250	9	friend+less+ness	friend+less+ness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	250	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	251	1	To	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	251	2	nevertheless	nevertheless	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	251	3	allow	allow	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	251	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	5	fair	fair	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	251	6	comparison	comparison	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	251	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	251	9	pro-	pro-	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	251	10	vide	vide	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	251	11	versions	version	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	251	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	13	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	251	14	experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	251	15	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	16	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	251	17	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	18	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	251	19	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	251	20	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	251	21	morpheme	morpheme	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	22	assumption	assumption	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	251	23	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	251	24	appropriate	appropriate	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	251	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	252	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	252	2	ii	ii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	252	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	252	4	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	252	5	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	252	6	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	252	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	253	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	253	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	253	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	253	4	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	253	5	provide	provide	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	253	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	253	7	single	single	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	253	8	train	train	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	253	9	/	/	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	253	10	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	253	11	split	split	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	253	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	254	1	As	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	2	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	254	3	require	require	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	254	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	5	held	hold	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	6	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	254	7	out	out	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	8	develop-	develop-	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	254	9	ment	ment	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	254	10	set	set	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	12	hyperparameter	hyperparameter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	13	tuning	tuning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	254	15	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	254	16	randomly	randomly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	254	17	allocate	allocate	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	254	18	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	19	portion	portion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	20	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	22	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	23	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	254	24	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	254	25	select	select	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	254	26	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	27	hyperparameters	hyperparameter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	254	28	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	254	29	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	254	30	retrain	retrain	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	254	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	32	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	33	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	254	34	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	35	parameters	parameter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	254	36	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	37	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	254	38	original	original	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	254	39	train	train	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	40	split	split	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	254	41	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	255	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	255	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	255	3	report	report	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	255	4	10-fold	10-fold	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	255	5	cross	cross	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	6	validation	validation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	7	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	255	8	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	9	addition	addition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	10	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	11	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	255	12	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	255	13	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	255	14	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	255	15	’s	’s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	255	16	train	train	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	17	/	/	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	255	18	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	19	split	split	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	255	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	256	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	256	2	German	german	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	256	3	dataset	dataset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	256	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	256	5	taken	take	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	256	6	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	256	7	Zeller	Zeller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	256	8	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	256	9	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	256	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	257	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	257	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	257	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	257	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	257	5	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	257	6	described	describe	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	257	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	257	8	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	257	9	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	257	10	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	257	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	258	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	258	2	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	258	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	258	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	259	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	259	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	259	3	again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	259	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	259	5	consists	consist	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	259	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	259	7	10,000	10,000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	259	8	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	259	9	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	259	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	260	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	260	2	report	report	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	260	3	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	260	4	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	260	5	10-fold	10-fold	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	260	6	cross	cross	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	260	7	validation	validation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	260	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	261	1	6.1	6.1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	261	2	Experiment	experiment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	261	3	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	261	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	261	5	Canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	261	6	Segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	261	7	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	261	8	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	261	9	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	261	10	experiment	experiment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	261	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	261	12	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	261	13	test	test	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	261	14	whether	whether	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	261	15	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	261	16	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	261	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	261	18	continuous	continuous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	261	19	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	261	20	allows	allow	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	261	21	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	261	22	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	261	23	segment	segment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	261	24	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	261	25	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	261	26	accurately	accurately	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	261	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	262	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	262	2	assume	assume	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	262	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	262	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	262	5	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	262	6	given	give	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	262	7	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	262	8	embedding	embedding	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	262	9	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	262	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	262	11	target	target	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	262	12	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	262	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	263	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	2	estimate	estimate	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	263	4	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	263	5	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	263	7	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	263	9	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	263	11	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	12	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	13	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	263	15	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	263	16	described	describe	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	263	17	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	263	18	§	§	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	263	19	4	4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	263	20	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	263	21	L2	l2	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	263	22	regularization	regularization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	263	23	λ||θ||22	λ||θ||22	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	263	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	264	1	To	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	264	2	evaluate	evaluate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	264	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	264	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	5	decode	decode	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	264	7	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	264	8	p(s	p(s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	264	10	l	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	264	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	264	12	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	13	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	14	v	v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	264	16	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	264	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	264	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	265	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	265	2	per-	per-	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	265	3	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	4	approximate	approximate	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	265	5	MAP	MAP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	265	6	inference	inference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	7	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	8	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	9	sampling	sampling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	10	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	265	11	taking	take	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	265	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	265	13	sample	sample	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	14	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	265	16	highest	high	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	265	17	score	score	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	265	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	266	1	40	40	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	2	EN	EN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	3	DE	de	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	4	BOW2	BOW2	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	266	5	BOW5	BOW5	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	6	DEPs	DEPs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	7	SG	SG	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	8	dev	dev	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	9	test	test	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	10	dev	dev	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	11	test	test	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	12	dev	dev	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	13	test	test	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	14	dev	dev	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	15	test	test	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	16	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	266	17	ac	ac	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	18	le	le	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	19	stem	stem	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	20	.403	.403	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	21	.402	.402	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	22	.374	.374	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	23	.376	.376	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	24	.422	.422	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	25	.422	.422	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	26	.400	.400	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	27	.405	.405	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	28	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	266	29	.635	.635	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	30	.635	.635	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	31	.541	.541	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	32	.542	.542	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	33	.787	.787	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	34	.785	.785	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	35	.712	.712	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	36	.711	.711	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	37	LDS	lds	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	38	.660	.660	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	39	.660	.660	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	40	.566	.566	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	41	.568	.568	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	42	.806	.806	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	43	.804	.804	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	44	.717	.717	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	45	.718	.718	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	46	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	47	.660	.660	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	48	.660	.660	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	49	.565	.565	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	50	.567	.567	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	51	.807	.807	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	52	.806	.806	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	53	.707	.707	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	54	.712	.712	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	55	jo	jo	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	266	56	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	57	t	t	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	58	stem	stem	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	59	.399	.399	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	60	.400	.400	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	61	.371	.371	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	62	.372	.372	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	63	.411	.411	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	64	.412	.412	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	65	.394	.394	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	66	.398	.398	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	67	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	266	68	.625	.625	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	69	.625	.625	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	70	.524	.524	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	71	.525	.525	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	72	.782	.782	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	73	.781	.781	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	74	.705	.705	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	75	.704	.704	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	76	LDS	lds	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	77	.648	.648	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	78	.648	.648	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	79	.547	.547	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	80	.547	.547	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	81	.799	.799	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	82	.797	.797	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	83	.712	.712	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	84	.711	.711	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	85	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	86	.649	.649	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	87	.647	.647	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	88	.547	.547	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	89	.546	.546	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	90	.801	.801	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	91	.799	.799	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	92	.706	.706	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	93	.708	.708	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	94	ch	ch	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	95	ar	ar	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	96	GRU	GRU	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	97	.586	.586	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	98	.585	.585	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	99	.452	.452	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	100	.452	.452	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	101	.769	.769	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	102	.768	.768	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	103	.675	.675	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	104	.667	.667	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	105	LSTM	LSTM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	106	.586	.586	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	107	.586	.586	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	108	.455	.455	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	109	.455	.455	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	110	.768	.768	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	111	.767	.767	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	112	.677	.677	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	113	.666	.666	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	114	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	115	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	266	116	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	266	117	Vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	118	approximation	approximation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	119	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	266	120	measured	measure	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	121	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	122	mean	mean	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	123	co-	co-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	124	sine	sine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	125	similarity	similarity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	126	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	266	127	both	both	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	266	128	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	129	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	266	130	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	266	131	oracle	oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	132	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	266	133	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	266	134	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	266	135	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	136	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	266	137	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	266	138	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	266	139	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	266	140	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	266	141	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	266	142	char	char	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	266	143	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	266	144	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	266	145	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	146	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	266	147	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	267	1	Surprisingly	surprisingly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	267	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	267	3	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	267	4	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	267	5	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	267	6	close	close	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	267	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	267	8	performance	performance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	267	9	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	267	10	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	267	11	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	267	12	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	267	13	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	267	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	268	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	2	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	268	3	experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	268	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	268	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	268	6	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	268	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	268	8	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	268	9	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	268	11	de-	de-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	12	pendency	pendency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	13	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	268	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	268	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	268	16	combination	combination	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	18	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	268	19	per-	per-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	20	forms	form	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	268	21	best	well	RBS
work_cxyxl5vs6zd6xlanudcx46qoqy	268	22	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	23	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	24	approximation	approximation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	25	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	268	26	§	§	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	268	27	6.2	6.2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	268	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	269	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	269	2	follow	follow	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	269	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	269	4	experimental	experimental	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	269	5	design	design	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	269	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	269	7	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	269	8	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	269	9	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	269	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	270	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	270	2	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	270	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	270	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	271	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	2	compare	compare	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	3	against	against	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	4	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	271	5	base-	base-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	6	lines	line	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	271	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	271	8	marked	mark	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	9	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	271	10	Baseline	Baseline	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	11	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	271	12	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	13	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	14	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	271	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	271	16	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	271	17	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	271	18	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	271	20	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	271	21	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	271	22	Semi-	Semi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	23	CRF	CRF	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	24	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	271	25	segmenter	segmenter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	26	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	271	27	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	271	28	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	271	29	account	account	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	271	30	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	31	ortho-	ortho-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	271	32	graphic	graphic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	271	33	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	271	34	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	271	35	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	271	36	ii	ii	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	271	37	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	271	38	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	271	39	full	full	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	271	40	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	271	41	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	42	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	271	43	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	44	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	271	45	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	46	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	47	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	271	48	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	272	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	272	2	2016b).13	2016b).13	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	272	3	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	272	4	additionally	additionally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	272	5	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	272	6	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	272	7	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	272	8	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	272	9	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	272	10	setting	setting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	272	12	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	272	13	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	272	14	give	give	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	272	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	272	16	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	272	18	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	19	underlying	underlie	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	272	20	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	272	21	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	22	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	272	23	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	272	24	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	272	25	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	272	26	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	272	27	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	28	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	272	29	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	30	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	272	31	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	32	time	time	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	272	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	273	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	273	2	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	273	3	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	273	4	insight	insight	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	5	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	273	7	performance	performance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	273	10	transduction	transduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	11	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	13	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	273	14	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	273	16	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	273	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	273	18	how	how	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	273	19	much	much	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	273	20	could	could	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	273	21	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	273	22	benefit	benefit	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	273	23	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	24	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	273	25	richer	rich	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	273	26	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	273	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	274	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	274	2	hyperparameters	hyperparameter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	274	3	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	274	4	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	274	5	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	274	6	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	274	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	274	8	regularization	regularization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	274	9	coefficient	coefficient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	274	10	λ	λ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	274	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	274	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	274	13	ii	ii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	274	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	274	15	σ2	σ2	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	274	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	274	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	274	18	variance	variance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	274	19	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	274	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	274	21	Gaussian	gaussian	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	274	22	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	274	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	275	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	275	2	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	275	3	grid	grid	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	275	4	search	search	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	275	5	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	275	6	tune	tune	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	275	7	them	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	275	8	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	275	9	λ	λ	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	275	10	∈	∈	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	275	11	{	{	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	275	12	0.0	0.0	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	14	101	101	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	16	102	102	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	18	103	103	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	20	104	104	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	22	105	105	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	23	}	}	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	275	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	25	σ2	σ2	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	275	26	∈	∈	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	275	27	{	{	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	275	28	0.25	0.25	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	30	0.5	0.5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	32	0.75	0.75	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	275	34	1.0	1.0	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	275	35	}	}	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	275	36	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	276	1	Metrics	metric	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	276	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	277	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	277	2	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	277	3	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	277	4	metrics	metric	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	277	5	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	277	6	evaluate	evaluate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	277	7	segmen-	segmen-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	277	8	tation	tation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	277	9	accuracy	accuracy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	277	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	278	1	Note	note	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	278	2	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	278	4	evaluation	evaluation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	6	canon-	canon-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	278	7	ical	ical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	278	8	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	278	10	hard	hard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	278	11	since	since	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	278	13	system	system	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	14	may	may	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	278	15	re-	re-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	278	16	turn	turn	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	278	17	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	278	18	sequence	sequence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	19	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	20	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	278	21	whose	whose	WP$
work_cxyxl5vs6zd6xlanudcx46qoqy	278	22	concatenation	concatenation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	23	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	278	24	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	278	25	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	278	26	same	same	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	278	27	length	length	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	28	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	29	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	278	30	concatenation	concatenation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	31	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	32	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	278	33	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	278	34	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	278	35	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	279	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	279	2	rules	rule	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	279	3	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	4	metrics	metric	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	279	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	279	6	surface	surface	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	279	7	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	279	8	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	279	9	border	border	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	10	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	11	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	279	12	Kurimo	Kurimo	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	13	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	14	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	15	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	279	17	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	279	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	279	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	279	20	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	279	21	require	require	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	279	22	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	279	23	strings	string	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	279	24	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	279	25	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	279	26	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	279	27	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	279	28	same	same	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	279	29	length	length	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	279	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	280	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	280	2	now	now	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	280	3	define	define	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	280	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	280	5	metrics	metric	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	280	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	281	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	281	2	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	281	4	Segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	5	accuracy	accuracy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	6	measures	measure	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	281	7	whether	whether	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	8	every	every	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	281	9	single	single	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	281	10	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	281	11	morpheme	morpheme	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	12	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	281	14	returned	return	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	15	sequence	sequence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	281	16	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	281	17	correct	correct	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	281	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	282	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	282	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	282	3	inflexible	inflexible	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	282	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	282	5	closer	close	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	282	6	answers	answer	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	282	7	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	282	8	penalized	penalize	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	282	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	282	10	same	same	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	282	11	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	282	12	13i.e	13i.e	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	282	13	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	282	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	282	15	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	282	16	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	282	17	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	282	18	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	282	19	Gaussian	gaussian	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	282	20	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	282	21	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	282	22	scores	score	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	282	23	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	282	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	283	1	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	283	2	distant	distant	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	283	3	answers	answer	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	283	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	284	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	284	2	ii	ii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	284	4	Morpheme	Morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	5	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	6	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	284	7	van	van	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	8	den	den	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	9	Bosch	Bosch	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	284	11	Daelemans	Daelemans	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	284	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	284	13	1999	1999	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	284	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	284	15	takes	take	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	284	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	284	17	predicted	predict	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	18	se-	se-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	19	quence	quence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	20	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	21	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	284	22	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	284	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	284	24	turns	turn	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	284	25	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	26	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	284	28	set	set	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	284	30	computes	compute	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	284	31	precision	precision	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	32	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	284	33	recall	recall	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	34	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	35	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	284	36	standard	standard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	284	37	way	way	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	38	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	284	39	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	40	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	284	41	that	that	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	284	42	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	284	43	computes	compute	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	284	44	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	284	45	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	285	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	285	2	metric	metric	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	285	3	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	285	4	credit	credit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	285	5	if	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	285	6	some	some	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	285	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	285	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	285	9	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	285	10	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	285	11	were	be	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	285	12	correct	correct	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	285	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	286	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	286	2	iii	iii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	286	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	286	4	Levenshtein	Levenshtein	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	286	5	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	6	joins	join	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	286	8	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	9	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	286	10	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	286	12	special	special	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	13	symbol	symbol	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	14	#	#	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	15	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	16	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	286	17	single	single	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	18	string	string	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	19	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	286	20	computes	compute	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	286	22	Levenshtein	Levenshtein	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	286	23	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	24	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	286	25	predicted	predicted	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	26	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	286	27	gold	gold	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	286	28	strings	string	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	286	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	287	1	Discussion	discussion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	287	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	288	1	Results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	288	2	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	288	3	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	288	4	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	288	5	show	show	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	288	6	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	288	7	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	288	8	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	288	9	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	288	10	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	288	11	improves	improve	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	288	12	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	288	13	ability	ability	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	288	14	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	288	15	analyze	analyze	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	288	16	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	288	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	289	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	2	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	289	4	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	289	5	proposed	propose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	6	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	289	7	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	289	9	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	289	10	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	289	11	Work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	12	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	289	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	289	14	outperforms	outperform	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	289	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	289	16	baseline	baseline	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	17	supervised	supervise	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	289	18	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	289	19	segmenter	segmenter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	289	21	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	289	22	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	289	23	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	24	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	289	25	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	26	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	289	27	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	289	28	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	289	29	art	art	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	30	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	289	32	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	289	34	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	35	.05	.05	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	289	36	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	289	37	resp	resp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	289	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	290	1	.03	.03	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	290	2	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	290	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	290	4	accuracy	accuracy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	290	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	290	6	.03	.03	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	290	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	290	8	resp	resp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	290	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	291	1	.03	.03	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	291	2	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	291	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	291	4	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	291	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	291	6	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	291	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	291	8	resp	resp	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	291	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	292	1	German	German	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	292	2	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	292	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	293	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	293	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	293	3	find	find	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	293	4	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	293	5	when	when	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	293	6	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	293	7	give	give	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	293	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	293	9	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	293	10	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	293	11	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	293	12	oracle	oracle	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	293	13	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	293	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	293	15	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	293	16	generally	generally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	293	17	help	help	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	293	18	less	less	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	293	19	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	293	20	.01	.01	.
work_cxyxl5vs6zd6xlanudcx46qoqy	293	21	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	293	22	resp	resp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	293	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	294	1	.02	.02	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	294	2	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	294	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	294	4	ac-	ac-	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	294	5	curacy	curacy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	294	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	294	7	.01	.01	.
work_cxyxl5vs6zd6xlanudcx46qoqy	294	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	294	9	resp	resp	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	294	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	295	1	.03	.03	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	295	2	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	295	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	295	4	F1	F1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	295	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	295	6	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	295	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	295	8	resp	resp	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	295	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	296	1	German	German	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	296	2	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	296	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	297	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	297	2	indicates	indicate	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	297	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	297	5	chief	chief	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	6	boon	boon	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	297	8	vec-	vec-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	297	9	tor	tor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	10	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	11	factor	factor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	12	provides	provide	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	297	13	lies	lie	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	297	14	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	15	selection	selection	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	16	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	297	17	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	297	18	appropriate	appropriate	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	297	19	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	297	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	298	1	Moreover	moreover	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	298	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	298	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	298	4	up	up	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	5	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	298	6	.15	.15	.
work_cxyxl5vs6zd6xlanudcx46qoqy	298	7	differ-	differ-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	8	ence	ence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	9	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	10	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	298	11	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	12	systems	system	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	298	13	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	298	15	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	16	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	298	17	oracle	oracle	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	18	UR	UR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	298	19	suggests	suggest	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	298	20	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	21	reversing	reverse	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	298	22	orthographic	orthographic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	298	23	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	298	24	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	298	25	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	298	26	particularly	particularly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	298	27	difficult	difficult	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	298	28	part	part	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	29	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	30	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	298	31	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	298	33	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	34	least	least	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	298	35	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	298	36	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	298	37	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	299	1	6.2	6.2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	299	2	Experiment	experiment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	299	3	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	299	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	299	5	Vector	Vector	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	299	6	Approximation	Approximation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	299	7	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	299	8	adopt	adopt	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	299	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	299	10	experimental	experimental	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	299	11	design	design	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	299	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	299	13	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	299	14	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	299	15	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	299	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	300	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	300	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	300	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	300	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	301	1	Its	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	301	2	aim	aim	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	301	4	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	301	5	approximate	approximate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	301	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	301	7	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	301	10	derivationally	derivationally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	301	11	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	301	12	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	13	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	301	14	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	301	15	learned	learn	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	16	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	17	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	18	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	301	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	302	1	As	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	302	2	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	302	3	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	302	4	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	302	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	303	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	303	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	303	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	303	4	assume	assume	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	303	6	gold	gold	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	303	7	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	303	8	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	303	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	303	10	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	11	compare	compare	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	12	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	303	13	set-	set-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	14	tings	ting	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	303	15	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	303	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	303	17	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	303	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	303	19	oracle	oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	20	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	303	21	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	303	22	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	303	23	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	303	24	ii	ii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	25	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	303	26	in-	in-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	303	27	ferred	ferre	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	303	28	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	303	29	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	303	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	304	1	To	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	304	3	best	good	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	304	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	5	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	304	6	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	304	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	304	9	ii	ii	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	304	11	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	304	12	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	304	13	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	304	14	experimental	experimental	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	304	15	condition	condition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	16	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	17	no	no	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	304	18	previous	previous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	304	19	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	20	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	304	21	addressed	address	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	304	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	305	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	305	2	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	305	3	four	four	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	305	4	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	305	5	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	305	6	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	305	7	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	305	8	Ta-	Ta-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	305	9	ble	ble	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	305	10	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	305	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	305	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	306	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	306	2	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	306	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	306	4	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	306	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	306	6	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	306	7	just	just	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	306	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	306	9	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	306	10	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	306	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	307	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	307	2	baseline	baseline	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	3	tells	tell	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	307	4	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	307	5	what	what	WP
work_cxyxl5vs6zd6xlanudcx46qoqy	307	6	happens	happen	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	307	7	if	if	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	307	9	make	make	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	307	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	307	11	incor-	incor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	12	rect	rect	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	13	assumption	assumption	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	14	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	15	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	16	behaves	behave	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	307	17	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	307	18	inflec-	inflec-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	307	19	tion	tion	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	307	20	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	307	21	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	307	22	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	307	23	meaning	mean	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	307	24	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	307	25	changing	change	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	307	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	308	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	308	2	ii	ii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	308	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	308	4	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	308	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	308	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	308	7	purely	purely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	308	8	additive	additive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	308	9	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	308	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	309	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	309	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	309	3	arguably	arguably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	309	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	309	5	simplest	simple	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	309	6	way	way	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	309	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	309	8	combining	combine	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	309	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	309	10	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	309	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	309	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	309	13	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	309	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	310	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	310	2	iii	iii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	310	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	310	4	LDS	LDS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	310	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	310	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	310	7	linear	linear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	310	8	dynamical	dynamical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	310	9	system	system	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	310	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	311	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	311	2	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	311	3	arguably	arguably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	311	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	311	5	sim-	sim-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	311	6	plest	pl	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	311	7	sequence	sequence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	311	8	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	311	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	312	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	312	2	iv	iv	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	312	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	312	4	A	a	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	312	5	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	312	6	simple	simple	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	312	7	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	312	8	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	312	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	313	1	Recur-	Recur-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	2	41	41	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	313	3	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	313	4	HR	hr	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	313	5	LR	LR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	6	-less	-less	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	7	in-	in-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	8	un-	un-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	313	9	L	l	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	313	10	az	az	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	313	11	ar	ar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	12	i	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	13	d	d	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	14	ou	ou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	15	stem	stem	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	16	.47	.47	.
work_cxyxl5vs6zd6xlanudcx46qoqy	313	17	.52	.52	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	313	18	.32	.32	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	313	19	.22	.22	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	313	20	.39	.39	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	21	.33	.33	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	22	mult	mult	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	23	.39	.39	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	24	.43	.43	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	313	25	.28	.28	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	313	26	.23	.23	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	313	27	.34	.34	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	28	.33	.33	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	29	dil	dil	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	313	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	314	1	.48	.48	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	2	.53	.53	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	3	.33	.33	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	4	.30	.30	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	5	.45	.45	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	314	6	.41	.41	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	7	wadd	wadd	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	8	.50	.50	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	9	.55	.55	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	10	.38	.38	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	314	11	.24	.24	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	12	.40	.40	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	13	.34	.34	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	14	fulladd	fulladd	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	15	.56	.56	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	16	.61	.61	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	17	.41	.41	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	18	.38	.38	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	19	.47	.47	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	20	.44	.44	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	21	lexfunc	lexfunc	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	22	.54	.54	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	23	.58	.58	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	24	.42	.42	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	25	.44	.44	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	26	.45	.45	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	314	27	.46	.46	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	28	B	b	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	29	O	o	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	30	W	w	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	31	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	32	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	33	.43	.43	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	34	.44	.44	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	35	.38	.38	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	36	.32	.32	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	37	.43	.43	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	38	.51	.51	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	39	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	314	40	.65	.65	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	41	.67	.67	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	42	.61	.61	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	43	.60	.60	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	44	.64	.64	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	45	.67	.67	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	46	LDS	lds	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	47	.67	.67	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	48	.69	.69	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	49	.62	.62	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	50	.61	.61	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	51	.66	.66	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	52	.67	.67	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	53	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	54	.67	.67	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	55	.69	.69	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	56	.60	.60	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	57	.60	.60	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	58	.65	.65	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	59	.66	.66	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	60	c	c	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	314	61	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	62	GRU	gru	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	63	.59	.59	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	64	.60	.60	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	65	.55	.55	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	66	.59	.59	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	314	67	.55	.55	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	68	.57	.57	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	69	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	70	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	71	LSTM	LSTM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	72	.52	.52	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	73	.53	.53	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	314	74	.50	.50	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	75	.55	.55	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	76	.50	.50	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	77	.50	.50	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	78	B	b	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	79	O	O	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	80	W	w	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	81	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	82	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	83	.40	.40	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	84	.43	.43	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	85	.33	.33	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	86	.27	.27	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	87	.37	.37	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	88	.46	.46	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	89	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	314	90	.56	.56	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	91	.59	.59	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	92	.51	.51	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	93	.46	.46	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	94	.55	.55	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	95	.59	.59	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	96	LDS	LDS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	97	.58	.58	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	98	.61	.61	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	99	.51	.51	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	100	.48	.48	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	101	.57	.57	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	102	.60	.60	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	103	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	104	.58	.58	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	105	.61	.61	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	106	.50	.50	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	107	.48	.48	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	108	.56	.56	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	109	.58	.58	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	110	c	c	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	111	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	112	GRU	gru	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	113	.45	.45	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	114	.47	.47	.
work_cxyxl5vs6zd6xlanudcx46qoqy	314	115	.42	.42	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	116	.42	.42	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	117	.43	.43	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	118	.45	.45	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	119	c	c	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	120	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	121	LSTM	LSTM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	122	.46	.46	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	123	.47	.47	.
work_cxyxl5vs6zd6xlanudcx46qoqy	314	124	.43	.43	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	125	.43	.43	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	126	.45	.45	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	127	.46	.46	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	128	D	d	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	129	E	e	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	130	P	p	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	131	s	s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	132	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	133	.46	.46	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	134	.45	.45	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	135	.49	.49	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	136	.38	.38	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	137	.57	.57	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	138	.67	.67	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	139	add	add	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	314	140	.79	.79	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	141	.79	.79	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	142	.77	.77	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	143	.78	.78	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	144	.80	.80	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	145	.80	.80	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	146	LDS	lds	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	147	.80	.80	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	148	.81	.81	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	149	.77	.77	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	150	.79	.79	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	151	.81	.81	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	152	.81	.81	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	153	RNN	rnn	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	314	154	.81	.81	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	155	.82	.82	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	156	.77	.77	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	157	.79	.79	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	158	.80	.80	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	159	.81	.81	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	160	c	c	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	161	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	162	GRU	gru	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	163	.75	.75	.
work_cxyxl5vs6zd6xlanudcx46qoqy	314	164	.76	.76	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	165	.72	.72	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	166	.78	.78	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	167	.74	.74	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	168	.75	.75	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	169	c	c	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	170	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	171	LSTM	LSTM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	172	.75	.75	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	173	.76	.76	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	174	.71	.71	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	175	.77	.77	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	176	.72	.72	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	177	.73	.73	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	178	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	179	4	4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	314	180	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	314	181	Vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	182	approximation	approximation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	183	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	314	184	measured	measure	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	185	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	186	mean	mean	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	314	187	cosine	cosine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	188	similarity	similarity	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	189	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	314	190	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	191	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	192	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	193	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	194	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	314	195	train	train	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	196	/	/	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	314	197	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	198	split	split	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	199	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	314	200	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	201	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	314	202	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	314	203	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	315	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	315	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	315	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	315	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	316	1	HR	HR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	316	2	/	/	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	316	3	LR	LR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	316	4	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	316	5	high	high	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	316	6	/	/	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	316	7	low	low	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	316	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	316	9	relatedness	relatedness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	316	10	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	316	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	317	1	See	see	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	317	2	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	317	3	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	317	4	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	317	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	318	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	318	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	318	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	318	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	318	5	details	detail	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	318	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	319	1	rent	rent	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	319	2	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	319	3	networks	network	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	319	4	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	319	5	currently	currently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	319	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	319	7	most	most	RBS
work_cxyxl5vs6zd6xlanudcx46qoqy	319	8	widely	widely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	319	9	used	use	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	319	10	nonlinear	nonlinear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	319	11	sequence	sequence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	319	12	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	319	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	319	14	simple	simple	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	319	15	RNNs	rnn	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	319	16	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	319	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	319	18	simplest	simple	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	319	19	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	319	20	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	319	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	320	1	Part	part	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	320	4	motivation	motivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	6	considering	consider	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	320	7	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	320	8	richer	rich	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	320	9	class	class	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	11	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	320	12	lies	lie	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	320	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	14	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	320	15	removal	removal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	16	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	320	18	two-	two-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	320	19	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	320	20	assumption	assumption	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	320	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	321	1	Indeed	indeed	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	321	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	321	3	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	5	unclear	unclear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	6	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	321	8	wadd	wadd	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	321	10	fulladd	fulladd	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	11	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	321	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	321	13	Mitchell	Mitchell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	321	15	Lapata	Lapata	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	321	17	2008	2008	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	321	18	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	321	19	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	20	useful	useful	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	21	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	321	22	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	23	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	321	24	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	25	case	case	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	26	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	27	multi-	multi-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	28	morphemic	morphemic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	29	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	321	30	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	321	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	321	32	weights	weight	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	321	33	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	34	tied	tie	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	35	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	36	position	position	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	37	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	321	38	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	321	39	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	321	40	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	321	41	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	42	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	43	’s	’s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	321	44	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	45	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	321	46	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	321	47	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	321	48	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	321	49	prefix	prefix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	50	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	321	51	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	52	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	321	53	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	54	always	always	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	321	55	multiplied	multiply	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	56	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	57	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	321	58	same	same	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	321	59	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	321	60	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	322	1	Comparison	comparison	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	322	2	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	322	3	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	322	4	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	322	5	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	322	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	323	1	To	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	323	2	compare	compare	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	323	3	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	323	4	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	323	5	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	323	6	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	323	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	324	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	324	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	324	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	324	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	324	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	324	6	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	324	7	their	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	324	8	exact	exact	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	324	9	train	train	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	324	10	/	/	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	324	11	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	324	12	split	split	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	324	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	325	1	Those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	325	2	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	325	3	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	325	4	reported	report	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	325	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	325	6	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	325	7	4	4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	325	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	326	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	326	2	dataset	dataset	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	326	3	enforces	enforce	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	326	4	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	326	5	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	326	6	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	326	7	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	326	8	composed	compose	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	326	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	326	10	exactly	exactly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	326	11	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	326	12	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	326	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	327	1	Thus	thus	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	327	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	327	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	327	4	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	327	5	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	327	6	unques-	unques-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	327	7	tionably	tionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	327	8	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	327	9	segmented	segment	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	327	10	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	327	11	un+questionably	un+questionably	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	327	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	327	13	with-	with-	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	327	14	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	327	15	further	further	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	327	16	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	327	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	328	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	328	2	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	328	3	employed	employ	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	328	4	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	328	5	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	328	6	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	328	7	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	328	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	329	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	329	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	329	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	329	4	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	329	5	high	high	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	329	6	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	329	7	dimensional	dimensional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	329	8	count	count	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	9	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	329	10	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	11	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	12	lemmatized	lemmatized	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	329	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	329	14	POS	POS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	329	15	tagged	tag	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	16	text	text	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	17	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	18	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	329	19	before	before	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	20	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	329	21	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	329	22	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	329	23	after	after	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	24	window	window	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	25	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	26	size	size	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	329	27	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	329	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	330	1	They	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	330	2	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	330	3	apply	apply	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	330	4	pointwise	pointwise	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	5	mutual	mutual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	330	6	informa-	informa-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	7	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	330	9	PMI	PMI	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	330	10	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	330	11	weighting	weighting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	330	13	dimensionality	dimensionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	14	reduction	reduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	15	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	16	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	330	17	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	330	18	negative	negative	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	330	19	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	20	factorization	factorization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	330	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	331	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	2	contrast	contrast	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	331	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	5	employ	employ	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	6	WORD2VEC	word2vec	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	331	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	331	8	Mikolov	Mikolov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	9	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	331	10	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	11	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	331	13	2013a	2013a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	331	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	331	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	331	16	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	331	17	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	18	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	331	19	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	331	20	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	331	21	interpretable	interpretable	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	331	22	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	23	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	331	24	factorization	factorization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	25	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	26	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	331	27	PMI	PMI	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	28	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	331	29	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	331	30	Levy	Levy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	31	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	331	32	Goldberg	Goldberg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	331	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	331	34	2014b	2014b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	331	35	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	331	36	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	332	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	2	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	3	sider	sider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	4	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	332	5	WORD2VEC	WORD2VEC	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	6	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	332	7	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	332	8	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	332	9	bag	bag	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	10	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	332	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	12	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	332	13	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	332	15	BOW	BOW	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	332	17	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	332	18	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	19	before	before	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	332	20	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	332	21	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	332	22	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	332	23	after	after	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	24	windows	window	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	332	25	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	26	size	size	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	27	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	332	28	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	332	29	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	332	30	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	332	31	DEPs	dep	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	332	32	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	332	33	Levy	Levy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	34	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	332	35	Goldberg	Goldberg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	36	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	332	37	2014a	2014a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	332	38	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	332	39	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	332	40	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	332	41	dependency	dependency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	42	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	332	43	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	44	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	45	whose	whose	WP$
work_cxyxl5vs6zd6xlanudcx46qoqy	332	46	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	47	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	332	48	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	49	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	50	dependency	dependency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	51	parses	parse	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	332	52	rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	332	53	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	332	54	BOW	BOW	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	332	55	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	333	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	2	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	333	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	333	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	333	5	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	333	6	indicate	indicate	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	333	7	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	333	9	key	key	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	10	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	11	better	well	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	333	12	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	13	approximation	approximation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	14	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	333	15	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	333	16	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	333	17	richer	rich	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	333	18	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	19	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	20	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	333	22	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	333	23	rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	333	24	lies	lie	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	333	25	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	26	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	333	27	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	333	28	them-	them-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	333	29	selves	self	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	333	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	334	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	334	2	find	find	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	334	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	334	4	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	334	5	best	good	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	334	6	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	334	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	334	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	334	9	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	334	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	334	11	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	334	12	marginally	marginally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	334	13	edges	edge	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	334	14	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	334	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	334	16	LDS	LDS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	334	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	335	1	Additionally	additionally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	335	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	335	3	looking	look	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	335	4	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	335	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	335	6	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	335	7	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	335	8	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	335	9	column	column	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	335	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	335	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	335	12	DEPs	DEPs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	335	13	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	335	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	335	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	335	16	sim-	sim-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	335	17	ple	ple	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	335	18	additive	additive	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	335	19	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	335	20	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	335	21	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	335	22	≤.02	≤.02	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	335	23	lower	low	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	335	24	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	335	25	LDS	LDS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	335	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	336	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	336	2	comparison	comparison	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	336	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	336	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	336	5	observe	observe	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	336	6	large	large	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	336	7	differences	difference	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	336	8	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	336	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	336	10	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	336	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	337	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	337	2	RNN+DEPs	rnn+deps	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	337	3	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	337	5	.23	.23	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	6	bet-	bet-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	7	ter	ter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	8	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	337	10	BOW5	BOW5	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	11	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	337	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	337	13	.81	.81	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	14	vs.	vs.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	337	15	.58	.58	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	337	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	337	18	.14	.14	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	19	better	well	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	337	20	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	337	22	BOW2	BOW2	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	23	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	337	24	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	337	25	.81	.81	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	26	vs.	vs.	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	27	.67	.67	.
work_cxyxl5vs6zd6xlanudcx46qoqy	337	28	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	337	29	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	337	30	.25	.25	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	337	31	bet-	bet-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	32	ter	ter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	33	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	34	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	35	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	337	36	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	37	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	337	38	’s	’s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	337	39	best	good	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	337	40	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	41	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	337	42	.81	.81	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	43	vs.	vs.	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	337	44	.56	.56	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	337	45	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	337	46	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	338	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	338	2	wider	wide	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	338	3	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	338	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	338	5	BOW	BOW	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	338	6	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	338	7	5	5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	338	8	instead	instead	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	338	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	338	10	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	338	11	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	338	12	yields	yield	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	338	13	worse	bad	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	338	14	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	338	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	339	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	339	2	suggests	suggest	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	339	3	that	that	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	339	4	syntactic	syntactic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	339	5	infor-	infor-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	339	6	mation	mation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	339	7	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	339	8	at	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	339	9	least	least	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	339	10	positional	positional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	339	11	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	339	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	339	13	neces-	neces-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	339	14	sary	sary	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	339	15	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	339	16	improved	improved	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	339	17	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	339	18	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	339	19	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	339	20	composi-	composi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	339	21	tion	tion	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	339	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	340	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	340	2	test	test	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	3	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	340	4	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	340	5	annotated	annotate	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	7	relatedness	relatedness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	340	9	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	340	10	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	340	11	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	340	12	proxy	proxy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	13	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	14	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	340	15	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	340	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	341	1	HR	HR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	341	2	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	341	3	high-	high-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	341	4	relatedness	relatedness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	341	5	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	341	6	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	341	7	were	be	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	341	8	judged	judge	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	341	9	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	341	10	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	341	11	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	341	12	compo-	compo-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	341	13	sitional	sitional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	341	14	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	341	15	LR	LR	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	341	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	341	17	low	low	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	341	18	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	341	19	relatedness	relatedness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	341	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	341	21	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	341	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	342	1	Character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	342	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	342	3	Level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	342	4	Neural	Neural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	342	5	Retrofitting	Retrofitting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	342	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	343	1	As	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	2	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	343	3	fur-	fur-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	4	ther	ther	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	343	5	strong	strong	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	343	6	baseline	baseline	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	343	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	343	9	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	343	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	343	11	retrofitting	retrofitting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	343	13	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	343	14	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	343	15	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	343	16	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	343	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	343	18	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	343	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	343	20	approach	approach	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	21	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	22	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	23	character-	character-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	343	24	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	343	25	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	343	26	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	343	27	networks	network	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	343	28	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	344	1	Recently	recently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	344	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	344	3	running	run	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	344	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	344	5	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	344	6	net	net	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	7	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	344	9	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	10	stream	stream	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	11	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	344	12	become	become	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	344	14	popular	popular	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	344	15	way	way	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	16	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	17	incorporating	incorporate	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	344	18	subword	subword	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	344	19	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	20	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	21	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	344	22	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	23	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	344	24	empirical	empirical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	344	25	gains	gain	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	344	26	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	27	been	be	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	28	observed	observe	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	29	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	30	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	344	31	diverse	diverse	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	344	32	set	set	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	33	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	34	NLP	NLP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	35	tasks	task	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	344	36	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	344	37	POS	POS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	38	tagging	tag	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	344	39	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	344	40	dos	dos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	41	Santos	Santos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	42	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	344	43	Zadrozny	Zadrozny	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	44	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	344	45	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	344	46	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	344	47	Ling	Ling	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	48	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	49	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	50	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	51	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	344	52	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	344	53	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	344	54	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	344	55	pars-	pars-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	56	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	57	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	344	58	Ballesteros	Ballesteros	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	59	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	60	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	61	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	62	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	344	63	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	344	64	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	344	65	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	344	66	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	344	67	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	344	68	42	42	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	344	69	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	344	70	Kim	Kim	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	71	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	344	72	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	73	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	344	74	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	344	75	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	344	76	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	344	77	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	345	1	To	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	345	3	best	good	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	345	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	5	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	345	6	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	345	8	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	9	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	345	10	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	11	retrofitting	retrofitting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	12	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	345	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	345	14	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	345	15	approach	approach	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	345	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	346	1	Given	give	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	2	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	3	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	4	v	v	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	7	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	8	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	9	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	346	11	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	12	seek	seek	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	14	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	15	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	346	16	minimize	minimize	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	346	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	18	following	follow	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	346	19	objective	objective	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	20	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	346	21	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	346	22	||v	||v	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	346	23	−hN||22	−hN||22	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	346	25	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	346	26	10	10	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	346	27	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	346	28	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	346	29	hN	hN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	30	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	31	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	32	final	final	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	33	hidden	hidden	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	34	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	35	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	36	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	37	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	38	neu-	neu-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	39	ral	ral	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	40	architecture	architecture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	41	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	346	42	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	346	43	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	346	44	hi	hi	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	346	45	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	346	46	σ(Ahi−1	σ(Ahi−1	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	346	47	+	+	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	48	Bwi	Bwi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	49	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	346	50	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	346	51	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	346	52	11	11	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	346	53	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	346	54	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	346	55	σ	σ	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	56	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	57	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	58	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	59	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	60	linearity	linearity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	61	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	346	62	wi	wi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	63	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	64	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	65	ith	ith	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	66	char-	char-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	346	67	acter	acter	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	346	68	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	69	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	70	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	346	71	hi−1	hi−1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	72	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	73	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	346	74	previous	previous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	346	75	hidden	hide	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	76	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	77	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	346	78	A	a	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	79	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	346	80	B	b	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	346	81	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	346	82	matrices	matrix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	346	83	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	347	1	While	while	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	2	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	3	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	4	defined	define	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	347	6	architecture	architecture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	7	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	8	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	347	9	vanilla	vanilla	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	10	RNN	rnn	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	347	12	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	13	experiment	experiment	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	14	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	15	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	347	16	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	347	17	advanced	advanced	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	347	18	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	347	19	architectures	architecture	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	347	20	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	347	21	GRUs	GRUs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	22	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	347	23	Cho	Cho	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	24	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	25	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	26	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	347	28	2014b	2014b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	347	29	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	347	30	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	347	31	LSTMs	LSTMs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	32	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	347	33	Hochreiter	Hochreiter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	34	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	347	35	Schmidhuber	Schmidhuber	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	36	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	347	37	1997	1997	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	347	38	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	347	39	as	as	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	347	40	well	well	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	347	41	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	347	42	deep	deep	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	347	43	variants	variant	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	347	44	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	347	45	Sutskever	Sutskever	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	46	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	47	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	48	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	49	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	347	50	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	347	51	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	347	52	Gillick	Gillick	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	53	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	54	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	55	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	56	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	347	57	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	347	58	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	347	59	Firat	Firat	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	60	et	et	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	347	61	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	62	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	347	63	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	347	64	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	347	65	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	347	66	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	348	1	Importantly	importantly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	348	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	348	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	348	4	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	348	5	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	348	6	no	no	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	348	7	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	348	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	348	9	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	348	10	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	348	11	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	348	12	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	348	13	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	348	14	rely	rely	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	348	15	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	348	16	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	348	17	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	348	18	extracts	extract	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	348	19	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	348	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	348	21	characters	character	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	348	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	349	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	349	2	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	349	3	us	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	349	4	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	349	5	clear	clear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	349	6	ablation	ablation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	349	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	349	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	349	9	benefit	benefit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	349	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	349	11	adding	add	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	349	12	structured	structure	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	349	13	morpho-	morpho-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	349	14	logical	logical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	349	15	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	349	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	350	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	350	2	optimize	optimize	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	350	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	350	4	depth	depth	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	350	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	350	7	size	size	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	350	10	hidden	hide	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	11	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	350	12	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	13	development	development	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	14	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	350	15	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	350	16	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	350	17	coarse	coarse	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	350	18	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	350	19	grained	grain	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	20	grid	grid	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	21	search	search	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	350	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	351	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	351	2	found	find	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	351	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	351	4	depth	depth	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	351	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	351	6	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	351	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	351	8	hidden	hide	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	351	9	units	unit	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	351	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	351	11	size	size	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	351	12	100	100	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	351	13	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	351	14	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	351	15	both	both	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	351	16	LSTM	LSTM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	351	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	351	18	GRU	GRU	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	351	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	351	20	performed	perform	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	351	21	best	well	RBS
work_cxyxl5vs6zd6xlanudcx46qoqy	351	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	352	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	352	2	trained	train	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	352	3	all	all	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	352	4	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	352	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	6	100	100	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	352	7	iterations	iteration	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	352	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	9	Adam	Adam	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	352	10	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	352	11	Kingma	Kingma	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	352	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	352	13	Ba	Ba	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	352	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	352	15	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	352	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	352	17	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	18	L2	l2	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	352	19	regularization	regularization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	20	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	21	regularization	regularization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	22	coefficient	coefficient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	352	23	0.01	0.01	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	352	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	353	1	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	353	2	4	4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	353	3	shows	show	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	353	4	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	353	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	353	6	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	353	7	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	353	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	353	9	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	353	10	mod-	mod-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	353	11	els	el	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	353	12	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	353	13	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	353	14	c	c	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	353	15	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	353	16	GRU	GRU	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	353	17	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	353	18	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	353	19	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	353	20	c	c	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	353	21	-	-	:
work_cxyxl5vs6zd6xlanudcx46qoqy	353	22	LSTM	LSTM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	353	23	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	353	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	353	25	perform	perform	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	353	26	much	much	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	353	27	worse	bad	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	353	28	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	353	29	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	353	30	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	353	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	354	1	This	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	354	2	indicates	indicate	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	354	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	4	supervised	supervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	354	5	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	6	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	354	7	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	8	produces	produce	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	354	9	higher	high	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	354	10	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	354	11	quality	quality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	12	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	13	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	354	14	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	15	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	354	16	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	17	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	354	18	poor	poor	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	354	19	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	354	20	character-	character-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	354	21	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	354	22	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	354	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	355	1	However	however	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	355	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	355	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	355	4	note	note	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	355	5	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	6	these	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	355	7	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	355	9	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	10	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	355	11	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	355	12	fewer	few	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	355	13	parameters	parameter	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	355	14	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	15	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	355	16	morpheme	morpheme	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	355	17	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	355	18	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	19	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	355	20	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	355	21	there	there	EX
work_cxyxl5vs6zd6xlanudcx46qoqy	355	22	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	355	23	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	355	24	more	more	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	355	25	morphemes	morpheme	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	355	26	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	27	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	355	28	languages	language	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	355	29	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	355	30	characters	character	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	355	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	356	1	Oracle	Oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	356	2	Morphology	Morphology	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	356	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	357	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	357	2	general	general	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	357	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	357	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	357	5	two-	two-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	357	6	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	357	7	assumption	assumption	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	357	8	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	357	9	incorrect	incorrect	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	357	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	358	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	358	2	consider	consider	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	358	3	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	358	4	expanded	expand	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	358	5	setting	setting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	358	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	358	7	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	358	8	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	358	9	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	358	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	359	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	359	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	359	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	359	4	’s	’s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	359	5	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	359	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	359	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	359	8	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	359	9	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	359	10	fully	fully	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	359	11	decompose	decompose	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	359	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	359	13	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	359	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	359	15	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	359	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	359	17	unquestionably7→un+question+able+ly	unquestionably7→un+question+able+ly	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	359	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	360	1	These	these	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	360	2	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	360	3	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	360	4	reported	report	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	360	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	360	6	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	360	7	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	360	8	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	360	9	top	top	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	360	10	block	block	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	360	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	360	12	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	360	13	oracle	oracle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	360	14	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	360	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	360	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	361	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	361	2	report	report	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	361	3	mean	mean	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	361	4	cosine	cosine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	361	5	similarity	similarity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	361	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	362	1	Standard	standard	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	362	2	devia-	devia-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	362	3	tions	tion	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	362	4	s	s	POS
work_cxyxl5vs6zd6xlanudcx46qoqy	362	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	6	10-fold	10-fold	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	362	7	cross	cross	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	8	-	-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	9	validation	validation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	10	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	362	11	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	362	12	shown	show	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	13	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	362	14	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	362	15	small	small	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	362	16	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	362	17	≤	≤	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	362	18	.012	.012	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	362	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	362	20	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	21	two	two	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	362	22	exceptions	exception	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	362	23	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	362	24	s	s	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	362	25	=	=	SYM
work_cxyxl5vs6zd6xlanudcx46qoqy	362	26	.044	.044	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	362	27	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	28	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	362	29	DEPs	deps	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	362	30	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	362	31	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	362	32	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	362	33	stem	stem	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	362	34	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	362	35	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	362	36	.411	.411	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	362	37	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	362	38	.412	.412	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	362	39	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	362	40	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	363	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	363	2	multi	multi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	363	3	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	363	4	morphemic	morphemic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	363	5	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	363	6	mirror	mirror	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	363	7	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	363	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	363	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	363	10	bi	bi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	363	11	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	363	12	morphemic	morphemic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	363	13	setting	setting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	363	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	363	15	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	363	16	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	363	17	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	363	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	364	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	364	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	364	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	364	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	365	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	365	2	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	365	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	365	4	RNN+DEPs	rnn+deps	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	365	5	attains	attain	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	365	6	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	365	7	average	average	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	365	8	cosine	cosine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	365	9	similarity	similarity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	365	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	365	11	around	around	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	365	12	.80	.80	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	365	13	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	365	14	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	365	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	366	1	Numbers	number	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	366	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	366	3	German	German	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	366	4	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	366	5	lower	low	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	366	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	366	7	around	around	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	366	8	.70	.70	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	366	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	367	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	367	2	ii	ii	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	367	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	367	4	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	367	5	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	367	6	only	only	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	367	7	marginally	marginally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	367	8	edges	edge	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	367	9	out	out	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	367	10	LDS	LDS	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	367	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	367	12	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	367	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	367	14	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	367	15	slightly	slightly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	367	16	worse	bad	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	367	17	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	367	18	German	German	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	367	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	368	1	Again	again	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	368	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	368	3	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	368	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	368	5	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	368	6	surprising	surprising	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	368	7	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	368	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	368	9	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	368	10	mod-	mod-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	368	11	eling	ele	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	368	12	short	short	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	368	13	sequences	sequence	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	368	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	369	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	369	2	iii	iii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	369	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	369	4	Certain	certain	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	369	5	embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	369	6	lend	lend	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	369	7	themselves	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	369	8	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	369	9	naturally	naturally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	369	10	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	369	11	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	369	12	composi-	composi-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	369	13	tionality	tionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	369	14	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	369	15	BOW2	BOW2	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	369	16	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	369	17	better	well	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	369	18	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	369	19	BOW5	BOW5	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	369	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	369	21	DEPs	DEPs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	369	22	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	369	23	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	369	24	clear	clear	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	369	25	winner	winner	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	369	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	370	1	Inferred	Inferred	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	370	2	Morphology	Morphology	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	370	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	371	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	371	2	final	final	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	371	3	setting	setting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	371	5	con-	con-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	6	sider	sider	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	7	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	371	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	371	9	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	10	approximation	approximation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	11	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	12	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	13	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	14	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	371	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	372	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	372	2	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	372	3	case	case	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	372	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	372	5	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	6	rely	rely	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	372	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	372	9	full	full	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	372	10	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	372	11	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	372	12	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	372	14	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	372	16	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	372	18	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	19	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	20	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	372	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	372	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	373	1	At	at	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	373	2	evaluation	evaluation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	373	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	373	4	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	5	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	6	in-	in-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	373	7	terested	tereste	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	373	8	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	373	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	373	10	marginal	marginal	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	373	11	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	373	12	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	13	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	14	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	15	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	373	16	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	17	∑	∑	.
work_cxyxl5vs6zd6xlanudcx46qoqy	373	18	s	s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	373	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	373	20	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	373	22	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	23	p(v	p(v	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	373	25	s	s	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	373	27	l	l	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	373	29	u	u	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	30	|	|	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	31	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	373	32	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	373	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	374	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	2	then	then	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	374	3	use	use	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	4	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	5	sampling	sampling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	6	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	374	7	approximate	approximate	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	374	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	374	9	mean	mean	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	11	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	374	12	marginal	marginal	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	374	13	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	14	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	374	16	predicted	predict	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	17	embedding	embedding	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	374	19	i.e.	i.e.	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	374	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	374	21	v̂	v̂	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	22	=	=	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	23	∫	∫	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	24	vp(v	vp(v	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	374	25	|	|	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	374	26	w)dv	w)dv	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	27	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	28	12	12	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	374	29	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	30	≈	≈	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	31	1∑M	1∑m	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	374	32	i=1	i=1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	33	w	w	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	34	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	35	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	36	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	37	M∑	M∑	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	38	i=1	i=1	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	39	w(i)Cβ(l(i),s(i	w(i)cβ(l(i),s(i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	40	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	41	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	42	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	374	43	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	44	13	13	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	374	45	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	46	where	where	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	374	47	w(i	w(i	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	48	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	49	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	50	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	374	51	importance	importance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	52	weights	weight	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	374	53	defined	define	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	54	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	55	Equation	Equation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	56	8	8	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	374	57	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	374	58	l(i	l(i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	59	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	60	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	374	61	s(i	s(i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	62	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	374	63	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	64	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	374	65	ith	ith	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	374	66	sampled	sample	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	67	la-	la-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	374	68	beling	beling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	69	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	374	70	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	374	71	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	374	72	respectively	respectively	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	374	73	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	375	1	Discussion	discussion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	375	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	376	1	Surprisingly	surprisingly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	376	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	376	3	Table	table	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	376	4	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	376	5	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	376	6	joint	joint	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	376	7	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	376	8	shows	show	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	376	9	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	376	10	relying	rely	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	376	11	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	376	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	376	13	inferred	infer	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	376	14	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	376	15	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	376	16	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	376	17	drastically	drastically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	376	18	affect	affect	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	376	19	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	376	20	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	376	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	377	1	Indeed	indeed	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	377	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	377	3	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	377	4	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	377	5	often	often	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	377	6	within	within	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	7	.01	.01	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	377	10	result	result	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	11	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	12	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	13	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	377	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	378	1	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	378	2	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	3	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	378	4	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	378	5	viewed	view	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	6	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	7	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	378	8	retrofitting	retrofitting	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	9	procedure	procedure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	10	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	378	11	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	378	12	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	378	13	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	378	14	.	.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	378	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	378	16	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	378	17	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	378	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	378	19	so	so	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	378	20	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	378	21	result	result	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	22	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	23	useful	useful	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	24	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	378	25	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	378	26	indi-	indi-	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	378	27	cates	cat	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	28	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	29	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	30	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	31	synthesis	synthesis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	32	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	378	33	morphologi-	morphologi-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	34	cal	cal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	378	35	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	36	produces	produce	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	37	high	high	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	378	38	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	378	39	quality	quality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	378	40	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	378	41	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	379	1	6.3	6.3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	379	2	Experiment	experiment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	379	3	3	3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	379	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	379	5	Derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	379	6	Productivity	Productivity	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	379	7	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	379	8	now	now	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	379	9	delve	delve	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	379	10	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	379	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	379	12	relation	relation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	379	13	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	379	14	distribu-	distribu-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	379	15	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	379	16	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	379	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	379	18	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	379	19	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	379	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	380	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	380	2	extent	extent	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	3	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	380	4	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	380	5	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	380	6	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	380	7	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	380	8	aids	aid	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	380	9	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	380	10	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	11	will	will	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	380	12	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	380	13	determined	determine	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	14	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	380	16	in-	in-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	380	17	herent	herent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	380	18	compositionality	compositionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	19	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	380	21	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	380	22	within	within	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	23	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	380	24	vec-	vec-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	380	25	tor	tor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	26	space	space	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	380	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	381	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	2	break	break	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	3	down	down	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	4	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	381	5	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	381	6	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	8	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	9	approximation	approximation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	10	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	11	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	12	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	13	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	14	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	381	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	16	43	43	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	17	-ly	-ly	:
work_cxyxl5vs6zd6xlanudcx46qoqy	381	18	-ness	-ness	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	381	19	-ize	-ize	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	381	20	-able	-able	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	381	21	in-	in-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	22	-ful	-ful	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	381	23	-ous	-ous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	24	-less	-less	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	25	-ic	-ic	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	381	26	-ist	-ist	:
work_cxyxl5vs6zd6xlanudcx46qoqy	381	27	-y	-y	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	381	28	un-	un-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	29	-ity	-ity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	30	re-	re-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	31	-ion	-ion	''
work_cxyxl5vs6zd6xlanudcx46qoqy	381	32	-ment	-ment	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	381	33	-al	-al	''
work_cxyxl5vs6zd6xlanudcx46qoqy	381	34	-er	-er	:
work_cxyxl5vs6zd6xlanudcx46qoqy	381	35	Affix	Affix	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	36	0.2	0.2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	37	0.3	0.3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	38	0.4	0.4	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	39	0.5	0.5	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	40	0.6	0.6	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	41	0.7	0.7	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	42	0.8	0.8	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	43	0.9	0.9	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	44	1.0	1.0	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	45	A	a	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	46	v	v	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	47	er	er	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	381	48	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	49	g	g	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	50	e	e	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	51	C	c	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	52	o	o	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	53	si	si	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	54	n	n	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	55	e	e	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	56	S	s	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	57	i	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	58	m	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	59	il	il	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	60	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	61	ri	ri	FW
work_cxyxl5vs6zd6xlanudcx46qoqy	381	62	ty	ty	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	63	Figure	figure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	64	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	381	65	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	381	66	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	67	boxplot	boxplot	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	68	breaks	break	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	69	down	down	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	70	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	71	cosine	cosine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	72	similarity	similarity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	73	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	74	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	75	approximated	approximated	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	381	76	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	77	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	381	78	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	381	79	target	target	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	80	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	81	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	82	affix	affix	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	381	83	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	381	84	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	381	85	gold	gold	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	86	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	381	87	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	381	88	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	382	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	382	2	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	382	3	ordered	order	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	382	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	382	5	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	382	6	such	such	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	382	7	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	382	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	382	9	better	well	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	382	10	approximated	approximate	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	382	11	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	382	12	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	382	13	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	382	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	382	15	left	left	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	382	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	383	1	dependency	dependency	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	383	2	vectors	vectors	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	383	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	383	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	383	5	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	383	6	composer	composer	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	383	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	383	8	Fig-	Fig-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	383	9	ure	ure	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	383	10	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	383	11	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	383	12	selected	select	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	383	13	affixes	affix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	383	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	384	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	384	2	observe	observe	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	384	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	384	4	wide	wide	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	384	5	range	range	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	7	scores	score	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	384	8	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	384	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	384	10	most	most	RBS
work_cxyxl5vs6zd6xlanudcx46qoqy	384	11	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	384	12	ending	ending	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	13	ly	ly	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	384	14	gives	give	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	384	15	rise	rise	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	16	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	17	cosine	cosine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	18	similarities	similarity	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	384	19	that	that	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	384	20	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	384	21	20	20	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	384	22	points	point	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	384	23	higher	high	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	384	24	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	25	those	those	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	384	26	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	384	27	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	384	28	least	least	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	384	29	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	384	30	er	er	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	384	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	385	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	385	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	385	3	left	left	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	385	4	end	end	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	385	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	385	6	Figure	figure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	385	7	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	385	8	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	385	9	see	see	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	385	10	extremely	extremely	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	385	11	pro-	pro-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	385	12	ductive	ductive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	385	13	suffixes	suffix	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	385	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	386	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	386	2	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	386	3	ize	ize	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	386	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	386	5	used	use	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	386	6	productively	productively	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	386	7	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	386	8	relatively	relatively	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	386	9	obscure	obscure	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	386	10	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	386	11	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	386	12	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	386	13	sciences	science	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	386	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	386	15	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	386	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	386	17	Rao	Rao	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	386	18	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	386	19	Blackwellize	Blackwellize	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	386	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	387	1	Likewise	likewise	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	387	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	387	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	387	4	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	5	ness	ness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	6	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	387	7	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	387	8	applied	apply	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	9	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	10	almost	almost	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	387	11	any	any	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	387	12	adjective	adjective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	13	without	without	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	14	restriction	restriction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	387	16	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	387	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	387	18	Poissonness	Poissonness	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	387	19	‘	'	``
work_cxyxl5vs6zd6xlanudcx46qoqy	387	20	degree	degree	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	21	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	387	22	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	387	23	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	387	24	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	387	25	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	387	26	Pois-	Pois-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	387	27	son	son	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	28	distribution	distribution	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	387	29	’	'	''
work_cxyxl5vs6zd6xlanudcx46qoqy	387	30	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	388	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	388	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	388	3	right	right	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	388	4	end	end	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	388	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	388	6	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	388	7	find	find	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	388	8	-ment	-ment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	388	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	388	10	-er	-er	:
work_cxyxl5vs6zd6xlanudcx46qoqy	388	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	388	12	re-	re-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	388	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	389	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	389	2	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	389	3	-ment	-ment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	389	4	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	389	5	borderline	borderline	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	389	6	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	389	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	389	8	Bauer	Bauer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	389	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	389	10	1983)—modern	1983)—modern	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	389	11	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	389	12	tends	tend	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	389	13	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	389	14	form	form	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	389	15	novel	novel	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	389	16	nominalizations	nominalization	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	389	17	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	389	18	ness	ness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	389	19	or	or	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	389	20	ity	ity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	389	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	390	1	More	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	390	2	interesting	interesting	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	390	3	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	390	4	re-	re-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	390	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	390	6	er	er	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	390	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	390	8	both	both	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	390	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	390	10	which	which	WDT
work_cxyxl5vs6zd6xlanudcx46qoqy	390	11	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	390	12	very	very	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	390	13	productive	productive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	390	14	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	390	15	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	390	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	391	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	391	2	er	er	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	391	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	391	4	many	many	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	391	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	391	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	391	7	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	391	8	bringing	bring	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	391	9	down	down	RP
work_cxyxl5vs6zd6xlanudcx46qoqy	391	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	391	11	average	average	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	391	12	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	391	13	simply	simply	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	391	14	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	391	15	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	391	16	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	391	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	392	1	For	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	2	ex-	ex-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	392	3	ample	ample	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	392	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	392	5	homer	homer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	392	6	‘	'	``
work_cxyxl5vs6zd6xlanudcx46qoqy	392	7	homerun	homerun	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	8	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	9	baseball	baseball	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	10	’	'	''
work_cxyxl5vs6zd6xlanudcx46qoqy	392	11	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	392	12	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	392	13	derived	derive	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	14	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	15	home+er	home+er	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	392	16	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	392	17	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	392	18	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	392	19	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	392	20	error	error	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	21	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	392	22	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	392	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	393	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	393	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	393	3	see	see	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	393	4	examples	example	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	393	5	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	393	6	cutter	cutter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	393	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	394	1	It	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	394	2	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	394	3	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	394	4	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	394	5	read-	read-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	6	ing	ing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	7	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	394	8	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	394	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	394	10	“	"	``
work_cxyxl5vs6zd6xlanudcx46qoqy	394	11	box	box	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	394	12	cutter	cutter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	394	13	”	"	''
work_cxyxl5vs6zd6xlanudcx46qoqy	394	14	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	394	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	394	16	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	394	17	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	394	18	frequently	frequently	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	394	19	occurs	occur	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	394	20	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	394	22	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	394	23	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	394	24	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	394	25	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	26	‘	'	``
work_cxyxl5vs6zd6xlanudcx46qoqy	394	27	type	type	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	28	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	29	boat	boat	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	394	30	’	'	''
work_cxyxl5vs6zd6xlanudcx46qoqy	394	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	395	1	Fi-	fi-	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	395	2	nally	nally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	395	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	395	4	proper	proper	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	395	5	nouns	noun	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	395	6	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	395	7	Homer	Homer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	395	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	395	9	Turner	Turner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	395	10	end	end	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	395	11	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	395	12	er	er	UH
work_cxyxl5vs6zd6xlanudcx46qoqy	395	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	395	14	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	395	15	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	395	16	experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	395	17	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	395	18	computed	compute	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	395	19	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	395	20	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	395	21	lowercased	lowercase	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	395	22	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	395	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	396	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	396	2	affix	affix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	396	3	re-	re-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	396	4	similarly	similarly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	396	5	has	have	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	396	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	396	7	large	large	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	396	8	number	number	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	396	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	396	10	non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	396	11	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	396	12	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	396	13	cases	case	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	396	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	396	15	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	396	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	396	17	remove	remove	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	396	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	396	19	relocate	relocate	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	396	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	396	21	remark	remark	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	396	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	397	1	Indeed	indeed	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	397	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	397	3	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	397	4	get	get	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	397	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	397	6	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	397	7	reading	reading	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	9	remove	remove	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	397	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	397	12	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	397	13	syllable	syllable	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	397	15	rather	rather	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	397	16	than	than	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	17	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	397	18	second	second	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	19	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	397	20	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	397	21	typically	typically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	397	22	stressed	stress	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	23	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	397	24	emphasize	emphasize	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	397	25	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	397	26	prefix	prefix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	397	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	398	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	398	2	finally	finally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	398	3	note	note	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	398	4	several	several	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	398	5	limitations	limitation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	398	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	398	7	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	398	8	exper-	exper-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	398	9	iment	iment	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	398	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	399	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	399	2	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	399	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	399	4	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	399	5	ability	ability	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	399	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	399	7	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	399	8	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	399	9	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	399	10	even	even	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	399	11	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	399	12	re-	re-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	399	13	current	current	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	399	14	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	399	15	network	network	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	399	16	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	399	17	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	399	18	model	model	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	399	19	transformations	transformation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	399	20	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	399	21	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	399	22	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	399	23	limited	limited	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	399	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	400	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	400	2	ii	ii	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	400	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	400	4	Our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	400	5	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	400	6	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	400	7	far	far	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	400	8	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	9	perfect	perfect	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	400	10	;	;	:
work_cxyxl5vs6zd6xlanudcx46qoqy	400	11	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	400	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	400	13	sparseness	sparseness	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	14	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	400	16	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	17	data	datum	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	400	18	af-	af-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	400	19	fects	fect	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	400	20	quality	quality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	21	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	400	22	some	some	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	400	23	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	24	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	400	25	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	400	26	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	27	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	400	28	corpus	corpus	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	400	29	are	be	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	400	30	rare	rare	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	400	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	401	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	401	2	iii	iii	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	401	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	401	4	Semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	401	5	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	401	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	401	7	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	401	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	401	9	only	only	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	401	10	crite-	crite-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	401	11	rion	rion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	401	12	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	401	13	productivity	productivity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	401	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	402	1	An	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	402	2	example	example	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	402	3	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	402	4	-th	-th	:
work_cxyxl5vs6zd6xlanudcx46qoqy	402	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	402	6	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	402	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	403	1	As	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	2	noted	note	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	3	earlier	early	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	403	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	403	5	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	403	6	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	403	7	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	403	8	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	403	10	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	11	like	like	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	12	warmth	warmth	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	403	14	but	but	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	403	15	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	403	16	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	403	17	not	not	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	403	18	be	be	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	403	19	used	use	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	403	20	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	403	21	form	form	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	403	22	new	new	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	403	23	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	403	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	404	1	7	7	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	404	2	Conclusion	Conclusion	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	404	3	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	404	4	have	have	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	404	5	presented	present	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	404	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	404	7	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	404	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	404	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	404	10	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	404	11	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	404	12	structure	structure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	404	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	404	14	derivationally	derivationally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	404	15	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	404	16	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	404	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	405	1	To	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	3	best	good	JJS
work_cxyxl5vs6zd6xlanudcx46qoqy	405	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	5	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	405	6	knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	405	8	this	this	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	9	is	be	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	405	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	11	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	405	12	attempt	attempt	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	13	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	405	14	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	405	15	consider	consider	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	405	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	405	17	within	within	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	18	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	19	single	single	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	405	20	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	405	22	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	405	23	i	i	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	24	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	405	25	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	26	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	27	phological	phological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	405	28	decomposition	decomposition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	29	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	30	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	31	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	32	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	33	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	405	34	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	405	35	ii	ii	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	405	36	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	405	37	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	38	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	405	39	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	40	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	41	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	405	42	resulting	result	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	405	43	anal-	anal-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	44	ysis	ysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	405	45	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	406	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	406	2	found	find	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	406	3	that	that	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	406	4	directly	directly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	406	5	modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	406	6	coherence	coherence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	406	7	increases	increase	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	406	8	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	406	9	accuracy	accuracy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	406	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	406	11	improving	improve	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	406	12	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	406	13	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	406	14	strong	strong	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	406	15	baseline	baseline	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	406	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	407	1	Also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	407	2	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	407	3	our	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	407	4	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	407	5	show	show	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	407	6	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	7	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	407	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	9	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	407	10	the-	the-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	11	art	art	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	12	performance	performance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	13	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	407	15	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	407	16	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	17	approxi-	approxi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	407	18	mation	mation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	407	19	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	20	introduced	introduce	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	21	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	407	22	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	407	23	et	et	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	407	24	al	al	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	407	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	408	1	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	408	2	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	408	3	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	408	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	409	1	Future	future	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	409	2	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	3	will	will	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	409	4	focus	focus	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	409	5	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	409	7	extension	extension	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	409	10	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	11	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	12	more	more	RBR
work_cxyxl5vs6zd6xlanudcx46qoqy	409	13	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	409	14	instances	instance	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	409	15	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	16	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	409	17	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	409	19	e.g.	e.g.	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	409	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	409	21	compounding	compounding	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	22	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	409	23	reduplication	reduplication	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	409	25	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	409	26	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	27	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	409	28	extension	extension	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	29	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	409	30	additional	additional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	409	31	languages	language	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	409	32	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	410	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	410	2	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	410	3	plan	plan	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	410	4	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	410	5	explore	explore	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	410	6	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	410	7	relation	relation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	410	8	between	between	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	410	9	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	410	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	410	11	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	410	12	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	410	13	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	410	14	greater	great	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	410	15	detail	detail	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	410	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	411	1	44	44	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	411	2	Acknowledgments	acknowledgment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	411	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	412	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	412	2	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	412	3	author	author	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	4	was	be	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	412	5	sup-	sup-	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	412	6	ported	port	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	7	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	8	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	412	9	DAAD	DAAD	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	10	Long	Long	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	11	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	412	12	Term	Term	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	13	Research	Research	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	14	Grant	Grant	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	15	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	412	16	an	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	412	17	NDSEG	NDSEG	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	18	fellowship	fellowship	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	19	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	412	20	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	412	21	second	second	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	412	22	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	23	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	412	24	Volk-	Volk-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	25	swagenstiftung	swagenstiftung	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	26	Opus	Opus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	27	Magnum	Magnum	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	412	28	grant	grant	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	412	29	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	413	1	We	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	413	2	would	would	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	413	3	also	also	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	413	4	like	like	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	413	5	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	413	6	thank	thank	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	413	7	action	action	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	413	8	editor	editor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	413	9	Regina	Regina	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	413	10	Barzilay	Barzilay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	413	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	413	12	suggesting	suggest	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	413	13	several	several	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	413	14	changes	change	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	413	15	we	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	413	16	incorporated	incorporate	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	413	17	into	into	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	413	18	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	413	19	work	work	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	413	20	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	413	21	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	413	22	three	three	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	413	23	anonymous	anonymous	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	413	24	reviewers	reviewer	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	413	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	414	1	References	References	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	2	Mohamed	Mohamed	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	3	Afify	Afify	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	414	5	Ruhi	Ruhi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	6	Sarikaya	Sarikaya	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	414	8	Hong	Hong	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	9	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	414	10	Kwang	Kwang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	11	Jeff	Jeff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	12	Kuo	Kuo	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	414	14	Laurent	Laurent	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	15	Besacier	Besacier	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	414	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	414	18	Yuqing	Yuqing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	19	Gao	Gao	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	414	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	415	1	2006	2006	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	415	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	416	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	416	3	use	use	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	5	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	416	6	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	7	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	8	dialectal	dialectal	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	416	9	Arabic	arabic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	416	10	speech	speech	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	11	recognition	recognition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	416	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	417	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	417	2	Ninth	Ninth	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	417	3	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	417	4	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	417	5	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	417	6	Spoken	Spoken	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	417	7	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	417	8	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	417	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	418	1	Mark	Mark	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	418	2	Aronoff	Aronoff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	418	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	419	1	1976	1976	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	419	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	420	1	Word	Word	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	420	2	Formation	Formation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	420	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	420	4	Generative	Generative	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	420	5	Grammar	Grammar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	420	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	421	1	MIT	MIT	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	421	2	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	421	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	422	1	Harald	Harald	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	2	Baayen	Baayen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	422	4	Richard	Richard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	5	Piepenbrock	Piepenbrock	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	422	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	422	8	Hedderik	Hedderik	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	9	van	van	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	10	Rijn	Rijn	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	422	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	423	1	1993	1993	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	423	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	424	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	424	2	CELEX	CELEX	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	424	3	lexical	lexical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	424	4	data	data	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	424	5	base	base	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	424	6	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	424	7	CD-	CD-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	424	8	ROM	ROM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	424	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	425	1	Miguel	Miguel	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	2	Ballesteros	Ballesteros	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	425	4	Chris	Chris	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	5	Dyer	Dyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	425	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	425	8	Noah	Noah	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	9	A.	A.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	10	Smith	Smith	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	425	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	426	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	426	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	427	1	Improved	improved	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	427	2	transition	transition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	427	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	427	4	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	427	5	parsing	parsing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	427	6	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	427	7	model-	model-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	427	8	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	427	9	characters	character	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	427	10	instead	instead	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	427	11	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	427	12	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	427	13	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	427	14	LSTMs	LSTMs	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	427	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	428	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	428	2	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	3	ceedings	ceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	428	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	428	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	428	6	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	428	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	428	9	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	10	Meth-	Meth-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	11	ods	od	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	428	12	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	428	13	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	14	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	15	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	428	17	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	428	18	349–359	349–359	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	428	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	428	20	Lisbon	Lisbon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	428	22	Portugal	Portugal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	428	24	September	September	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	428	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	429	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	429	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	429	3	Compu-	Compu-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	429	4	tational	tational	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	429	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	429	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	430	1	Laurie	Laurie	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	430	2	Bauer	Bauer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	430	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	431	1	1983	1983	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	431	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	432	1	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	432	2	Word	Word	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	432	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	432	4	Formation	Formation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	432	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	433	1	Cam-	Cam-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	433	2	bridge	bridge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	433	3	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	433	4	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	433	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	434	1	Yoshua	Yoshua	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	434	2	Bengio	Bengio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	434	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	434	4	Jean	Jean	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	434	5	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	434	6	Sébastien	Sébastien	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	434	7	Senecal	Senecal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	434	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	435	1	2003	2003	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	435	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	436	1	Quick	quick	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	436	2	training	training	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	436	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	436	4	probabilistic	probabilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	436	5	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	436	6	nets	net	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	436	7	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	436	8	impor-	impor-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	436	9	tance	tance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	436	10	sampling	sampling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	436	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	437	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	437	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	437	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	437	5	Ninth	Ninth	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	6	Interna-	Interna-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	7	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	437	8	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	9	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	437	10	Artificial	Artificial	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	11	Intelligence	Intelligence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	437	13	Statis-	Statis-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	437	14	tics	tic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	437	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	438	1	Jan	Jan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	438	2	A.	A.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	438	3	Botha	Botha	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	438	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	438	5	Phil	Phil	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	438	6	Blunsom	Blunsom	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	438	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	439	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	439	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	440	1	Compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	440	2	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	440	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	440	4	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	440	5	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	440	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	440	7	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	440	8	modelling	modelling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	440	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	441	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	441	2	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	441	3	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	441	4	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	441	5	Machine	Machine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	441	6	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	441	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	441	8	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	441	9	1899–1907	1899–1907	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	441	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	442	1	Kyunghyun	Kyunghyun	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	2	Cho	Cho	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	442	4	Bart	Bart	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	5	van	van	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	6	Merriënboer	merriënboer	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	442	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	442	8	Dzmitry	Dzmitry	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	9	Bah-	Bah-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	10	danau	danau	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	442	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	442	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	442	13	Yoshua	Yoshua	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	14	Bengio	Bengio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	442	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	443	1	2014a	2014a	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	443	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	444	1	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	444	2	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	444	3	properties	property	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	444	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	444	5	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	444	6	machine	machine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	444	7	translation	translation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	444	8	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	444	9	Encoder	encoder	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	444	10	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	444	11	decoder	decoder	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	444	12	ap-	ap-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	444	13	proaches	proache	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	444	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	445	1	Workshop	Workshop	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	445	2	On	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	445	3	Syntax	Syntax	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	445	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	445	5	Semantics	Semantics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	445	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	445	7	Struc-	Struc-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	445	8	ture	ture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	445	9	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	445	10	Statistical	Statistical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	445	11	Translation	Translation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	445	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	446	1	Kyunghyun	Kyunghyun	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	2	Cho	Cho	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	446	4	Bart	Bart	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	5	Van	Van	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	6	Merriënboer	merriënboer	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	446	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	446	8	Caglar	Caglar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	9	Gul-	Gul-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	10	cehre	cehre	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	446	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	446	12	Dzmitry	Dzmitry	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	13	Bahdanau	Bahdanau	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	446	15	Fethi	Fethi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	16	Bougares	Bougares	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	446	18	Holger	Holger	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	19	Schwenk	Schwenk	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	446	21	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	446	22	Yoshua	Yoshua	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	23	Bengio	Bengio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	446	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	447	1	2014b	2014b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	447	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	448	1	Learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	448	2	phrase	phrase	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	448	3	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	448	4	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	448	5	RNN	RNN	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	448	6	encoder	encoder	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	448	7	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	448	8	decoder	decoder	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	448	9	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	448	10	statistical	statistical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	448	11	machine	machine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	448	12	translation	translation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	448	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	449	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	449	2	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	449	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	449	4	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	449	5	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	449	6	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	449	7	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	449	8	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	449	9	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	449	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	450	1	Noam	Noam	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	450	2	Chomsky	Chomsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	450	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	450	4	Morris	Morris	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	450	5	Halle	Halle	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	450	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	451	1	1968	1968	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	451	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	452	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	452	2	sound	sound	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	452	3	pat-	pat-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	452	4	tern	tern	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	452	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	452	6	English	English	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	452	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	453	1	Harper	Harper	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	453	2	&	&	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	453	3	Row	Row	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	453	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	454	1	Noam	Noam	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	454	2	Chomsky	Chomsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	454	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	455	1	1965	1965	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	455	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	456	1	Aspects	aspect	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	456	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	456	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	456	4	Theory	Theory	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	456	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	456	6	Syntax	Syntax	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	456	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	457	1	MIT	MIT	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	457	2	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	457	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	458	1	Ann	Ann	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	458	2	Clifton	Clifton	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	458	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	458	4	Anoop	Anoop	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	458	5	Sarkar	Sarkar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	458	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	459	1	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	459	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	460	1	Combin-	Combin-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	460	2	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	460	3	morpheme	morpheme	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	4	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	460	5	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	6	machine	machine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	7	translation	translation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	8	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	9	post-	post-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	460	10	processing	processing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	11	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	460	12	prediction	prediction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	460	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	461	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	461	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	461	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	461	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	461	5	49th	49th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	461	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	461	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	461	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	461	12	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	13	putational	putational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	461	14	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	15	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	461	16	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	17	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	18	Technolo-	Technolo-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	19	gies	gy	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	461	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	461	21	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	461	22	32–42	32–42	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	461	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	461	24	Portland	Portland	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	461	26	Oregon	Oregon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	461	28	USA	USA	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	461	30	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	461	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	462	1	As-	as-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	462	2	sociation	sociation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	462	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	462	4	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	462	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	462	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	463	1	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	463	2	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	463	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	463	4	Nanyun	Nanyun	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	463	5	Peng	Peng	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	463	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	463	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	463	8	Jason	Jason	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	463	9	Eisner	Eisner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	463	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	464	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	464	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	465	1	Stochastic	stochastic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	465	2	contextual	contextual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	465	3	edit	edit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	465	4	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	465	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	465	6	probabilistic	probabilistic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	465	7	FSTs	fst	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	465	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	466	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	466	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	466	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	466	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	466	5	52nd	52nd	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	466	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	466	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	466	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	466	12	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	13	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	466	15	Vol-	Vol-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	16	ume	ume	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	17	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	466	18	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	466	19	Short	Short	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	20	Papers	Papers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	466	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	466	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	466	23	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	466	24	625–630	625–630	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	466	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	466	26	Baltimore	Baltimore	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	466	28	Maryland	Maryland	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	466	30	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	466	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	467	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	467	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	467	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	467	4	Lin-	Lin-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	467	5	guistics	guistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	467	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	468	1	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	2	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	468	4	Thomas	Thomas	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	5	Müller	Müller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	468	7	Alexander	Alexander	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	8	Fraser	Fraser	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	468	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	468	11	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	12	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	468	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	469	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	469	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	470	1	Labeled	label	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	470	2	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	470	3	seg-	seg-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	470	4	mentation	mentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	470	5	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	470	6	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	470	7	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	470	8	Markov	markov	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	470	9	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	470	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	471	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	471	2	Proceed-	Proceed-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	3	ings	ing	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	471	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	471	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	471	6	Nineteenth	Nineteenth	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	471	9	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	10	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	11	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	12	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	471	14	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	471	15	164–174	164–174	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	471	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	471	17	Beijing	Beijing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	471	19	China	China	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	471	21	July	July	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	471	22	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	472	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	472	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	472	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	472	4	Linguis-	Linguis-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	472	5	tics	tic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	472	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	473	1	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	473	2	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	473	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	473	4	Arun	Arun	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	473	5	Kumar	Kumar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	473	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	473	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	473	8	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	473	9	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	473	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	474	1	2016a	2016a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	474	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	475	1	Morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	475	2	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	475	3	inside	inside	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	475	4	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	475	5	out	out	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	475	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	476	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	476	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	476	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	476	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	476	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	476	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	476	8	Empiri-	Empiri-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	9	cal	cal	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	476	10	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	476	11	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	476	12	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	13	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	14	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	476	16	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	476	17	2325–2330	2325–2330	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	476	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	476	19	Austin	Austin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	476	21	Texas	Texas	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	476	23	November	November	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	476	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	477	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	477	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	477	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	477	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	477	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	478	1	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	478	2	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	478	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	478	4	Tim	Tim	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	478	5	Vieira	Vieira	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	478	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	478	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	478	8	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	478	9	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	478	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	479	1	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	479	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	480	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	480	2	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	480	3	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	480	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	480	5	orthography	orthography	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	480	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	480	7	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	480	8	seg-	seg-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	480	9	mentation	mentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	480	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	481	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	481	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	481	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	481	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	481	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	481	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	481	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	481	9	North	north	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	481	10	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	11	Chapter	Chapter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	481	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	481	14	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	15	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	481	16	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	17	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	481	18	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	481	19	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	20	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	21	Tech-	Tech-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	22	nologies	nologie	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	481	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	481	24	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	481	25	664–669	664–669	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	481	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	481	27	San	San	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	28	Diego	Diego	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	481	30	California	California	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	481	32	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	481	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	482	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	482	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	482	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	482	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	482	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	483	1	Mathias	Mathias	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	483	2	Creutz	Creutz	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	483	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	483	4	Krista	Krista	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	483	5	Lagus	Lagus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	483	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	484	1	2007	2007	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	484	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	485	1	Unsupervised	unsupervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	485	2	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	485	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	485	4	morpheme	morpheme	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	485	5	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	485	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	485	7	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	485	8	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	485	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	486	1	ACM	ACM	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	486	2	Transactions	Transactions	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	486	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	486	4	Speech	Speech	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	486	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	486	6	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	486	7	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	486	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	486	9	4(1):3	4(1):3	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	486	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	487	1	Cı́cero	Cı́cero	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	487	2	Nogueira	Nogueira	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	487	3	dos	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	487	4	Santos	Santos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	487	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	487	6	Bianca	Bianca	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	487	7	Zadrozny	Zadrozny	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	487	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	488	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	488	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	489	1	Learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	489	2	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	489	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	489	4	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	489	5	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	489	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	489	7	part	part	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	489	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	489	9	of-	of-	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	489	10	speech	speech	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	489	11	tagging	tagging	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	489	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	490	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	490	2	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	490	3	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	490	4	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	490	5	Ma-	Ma-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	490	6	chine	chine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	490	7	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	490	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	490	9	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	490	10	1818–1826	1818–1826	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	490	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	491	1	Markus	Markus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	491	2	Dreyer	Dreyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	491	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	491	4	Jason	Jason	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	491	5	Eisner	Eisner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	491	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	492	1	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	492	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	493	1	Discover-	Discover-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	493	2	ing	e	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	493	3	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	493	4	paradigms	paradigms	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	493	5	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	493	6	plain	plain	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	493	7	text	text	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	493	8	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	493	9	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	493	10	Dirichlet	Dirichlet	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	493	11	process	process	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	493	12	mixture	mixture	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	493	13	model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	493	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	494	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	494	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	494	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	494	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	494	5	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	494	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	494	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	494	8	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	494	9	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	494	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	494	11	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	494	12	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	494	13	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	494	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	494	15	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	494	16	616–627	616–627	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	494	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	495	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	495	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	495	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	495	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	495	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	495	6	July	July	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	495	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	496	1	Markus	Markus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	496	2	Dreyer	Dreyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	496	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	497	1	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	497	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	498	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	498	2	Non	non	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	498	3	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	498	4	parametric	parametric	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	498	5	Model	model	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	498	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	498	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	498	8	Discovery	Discovery	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	498	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	498	10	Inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	498	11	Paradigms	Paradigms	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	498	12	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	498	13	Plain	Plain	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	498	14	Text	Text	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	498	15	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	498	16	Graphical	Graphical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	498	17	Models	Models	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	498	18	over	over	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	498	19	Strings	Strings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	498	20	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	499	1	Ph.D.	Ph.D.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	499	2	thesis	thesis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	499	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	499	4	Johns	Johns	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	499	5	Hopkins	Hopkins	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	499	6	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	499	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	500	1	John	John	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	500	2	Duchi	Duchi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	500	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	500	4	Elad	Elad	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	500	5	Hazan	Hazan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	500	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	500	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	500	8	Yoram	Yoram	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	500	9	Singer	Singer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	500	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	501	1	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	501	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	502	1	Adaptive	adaptive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	502	2	subgradient	subgradient	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	502	3	methods	method	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	502	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	502	5	online	online	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	502	6	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	502	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	502	8	45	45	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	502	9	stochastic	stochastic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	502	10	optimization	optimization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	502	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	503	1	Journal	Journal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	503	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	503	3	Machine	Machine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	503	4	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	503	5	Research	Research	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	503	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	503	7	12:2121–2159	12:2121–2159	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	503	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	504	1	Jeffrey	Jeffrey	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	504	2	L.	L.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	504	3	Elman	Elman	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	504	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	505	1	1990	1990	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	505	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	506	1	Finding	find	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	506	2	structure	structure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	506	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	506	4	time	time	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	506	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	507	1	Cog-	cog-	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	507	2	nitive	nitive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	507	3	Science	Science	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	507	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	507	5	14(2):179–211	14(2):179–211	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	507	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	508	1	Manaal	Manaal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	2	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	508	4	Jesse	Jesse	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	5	Dodge	Dodge	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	508	7	Sujay	Sujay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	8	Kumar	Kumar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	9	Jauhar	Jauhar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	508	11	Chris	Chris	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	12	Dyer	Dyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	508	14	Eduard	Eduard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	15	Hovy	Hovy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	508	17	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	508	18	Noah	Noah	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	19	A.	A.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	20	Smith	Smith	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	508	21	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	509	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	509	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	510	1	Retrofitting	retrofit	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	510	2	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	510	3	vectors	vector	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	510	4	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	510	5	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	510	6	lexicons	lexicon	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	510	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	511	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	511	2	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	3	ceedings	ceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	511	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	511	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	511	6	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	511	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	511	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	511	10	North	North	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	11	Ameri-	Ameri-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	12	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	511	13	Chapter	chapter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	511	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	511	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	511	16	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	17	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	511	18	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	19	Lin-	Lin-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	20	guistics	guistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	511	21	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	511	22	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	23	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	24	Technologies	Technologies	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	511	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	511	26	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	511	27	1606	1606	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	511	28	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	511	29	1615	1615	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	511	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	511	31	Denver	Denver	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	511	33	Colorado	Colorado	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	511	35	May	May	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	36	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	511	37	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	511	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	512	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	512	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	512	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	512	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	512	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	513	1	Manaal	Manaal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	513	2	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	513	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	513	4	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	513	5	McDonald	McDonald	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	513	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	513	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	513	8	Radu	Radu	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	513	9	Soricut	Soricut	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	513	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	514	1	2016a	2016a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	514	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	515	1	Morpho	morpho	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	515	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	515	3	syntactic	syntactic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	515	4	lexicon	lexicon	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	515	5	generation	generation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	515	6	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	515	7	graph	graph	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	515	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	515	9	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	515	10	semi	semi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	515	11	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	515	12	supervised	supervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	515	13	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	515	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	516	1	Transactions	transaction	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	516	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	516	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	516	4	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	516	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	516	6	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	516	7	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	516	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	516	9	4:1	4:1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	516	10	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	516	11	16	16	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	516	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	517	1	Manaal	Manaal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	2	Faruqui	Faruqui	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	517	4	Yulia	Yulia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	5	Tsvetkov	Tsvetkov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	517	7	Graham	Graham	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	8	Neubig	Neubig	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	517	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	517	11	Chris	Chris	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	12	Dyer	Dyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	517	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	518	1	2016b	2016b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	518	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	519	1	Morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	519	2	inflection	inflection	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	519	3	gen-	gen-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	519	4	eration	eration	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	519	5	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	519	6	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	519	7	sequence	sequence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	519	8	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	519	9	sequence	sequence	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	519	10	learn-	learn-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	519	11	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	519	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	520	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	520	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	520	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	520	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	520	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	520	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	520	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	520	9	North	north	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	520	10	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	11	Chapter	Chapter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	520	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	520	14	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	15	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	520	16	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	17	putational	putational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	520	18	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	19	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	520	20	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	21	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	22	Technolo-	Technolo-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	23	gies	gy	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	520	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	520	25	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	520	26	634–643	634–643	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	520	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	520	28	San	San	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	29	Diego	Diego	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	520	31	California	California	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	520	33	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	520	34	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	521	1	As-	as-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	521	2	sociation	sociation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	521	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	521	4	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	521	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	521	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	522	1	Orhan	Orhan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	522	2	Firat	Firat	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	522	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	522	4	Kyunghyun	Kyunghyun	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	522	5	Cho	Cho	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	522	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	522	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	522	8	Yoshua	Yoshua	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	522	9	Bengio	Bengio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	522	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	523	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	523	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	524	1	Multi	multi	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	524	2	-	-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	524	3	way	way	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	524	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	524	5	multilingual	multilingual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	524	6	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	524	7	machine	machine	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	524	8	translation	translation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	524	9	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	524	10	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	524	11	shared	share	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	524	12	attention	attention	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	524	13	mechanism	mechanism	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	524	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	525	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	525	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	525	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	525	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	525	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	525	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	525	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	525	9	North	North	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	10	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	11	Chap-	Chap-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	12	ter	ter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	525	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	525	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	525	15	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	16	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	525	17	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	18	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	525	19	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	525	20	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	21	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	22	Technologies	Technologies	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	525	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	525	24	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	525	25	866–875	866–875	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	525	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	525	27	San	San	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	28	Diego	Diego	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	525	30	California	California	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	525	32	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	525	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	526	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	526	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	526	3	Computa-	Computa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	526	4	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	526	5	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	526	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	527	1	Gottlob	Gottlob	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	527	2	Frege	Frege	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	527	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	528	1	1892	1892	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	528	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	529	1	Über	Über	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	529	2	Begriff	Begriff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	529	3	und	und	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	529	4	Gegenstand	Gegenstand	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	529	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	530	1	Vierteljahresschrift	Vierteljahresschrift	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	530	2	für	für	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	530	3	wissenschaftliche	wissenschaftliche	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	530	4	Philosophie	Philosophie	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	530	5	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	530	6	16:192–205	16:192–205	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	530	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	531	1	Dan	Dan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	2	Gillick	Gillick	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	531	4	Cliff	Cliff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	5	Brunk	Brunk	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	531	7	Oriol	Oriol	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	8	Vinyals	Vinyals	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	531	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	531	11	Amarnag	Amarnag	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	12	Subramanya	Subramanya	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	531	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	532	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	532	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	533	1	Multilingual	multilingual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	533	2	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	533	3	process-	process-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	533	4	ing	ing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	533	5	from	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	533	6	bytes	byte	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	533	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	534	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	534	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	534	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	534	6	Confer-	Confer-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	7	ence	ence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	534	10	North	north	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	534	11	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	12	Chapter	Chapter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	13	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	14	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	534	15	Associa-	Associa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	16	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	17	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	18	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	19	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	534	20	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	534	21	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	22	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	23	Technologies	Technologies	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	534	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	534	25	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	534	26	1296–1306	1296–1306	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	534	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	534	28	San	San	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	29	Diego	Diego	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	534	31	Califor-	Califor-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	32	nia	nia	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	534	33	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	534	34	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	534	35	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	535	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	535	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	535	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	535	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	535	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	536	1	Christoph	Christoph	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	536	2	Goller	Goller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	536	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	536	4	Andreas	Andreas	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	536	5	Küchler	Küchler	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	536	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	537	1	1996	1996	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	537	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	538	1	Learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	538	2	task	task	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	538	4	dependent	dependent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	538	5	distributed	distribute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	6	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	538	7	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	8	back-	back-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	9	propagation	propagation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	10	through	through	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	11	structure	structure	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	538	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	539	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	539	2	IEEE	IEEE	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	539	3	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	539	4	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	539	5	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	539	6	Neural	Neural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	539	7	Networks	Networks	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	539	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	540	1	Kazuma	Kazuma	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	540	2	Hashimoto	Hashimoto	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	540	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	540	4	Yoshimasa	Yoshimasa	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	540	5	Tsuruoka	Tsuruoka	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	540	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	541	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	541	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	542	1	Adaptive	adaptive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	542	2	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	542	3	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	542	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	542	5	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	542	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	542	7	non-	non-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	542	8	compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	542	9	phrase	phrase	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	542	10	embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	542	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	543	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	543	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	543	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	543	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	543	5	54th	54th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	543	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	543	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	543	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	543	12	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	13	putational	putational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	543	14	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	543	16	Volume	volume	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	543	17	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	543	18	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	543	19	Long	Long	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	20	Papers	Papers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	543	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	543	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	543	23	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	543	24	205–215	205–215	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	543	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	543	26	Berlin	Berlin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	543	28	Germany	Germany	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	543	30	August	August	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	543	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	544	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	544	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	544	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	544	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	544	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	545	1	Martin	Martin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	545	2	Haspelmath	Haspelmath	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	545	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	545	4	Andrea	Andrea	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	545	5	Sims	Sims	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	545	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	546	1	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	546	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	547	1	Under-	Under-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	547	2	standing	stand	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	547	3	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	547	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	548	1	Routledge	Routledge	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	548	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	549	1	Irene	Irene	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	549	2	Heim	Heim	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	549	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	549	4	Angelika	Angelika	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	549	5	Kratzer	Kratzer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	549	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	550	1	1998	1998	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	550	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	551	1	Semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	551	2	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	551	3	Generative	Generative	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	551	4	Grammar	Grammar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	551	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	552	1	Blackwell	Blackwell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	552	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	553	1	Sepp	Sepp	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	553	2	Hochreiter	Hochreiter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	553	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	553	4	Jürgen	Jürgen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	553	5	Schmidhuber	Schmidhuber	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	553	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	554	1	1997	1997	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	554	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	555	1	Long	long	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	555	2	short	short	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	555	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	555	4	term	term	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	555	5	memory	memory	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	555	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	556	1	Neural	Neural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	556	2	Computation	Computation	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	556	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	556	4	9(8):1735	9(8):1735	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	556	5	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	556	6	1780	1780	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	556	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	557	1	Katharina	Katharina	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	557	2	Kann	Kann	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	557	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	557	4	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	557	5	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	557	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	557	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	557	8	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	557	9	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	557	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	558	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	558	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	559	1	Neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	559	2	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	559	3	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	559	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	559	5	Encoding-	Encoding-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	559	6	decoding	decode	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	559	7	canonical	canonical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	559	8	segments	segment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	559	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	560	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	560	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	560	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	560	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	560	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	560	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	560	8	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	9	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	560	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	560	11	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	12	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	13	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	560	15	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	560	16	961–967	961–967	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	560	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	560	18	Austin	Austin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	560	20	Texas	Texas	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	560	22	November	November	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	560	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	561	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	561	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	561	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	561	4	Linguis-	Linguis-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	561	5	tics	tic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	561	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	562	1	Ronald	Ronald	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	562	2	M.	M.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	562	3	Kaplan	Kaplan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	562	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	562	5	Martin	Martin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	562	6	Kay	Kay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	562	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	563	1	1994	1994	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	563	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	564	1	Regular	regular	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	564	2	mod-	mod-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	564	3	els	el	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	564	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	564	5	phonological	phonological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	564	6	rule	rule	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	564	7	systems	system	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	564	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	565	1	Computational	computational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	565	2	Lin-	Lin-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	565	3	guistics	guistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	565	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	565	5	20(3):331–378	20(3):331–378	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	565	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	566	1	Martin	Martin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	566	2	Kay	Kay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	566	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	567	1	1977	1977	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	567	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	568	1	Morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	568	2	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	568	3	syntactic	syntactic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	568	4	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	568	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	569	1	Linguistic	Linguistic	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	569	2	Structures	Structures	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	569	3	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	569	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	569	5	5:131	5:131	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	569	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	570	1	Yoon	Yoon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	2	Kim	Kim	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	570	4	Yacine	Yacine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	5	Jernite	Jernite	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	570	7	David	David	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	8	Sontag	Sontag	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	570	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	570	11	Alexan-	Alexan-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	12	der	der	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	13	M.	M.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	14	Rush	Rush	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	570	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	571	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	571	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	572	1	Character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	572	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	572	3	aware	aware	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	572	4	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	572	5	language	language	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	572	6	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	572	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	573	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	573	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	573	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	573	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	573	5	Thirtieth	Thirtieth	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	573	6	AAAI	AAAI	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	573	7	Confer-	Confer-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	573	8	ence	ence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	573	9	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	573	10	Artificial	Artificial	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	573	11	Intelligence	Intelligence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	573	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	573	13	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	573	14	2741–2749	2741–2749	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	573	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	574	1	Diederik	Diederik	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	574	2	Kingma	Kingma	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	574	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	574	4	Jimmy	Jimmy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	574	5	Ba	Ba	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	574	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	575	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	575	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	576	1	Adam	Adam	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	576	2	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	576	3	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	576	4	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	576	5	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	576	6	stochastic	stochastic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	576	7	optimization	optimization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	576	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	577	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	577	2	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	577	3	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	577	4	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	577	5	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	577	6	Representations	Representations	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	577	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	578	1	Max	Max	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	2	Kisselew	Kisselew	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	578	4	Sebastian	Sebastian	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	5	Padó	Padó	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	578	7	Alexis	Alexis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	8	Palmer	Palmer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	578	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	578	11	Jan	Jan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	12	Šnajder	Šnajder	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	578	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	579	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	579	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	580	1	Obtaining	obtain	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	580	2	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	580	3	better	well	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	580	4	understanding	understanding	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	580	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	580	6	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	580	7	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	580	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	580	9	German	german	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	580	10	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	580	11	morphol-	morphol-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	580	12	ogy	ogy	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	580	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	581	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	581	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	581	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	581	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	581	5	11th	11th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	581	6	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	581	7	Confer-	Confer-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	581	8	ence	ence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	581	9	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	581	10	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	581	11	Semantics	Semantics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	581	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	581	13	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	581	14	58–63	58–63	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	581	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	582	1	Daphne	Daphne	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	582	2	Koller	Koller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	582	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	582	4	Nir	Nir	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	582	5	Friedman	Friedman	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	582	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	583	1	2009	2009	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	583	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	584	1	Probabilistic	Probabilistic	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	584	2	Graphical	Graphical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	584	3	Models	Models	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	584	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	584	5	Principles	Principles	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	584	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	584	7	Techniques	technique	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	584	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	585	1	MIT	MIT	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	585	2	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	585	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	586	1	Lingpeng	Lingpeng	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	2	Kong	Kong	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	586	4	Chris	Chris	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	5	Dyer	Dyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	586	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	586	8	Noah	Noah	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	9	A	A	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	10	Smith	Smith	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	586	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	587	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	587	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	588	1	Segmental	segmental	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	588	2	recurrent	recurrent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	588	3	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	588	4	networks	network	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	588	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	589	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	589	2	4th	4th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	589	3	Interna-	Interna-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	589	4	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	589	5	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	589	6	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	589	7	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	589	8	Representations	Representations	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	589	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	590	1	Mikko	Mikko	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	2	Kurimo	Kurimo	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	590	4	Sami	Sami	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	5	Virpioja	Virpioja	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	590	7	Ville	Ville	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	8	Turunen	Turunen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	590	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	590	11	Krista	Krista	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	12	Lagus	Lagus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	590	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	591	1	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	591	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	592	1	Morpho	Morpho	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	592	2	Challenge	Challenge	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	592	3	competition	competition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	592	4	2005	2005	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	592	5	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	592	6	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	592	7	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	592	8	Evaluations	evaluation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	592	9	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	592	10	results	result	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	592	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	593	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	593	2	Special	Special	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	593	3	Interest	Interest	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	593	4	Group	Group	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	593	5	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	593	6	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	593	7	Morphology	Morphology	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	593	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	593	9	Phonology	Phonology	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	593	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	594	1	Angeliki	Angeliki	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	2	Lazaridou	Lazaridou	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	594	4	Marco	Marco	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	5	Marelli	Marelli	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	594	7	Roberto	Roberto	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	8	Zamparelli	Zamparelli	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	594	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	594	11	Marco	Marco	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	12	Baroni	Baroni	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	594	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	595	1	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	595	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	596	1	Compositional	compositional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	596	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	596	3	ly	ly	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	596	4	derived	derive	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	596	5	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	596	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	596	7	morphologically	morphologically	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	596	8	complex	complex	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	596	9	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	596	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	596	11	distributional	distributional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	596	12	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	596	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	597	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	597	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	597	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	597	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	597	5	51st	51st	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	597	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	597	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	597	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	597	12	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	13	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	597	15	Volume	volume	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	597	16	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	597	17	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	597	18	Long	Long	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	19	Papers	Papers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	597	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	597	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	597	22	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	597	23	1517	1517	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	597	24	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	597	25	1526	1526	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	597	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	597	27	Sofia	Sofia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	28	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	597	29	Bulgaria	Bulgaria	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	597	31	August	August	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	597	32	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	598	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	598	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	598	3	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	598	4	putational	putational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	598	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	598	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	599	1	Omer	Omer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	599	2	Levy	Levy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	599	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	599	4	Yoav	Yoav	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	599	5	Goldberg	Goldberg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	599	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	600	1	2014a	2014a	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	600	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	601	1	Dependency-	Dependency-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	601	2	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	601	3	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	601	4	embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	601	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	602	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	602	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	602	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	602	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	602	5	52nd	52nd	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	602	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	602	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	602	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	602	12	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	13	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	602	15	Volume	volume	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	602	16	2	2	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	602	17	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	602	18	Short	Short	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	19	Papers	Papers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	602	20	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	602	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	602	22	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	602	23	302–308	302–308	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	602	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	602	25	46	46	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	602	26	Baltimore	Baltimore	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	602	28	Maryland	Maryland	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	602	30	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	602	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	603	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	603	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	603	3	Computa-	Computa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	603	4	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	603	5	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	603	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	604	1	Omer	Omer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	604	2	Levy	Levy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	604	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	604	4	Yoav	Yoav	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	604	5	Goldberg	Goldberg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	604	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	605	1	2014b	2014b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	605	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	606	1	Neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	606	2	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	606	3	embedding	embed	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	606	4	as	as	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	606	5	implicit	implicit	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	606	6	matrix	matrix	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	606	7	factorization	factorization	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	606	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	607	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	607	2	Ad-	Ad-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	607	3	vances	vance	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	607	4	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	607	5	Neural	Neural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	607	6	Information	Information	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	607	7	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	607	8	Systems	Systems	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	607	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	607	10	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	607	11	2177–2185	2177–2185	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	607	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	608	1	Marc	Marc	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	608	2	Light	Light	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	608	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	609	1	1996	1996	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	609	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	610	1	Morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	610	2	cues	cue	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	610	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	610	4	lexical	lexical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	610	5	se-	se-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	610	6	mantics	mantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	610	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	611	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	611	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	611	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	611	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	611	5	34th	34th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	611	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	611	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	611	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	611	12	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	13	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	611	15	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	611	16	25–31	25–31	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	611	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	611	18	Santa	Santa	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	19	Cruz	Cruz	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	611	21	California	California	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	611	23	USA	USA	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	611	25	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	611	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	612	1	Associa-	Associa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	612	2	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	612	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	612	4	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	612	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	612	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	613	1	Wang	Wang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	2	Ling	Ling	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	4	Chris	Chris	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	5	Dyer	Dyer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	7	Alan	Alan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	8	W.	W.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	9	Black	Black	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	11	Isabel	Isabel	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	12	Trancoso	Trancoso	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	14	Ramon	Ramon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	15	Fermandez	Fermandez	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	17	Silvio	Silvio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	18	Amir	Amir	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	20	Luis	Luis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	21	Marujo	Marujo	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	613	23	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	613	24	Tiago	Tiago	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	25	Luis	Luis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	613	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	614	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	614	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	615	1	Finding	find	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	615	2	function	function	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	4	form	form	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	5	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	615	6	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	615	7	positional	positional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	615	8	character	character	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	9	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	615	10	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	11	open	open	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	615	12	vocabulary	vocabulary	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	615	13	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	14	representation	representation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	615	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	616	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	616	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	616	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	616	5	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	616	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	616	8	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	9	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	616	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	616	11	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	12	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	13	Process-	Process-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	14	ing	ing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	616	16	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	616	17	1520–1530	1520–1530	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	616	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	616	19	Lisbon	Lisbon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	616	21	Portugal	Portugal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	616	23	September	September	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	616	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	617	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	617	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	617	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	617	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	617	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	618	1	Thang	Thang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	618	2	Luong	Luong	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	618	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	618	4	Richard	Richard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	618	5	Socher	Socher	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	618	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	618	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	618	8	Christopher	Christopher	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	618	9	Man-	Man-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	618	10	ning	ning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	618	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	619	1	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	619	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	620	1	Better	well	JJR
work_cxyxl5vs6zd6xlanudcx46qoqy	620	2	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	620	3	representations	representation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	620	4	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	620	5	recur-	recur-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	620	6	sive	sive	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	620	7	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	620	8	networks	network	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	620	9	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	620	10	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	620	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	621	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	621	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	621	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	621	5	Seventeenth	Seventeenth	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	621	8	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	9	Nat-	Nat-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	10	ural	ural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	621	11	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	12	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	621	14	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	621	15	104–113	104–113	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	621	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	621	17	Sofia	Sofia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	621	19	Bul-	Bul-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	20	garia	garia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	621	22	August	August	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	621	23	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	622	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	622	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	622	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	622	4	Linguis-	Linguis-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	622	5	tics	tic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	622	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	623	1	John	John	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	623	2	Lyons	Lyons	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	623	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	624	1	1977	1977	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	624	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	625	1	Semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	625	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	626	1	Cambridge	Cambridge	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	626	2	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	626	3	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	626	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	627	1	Tomas	Tomas	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	2	Mikolov	Mikolov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	627	4	Kai	Kai	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	5	Chen	Chen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	627	7	Greg	Greg	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	8	Corrado	Corrado	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	627	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	627	11	Jeffrey	Jeffrey	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	12	Dean	Dean	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	627	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	628	1	2013a	2013a	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	628	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	629	1	Efficient	efficient	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	629	2	estimation	estimation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	629	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	629	4	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	629	5	representa-	representa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	629	6	tions	tion	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	629	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	629	8	vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	629	9	space	space	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	629	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	630	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	630	2	2th	2th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	630	3	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	630	4	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	630	5	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	630	6	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	630	7	Representations	Representations	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	630	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	631	1	Tomas	Tomas	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	2	Mikolov	Mikolov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	631	4	Ilya	Ilya	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	5	Sutskever	Sutskever	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	631	7	Kai	Kai	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	8	Chen	Chen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	631	10	Gregory	Gregory	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	11	S.	S.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	12	Corrado	Corrado	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	631	14	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	631	15	Jeffrey	Jeffrey	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	16	Dean	Dean	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	631	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	632	1	2013b	2013b	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	632	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	633	1	Distributed	distribute	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	633	2	rep-	rep-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	633	3	resentations	resentation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	633	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	633	5	words	word	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	633	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	633	7	phrases	phrase	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	633	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	633	9	their	-PRON-	PRP$
work_cxyxl5vs6zd6xlanudcx46qoqy	633	10	composi-	composi-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	633	11	tionality	tionality	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	633	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	634	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	634	2	Advances	advance	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	634	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	634	4	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	634	5	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	634	6	process-	process-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	634	7	ing	ing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	634	8	systems	system	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	634	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	634	10	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	634	11	3111–3119	3111–3119	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	634	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	635	1	Jeff	Jeff	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	635	2	Mitchell	Mitchell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	635	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	635	4	Mirella	Mirella	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	635	5	Lapata	Lapata	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	635	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	636	1	2008	2008	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	636	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	637	1	Vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	637	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	637	3	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	637	4	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	637	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	637	6	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	637	7	composition	composition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	637	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	638	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	638	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	638	4	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	638	6	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	7	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	638	9	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	638	10	236	236	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	638	11	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	638	12	244	244	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	638	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	638	14	Columbus	Columbus	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	638	16	Ohio	Ohio	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	638	18	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	638	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	639	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	639	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	639	3	Computa-	Computa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	639	4	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	639	5	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	639	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	640	1	Kevin	Kevin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	640	2	P.	P.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	640	3	Murphy	Murphy	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	640	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	641	1	2012	2012	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	641	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	642	1	Machine	Machine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	642	2	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	642	3	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	642	4	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	642	5	Proba-	Proba-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	642	6	bilistic	bilistic	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	642	7	Perspective	perspective	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	642	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	643	1	MIT	MIT	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	643	2	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	643	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	644	1	Jason	Jason	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	644	2	Naradowsky	Naradowsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	644	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	644	4	Sharon	Sharon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	644	5	Goldwater	Goldwater	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	644	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	645	1	2009	2009	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	645	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	646	1	Im-	im-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	646	2	proving	prove	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	646	3	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	646	4	induction	induction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	646	5	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	646	6	learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	646	7	spelling	spelling	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	646	8	rules	rule	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	646	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	647	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	647	2	Twenty	twenty	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	647	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	647	4	first	first	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	647	5	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	647	6	Joint	Joint	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	647	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	647	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	647	9	Artificial	Artificial	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	647	10	Intelligence	Intelligence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	647	11	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	647	12	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	647	13	1531–1536	1531–1536	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	647	14	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	648	1	Karthik	Karthik	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	2	Narasimhan	Narasimhan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	648	4	Damianos	Damianos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	5	Karakos	Karakos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	648	7	Richard	Richard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	8	Schwartz	Schwartz	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	648	10	Stavros	Stavros	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	11	Tsakalidis	Tsakalidis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	648	13	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	648	14	Regina	Regina	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	15	Barzilay	Barzilay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	648	16	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	649	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	649	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	650	1	Morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	650	2	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	650	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	650	4	keyword	keyword	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	650	5	spot-	spot-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	650	6	ting	ting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	650	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	651	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	651	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	651	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	651	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	651	5	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	651	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	651	8	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	9	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	651	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	651	11	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	12	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	13	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	651	15	EMNLP	EMNLP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	651	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	651	18	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	651	19	880–885	880–885	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	651	21	Doha	Doha	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	651	23	Qatar	Qatar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	651	25	October	October	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	651	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	652	1	As-	as-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	652	2	sociation	sociation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	652	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	652	4	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	652	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	652	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	653	1	Karthik	Karthik	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	653	2	Narasimhan	Narasimhan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	653	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	653	4	Regina	Regina	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	653	5	Barzilay	Barzilay	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	653	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	653	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	653	8	Tommi	Tommi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	653	9	Jaakkola	Jaakkola	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	653	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	654	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	654	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	655	1	An	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	655	2	unsupervised	unsupervised	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	655	3	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	655	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	655	5	uncov-	uncov-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	655	6	ering	ere	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	655	7	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	655	8	chains	chain	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	655	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	656	1	Transactions	transaction	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	656	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	656	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	656	4	Asso-	Asso-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	656	5	ciation	ciation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	656	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	656	7	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	656	8	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	656	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	656	10	3:157–167	3:157–167	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	656	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	657	1	OUP	OUP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	657	2	editors	editor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	657	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	658	1	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	658	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	659	1	New	New	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	659	2	Oxford	Oxford	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	659	3	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	659	4	Dictionary	Dictionary	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	659	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	660	1	Oxford	Oxford	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	660	2	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	660	3	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	660	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	661	1	Sebastian	Sebastian	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	2	Padó	Padó	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	661	4	Alexis	Alexis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	5	Palmer	Palmer	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	661	7	Max	Max	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	8	Kisselew	Kisselew	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	661	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	661	11	Jan	Jan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	12	Šnajder	Šnajder	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	661	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	662	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	662	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	663	1	Measuring	measure	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	663	2	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	663	3	content	content	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	663	4	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	663	5	as-	as-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	663	6	sess	sess	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	663	7	asymmetry	asymmetry	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	663	8	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	663	9	derivation	derivation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	663	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	664	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	664	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	664	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	664	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	664	5	11th	11th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	664	6	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	664	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	664	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	664	9	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	664	10	Se-	Se-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	664	11	mantics	mantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	664	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	665	1	Lawrence	Lawrence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	665	2	R.	R.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	665	3	Rabiner	Rabiner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	665	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	666	1	1989	1989	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	666	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	667	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	667	2	tutorial	tutorial	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	667	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	667	4	hidden	hidden	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	667	5	Markov	Markov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	667	6	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	667	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	667	8	selected	select	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	667	9	applications	application	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	667	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	667	11	speech	speech	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	667	12	recognition	recognition	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	667	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	668	1	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	668	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	668	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	668	4	Institute	Institute	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	668	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	668	6	Electrical	Electrical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	668	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	668	8	Electronics	Electronics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	668	9	Engineers	Engineers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	668	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	668	11	77(2):257–286	77(2):257–286	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	668	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	669	1	Pushpendre	Pushpendre	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	669	2	Rastogi	Rastogi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	669	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	669	4	Ryan	Ryan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	669	5	Cotterell	Cotterell	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	669	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	669	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	669	8	Jason	Jason	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	669	9	Eisner	Eisner	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	669	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	670	1	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	670	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	671	1	Weighting	weight	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	671	2	finite	finite	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	671	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	671	4	state	state	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	671	5	transductions	transduction	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	671	6	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	671	7	neu-	neu-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	671	8	ral	ral	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	671	9	context	context	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	671	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	672	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	672	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	672	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	672	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	672	5	2016	2016	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	672	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	7	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	672	8	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	672	9	North	north	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	672	10	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	11	Chapter	Chapter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	12	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	672	13	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	672	14	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	15	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	672	16	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	17	Linguistics	Linguistics	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	672	18	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	672	19	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	20	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	21	Tech-	Tech-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	22	nologies	nologie	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	672	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	672	24	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	672	25	623–633	623–633	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	672	26	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	672	27	San	San	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	28	Diego	Diego	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	672	30	California	California	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	31	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	672	32	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	672	33	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	673	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	673	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	673	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	673	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	673	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	674	1	Eric	Eric	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	674	2	S.	S.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	674	3	Ristad	Ristad	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	674	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	674	5	Peter	Peter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	674	6	N.	N.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	674	7	Yianilos	Yianilos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	674	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	675	1	1998	1998	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	675	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	676	1	Learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	676	2	string	string	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	676	3	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	676	4	edit	edit	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	676	5	distance	distance	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	676	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	677	1	IEEE	IEEE	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	677	2	Transactions	transaction	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	677	3	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	677	4	Pattern	Pattern	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	677	5	Analysis	Analysis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	677	6	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	677	7	Machine	Machine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	677	8	Intelligence	Intelligence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	677	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	677	10	20(5):522–532	20(5):522–532	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	677	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	678	1	Reuven	Reuven	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	678	2	Y.	Y.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	678	3	Rubinstein	Rubinstein	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	678	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	678	5	Dirk	Dirk	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	678	6	P.	P.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	678	7	Kroese	Kroese	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	678	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	679	1	2011	2011	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	679	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	680	1	Sim-	Sim-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	680	2	ulation	ulation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	680	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	680	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	680	5	Monte	Monte	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	680	6	Carlo	Carlo	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	680	7	method	method	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	680	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	681	1	John	John	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	681	2	Wiley	Wiley	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	681	3	&	&	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	681	4	Sons	Sons	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	681	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	682	1	Teemu	Teemu	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	2	Ruokolainen	Ruokolainen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	682	4	Oskar	Oskar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	5	Kohonen	Kohonen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	682	7	Sami	Sami	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	8	Virpioja	Virpioja	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	682	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	682	11	Mikko	Mikko	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	12	Kurimo	Kurimo	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	682	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	683	1	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	683	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	684	1	Supervised	supervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	684	2	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	684	3	seg-	seg-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	684	4	mentation	mentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	684	5	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	684	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	684	7	low	low	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	684	8	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	684	9	resource	resource	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	684	10	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	684	11	setting	set	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	684	12	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	684	13	con-	con-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	684	14	ditional	ditional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	684	15	random	random	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	684	16	fields	field	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	684	17	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	685	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	685	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	685	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	685	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	685	5	Sev-	Sev-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	6	enteenth	enteenth	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	685	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	685	9	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	10	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	11	Lan-	Lan-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	12	guage	guage	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	685	13	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	14	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	685	15	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	685	16	29–37	29–37	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	685	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	685	18	Sofia	Sofia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	685	20	Bulgaria	Bulgaria	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	685	22	Au-	Au-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	685	23	gust	gust	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	685	24	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	686	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	686	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	686	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	686	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	686	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	687	1	Gerard	Gerard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	687	2	Salton	Salton	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	687	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	687	4	editor	editor	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	687	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	688	1	1971	1971	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	688	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	689	1	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	689	2	SMART	SMART	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	689	3	Retrieval	Retrieval	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	689	4	System	System	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	689	5	—	—	:
work_cxyxl5vs6zd6xlanudcx46qoqy	689	6	Experiments	experiment	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	689	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	689	8	Automatic	Automatic	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	689	9	Document	Document	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	689	10	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	689	11	cessing	cessing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	689	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	690	1	Prentice	Prentice	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	690	2	Hall	Hall	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	690	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	691	1	Sunita	Sunita	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	691	2	Sarawagi	Sarawagi	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	691	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	691	4	William	William	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	691	5	W.	W.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	691	6	Cohen	Cohen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	691	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	692	1	2005	2005	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	692	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	693	1	Semi-	Semi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	693	2	Markov	Markov	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	693	3	conditional	conditional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	693	4	random	random	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	693	5	fields	field	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	693	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	693	7	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	693	8	ex-	ex-	XX
work_cxyxl5vs6zd6xlanudcx46qoqy	693	9	traction	traction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	693	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	694	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	694	2	Advances	advance	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	694	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	694	4	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	694	5	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	694	6	process-	process-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	694	7	ing	ing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	694	8	systems	system	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	694	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	694	10	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	694	11	1185–1192	1185–1192	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	694	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	695	1	Patrick	Patrick	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	695	2	Schone	Schone	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	695	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	695	4	Daniel	Daniel	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	695	5	Jurafsky	Jurafsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	695	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	696	1	2000	2000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	696	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	697	1	Knowledge-	Knowledge-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	697	2	free	free	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	697	3	induction	induction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	697	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	697	5	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	697	6	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	697	7	latent	latent	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	697	8	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	697	9	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	697	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	698	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	698	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	698	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	698	4	4th	4th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	698	5	Conference	conference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	698	6	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	698	7	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	698	8	putational	putational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	698	9	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	698	10	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	698	11	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	698	12	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	698	13	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	698	14	67–72	67–72	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	698	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	699	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	699	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	699	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	699	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	699	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	700	1	Patrick	Patrick	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	700	2	Schone	Schone	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	700	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	700	4	Daniel	Daniel	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	700	5	Jurafsky	Jurafsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	700	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	701	1	2001	2001	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	701	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	702	1	Knowledge-	Knowledge-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	702	2	free	free	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	702	3	induction	induction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	702	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	702	5	inflectional	inflectional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	702	6	morphologies	morphology	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	702	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	703	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	703	2	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	3	ceedings	ceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	703	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	703	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	703	6	Second	Second	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	703	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	703	10	North	North	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	11	American	American	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	12	47	47	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	703	13	Chapter	Chapter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	703	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	703	16	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	17	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	703	18	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	19	Linguis-	Linguis-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	20	tics	tic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	703	21	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	703	22	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	23	Technologies	Technologies	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	703	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	703	25	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	703	26	1–9	1–9	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	703	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	704	1	Associa-	Associa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	704	2	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	704	3	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	704	4	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	704	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	704	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	705	1	Wolfgang	Wolfgang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	705	2	Seeker	Seeker	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	705	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	705	4	Özlem	Özlem	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	705	5	Çetinoğlu	Çetinoğlu	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	705	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	706	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	706	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	707	1	A	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	707	2	graph-	graph-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	3	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	4	lattice	lattice	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	5	dependency	dependency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	6	parser	parser	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	7	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	8	joint	joint	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	707	9	morpholog-	morpholog-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	707	10	ical	ical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	707	11	segmentation	segmentation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	12	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	707	13	syntactic	syntactic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	707	14	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	707	15	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	708	1	Transac-	transac-	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	708	2	tions	tion	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	708	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	708	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	708	5	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	708	6	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	708	7	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	708	8	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	708	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	708	10	3:359–373	3:359–373	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	708	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	709	1	Radu	Radu	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	709	2	Soricut	Soricut	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	709	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	709	4	Franz	Franz	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	709	5	Och	Och	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	709	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	710	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	710	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	711	1	Unsupervised	unsupervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	711	2	mor-	mor-	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	711	3	phology	phology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	711	4	induction	induction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	711	5	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	711	6	word	word	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	711	7	embeddings	embedding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	711	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	712	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	712	2	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	3	ceedings	ceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	712	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	712	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	712	6	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	712	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	712	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	712	10	North	North	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	11	Ameri-	Ameri-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	12	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	712	13	Chapter	chapter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	712	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	712	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	712	16	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	17	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	712	18	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	19	Lin-	Lin-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	20	guistics	guistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	712	21	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	712	22	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	23	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	24	Technologies	Technologies	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	712	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	712	26	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	712	27	1627	1627	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	712	28	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	712	29	1637	1637	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	712	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	712	31	Denver	Denver	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	712	33	Colorado	Colorado	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	712	35	May	May	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	36	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	712	37	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	712	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	713	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	713	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	713	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	713	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	713	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	714	1	Ilya	Ilya	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	2	Sutskever	Sutskever	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	714	4	Oriol	Oriol	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	5	Vinyals	Vinyals	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	714	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	714	8	Quoc	Quoc	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	9	V.	V.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	10	Le	Le	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	714	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	715	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	715	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	716	1	Se-	se-	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	716	2	quence	quence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	716	3	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	716	4	sequence	sequence	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	716	5	learning	learn	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	716	6	with	with	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	716	7	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	716	8	networks	network	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	716	9	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	717	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	717	2	Advances	advance	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	717	3	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	717	4	Neural	Neural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	717	5	Information	Information	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	717	6	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	717	7	Systems	Systems	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	717	8	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	717	9	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	717	10	3104–3112	3104–3112	ADD
work_cxyxl5vs6zd6xlanudcx46qoqy	717	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	718	1	Charles	Charles	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	718	2	Sutton	Sutton	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	718	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	718	4	Andrew	Andrew	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	718	5	McCallum	McCallum	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	718	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	719	1	2006	2006	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	719	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	720	1	An	an	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	720	2	in-	in-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	720	3	troduction	troduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	720	4	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	720	5	conditional	conditional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	720	6	random	random	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	720	7	fields	field	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	720	8	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	720	9	relational	relational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	720	10	learning	learning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	720	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	721	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	721	2	Lise	Lise	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	3	Getoor	Getoor	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	721	5	Ben	Ben	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	6	Taskar	Taskar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	721	8	editors	editor	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	721	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	721	10	Introduction	introduction	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	721	11	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	721	12	Statistical	Statistical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	13	Relational	Relational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	14	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	721	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	721	16	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	721	17	93–128	93–128	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	721	18	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	722	1	MIT	MIT	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	722	2	Press	Press	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	722	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	723	1	Peter	Peter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	723	2	D.	D.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	723	3	Turney	Turney	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	723	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	723	5	Patrick	Patrick	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	723	6	Pantel	Pantel	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	723	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	724	1	2010	2010	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	724	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	725	1	From	from	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	2	fre-	fre-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	725	3	quency	quency	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	4	to	to	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	5	meaning	meaning	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	6	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	725	7	Vector	vector	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	8	space	space	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	9	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	725	10	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	725	11	semantics	semantic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	725	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	726	1	Journal	Journal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	726	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	726	3	Artificial	Artificial	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	726	4	Intelligence	Intelligence	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	726	5	Research	Research	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	726	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	726	7	37(1):141	37(1):141	LS
work_cxyxl5vs6zd6xlanudcx46qoqy	726	8	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	726	9	188	188	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	726	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	727	1	Antal	Antal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	727	2	van	van	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	727	3	den	den	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	727	4	Bosch	Bosch	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	727	5	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	727	6	Walter	Walter	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	727	7	Daelemans	Daelemans	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	727	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	728	1	1999	1999	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	728	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	729	1	Memory	memory	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	729	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	729	3	based	base	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	729	4	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	729	5	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	729	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	730	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	730	2	Proceed-	Proceed-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	3	ings	ing	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	730	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	730	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	730	6	37th	37th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	730	7	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	8	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	9	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	730	10	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	730	11	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	12	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	730	13	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	14	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	730	16	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	730	17	285–292	285–292	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	730	18	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	730	19	College	College	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	20	Park	Park	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	730	22	Maryland	Maryland	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	730	24	USA	USA	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	730	26	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	730	27	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	731	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	731	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	731	3	Compu-	Compu-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	731	4	tational	tational	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	731	5	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	731	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	732	1	Martin	Martin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	732	2	J.	J.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	732	3	Wainwright	Wainwright	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	732	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	732	5	Michael	Michael	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	732	6	I.	I.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	732	7	Jordan	Jordan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	732	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	733	1	2008	2008	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	733	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	734	1	Graphical	graphical	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	734	2	models	model	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	734	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	734	4	exponential	exponential	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	734	5	families	family	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	734	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	734	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	734	8	varia-	varia-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	734	9	tional	tional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	734	10	inference	inference	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	734	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	735	1	Foundations	foundation	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	735	2	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	735	3	Trends	Trends	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	735	4	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	735	5	Machine	Machine	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	735	6	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	735	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	735	8	1(1	1(1	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	735	9	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	735	10	2):1–305	2):1–305	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	735	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	736	1	Zhen	Zhen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	2	Wang	Wang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	736	4	Jianwen	Jianwen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	5	Zhang	Zhang	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	6	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	736	7	Jianlin	Jianlin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	8	Feng	Feng	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	9	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	736	10	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	736	11	Zheng	Zheng	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	12	Chen	Chen	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	736	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	737	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	737	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	738	1	Knowledge	knowledge	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	738	2	graph	graph	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	738	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	738	4	text	text	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	738	5	jointly	jointly	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	738	6	em-	em-	VBP
work_cxyxl5vs6zd6xlanudcx46qoqy	738	7	bedding	bed	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	738	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	739	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	739	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	739	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	739	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	739	5	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	739	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	739	8	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	9	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	739	10	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	739	11	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	12	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	13	Processing	Processing	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	14	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	739	15	EMNLP	EMNLP	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	16	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	739	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	739	18	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	739	19	1591–1601	1591–1601	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	739	20	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	739	21	Doha	Doha	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	739	23	Qatar	Qatar	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	24	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	739	25	October	October	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	739	26	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	740	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	740	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	740	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	740	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	740	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	741	1	Paul	Paul	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	741	2	J.	J.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	741	3	Werbos	Werbos	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	741	4	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	742	1	1990	1990	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	742	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	743	1	Backpropagation	backpropagation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	743	2	through	through	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	743	3	time	time	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	743	4	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	743	5	What	what	WP
work_cxyxl5vs6zd6xlanudcx46qoqy	743	6	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	743	7	does	do	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	743	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	743	9	how	how	WRB
work_cxyxl5vs6zd6xlanudcx46qoqy	743	10	to	to	TO
work_cxyxl5vs6zd6xlanudcx46qoqy	743	11	do	do	VB
work_cxyxl5vs6zd6xlanudcx46qoqy	743	12	it	-PRON-	PRP
work_cxyxl5vs6zd6xlanudcx46qoqy	743	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	744	1	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	744	2	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	744	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	744	4	Institute	Institute	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	744	5	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	744	6	Electrical	Electrical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	744	7	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	744	8	Electronics	Electronics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	744	9	Engineers	Engineers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	744	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	744	11	78(10):1550–1560	78(10):1550–1560	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	744	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	745	1	Richard	Richard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	745	2	Wicentowski	Wicentowski	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	745	3	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	746	1	2002	2002	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	746	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	747	1	Modeling	model	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	747	2	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	747	3	Learning	Learning	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	747	4	Multilingual	Multilingual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	747	5	Inflectional	Inflectional	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	747	6	Morphology	Morphology	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	747	7	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	747	8	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	747	9	Minimally	minimally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	747	10	Supervised	supervise	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	747	11	Framework	Framework	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	747	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	748	1	Ph.D.	Ph.D.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	748	2	thesis	thesis	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	748	3	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	748	4	Johns	Johns	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	748	5	Hopkins	Hopkins	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	748	6	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	748	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	749	1	Yadollah	Yadollah	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	749	2	Yaghoobzadeh	Yaghoobzadeh	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	749	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	749	4	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	749	5	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	749	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	750	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	750	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	751	1	Corpus	corpus	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	751	2	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	751	3	level	level	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	751	4	fine	fine	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	751	5	-	-	HYPH
work_cxyxl5vs6zd6xlanudcx46qoqy	751	6	grained	grain	VBN
work_cxyxl5vs6zd6xlanudcx46qoqy	751	7	entity	entity	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	751	8	typing	typing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	751	9	using	use	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	751	10	contex-	contex-	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	751	11	tual	tual	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	751	12	information	information	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	751	13	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	752	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	752	2	Proceedings	Proceedings	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	752	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	752	5	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	752	6	Confer-	Confer-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	7	ence	ence	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	752	8	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	752	9	Empirical	Empirical	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	10	Methods	Methods	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	752	11	in	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	752	12	Natural	Natural	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	13	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	14	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	15	cessing	cessing	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	752	16	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	752	17	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	752	18	715–725	715–725	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	752	19	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	752	20	Lisbon	Lisbon	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	21	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	752	22	Portugal	Portugal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	23	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	752	24	September	September	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	752	25	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	753	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	753	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	753	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	753	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	753	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	754	1	David	David	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	754	2	Yarowsky	Yarowsky	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	754	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	754	4	Richard	Richard	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	754	5	Wicentowski	Wicentowski	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	754	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	755	1	2000	2000	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	755	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	756	1	Min-	Min-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	756	2	imally	imally	RB
work_cxyxl5vs6zd6xlanudcx46qoqy	756	3	supervised	supervise	VBD
work_cxyxl5vs6zd6xlanudcx46qoqy	756	4	morphological	morphological	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	756	5	analysis	analysis	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	756	6	by	by	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	756	7	multi-	multi-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	756	8	modal	modal	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	756	9	alignment	alignment	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	756	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	757	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	757	2	The	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	757	3	38th	38th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	757	4	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	757	5	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	757	6	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	757	7	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	757	8	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	757	9	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	757	10	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	757	11	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	757	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	758	1	Wenpeng	Wenpeng	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	758	2	Yin	Yin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	758	3	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	758	4	Hinrich	Hinrich	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	758	5	Schütze	Schütze	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	758	6	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	759	1	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	759	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	760	1	Convolutional	convolutional	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	760	2	neural	neural	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	760	3	network	network	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	760	4	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	760	5	paraphrase	paraphrase	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	760	6	identification	identification	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	760	7	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	761	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	761	2	Pro-	Pro-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	3	ceedings	ceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	761	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	761	5	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	761	6	2015	2015	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	761	7	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	761	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	761	10	North	North	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	11	Ameri-	Ameri-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	12	can	can	MD
work_cxyxl5vs6zd6xlanudcx46qoqy	761	13	Chapter	chapter	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	761	14	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	761	15	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	761	16	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	17	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	761	18	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	19	Lin-	Lin-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	20	guistics	guistic	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	761	21	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	761	22	Human	Human	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	23	Language	Language	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	24	Technologies	Technologies	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	761	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	761	26	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	761	27	901	901	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	761	28	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	761	29	911	911	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	761	30	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	761	31	Denver	Denver	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	32	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	761	33	Colorado	Colorado	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	34	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	761	35	May	May	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	36	–	–	:
work_cxyxl5vs6zd6xlanudcx46qoqy	761	37	June	June	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	761	38	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	762	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	762	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	762	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	762	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	762	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	763	1	Britta	Britta	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	763	2	D.	D.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	763	3	Zeller	Zeller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	763	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	763	5	Jan	Jan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	763	6	Šnajder	Šnajder	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	763	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	763	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	763	9	Sebastian	sebastian	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	763	10	Padó.	padó.	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	764	1	2013	2013	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	764	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	765	1	DErivBase	DErivBase	NFP
work_cxyxl5vs6zd6xlanudcx46qoqy	765	2	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	765	3	Inducing	induce	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	765	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	765	5	evaluating	evaluate	VBG
work_cxyxl5vs6zd6xlanudcx46qoqy	765	6	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	765	7	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	765	8	morphology	morphology	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	765	9	resource	resource	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	765	10	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	765	11	German	German	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	765	12	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	766	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	766	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	766	3	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	766	4	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	766	5	51st	51st	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	766	6	Annual	Annual	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	7	Meeting	Meeting	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	8	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	766	9	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	766	10	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	11	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	766	12	Com-	Com-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	13	putational	putational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	766	14	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	15	(	(	-LRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	766	16	Volume	volume	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	766	17	1	1	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	766	18	:	:	:
work_cxyxl5vs6zd6xlanudcx46qoqy	766	19	Long	Long	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	20	Papers	Papers	NNPS
work_cxyxl5vs6zd6xlanudcx46qoqy	766	21	)	)	-RRB-
work_cxyxl5vs6zd6xlanudcx46qoqy	766	22	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	766	23	pages	page	VBZ
work_cxyxl5vs6zd6xlanudcx46qoqy	766	24	1201–1211	1201–1211	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	766	25	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	766	26	Sofia	Sofia	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	27	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	766	28	Bulgaria	Bulgaria	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	29	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	766	30	August	August	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	766	31	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	767	1	Association	Association	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	767	2	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	767	3	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	767	4	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	767	5	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	768	1	Britta	Britta	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	2	D.	D.	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	3	Zeller	Zeller	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	4	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	768	5	Sebastian	Sebastian	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	6	Padó	Padó	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	7	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	768	8	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	768	9	Jan	Jan	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	10	Šnajder	Šnajder	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	768	11	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	769	1	2014	2014	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	769	2	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	770	1	Towards	towards	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	770	2	semantic	semantic	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	770	3	validation	validation	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	770	4	of	of	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	770	5	a	a	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	770	6	derivational	derivational	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	770	7	lexicon	lexicon	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	770	8	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	771	1	In	in	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	771	2	Proceedings	proceeding	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	771	3	the	the	DT
work_cxyxl5vs6zd6xlanudcx46qoqy	771	4	25th	25th	JJ
work_cxyxl5vs6zd6xlanudcx46qoqy	771	5	International	International	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	6	Conference	Conference	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	7	on	on	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	771	8	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	9	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	10	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	771	11	pages	page	NNS
work_cxyxl5vs6zd6xlanudcx46qoqy	771	12	1728–1739	1728–1739	CD
work_cxyxl5vs6zd6xlanudcx46qoqy	771	13	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	771	14	Dublin	Dublin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	15	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	771	16	Ireland	Ireland	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	17	,	,	,
work_cxyxl5vs6zd6xlanudcx46qoqy	771	18	August	August	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	771	19	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	772	1	Dublin	Dublin	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	772	2	City	City	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	772	3	University	University	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	772	4	and	and	CC
work_cxyxl5vs6zd6xlanudcx46qoqy	772	5	Associa-	Associa-	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	772	6	tion	tion	NN
work_cxyxl5vs6zd6xlanudcx46qoqy	772	7	for	for	IN
work_cxyxl5vs6zd6xlanudcx46qoqy	772	8	Computational	Computational	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	772	9	Linguistics	Linguistics	NNP
work_cxyxl5vs6zd6xlanudcx46qoqy	772	10	.	.	.
work_cxyxl5vs6zd6xlanudcx46qoqy	773	1	48	48	CD