id	sid	tid	token	lemma	pos
work_wgi74n6oavhdxhlomowlxqjdhu	1	1	Transactions	transaction	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	1	2	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	1	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	1	4	Association	Association	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	1	5	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	1	6	Computational	Computational	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	1	7	Linguistics	Linguistics	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	1	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	1	9	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	1	10	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	1	11	2013	2013	CD
work_wgi74n6oavhdxhlomowlxqjdhu	1	12	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	1	13	1–12	1–12	CD
work_wgi74n6oavhdxhlomowlxqjdhu	1	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	2	1	Action	Action	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	2	2	Editor	Editor	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	2	3	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	2	4	Sharon	Sharon	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	2	5	Goldwater	Goldwater	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	2	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	3	1	Submitted	submit	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	3	2	11/2012	11/2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	3	3	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	3	4	Revised	revise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	3	5	1/2013	1/2013	CD
work_wgi74n6oavhdxhlomowlxqjdhu	3	6	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	3	7	Published	publish	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	3	8	3/2013	3/2013	CD
work_wgi74n6oavhdxhlomowlxqjdhu	3	9	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	4	1	c	c	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	4	2	©	©	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	4	3	2013	2013	CD
work_wgi74n6oavhdxhlomowlxqjdhu	4	4	Association	Association	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	4	5	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	4	6	Computational	Computational	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	4	7	Linguistics	Linguistics	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	4	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	5	1	Token	Token	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	2	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	5	3	Type	Type	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	4	Constraints	Constraints	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	5	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	6	Cross	Cross	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	7	-	-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	8	Lingual	Lingual	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	9	Part	Part	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	5	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	12	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	5	13	Speech	Speech	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	14	Tagging	Tagging	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	15	Oscar	Oscar	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	16	Täckström	Täckström	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	17	�	�	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	18	†∗	†∗	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	5	19	Dipanjan	Dipanjan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	20	Das‡	das‡	NN
work_wgi74n6oavhdxhlomowlxqjdhu	5	21	Slav	Slav	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	22	Petrov‡	Petrov‡	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	5	23	Ryan	Ryan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	24	McDonald‡	mcdonald‡	FW
work_wgi74n6oavhdxhlomowlxqjdhu	5	25	Joakim	Joakim	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	26	Nivre†∗	Nivre†∗	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	27	�	�	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	28	Swedish	Swedish	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	29	Institute	Institute	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	30	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	31	Computer	Computer	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	32	Science	Science	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	33	†	†	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	34	Department	Department	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	35	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	36	Linguistics	Linguistics	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	37	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	5	38	Philology	Philology	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	39	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	5	40	Uppsala	Uppsala	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	41	University	University	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	42	‡	‡	,
work_wgi74n6oavhdxhlomowlxqjdhu	5	43	Google	Google	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	44	Research	Research	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	45	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	5	46	New	New	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	47	York	York	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	48	oscar@sics.se	oscar@sics.se	CD
work_wgi74n6oavhdxhlomowlxqjdhu	5	49	{	{	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	5	50	dipanjand|slav|ryanmcd}@google.com	dipanjand|slav|ryanmcd}@google.com	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	51	joakim.nivre@lingfil.uu.se	joakim.nivre@lingfil.uu.se	NN
work_wgi74n6oavhdxhlomowlxqjdhu	5	52	Abstract	Abstract	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	5	53	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	5	54	consider	consider	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	5	55	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	5	56	construction	construction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	5	57	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	58	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	5	59	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	5	60	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	61	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	5	62	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	5	63	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	5	64	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	5	65	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	5	66	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	5	67	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	5	68	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	5	69	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	6	1	Recently	recently	RB
work_wgi74n6oavhdxhlomowlxqjdhu	6	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	6	3	manually	manually	RB
work_wgi74n6oavhdxhlomowlxqjdhu	6	4	constructed	construct	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	6	5	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	6	6	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	6	7	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	6	8	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	6	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	6	10	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	6	11	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	6	12	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	6	13	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	6	14	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	6	15	been	be	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	6	16	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	6	17	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	6	18	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	6	19	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	6	20	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	6	21	overcome	overcome	VB
work_wgi74n6oavhdxhlomowlxqjdhu	6	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	6	23	scarcity	scarcity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	6	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	6	25	annotated	annotate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	6	26	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	6	27	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	6	28	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	6	29	setting	setting	NN
work_wgi74n6oavhdxhlomowlxqjdhu	6	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	7	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	7	2	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	7	3	paper	paper	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	7	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	7	6	show	show	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	7	7	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	7	8	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	7	9	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	7	10	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	7	11	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	7	12	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	7	13	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	7	14	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	7	15	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	7	16	resource-	resource-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	7	17	rich	rich	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	7	18	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	19	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	20	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	7	21	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	7	22	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	23	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	7	24	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	7	25	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	26	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	27	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	7	28	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	29	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	7	30	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	7	31	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	7	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	8	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	8	2	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	8	3	several	several	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	8	4	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	8	5	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	8	6	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	8	7	end	end	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	8	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	8	9	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	8	10	particular	particular	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	8	11	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	8	12	par-	par-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	13	tially	tially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	8	14	observed	observe	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	8	15	conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	8	16	random	random	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	8	17	field	field	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	18	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	8	20	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	8	21	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	8	22	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	8	23	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	8	24	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	25	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	8	26	pro-	pro-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	27	vide	vide	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	8	28	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	8	29	partial	partial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	8	30	signal	signal	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	31	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	8	32	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	8	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	9	1	Averaged	average	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	9	2	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	9	3	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	9	4	previously	previously	RB
work_wgi74n6oavhdxhlomowlxqjdhu	9	5	studied	study	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	9	6	Indo	Indo	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	9	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	9	8	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	9	9	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	9	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	9	11	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	9	12	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	9	13	achieves	achieve	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	9	14	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	9	15	25	25	CD
work_wgi74n6oavhdxhlomowlxqjdhu	9	16	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	9	17	relative	relative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	9	18	error	error	NN
work_wgi74n6oavhdxhlomowlxqjdhu	9	19	reduction	reduction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	9	20	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	9	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	9	22	prior	prior	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	9	23	state	state	NN
work_wgi74n6oavhdxhlomowlxqjdhu	9	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	9	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	9	26	art	art	NN
work_wgi74n6oavhdxhlomowlxqjdhu	9	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	10	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	10	2	further	further	RB
work_wgi74n6oavhdxhlomowlxqjdhu	10	3	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	10	4	successful	successful	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	10	5	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	10	6	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	10	7	seven	seven	CD
work_wgi74n6oavhdxhlomowlxqjdhu	10	8	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	10	9	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	10	10	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	10	11	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	10	12	families	family	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	10	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	10	14	empirically	empirically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	10	15	demonstrating	demonstrate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	10	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	10	17	applicability	applicability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	10	18	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	10	19	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	10	20	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	10	21	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	10	22	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	10	23	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	10	24	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	10	25	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	10	26	diverse	diverse	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	10	27	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	10	28	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	10	29	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	10	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	11	1	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	11	2	Introduction	introduction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	11	3	Supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	11	4	part	part	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	11	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	11	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	11	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	11	8	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	11	9	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	11	10	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	11	11	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	11	12	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	11	13	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	11	14	avail-	avail-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	11	15	able	able	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	11	16	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	11	17	more	more	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	11	18	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	11	19	twenty	twenty	CD
work_wgi74n6oavhdxhlomowlxqjdhu	11	20	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	11	21	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	11	22	achieve	achieve	VB
work_wgi74n6oavhdxhlomowlxqjdhu	11	23	ac-	ac-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	11	24	curacies	curacy	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	11	25	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	11	26	around	around	RB
work_wgi74n6oavhdxhlomowlxqjdhu	11	27	95	95	CD
work_wgi74n6oavhdxhlomowlxqjdhu	11	28	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	11	29	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	11	30	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	11	31	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	11	32	domain	domain	NN
work_wgi74n6oavhdxhlomowlxqjdhu	11	33	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	11	34	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	11	35	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	11	36	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	11	37	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	11	38	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	11	39	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	11	40	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	11	41	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	11	42	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	12	1	Thanks	thank	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	12	2	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	12	3	their	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	12	4	efficiency	efficiency	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	12	6	robustness	robustness	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	12	8	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	12	9	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	12	10	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	12	11	routinely	routinely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	12	12	employed	employ	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	12	13	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	12	14	many	many	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	12	15	natural	natural	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	12	16	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	17	processing	processing	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	18	applications	application	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	12	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	12	20	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	12	21	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	12	22	syn-	syn-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	12	23	tactic	tactic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	12	24	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	12	25	semantic	semantic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	12	26	parsing	parsing	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	12	28	named	name	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	12	29	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	12	30	entity	entity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	31	recognition	recognition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	32	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	12	33	machine	machine	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	34	translation	translation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	12	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	13	1	Unfortunately	unfortunately	RB
work_wgi74n6oavhdxhlomowlxqjdhu	13	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	13	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	13	4	resources	resource	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	13	5	required	require	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	13	6	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	13	7	train	train	VB
work_wgi74n6oavhdxhlomowlxqjdhu	13	8	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	13	9	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	13	10	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	13	11	expensive	expensive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	13	12	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	13	13	create	create	VB
work_wgi74n6oavhdxhlomowlxqjdhu	13	14	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	13	15	unlikely	unlikely	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	13	16	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	13	17	exist	exist	VB
work_wgi74n6oavhdxhlomowlxqjdhu	13	18	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	13	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	13	20	majority	majority	NN
work_wgi74n6oavhdxhlomowlxqjdhu	13	21	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	13	22	written	write	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	13	23	∗Work	∗Work	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	13	24	primarily	primarily	RB
work_wgi74n6oavhdxhlomowlxqjdhu	13	25	carried	carry	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	13	26	out	out	RP
work_wgi74n6oavhdxhlomowlxqjdhu	13	27	while	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	13	28	at	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	13	29	Google	Google	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	13	30	Research	Research	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	13	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	14	1	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	14	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	15	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	15	2	necessity	necessity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	15	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	15	4	building	build	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	15	5	NLP	NLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	15	6	tools	tool	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	15	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	15	8	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	15	9	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	15	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	15	11	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	15	12	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	15	13	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	15	14	been	be	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	15	15	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	15	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	15	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	15	18	motivation	motivation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	15	19	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	15	20	research	research	NN
work_wgi74n6oavhdxhlomowlxqjdhu	15	21	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	15	22	unsupervised	unsupervised	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	15	23	learning	learning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	15	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	15	25	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	15	26	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	15	27	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	15	28	Christodoulopoulos	Christodoulopoulos	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	15	29	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	15	30	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	15	31	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	15	32	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	15	33	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	15	34	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	15	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	16	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	16	2	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	16	3	paper	paper	NN
work_wgi74n6oavhdxhlomowlxqjdhu	16	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	16	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	16	6	instead	instead	RB
work_wgi74n6oavhdxhlomowlxqjdhu	16	7	take	take	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	16	8	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	16	9	weakly	weakly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	16	10	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	16	11	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	16	12	towards	towards	IN
work_wgi74n6oavhdxhlomowlxqjdhu	16	13	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	16	14	problem	problem	NN
work_wgi74n6oavhdxhlomowlxqjdhu	16	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	17	1	Recently	recently	RB
work_wgi74n6oavhdxhlomowlxqjdhu	17	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	17	3	learning	learn	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	17	4	POS	pos	NN
work_wgi74n6oavhdxhlomowlxqjdhu	17	5	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	17	6	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	17	7	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	17	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	17	9	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	17	10	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	17	11	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	17	12	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	17	13	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	17	14	gained	gain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	17	15	popularity	popularity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	17	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	18	1	Tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	2	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	18	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	18	4	noisily	noisily	RB
work_wgi74n6oavhdxhlomowlxqjdhu	18	5	pro-	pro-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	6	jected	jecte	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	18	7	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	8	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	18	10	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	18	11	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	18	13	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	18	14	bridged	bridge	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	18	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	18	16	gap	gap	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	17	between	between	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	18	purely	purely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	18	19	unsupervised	unsupervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	18	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	18	21	fully	fully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	18	22	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	18	23	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	18	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	18	25	resulting	result	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	18	26	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	27	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	18	28	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	18	29	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	30	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	31	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	32	83	83	CD
work_wgi74n6oavhdxhlomowlxqjdhu	18	33	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	34	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	35	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	18	36	benchmark	benchmark	NN
work_wgi74n6oavhdxhlomowlxqjdhu	18	37	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	18	38	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	18	39	Indo	indo	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	18	40	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	18	41	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	18	42	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	18	43	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	18	44	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	18	45	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	18	46	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	18	47	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	18	48	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	18	49	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	18	50	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	19	1	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	19	2	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	19	3	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	19	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	20	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	20	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	20	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	20	4	further	further	RB
work_wgi74n6oavhdxhlomowlxqjdhu	20	5	im-	im-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	6	proved	prove	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	20	7	upon	upon	IN
work_wgi74n6oavhdxhlomowlxqjdhu	20	8	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	20	9	result	result	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	10	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	20	11	employing	employ	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	20	12	Wiktionary1	Wiktionary1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	20	13	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	20	14	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	20	15	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	16	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	20	17	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	18	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	20	19	resulting	result	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	20	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	20	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	20	22	hitherto	hitherto	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	20	23	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	20	24	published	publish	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	20	25	result	result	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	26	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	20	27	almost	almost	RB
work_wgi74n6oavhdxhlomowlxqjdhu	20	28	85	85	CD
work_wgi74n6oavhdxhlomowlxqjdhu	20	29	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	30	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	20	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	20	32	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	20	33	setup	setup	NN
work_wgi74n6oavhdxhlomowlxqjdhu	20	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	21	1	Although	although	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	21	3	aforementioned	aforementioned	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	4	weakly	weakly	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	21	5	supervised	supervise	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	21	6	approaches	approach	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	21	7	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	21	8	resulted	result	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	21	9	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	10	significant	significant	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	11	improvements	improvement	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	21	12	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	13	fully	fully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	21	14	unsupervised	unsupervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	21	15	approaches	approach	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	21	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	21	17	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	21	18	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	21	19	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	21	20	exploited	exploit	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	21	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	21	22	benefits	benefit	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	21	23	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	24	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	25	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	21	26	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	27	cross	cross	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	28	-	-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	29	lingual	lingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	30	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	31	methods	method	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	21	32	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	21	33	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	21	34	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	21	35	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	36	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	37	word-	word-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	38	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	21	39	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	40	between	between	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	41	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	21	42	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	43	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	44	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	45	interest	interest	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	46	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	21	47	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	21	48	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	49	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	21	50	rich	rich	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	51	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	52	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	21	53	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	21	54	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	21	55	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	21	56	English	English	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	21	57	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	22	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	22	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	22	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	22	4	setting	setting	NN
work_wgi74n6oavhdxhlomowlxqjdhu	22	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	22	6	consider	consider	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	22	7	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	22	8	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	22	9	paper	paper	NN
work_wgi74n6oavhdxhlomowlxqjdhu	22	10	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	22	11	§	§	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	22	12	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	22	13	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	22	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	23	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	23	2	prior	prior	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	3	work	work	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	4	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	23	5	successfully	successfully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	23	6	considered	consider	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	23	7	both	both	CC
work_wgi74n6oavhdxhlomowlxqjdhu	23	8	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	23	10	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	11	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	23	12	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	13	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	14	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	23	15	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	23	17	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	23	18	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	19	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	23	20	estimating	estimate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	23	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	23	22	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	23	parameters	parameter	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	23	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	23	25	genera-	genera-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	26	tive	tive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	27	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	28	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	23	29	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	23	30	Yarowsky	Yarowsky	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	23	31	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	23	32	Ngai	Ngai	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	23	33	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	23	34	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	23	35	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	23	36	Xi	Xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	23	37	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	23	38	Hwa	Hwa	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	23	39	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	23	40	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	23	41	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	23	42	inter	inter	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	23	43	alia	alia	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	23	44	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	23	45	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	23	46	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	23	47	key	key	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	48	observation	observation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	49	under-	under-	IN
work_wgi74n6oavhdxhlomowlxqjdhu	23	50	lying	lie	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	23	51	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	23	52	present	present	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	53	work	work	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	54	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	23	55	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	23	56	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	57	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	23	58	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	59	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	23	60	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	61	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	62	offer	offer	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	23	63	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	64	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	23	65	complementary	complementary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	23	66	sig-	sig-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	23	67	nals	nal	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	23	68	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	24	1	On	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	24	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	24	3	one	one	CD
work_wgi74n6oavhdxhlomowlxqjdhu	24	4	hand	hand	NN
work_wgi74n6oavhdxhlomowlxqjdhu	24	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	24	6	high	high	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	24	7	confidence	confidence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	24	8	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	24	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	24	10	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	24	11	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	24	12	offer	offer	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	24	13	precise	precise	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	24	14	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	24	15	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	24	16	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	24	17	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	24	18	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	24	19	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	24	20	particular	particular	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	24	21	context	context	NN
work_wgi74n6oavhdxhlomowlxqjdhu	24	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	25	1	On	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	25	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	25	3	other	other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	25	4	hand	hand	NN
work_wgi74n6oavhdxhlomowlxqjdhu	25	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	25	6	manually	manually	RB
work_wgi74n6oavhdxhlomowlxqjdhu	25	7	cre-	cre-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	25	8	1http://www.wiktionary.org/.	1http://www.wiktionary.org/.	CD
work_wgi74n6oavhdxhlomowlxqjdhu	26	1	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	26	2	ated	ated	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	26	3	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	4	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	26	5	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	6	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	26	7	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	26	8	have	have	VB
work_wgi74n6oavhdxhlomowlxqjdhu	26	9	broad	broad	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	26	10	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	26	12	do	do	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	26	13	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	26	14	suffer	suffer	VB
work_wgi74n6oavhdxhlomowlxqjdhu	26	15	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	26	16	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	17	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	26	18	alignment	alignment	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	19	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	26	20	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	26	21	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	26	22	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	26	23	therefore	therefore	RB
work_wgi74n6oavhdxhlomowlxqjdhu	26	24	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	26	25	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	26	26	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	26	27	filter	filter	VB
work_wgi74n6oavhdxhlomowlxqjdhu	26	28	systematic	systematic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	26	29	as	as	RB
work_wgi74n6oavhdxhlomowlxqjdhu	26	30	well	well	RB
work_wgi74n6oavhdxhlomowlxqjdhu	26	31	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	26	32	random	random	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	26	33	noise	noise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	34	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	26	35	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	26	36	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	26	37	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	26	38	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	26	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	27	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	27	2	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	27	3	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	27	4	reap	reap	VB
work_wgi74n6oavhdxhlomowlxqjdhu	27	5	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	27	6	potential	potential	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	27	7	benefits	benefit	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	27	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	27	9	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	27	10	pro-	pro-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	27	11	pose	pose	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	27	12	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	27	13	partially	partially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	27	14	observed	observe	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	27	15	conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	27	16	random	random	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	27	17	field	field	NN
work_wgi74n6oavhdxhlomowlxqjdhu	27	18	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	27	19	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	27	20	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	27	21	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	27	22	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	27	23	Lafferty	Lafferty	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	27	24	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	27	25	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	27	26	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	27	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	27	28	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	27	29	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	27	30	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	27	31	couples	couple	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	27	32	to-	to-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	27	33	ken	ken	NN
work_wgi74n6oavhdxhlomowlxqjdhu	27	34	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	27	35	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	27	36	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	27	37	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	27	38	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	27	39	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	27	40	guide	guide	VB
work_wgi74n6oavhdxhlomowlxqjdhu	27	41	learning	learn	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	27	42	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	27	43	§	§	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	27	44	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	27	45	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	27	46	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	28	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	28	2	essence	essence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	28	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	28	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	28	5	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	28	6	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	28	7	given	give	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	28	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	28	9	freedom	freedom	NN
work_wgi74n6oavhdxhlomowlxqjdhu	28	10	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	28	11	push	push	VB
work_wgi74n6oavhdxhlomowlxqjdhu	28	12	probability	probability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	28	13	mass	mass	NN
work_wgi74n6oavhdxhlomowlxqjdhu	28	14	towards	towards	IN
work_wgi74n6oavhdxhlomowlxqjdhu	28	15	hypotheses	hypothesis	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	28	16	consistent	consistent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	28	17	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	28	18	both	both	DT
work_wgi74n6oavhdxhlomowlxqjdhu	28	19	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	28	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	28	21	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	28	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	29	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	29	2	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	29	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	29	4	flex-	flex-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	29	5	ible	ible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	29	6	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	29	7	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	29	8	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	29	9	use	use	VB
work_wgi74n6oavhdxhlomowlxqjdhu	29	10	either	either	CC
work_wgi74n6oavhdxhlomowlxqjdhu	29	11	noisy	noisy	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	29	12	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	29	13	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	29	14	manually	manually	RB
work_wgi74n6oavhdxhlomowlxqjdhu	29	15	constructed	construct	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	29	16	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	29	17	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	29	18	generate	generate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	29	19	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	29	20	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	29	21	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	29	22	furthermore	furthermore	RB
work_wgi74n6oavhdxhlomowlxqjdhu	29	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	29	24	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	29	25	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	29	26	incorporate	incorporate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	29	27	arbitrary	arbitrary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	29	28	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	29	29	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	29	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	29	31	input	input	NN
work_wgi74n6oavhdxhlomowlxqjdhu	29	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	30	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	30	2	addition	addition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	30	3	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	30	4	standard	standard	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	30	5	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	30	6	contextual	contextual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	30	7	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	30	8	lexical	lexical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	30	9	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	30	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	30	11	transition	transition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	30	12	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	30	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	30	14	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	30	15	observe	observe	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	30	16	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	30	17	adding	add	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	30	18	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	30	19	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	30	20	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	30	21	monolingual	monolingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	30	22	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	30	23	cluster-	cluster-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	30	24	ing	ing	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	30	25	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	30	26	Uszkoreit	Uszkoreit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	30	27	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	30	28	Brants	Brants	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	30	29	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	30	30	2008	2008	CD
work_wgi74n6oavhdxhlomowlxqjdhu	30	31	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	30	32	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	30	33	significantly	significantly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	30	34	im-	im-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	30	35	prove	prove	VB
work_wgi74n6oavhdxhlomowlxqjdhu	30	36	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	30	37	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	31	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	31	2	most	most	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	31	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	31	4	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	31	5	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	31	6	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	31	7	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	31	8	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	31	9	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	31	10	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	31	11	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	31	12	generative	generative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	31	13	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	31	14	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	31	15	based	base	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	31	16	hidden	hide	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	31	17	Markov	Markov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	18	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	31	19	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	31	20	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	21	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	31	22	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	31	23	Berg	Berg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	24	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	31	25	Kirkpatrick	Kirkpatrick	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	26	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	31	27	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	28	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	29	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	31	30	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	31	31	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	31	32	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	31	33	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	31	34	achieve	achieve	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	31	35	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	31	36	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	31	37	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	31	38	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	31	39	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	31	40	globally	globally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	31	41	normalized	normalized	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	31	42	discriminative	discriminative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	31	43	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	31	44	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	31	45	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	32	1	To	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	32	2	evaluate	evaluate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	32	3	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	32	4	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	32	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	32	6	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	32	7	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	32	8	extensive	extensive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	32	9	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	32	10	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	32	11	standard	standard	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	32	12	publicly	publicly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	32	13	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	32	14	datasets	dataset	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	32	15	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	32	16	15	15	CD
work_wgi74n6oavhdxhlomowlxqjdhu	32	17	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	32	18	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	32	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	32	20	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	32	21	Indo	indo	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	32	22	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	32	23	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	32	24	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	32	25	pre-	pre-	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	32	26	viously	viously	RB
work_wgi74n6oavhdxhlomowlxqjdhu	32	27	studied	study	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	32	28	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	32	29	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	32	30	context	context	NN
work_wgi74n6oavhdxhlomowlxqjdhu	32	31	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	32	32	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	32	33	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	32	34	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	32	35	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	32	36	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	32	37	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	32	38	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	32	39	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	32	40	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	32	41	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	32	42	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	33	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	33	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	33	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	33	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	33	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	33	6	seven	seven	CD
work_wgi74n6oavhdxhlomowlxqjdhu	33	7	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	33	8	lan-	lan-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	33	9	guages	guage	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	33	10	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	33	11	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	33	12	families	family	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	33	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	33	14	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	33	15	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	33	16	no	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	33	17	compa-	compa-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	33	18	rable	rable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	33	19	study	study	NN
work_wgi74n6oavhdxhlomowlxqjdhu	33	20	exists	exist	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	33	21	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	34	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	34	2	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	34	3	4	4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	34	4	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	34	5	compare	compare	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	34	6	various	various	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	34	7	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	34	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	34	9	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	34	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	34	11	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	34	12	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	34	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	35	1	Our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	35	2	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	35	3	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	35	4	uses	use	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	35	5	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	35	6	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	35	7	derived	derive	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	35	8	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	35	9	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	35	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	35	11	together	together	RB
work_wgi74n6oavhdxhlomowlxqjdhu	35	12	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	35	13	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	35	14	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	35	15	derived	derive	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	35	16	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	35	17	high	high	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	35	18	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	35	19	confidence	confidence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	35	20	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	35	21	alignments	alignment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	35	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	36	1	When	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	36	2	averaged	average	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	36	3	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	36	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	36	5	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	36	6	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	36	7	studied	study	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	36	8	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	36	9	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	36	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	36	11	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	36	12	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	36	13	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	36	14	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	36	15	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	36	16	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	36	17	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	36	18	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	36	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	37	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	37	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	37	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	37	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	37	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	37	6	achieve	achieve	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	37	7	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	37	8	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	37	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	37	10	88.8	88.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	37	11	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	37	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	38	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	38	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	38	3	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	38	4	25	25	CD
work_wgi74n6oavhdxhlomowlxqjdhu	38	5	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	38	6	relative	relative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	38	7	error	error	NN
work_wgi74n6oavhdxhlomowlxqjdhu	38	8	reduction	reduction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	38	9	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	38	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	38	11	previous	previous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	38	12	state	state	NN
work_wgi74n6oavhdxhlomowlxqjdhu	38	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	38	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	38	15	art	art	NN
work_wgi74n6oavhdxhlomowlxqjdhu	38	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	39	1	Averaged	average	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	39	2	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	39	3	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	39	4	15	15	CD
work_wgi74n6oavhdxhlomowlxqjdhu	39	5	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	39	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	39	7	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	39	8	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	39	9	obtains	obtain	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	39	10	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	39	11	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	39	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	39	13	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	39	14	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	39	15	compared	compare	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	39	16	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	39	17	78.5	78.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	39	18	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	39	19	obtained	obtain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	39	20	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	39	21	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	39	22	strong	strong	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	39	23	generative	generative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	39	24	baseline	baseline	NN
work_wgi74n6oavhdxhlomowlxqjdhu	39	25	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	40	1	Fi-	fi-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	40	2	nally	nally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	40	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	40	4	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	40	5	provide	provide	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	40	6	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	40	7	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	40	8	depth	depth	NN
work_wgi74n6oavhdxhlomowlxqjdhu	40	9	analysis	analysis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	40	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	40	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	40	12	relative	relative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	40	13	contributions	contribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	40	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	40	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	40	16	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	40	17	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	40	18	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	40	19	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	40	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	40	21	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	40	22	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	40	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	41	1	2	2	LS
work_wgi74n6oavhdxhlomowlxqjdhu	41	2	Coupling	couple	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	41	3	Token	Token	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	4	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	5	Type	Type	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	6	Constraints	Constraints	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	7	Type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	41	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	41	9	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	41	10	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	41	11	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	41	12	been	be	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	41	13	amply	amply	RB
work_wgi74n6oavhdxhlomowlxqjdhu	41	14	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	41	15	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	41	16	weakly	weakly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	41	17	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	41	18	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	19	induction	induction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	41	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	21	either	either	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	22	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	41	23	pure	pure	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	41	24	manually	manually	RB
work_wgi74n6oavhdxhlomowlxqjdhu	41	25	crafted	craft	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	41	26	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	41	27	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	41	28	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	41	29	Smith	Smith	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	30	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	31	Eisner	Eisner	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	32	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	33	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	41	34	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	41	35	Ravi	Ravi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	36	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	37	Knight	Knight	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	38	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	39	2009	2009	CD
work_wgi74n6oavhdxhlomowlxqjdhu	41	40	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	41	41	Garrette	Garrette	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	42	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	43	Baldridge	Baldridge	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	44	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	45	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	41	46	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	41	47	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	48	noisily	noisily	RB
work_wgi74n6oavhdxhlomowlxqjdhu	41	49	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	41	50	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	41	51	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	41	52	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	41	53	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	54	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	55	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	56	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	57	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	41	58	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	41	59	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	41	60	through	through	IN
work_wgi74n6oavhdxhlomowlxqjdhu	41	61	crowdsourced	crowdsource	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	41	62	lexica	lexica	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	63	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	64	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	41	65	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	41	66	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	67	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	41	68	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	69	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	41	70	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	71	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	41	72	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	41	73	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	41	74	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	41	75	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	42	1	At	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	42	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	42	3	other	other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	42	4	end	end	NN
work_wgi74n6oavhdxhlomowlxqjdhu	42	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	42	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	42	7	spectrum	spectrum	NN
work_wgi74n6oavhdxhlomowlxqjdhu	42	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	42	9	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	42	10	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	42	11	been	be	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	42	12	efforts	effort	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	42	13	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	42	14	project	project	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	42	15	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	42	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	42	17	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	42	18	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	42	19	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	42	20	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	42	21	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	42	22	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	42	23	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	42	24	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	42	25	Yarowsky	Yarowsky	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	42	26	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	42	27	Ngai	Ngai	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	42	28	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	42	29	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	42	30	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	42	31	Xi	Xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	42	32	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	42	33	Hwa	Hwa	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	42	34	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	42	35	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	42	36	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	42	37	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	43	1	How-	How-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	43	2	ever	ever	RB
work_wgi74n6oavhdxhlomowlxqjdhu	43	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	43	4	systems	system	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	43	5	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	43	6	combine	combine	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	43	7	both	both	DT
work_wgi74n6oavhdxhlomowlxqjdhu	43	8	sources	source	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	43	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	43	10	informa-	informa-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	43	11	tion	tion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	43	12	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	43	13	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	43	14	single	single	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	43	15	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	43	16	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	43	17	yet	yet	RB
work_wgi74n6oavhdxhlomowlxqjdhu	43	18	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	43	19	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	43	20	fully	fully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	43	21	explored	explore	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	43	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	44	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	44	2	following	follow	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	44	3	three	three	CD
work_wgi74n6oavhdxhlomowlxqjdhu	44	4	subsections	subsection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	44	5	outline	outline	VB
work_wgi74n6oavhdxhlomowlxqjdhu	44	6	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	44	7	overall	overall	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	44	8	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	44	9	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	44	10	coupling	couple	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	44	11	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	44	12	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	44	13	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	44	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	44	15	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	44	16	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	44	17	build	build	VB
work_wgi74n6oavhdxhlomowlxqjdhu	44	18	robust	robust	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	44	19	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	44	20	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	44	21	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	44	22	do	do	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	44	23	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	44	24	require	require	VB
work_wgi74n6oavhdxhlomowlxqjdhu	44	25	any	any	DT
work_wgi74n6oavhdxhlomowlxqjdhu	44	26	direct	direct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	44	27	supervision	supervision	NN
work_wgi74n6oavhdxhlomowlxqjdhu	44	28	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	44	29	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	44	30	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	44	31	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	44	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	45	1	2.1	2.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	45	2	Token	Token	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	45	3	Constraints	Constraints	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	45	4	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	45	6	majority	majority	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	7	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	8	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	45	10	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	45	11	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	45	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	45	13	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	45	14	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	45	15	at	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	16	least	least	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	45	17	some	some	DT
work_wgi74n6oavhdxhlomowlxqjdhu	45	18	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	19	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	20	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	45	21	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	22	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	45	23	rich	rich	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	45	24	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	25	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	26	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	45	27	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	28	simplicity	simplicity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	29	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	45	30	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	45	31	choose	choose	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	45	32	English	English	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	45	33	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	34	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	45	35	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	36	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	45	37	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	45	38	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	45	39	experiments	experiment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	45	40	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	46	1	It	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	46	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	46	3	then	then	RB
work_wgi74n6oavhdxhlomowlxqjdhu	46	4	nat-	nat-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	46	5	ural	ural	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	46	6	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	46	7	consider	consider	VB
work_wgi74n6oavhdxhlomowlxqjdhu	46	8	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	46	9	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	46	10	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	46	11	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	12	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	46	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	46	14	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	46	15	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	16	tagger	tagger	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	17	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	46	18	predict	predict	VB
work_wgi74n6oavhdxhlomowlxqjdhu	46	19	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	20	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	46	21	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	46	22	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	46	23	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	24	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	46	25	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	46	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	46	27	English	english	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	46	28	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	29	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	46	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	46	31	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	46	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	47	1	These	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	47	2	predicted	predict	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	47	3	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	47	4	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	47	5	subse-	subse-	VB
work_wgi74n6oavhdxhlomowlxqjdhu	47	6	quently	quently	RB
work_wgi74n6oavhdxhlomowlxqjdhu	47	7	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	47	8	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	47	9	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	47	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	47	11	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	47	12	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	47	13	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	47	14	automatic	automatic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	47	15	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	47	16	alignments	alignment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	47	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	48	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	48	2	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	48	3	was	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	48	4	pioneered	pioneer	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	48	5	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	48	6	Yarowsky	Yarowsky	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	48	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	48	8	Ngai	Ngai	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	48	9	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	48	10	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	48	11	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	48	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	48	13	who	who	WP
work_wgi74n6oavhdxhlomowlxqjdhu	48	14	used	use	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	48	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	48	16	resulting	result	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	48	17	partial	partial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	48	18	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	48	19	annotation	annotation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	48	20	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	48	21	estimate	estimate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	48	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	48	23	parameters	parameter	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	48	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	48	25	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	48	26	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	48	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	49	1	However	however	RB
work_wgi74n6oavhdxhlomowlxqjdhu	49	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	49	3	due	due	IN
work_wgi74n6oavhdxhlomowlxqjdhu	49	4	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	49	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	49	6	automatic	automatic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	49	7	nature	nature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	49	8	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	49	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	49	10	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	49	11	alignments	alignment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	49	12	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	49	13	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	49	14	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	49	15	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	49	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	49	17	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	49	18	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	49	19	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	49	20	significant	significant	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	49	21	noise	noise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	49	22	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	49	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	49	24	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	49	25	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	49	26	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	50	1	To	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	50	2	conquer	conquer	VB
work_wgi74n6oavhdxhlomowlxqjdhu	50	3	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	50	4	noise	noise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	50	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	50	6	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	50	7	used	use	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	50	8	very	very	RB
work_wgi74n6oavhdxhlomowlxqjdhu	50	9	aggressive	aggressive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	50	10	smoothing	smooth	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	50	11	tech-	tech-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	50	12	niques	nique	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	50	13	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	50	14	training	train	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	50	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	50	16	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	50	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	51	1	Fossum	Fossum	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	51	2	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	51	3	Abney	Abney	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	51	4	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	51	5	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	51	6	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	51	7	used	use	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	51	8	similar	similar	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	9	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	51	11	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	12	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	51	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	51	14	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	51	15	in-	in-	DT
work_wgi74n6oavhdxhlomowlxqjdhu	51	16	stead	stead	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	51	17	combined	combine	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	51	18	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	51	19	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	51	20	multiple	multiple	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	21	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	22	lan-	lan-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	23	guages	guage	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	51	24	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	51	25	filter	filter	VB
work_wgi74n6oavhdxhlomowlxqjdhu	51	26	out	out	RP
work_wgi74n6oavhdxhlomowlxqjdhu	51	27	random	random	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	28	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	29	noise	noise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	30	as	as	RB
work_wgi74n6oavhdxhlomowlxqjdhu	51	31	well	well	RB
work_wgi74n6oavhdxhlomowlxqjdhu	51	32	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	51	33	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	51	34	systematic	systematic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	35	noise	noise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	36	arising	arise	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	51	37	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	51	38	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	39	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	40	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	51	41	annotations	annotation	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	51	42	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	51	43	syntactic	syntactic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	51	44	divergences	divergence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	51	45	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	52	1	2.2	2.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	52	2	Type	Type	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	3	Constraints	Constraints	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	52	4	It	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	52	5	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	52	6	well	well	RB
work_wgi74n6oavhdxhlomowlxqjdhu	52	7	known	know	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	52	8	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	52	9	given	give	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	52	10	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	52	11	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	52	12	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	52	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	52	14	even	even	RB
work_wgi74n6oavhdxhlomowlxqjdhu	52	15	if	if	IN
work_wgi74n6oavhdxhlomowlxqjdhu	52	16	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	52	17	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	52	18	incomplete	incomplete	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	52	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	52	20	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	52	21	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	52	22	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	52	23	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	52	24	learn	learn	VB
work_wgi74n6oavhdxhlomowlxqjdhu	52	25	accurate	accurate	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	52	26	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	27	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	52	28	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	52	29	Smith	Smith	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	30	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	52	31	Eisner	Eisner	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	32	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	52	33	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	52	34	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	52	35	Goldberg	Goldberg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	36	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	52	37	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	38	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	39	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	52	40	2008	2008	CD
work_wgi74n6oavhdxhlomowlxqjdhu	52	41	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	52	42	Ravi	Ravi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	43	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	52	44	Knight	Knight	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	45	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	52	46	2009	2009	CD
work_wgi74n6oavhdxhlomowlxqjdhu	52	47	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	52	48	Naseem	Naseem	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	49	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	50	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	51	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	52	52	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	52	53	2009	2009	CD
work_wgi74n6oavhdxhlomowlxqjdhu	52	54	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	52	55	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	53	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	53	2	widely	widely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	53	3	differing	differ	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	53	4	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	53	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	53	6	specific	specific	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	53	7	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	53	8	struc-	struc-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	53	9	ture	ture	NN
work_wgi74n6oavhdxhlomowlxqjdhu	53	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	53	11	learning	learning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	53	12	objective	objective	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	53	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	53	14	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	53	15	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	53	16	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	53	17	approaches	approach	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	53	18	achieve	achieve	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	53	19	excellent	excellent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	53	20	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	53	21	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	54	1	Unfortunately	unfortunately	RB
work_wgi74n6oavhdxhlomowlxqjdhu	54	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	54	3	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	54	4	rely	rely	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	54	5	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	54	6	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	54	7	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	54	8	extracted	extract	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	54	9	directly	directly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	54	10	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	54	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	54	12	un-	un-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	54	13	derlying	derlye	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	54	14	treebank	treebank	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	54	15	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	54	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	55	1	Such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	2	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	3	provide	provide	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	55	4	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	5	depth	depth	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	6	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	7	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	55	9	test	test	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	10	domain	domain	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	55	12	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	55	13	list	list	VB
work_wgi74n6oavhdxhlomowlxqjdhu	55	14	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	55	15	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	55	16	�	�	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	17	�	�	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	18	�	�	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	19	�	�	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	20	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	21	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	22	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	23	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	24	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	25	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	26	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	27	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	28	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	29	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	30	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	31	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	32	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	33	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	34	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	35	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	36	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	37	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	38	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	39	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	40	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	41	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	42	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	43	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	44	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	45	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	46	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	47	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	48	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	49	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	50	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	51	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	52	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	53	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	54	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	55	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	56	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	57	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	58	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	59	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	60	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	61	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	62	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	63	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	64	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	65	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	66	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	67	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	68	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	69	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	70	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	71	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	72	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	73	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	74	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	75	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	76	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	77	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	78	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	79	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	80	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	81	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	82	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	83	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	84	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	85	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	86	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	87	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	88	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	89	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	90	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	91	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	92	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	93	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	94	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	95	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	96	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	97	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	98	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	99	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	100	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	101	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	102	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	103	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	104	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	105	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	106	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	107	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	108	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	109	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	110	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	111	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	112	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	113	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	114	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	115	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	116	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	117	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	118	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	119	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	120	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	121	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	122	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	123	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	124	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	125	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	126	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	127	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	128	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	129	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	130	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	131	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	132	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	133	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	134	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	135	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	136	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	137	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	138	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	139	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	140	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	141	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	142	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	143	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	144	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	145	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	146	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	147	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	148	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	149	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	150	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	151	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	152	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	153	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	154	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	155	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	156	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	157	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	158	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	159	�	�	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	55	160	�	�	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	161	�	�	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	162	�	�	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	163	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	164	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	55	165	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	55	166	Lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	167	representation	representation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	168	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	169	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	55	170	inference	inference	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	171	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	172	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	173	Y(x	y(x	FW
work_wgi74n6oavhdxhlomowlxqjdhu	55	174	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	55	175	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	176	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	55	177	authentic	authentic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	178	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	179	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	180	Swedish	Swedish	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	55	181	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	55	182	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	55	183	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	55	184	farming	farming	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	185	products	product	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	186	must	must	MD
work_wgi74n6oavhdxhlomowlxqjdhu	55	187	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	55	188	pure	pure	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	189	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	55	190	must	must	MD
work_wgi74n6oavhdxhlomowlxqjdhu	55	191	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	55	192	contain	contain	VB
work_wgi74n6oavhdxhlomowlxqjdhu	55	193	any	any	DT
work_wgi74n6oavhdxhlomowlxqjdhu	55	194	additives	additive	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	195	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	55	196	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	55	197	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	55	198	after	after	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	199	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	55	200	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	55	201	Wiktionary	wiktionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	55	202	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	55	203	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	55	204	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	56	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	56	2	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	56	3	parts	part	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	56	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	56	5	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	56	6	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	56	7	listed	list	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	56	8	underneath	underneath	IN
work_wgi74n6oavhdxhlomowlxqjdhu	56	9	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	56	10	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	56	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	57	1	Bold	Bold	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	57	2	nodes	node	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	57	3	show	show	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	57	4	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	57	5	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	57	6	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	57	7	ỹ.	ỹ.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	58	1	Underlined	underline	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	58	2	text	text	NN
work_wgi74n6oavhdxhlomowlxqjdhu	58	3	indicates	indicate	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	58	4	incorrect	incorrect	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	58	5	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	58	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	59	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	59	2	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	59	3	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	59	4	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	59	5	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	59	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	59	7	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	59	8	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	59	9	consists	consist	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	59	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	59	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	59	12	bold	bold	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	59	13	nodes	node	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	59	14	together	together	RB
work_wgi74n6oavhdxhlomowlxqjdhu	59	15	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	59	16	nodes	node	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	59	17	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	59	18	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	59	19	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	59	20	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	59	21	lacking	lack	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	59	22	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	59	23	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	59	24	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	59	25	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	59	26	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	59	27	case	case	NN
work_wgi74n6oavhdxhlomowlxqjdhu	59	28	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	59	29	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	59	30	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	59	31	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	59	32	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	59	33	thus	thus	RB
work_wgi74n6oavhdxhlomowlxqjdhu	59	34	defines	define	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	59	35	exactly	exactly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	59	36	one	one	CD
work_wgi74n6oavhdxhlomowlxqjdhu	59	37	valid	valid	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	59	38	path	path	NN
work_wgi74n6oavhdxhlomowlxqjdhu	59	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	60	1	inflected	inflect	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	60	2	forms	form	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	60	3	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	60	4	both	both	DT
work_wgi74n6oavhdxhlomowlxqjdhu	60	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	60	6	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	60	7	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	60	8	difficult	difficult	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	60	9	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	60	10	obtain	obtain	VB
work_wgi74n6oavhdxhlomowlxqjdhu	60	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	60	12	unrealistic	unrealistic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	60	13	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	60	14	expect	expect	VB
work_wgi74n6oavhdxhlomowlxqjdhu	60	15	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	60	16	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	60	17	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	60	18	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	60	19	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	60	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	61	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	61	2	contrast	contrast	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	61	4	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	61	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	61	6	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	61	7	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	61	8	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	61	9	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	61	10	automatically	automatically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	61	11	create	create	VB
work_wgi74n6oavhdxhlomowlxqjdhu	61	12	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	13	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	61	14	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	15	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	16	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	61	17	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	61	18	aggregating	aggregate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	61	19	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	61	20	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	61	21	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	61	22	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	61	23	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	24	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	25	extracted	extract	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	61	26	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	61	27	bi-	bi-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	61	28	text	text	NN
work_wgi74n6oavhdxhlomowlxqjdhu	61	29	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	62	1	To	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	62	2	handle	handle	VB
work_wgi74n6oavhdxhlomowlxqjdhu	62	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	62	4	noise	noise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	5	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	62	6	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	62	7	automatic	automatic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	62	8	dictionar-	dictionar-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	9	ies	ie	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	62	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	62	11	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	62	12	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	62	13	label	label	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	14	propagation	propagation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	15	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	62	16	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	62	17	similarity	similarity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	18	graph	graph	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	19	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	62	20	smooth	smooth	VB
work_wgi74n6oavhdxhlomowlxqjdhu	62	21	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	62	22	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	62	23	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	62	24	expand	expand	VB
work_wgi74n6oavhdxhlomowlxqjdhu	62	25	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	62	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	62	27	label	label	NN
work_wgi74n6oavhdxhlomowlxqjdhu	62	28	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	62	29	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	63	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	63	2	their	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	63	3	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	63	4	produces	produce	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	63	5	good	good	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	6	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	63	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	63	8	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	63	9	applicable	applicable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	10	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	63	11	resource	resource	VB
work_wgi74n6oavhdxhlomowlxqjdhu	63	12	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	63	13	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	14	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	63	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	63	16	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	63	17	requires	require	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	63	18	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	63	19	complex	complex	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	20	multi	multi	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	21	-	-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	22	stage	stage	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	23	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	63	24	procedure	procedure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	63	25	including	include	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	63	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	63	27	construction	construction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	63	28	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	63	29	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	63	30	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	31	distributional	distributional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	63	32	similarity	similarity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	63	33	graph	graph	NN
work_wgi74n6oavhdxhlomowlxqjdhu	63	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	64	1	Recently	recently	RB
work_wgi74n6oavhdxhlomowlxqjdhu	64	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	64	3	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	64	4	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	64	5	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	64	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	65	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	65	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	65	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	65	4	presented	present	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	65	5	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	65	6	simple	simple	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	65	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	65	8	viable	viable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	65	9	alternative	alternative	NN
work_wgi74n6oavhdxhlomowlxqjdhu	65	10	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	65	11	crowdsourced	crowdsource	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	65	12	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	65	13	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	65	14	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	65	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	66	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	66	2	noisy	noisy	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	66	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	66	4	sparse	sparse	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	66	5	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	66	6	nature	nature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	66	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	66	8	Wik-	wik-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	66	9	tionary	tionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	66	10	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	66	11	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	66	12	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	66	13	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	66	14	170	170	CD
work_wgi74n6oavhdxhlomowlxqjdhu	66	15	languages.2	languages.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	66	16	Furthermore	furthermore	RB
work_wgi74n6oavhdxhlomowlxqjdhu	66	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	66	18	their	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	66	19	quality	quality	NN
work_wgi74n6oavhdxhlomowlxqjdhu	66	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	66	21	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	66	22	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	66	23	growing	grow	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	66	24	continuously	continuously	RB
work_wgi74n6oavhdxhlomowlxqjdhu	66	25	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	66	26	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	66	27	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	66	28	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	66	29	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	66	30	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	66	31	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	66	32	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	66	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	67	1	By	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	67	2	incorporating	incorporate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	67	3	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	67	4	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	67	5	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	67	6	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	67	7	into	into	IN
work_wgi74n6oavhdxhlomowlxqjdhu	67	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	67	9	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	67	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	67	11	based	base	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	67	12	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	67	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	67	14	Berg	Berg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	67	15	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	67	16	Kirkpatrick	Kirkpatrick	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	67	17	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	67	18	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	67	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	68	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	68	2	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	68	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	68	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	68	5	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	68	6	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	68	7	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	68	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	69	1	were	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	69	2	able	able	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	69	3	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	69	4	obtain	obtain	VB
work_wgi74n6oavhdxhlomowlxqjdhu	69	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	69	6	best	well	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	69	7	published	publish	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	69	8	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	69	9	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	69	10	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	69	11	setting	setting	NN
work_wgi74n6oavhdxhlomowlxqjdhu	69	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	69	13	surpassing	surpass	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	69	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	69	15	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	69	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	69	17	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	69	18	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	69	19	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	69	20	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	69	21	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	69	22	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	69	23	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	69	24	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	69	25	Indo	indo	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	69	26	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	69	27	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	69	28	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	69	29	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	70	1	2.3	2.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	70	2	Coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	70	3	Constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	70	4	Rather	rather	RB
work_wgi74n6oavhdxhlomowlxqjdhu	70	5	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	70	6	relying	rely	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	70	7	exclusively	exclusively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	70	8	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	70	9	either	either	DT
work_wgi74n6oavhdxhlomowlxqjdhu	70	10	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	70	11	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	70	12	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	70	13	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	70	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	70	15	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	70	16	propose	propose	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	70	17	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	70	18	complement	complement	VB
work_wgi74n6oavhdxhlomowlxqjdhu	70	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	70	20	one	one	NN
work_wgi74n6oavhdxhlomowlxqjdhu	70	21	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	70	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	70	23	other	other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	70	24	during	during	IN
work_wgi74n6oavhdxhlomowlxqjdhu	70	25	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	70	26	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	71	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	71	2	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	71	3	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	71	4	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	71	5	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	71	6	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	71	7	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	71	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	71	9	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	71	10	partially	partially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	71	11	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	71	12	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	71	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	71	14	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	71	15	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	71	16	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	71	17	constructed	construct	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	71	18	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	71	19	follows	follow	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	71	20	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	71	21	2http://meta.wikimedia.org/wiki/	2http://meta.wikimedia.org/wiki/	CD
work_wgi74n6oavhdxhlomowlxqjdhu	71	22	Wiktionary	wiktionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	71	23	—	—	:
work_wgi74n6oavhdxhlomowlxqjdhu	71	24	October	October	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	71	25	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	71	26	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	72	1	1	1	LS
work_wgi74n6oavhdxhlomowlxqjdhu	72	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	73	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	73	2	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	73	3	token	token	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	73	4	whose	whose	WP$
work_wgi74n6oavhdxhlomowlxqjdhu	73	5	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	73	6	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	73	7	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	73	8	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	73	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	73	10	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	73	11	dic-	dic-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	73	12	tionary	tionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	73	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	73	14	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	73	15	allow	allow	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	73	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	73	17	entire	entire	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	73	18	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	73	19	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	73	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	74	1	2	2	LS
work_wgi74n6oavhdxhlomowlxqjdhu	74	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	75	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	75	2	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	75	3	token	token	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	75	4	whose	whose	WP$
work_wgi74n6oavhdxhlomowlxqjdhu	75	5	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	75	6	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	75	7	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	75	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	75	9	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	75	10	dictio-	dictio-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	75	11	nary	nary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	75	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	75	13	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	75	14	prune	prune	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	75	15	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	75	16	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	75	17	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	75	18	licensed	license	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	75	19	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	75	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	75	21	dictio-	dictio-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	75	22	nary	nary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	75	23	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	75	24	mark	mark	VB
work_wgi74n6oavhdxhlomowlxqjdhu	75	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	75	26	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	75	27	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	75	28	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	75	29	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	75	30	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	75	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	76	1	3	3	LS
work_wgi74n6oavhdxhlomowlxqjdhu	76	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	77	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	2	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	3	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	4	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	77	5	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	77	6	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	7	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	8	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	77	9	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	10	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	11	high	high	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	77	12	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	77	13	confidence	confidence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	14	bidirectional	bidirectional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	77	15	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	16	alignment	alignment	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	17	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	77	18	if	if	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	20	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	77	21	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	22	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	77	23	still	still	RB
work_wgi74n6oavhdxhlomowlxqjdhu	77	24	present	present	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	77	25	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	27	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	28	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	77	29	then	then	RB
work_wgi74n6oavhdxhlomowlxqjdhu	77	30	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	77	31	prune	prune	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	77	32	every	every	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	33	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	34	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	77	35	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	36	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	77	37	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	38	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	39	that	that	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	40	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	41	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	77	42	if	if	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	43	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	44	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	77	45	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	46	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	77	47	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	77	48	present	present	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	77	49	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	50	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	51	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	52	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	77	53	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	77	54	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	77	55	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	77	56	happen	happen	VB
work_wgi74n6oavhdxhlomowlxqjdhu	77	57	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	77	58	dictionary-	dictionary-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	77	59	pruned	pruned	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	77	60	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	77	61	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	77	62	then	then	RB
work_wgi74n6oavhdxhlomowlxqjdhu	77	63	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	77	64	ignore	ignore	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	77	65	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	77	66	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	77	67	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	77	68	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	78	1	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	78	2	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	78	3	provides	provide	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	78	4	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	78	5	running	run	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	78	6	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	78	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	79	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	79	2	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	79	3	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	79	4	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	79	5	permitted	permit	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	79	6	after	after	IN
work_wgi74n6oavhdxhlomowlxqjdhu	79	7	constraining	constrain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	79	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	79	9	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	79	10	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	79	11	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	79	12	licensed	license	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	79	13	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	79	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	79	15	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	79	16	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	79	17	up	up	RB
work_wgi74n6oavhdxhlomowlxqjdhu	79	18	until	until	IN
work_wgi74n6oavhdxhlomowlxqjdhu	79	19	Step	step	NN
work_wgi74n6oavhdxhlomowlxqjdhu	79	20	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	79	21	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	79	22	above	above	RB
work_wgi74n6oavhdxhlomowlxqjdhu	79	23	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	79	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	80	1	There	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	80	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	80	3	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	80	4	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	80	5	single	single	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	80	6	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	80	7	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	80	8	Jordbruk-	Jordbruk-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	80	9	sprodukterna	sprodukterna	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	80	10	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	80	11	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	80	12	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	80	13	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	80	14	farming	farming	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	80	15	products	products	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	80	16	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	80	17	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	80	18	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	80	19	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	80	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	80	21	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	80	22	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	80	23	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	80	24	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	80	25	case	case	NN
work_wgi74n6oavhdxhlomowlxqjdhu	80	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	80	27	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	80	28	permits	permit	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	80	29	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	80	30	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	80	31	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	80	32	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	80	33	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	80	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	81	1	With	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	81	2	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	81	3	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	81	4	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	81	5	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	81	6	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	81	7	Step	step	NN
work_wgi74n6oavhdxhlomowlxqjdhu	81	8	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	81	9	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	81	10	nodes	node	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	81	11	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	81	12	bold	bold	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	81	13	border	border	NN
work_wgi74n6oavhdxhlomowlxqjdhu	81	14	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	81	15	Figure	Figure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	81	16	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	81	17	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	81	18	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	81	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	81	20	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	81	21	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	81	22	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	81	23	further	far	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	81	24	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	81	25	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	82	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	82	2	most	most	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	82	3	cases	case	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	82	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	82	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	82	6	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	82	7	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	82	8	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	82	9	both	both	CC
work_wgi74n6oavhdxhlomowlxqjdhu	82	10	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	82	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	82	12	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	82	13	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	82	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	82	15	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	82	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	82	17	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	82	18	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	82	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	83	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	83	2	thus	thus	RB
work_wgi74n6oavhdxhlomowlxqjdhu	83	3	successfully	successfully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	83	4	disambiguate	disambiguate	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	83	5	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	83	6	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	83	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	83	8	shrink	shrink	VB
work_wgi74n6oavhdxhlomowlxqjdhu	83	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	83	10	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	83	11	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	83	12	substantially	substantially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	83	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	84	1	There	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	84	2	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	84	3	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	84	4	cases	case	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	84	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	84	6	highlight	highlight	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	84	7	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	84	8	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	84	9	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	84	10	show	show	VB
work_wgi74n6oavhdxhlomowlxqjdhu	84	11	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	84	12	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	84	13	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	84	14	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	84	15	break	break	VB
work_wgi74n6oavhdxhlomowlxqjdhu	84	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	85	1	First	first	RB
work_wgi74n6oavhdxhlomowlxqjdhu	85	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	85	3	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	85	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	85	5	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	85	6	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	85	7	Jordbruksprodukterna	Jordbruksprodukterna	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	85	8	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	85	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	85	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	85	11	erroneously	erroneously	RB
work_wgi74n6oavhdxhlomowlxqjdhu	85	12	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	85	13	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	85	14	ADJ	ADJ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	85	15	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	85	16	eliminate	eliminate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	85	17	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	85	18	other	other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	85	19	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	85	20	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	85	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	85	22	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	85	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	85	24	including	include	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	85	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	85	26	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	85	27	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	85	28	NOUN	NOUN	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	85	29	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	86	1	Second	second	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	86	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	86	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	86	4	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	86	5	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	86	6	några	några	``
work_wgi74n6oavhdxhlomowlxqjdhu	86	7	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	86	8	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	86	9	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	86	10	any	any	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	86	11	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	86	12	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	86	13	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	86	14	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	86	15	single	single	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	86	16	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	86	17	entry	entry	NN
work_wgi74n6oavhdxhlomowlxqjdhu	86	18	PRON	PRON	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	86	19	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	86	20	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	86	21	missing	miss	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	86	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	86	23	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	86	24	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	86	25	DET	DET	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	86	26	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	87	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	87	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	87	3	case	case	NN
work_wgi74n6oavhdxhlomowlxqjdhu	87	4	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	87	5	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	87	6	DET	DET	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	87	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	87	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	87	9	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	87	10	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	87	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	87	12	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	87	13	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	87	14	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	87	15	add	add	VB
work_wgi74n6oavhdxhlomowlxqjdhu	87	16	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	87	17	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	87	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	87	19	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	87	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	87	21	simply	simply	RB
work_wgi74n6oavhdxhlomowlxqjdhu	87	22	ignore	ignore	VB
work_wgi74n6oavhdxhlomowlxqjdhu	87	23	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	87	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	88	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	88	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	88	3	because	because	IN
work_wgi74n6oavhdxhlomowlxqjdhu	88	4	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	88	5	hy-	hy-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	88	6	pothesize	pothesize	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	88	7	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	88	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	88	9	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	88	10	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	88	11	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	88	12	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	88	13	trusted	trust	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	88	14	more	more	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	88	15	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	88	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	88	17	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	88	18	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	88	19	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	88	20	noisy	noisy	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	88	21	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	88	22	alignments	alignment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	88	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	89	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	89	2	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	89	3	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	89	4	see	see	VB
work_wgi74n6oavhdxhlomowlxqjdhu	89	5	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	89	6	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	89	7	4	4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	89	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	89	9	taking	take	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	89	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	89	11	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	89	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	89	13	tags	tags	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	89	14	performs	perform	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	89	15	worse	bad	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	89	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	89	17	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	89	18	supports	support	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	89	19	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	89	20	hypothesis	hypothesis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	89	21	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	90	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	90	2	generative	generative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	90	3	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	90	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	90	5	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	90	6	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	90	7	HMMs	hmm	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	90	8	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	90	9	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	90	10	3.1	3.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	90	11	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	90	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	90	13	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	90	14	need	need	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	90	15	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	90	16	define	define	VB
work_wgi74n6oavhdxhlomowlxqjdhu	90	17	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	90	18	one	one	CD
work_wgi74n6oavhdxhlomowlxqjdhu	90	19	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	90	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	91	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	91	2	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	91	3	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	91	4	gen-	gen-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	5	erative	erative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	91	6	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	7	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	91	8	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	91	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	91	10	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	91	11	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	12	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	91	13	type-	type-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	91	14	constrained	constrain	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	91	15	lattice.3	lattice.3	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	91	16	At	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	91	17	prediction	prediction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	18	time	time	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	91	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	91	21	both	both	CC
work_wgi74n6oavhdxhlomowlxqjdhu	91	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	91	23	discriminative	discriminative	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	24	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	91	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	91	26	generative	generative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	91	27	cases	case	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	91	28	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	91	29	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	91	30	find	find	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	91	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	91	32	most	most	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	91	33	likely	likely	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	91	34	label	label	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	35	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	91	36	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	91	37	Viterbi	Viterbi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	91	38	decoding	decode	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	91	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	92	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	92	2	discriminative	discriminative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	92	3	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	92	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	92	5	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	92	6	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	92	7	CRFs	CRFs	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	92	8	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	92	9	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	92	10	3.2	3.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	92	11	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	92	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	92	13	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	92	14	need	need	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	92	15	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	92	16	define	define	VB
work_wgi74n6oavhdxhlomowlxqjdhu	92	17	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	92	18	lattices	lattice	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	92	19	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	92	20	one	one	CD
work_wgi74n6oavhdxhlomowlxqjdhu	92	21	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	92	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	92	23	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	24	moves	move	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	92	25	probability	probability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	26	mass	mass	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	27	towards	towards	IN
work_wgi74n6oavhdxhlomowlxqjdhu	92	28	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	92	29	another	another	DT
work_wgi74n6oavhdxhlomowlxqjdhu	92	30	one	one	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	31	defining	define	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	92	32	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	92	33	overall	overall	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	92	34	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	35	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	36	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	92	37	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	92	38	partition	partition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	39	func-	func-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	92	40	tion	tion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	92	41	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	92	42	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	93	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	93	2	traditional	traditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	3	supervised	supervised	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	4	learning	learning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	5	without	without	IN
work_wgi74n6oavhdxhlomowlxqjdhu	93	6	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	7	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	93	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	10	former	former	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	11	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	93	12	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	13	trivial	trivial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	14	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	15	containing	contain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	93	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	17	gold	gold	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	18	standard	standard	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	19	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	20	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	21	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	93	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	23	latter	latter	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	24	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	93	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	26	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	27	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	93	28	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	29	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	93	30	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	93	31	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	93	32	spanning	span	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	93	33	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	93	34	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	93	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	94	1	With	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	94	2	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	94	3	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	94	4	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	94	6	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	94	7	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	94	8	move	move	VB
work_wgi74n6oavhdxhlomowlxqjdhu	94	9	mass	mass	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	10	towards	towards	IN
work_wgi74n6oavhdxhlomowlxqjdhu	94	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	94	12	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	94	13	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	14	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	94	15	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	94	17	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	94	18	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	94	20	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	94	21	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	94	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	94	23	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	24	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	94	25	freely	freely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	94	26	distribute	distribute	VB
work_wgi74n6oavhdxhlomowlxqjdhu	94	27	mass	mass	NN
work_wgi74n6oavhdxhlomowlxqjdhu	94	28	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	94	29	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	94	30	paths	path	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	94	31	consistent	consistent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	94	32	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	94	33	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	94	34	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	94	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	95	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	2	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	3	defining	define	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	95	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	5	partition	partition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	6	function	function	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	7	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	95	8	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	95	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	10	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	95	11	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	95	13	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	95	14	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	15	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	95	16	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	95	17	no	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	18	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	19	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	95	20	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	95	21	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	95	22	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	95	23	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	24	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	25	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	95	26	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	95	27	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	95	28	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	95	29	consist	consist	VB
work_wgi74n6oavhdxhlomowlxqjdhu	95	30	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	95	31	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	32	dictionary-	dictionary-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	95	33	pruned	pruned	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	95	34	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	35	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	95	36	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	95	37	sans	san	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	95	38	Step	Step	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	95	39	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	95	40	above	above	RB
work_wgi74n6oavhdxhlomowlxqjdhu	95	41	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	95	42	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	95	43	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	95	44	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	45	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	95	46	possibilities	possibility	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	95	47	shown	show	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	95	48	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	95	49	Figure	Figure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	95	50	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	95	51	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	95	52	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	95	53	running	run	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	95	54	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	95	55	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	95	56	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	96	1	Figures	figure	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	96	2	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	96	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	96	4	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	96	5	provide	provide	VB
work_wgi74n6oavhdxhlomowlxqjdhu	96	6	statistics	statistic	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	96	7	regarding	regard	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	96	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	96	9	supervision	supervision	NN
work_wgi74n6oavhdxhlomowlxqjdhu	96	10	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	96	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	96	12	remaining	remain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	96	13	ambiguity	ambiguity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	96	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	97	1	Fig-	Fig-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	97	2	ure	ure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	97	3	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	97	4	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	97	5	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	97	6	more	more	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	97	7	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	97	8	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	97	9	thirds	third	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	97	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	97	11	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	97	12	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	97	13	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	97	14	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	97	15	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	97	16	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	97	17	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	97	18	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	97	19	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	97	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	98	1	However	however	RB
work_wgi74n6oavhdxhlomowlxqjdhu	98	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	98	3	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	98	4	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	98	5	considerable	considerable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	98	6	variation	variation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	7	between	between	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	8	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	98	9	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	98	10	Spanish	Spanish	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	98	11	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	98	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	98	13	highest	high	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	98	14	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	15	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	16	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	17	90	90	CD
work_wgi74n6oavhdxhlomowlxqjdhu	98	18	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	98	20	while	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	21	Turk-	Turk-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	98	22	ish	ish	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	98	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	98	24	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	98	25	agglutinative	agglutinative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	98	26	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	27	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	28	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	98	29	vast	vast	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	98	30	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	31	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	32	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	33	forms	form	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	98	34	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	98	35	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	98	36	less	less	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	98	37	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	98	38	50	50	CD
work_wgi74n6oavhdxhlomowlxqjdhu	98	39	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	40	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	98	41	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	99	1	Fig-	Fig-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	99	2	ure	ure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	99	3	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	99	4	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	99	5	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	6	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	99	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	99	8	substantial	substantial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	99	9	uncertainty	uncertainty	NN
work_wgi74n6oavhdxhlomowlxqjdhu	99	10	left	leave	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	99	11	after	after	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	12	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	99	13	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	14	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	99	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	99	16	since	since	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	17	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	99	18	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	99	19	rarely	rarely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	99	20	fully	fully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	99	21	disambiguated	disambiguate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	99	22	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	99	23	1.3	1.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	99	24	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	99	25	per	per	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	26	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	99	27	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	99	28	allowed	allow	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	99	29	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	30	average	average	NN
work_wgi74n6oavhdxhlomowlxqjdhu	99	31	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	32	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	99	33	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	99	34	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	99	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	100	1	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	2	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	100	3	further	further	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	100	4	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	100	5	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	6	high	high	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	100	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	100	8	confidence	confidence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	9	align-	align-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	10	ments	ment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	100	11	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	100	12	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	100	13	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	14	about	about	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	15	half	half	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	100	18	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	100	19	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	20	most	most	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	100	21	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	100	22	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	100	23	Japanese	Japanese	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	100	24	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	100	25	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	100	26	notable	notable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	100	27	exception	exception	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	28	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	29	3Other	3other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	100	30	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	31	methods	method	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	100	32	exist	exist	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	100	33	as	as	RB
work_wgi74n6oavhdxhlomowlxqjdhu	100	34	well	well	RB
work_wgi74n6oavhdxhlomowlxqjdhu	100	35	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	100	36	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	100	37	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	38	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	100	39	con-	con-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	100	40	trastive	trastive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	100	41	estimation	estimation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	100	42	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	100	43	Smith	Smith	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	100	44	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	100	45	Eisner	Eisner	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	100	46	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	100	47	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	100	48	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	100	49	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	101	1	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	101	2	25	25	CD
work_wgi74n6oavhdxhlomowlxqjdhu	101	3	50	50	CD
work_wgi74n6oavhdxhlomowlxqjdhu	101	4	75	75	CD
work_wgi74n6oavhdxhlomowlxqjdhu	101	5	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	101	6	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	7	bg	bg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	8	cs	cs	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	9	da	da	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	10	de	de	FW
work_wgi74n6oavhdxhlomowlxqjdhu	101	11	el	el	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	12	es	es	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	13	fr	fr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	101	14	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	101	15	ja	ja	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	16	nl	nl	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	17	pt	pt	UH
work_wgi74n6oavhdxhlomowlxqjdhu	101	18	sl	sl	UH
work_wgi74n6oavhdxhlomowlxqjdhu	101	19	sv	sv	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	20	tr	tr	UH
work_wgi74n6oavhdxhlomowlxqjdhu	101	21	zh	zh	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	22	P	P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	23	e	e	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	24	rc	rc	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	25	e	e	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	26	n	n	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	27	t	t	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	28	o	o	XX
work_wgi74n6oavhdxhlomowlxqjdhu	101	29	f	f	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	30	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	101	31	k	k	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	32	e	e	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	33	n	n	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	34	s	s	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	35	c	c	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	36	o	o	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	37	v	v	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	38	e	e	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	39	re	re	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	40	d	d	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	41	Token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	101	42	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	43	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	44	Projected	Projected	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	45	Projected+Filtered	Projected+Filtered	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	46	Figure	Figure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	101	47	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	101	48	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	101	49	Wiktionary	wiktionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	50	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	101	51	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	52	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	101	53	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	101	54	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	102	1	Shown	show	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	102	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	102	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	4	percentage	percentage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	102	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	102	6	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	102	7	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	102	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	9	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	102	10	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	102	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	102	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	13	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	102	14	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	102	15	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	102	16	covered	cover	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	102	17	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	102	18	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	102	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	102	20	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	102	21	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	102	22	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	23	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	102	24	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	102	25	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	102	26	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	102	27	that	that	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	28	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	102	29	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	30	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	102	31	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	102	32	after	after	IN
work_wgi74n6oavhdxhlomowlxqjdhu	102	33	intersecting	intersect	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	102	34	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	102	35	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	102	36	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	103	1	0.0	0.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	103	2	0.5	0.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	103	3	1.0	1.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	103	4	1.5	1.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	103	5	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	6	bg	bg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	7	cs	cs	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	8	da	da	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	9	de	de	FW
work_wgi74n6oavhdxhlomowlxqjdhu	103	10	el	el	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	11	es	es	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	12	fr	fr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	13	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	103	14	ja	ja	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	15	nl	nl	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	16	pt	pt	UH
work_wgi74n6oavhdxhlomowlxqjdhu	103	17	sl	sl	UH
work_wgi74n6oavhdxhlomowlxqjdhu	103	18	sv	sv	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	19	tr	tr	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	20	zh	zh	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	21	N	N	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	22	u	u	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	23	m	m	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	24	b	b	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	25	e	e	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	26	r	r	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	27	o	o	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	28	f	f	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	29	ta	ta	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	30	g	g	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	31	s	s	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	32	p	p	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	33	e	e	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	34	r	r	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	35	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	36	k	k	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	37	e	e	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	38	n	n	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	39	Figure	Figure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	40	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	103	41	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	103	42	Average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	103	43	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	44	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	45	licensed	license	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	103	46	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	103	47	per	per	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	48	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	49	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	50	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	103	51	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	52	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	53	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	54	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	103	55	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	103	56	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	103	57	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	58	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	103	59	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	103	60	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	103	61	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	104	1	less	less	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	104	2	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	104	3	30	30	CD
work_wgi74n6oavhdxhlomowlxqjdhu	104	4	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	104	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	104	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	104	7	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	104	8	covered	cover	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	104	9	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	104	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	105	1	Intersecting	intersect	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	105	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	3	Wiktionary	wiktionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	105	4	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	105	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	7	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	105	8	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	9	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	105	10	Step	step	NN
work_wgi74n6oavhdxhlomowlxqjdhu	105	11	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	105	12	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	105	13	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	105	14	above	above	IN
work_wgi74n6oavhdxhlomowlxqjdhu	105	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	105	16	filters	filter	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	17	out	out	RP
work_wgi74n6oavhdxhlomowlxqjdhu	105	18	some	some	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	19	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	105	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	21	potentially	potentially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	105	22	erroneous	erroneous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	105	23	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	105	25	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	105	26	preserves	preserve	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	105	27	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	28	majority	majority	NN
work_wgi74n6oavhdxhlomowlxqjdhu	105	29	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	105	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	31	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	105	32	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	33	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	105	34	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	35	remaining	remain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	105	36	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	105	37	presumably	presumably	RB
work_wgi74n6oavhdxhlomowlxqjdhu	105	38	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	105	39	accurate	accurate	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	105	40	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	105	41	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	42	cover	cover	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	105	43	almost	almost	RB
work_wgi74n6oavhdxhlomowlxqjdhu	105	44	half	half	NN
work_wgi74n6oavhdxhlomowlxqjdhu	105	45	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	105	46	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	47	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	105	48	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	105	49	greatly	greatly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	105	50	reducing	reduce	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	105	51	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	52	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	105	53	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	105	54	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	105	55	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	105	56	learner	learner	NN
work_wgi74n6oavhdxhlomowlxqjdhu	105	57	needs	need	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	105	58	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	105	59	explore	explore	VB
work_wgi74n6oavhdxhlomowlxqjdhu	105	60	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	106	1	3	3	LS
work_wgi74n6oavhdxhlomowlxqjdhu	106	2	Models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	106	3	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	106	4	Coupled	Coupled	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	106	5	Constraints	Constraints	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	106	6	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	106	7	now	now	RB
work_wgi74n6oavhdxhlomowlxqjdhu	106	8	formally	formally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	106	9	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	106	10	how	how	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	106	11	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	106	12	couple	couple	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	106	13	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	106	14	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	106	15	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	106	16	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	106	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	106	18	how	how	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	106	19	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	106	20	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	106	21	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	106	22	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	106	23	con-	con-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	106	24	straints	straint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	106	25	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	106	26	train	train	VB
work_wgi74n6oavhdxhlomowlxqjdhu	106	27	probabilistic	probabilistic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	106	28	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	106	29	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	106	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	107	1	Let	let	VB
work_wgi74n6oavhdxhlomowlxqjdhu	107	2	x	x	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	107	3	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	107	4	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	107	5	x1x2	x1x2	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	107	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	108	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	109	1	.x|x|	.x|x|	LS
work_wgi74n6oavhdxhlomowlxqjdhu	109	2	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	109	3	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	109	4	X	X	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	109	5	denote	denote	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	109	6	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	109	7	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	109	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	109	9	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	109	10	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	109	11	token	token	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	109	12	xi	xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	109	13	∈V	∈V	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	109	14	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	109	15	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	109	16	instance	instance	NN
work_wgi74n6oavhdxhlomowlxqjdhu	109	17	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	109	18	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	109	19	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	109	20	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	109	21	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	109	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	109	23	vocabulary	vocabulary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	109	24	V	v	NN
work_wgi74n6oavhdxhlomowlxqjdhu	109	25	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	109	26	let	let	VB
work_wgi74n6oavhdxhlomowlxqjdhu	109	27	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	109	28	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	109	29	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	109	30	y1y2	y1y2	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	109	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	110	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	111	1	.y|x|	.y|x|	LS
work_wgi74n6oavhdxhlomowlxqjdhu	111	2	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	111	3	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	111	4	Y	Y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	111	5	de-	de-	DT
work_wgi74n6oavhdxhlomowlxqjdhu	111	6	note	note	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	111	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	111	8	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	111	9	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	111	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	111	11	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	111	12	yi	yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	111	13	∈T	∈T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	111	14	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	111	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	111	16	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	111	17	assigned	assign	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	111	18	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	111	19	token	token	VB
work_wgi74n6oavhdxhlomowlxqjdhu	111	20	xi	xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	111	21	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	111	22	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	111	23	denotes	denote	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	111	24	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	111	25	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	111	26	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	111	27	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	111	28	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	111	29	part-	part-	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	111	30	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	111	31	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	111	32	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	111	33	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	111	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	112	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	112	2	denote	denote	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	112	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	112	4	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	112	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	112	6	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	112	7	admissible	admissible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	112	8	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	112	9	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	112	10	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	112	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	112	12	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	112	13	x	x	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	112	14	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	112	15	Y(x	y(x	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	112	16	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	112	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	113	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	113	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	113	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	113	4	4	4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	113	5	inference	inference	NN
work_wgi74n6oavhdxhlomowlxqjdhu	113	6	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	113	7	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	113	8	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	113	9	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	113	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	113	11	tagger	tagger	NN
work_wgi74n6oavhdxhlomowlxqjdhu	113	12	operates	operate	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	113	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	114	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	114	2	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	114	3	shall	shall	MD
work_wgi74n6oavhdxhlomowlxqjdhu	114	4	see	see	VB
work_wgi74n6oavhdxhlomowlxqjdhu	114	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	114	6	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	114	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	114	8	crucial	crucial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	114	9	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	114	10	constrain	constrain	VB
work_wgi74n6oavhdxhlomowlxqjdhu	114	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	114	12	size	size	NN
work_wgi74n6oavhdxhlomowlxqjdhu	114	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	114	14	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	114	15	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	114	16	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	114	17	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	114	18	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	114	19	simplify	simplify	VB
work_wgi74n6oavhdxhlomowlxqjdhu	114	20	learning	learn	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	114	21	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	114	22	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	114	23	incomplete	incomplete	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	114	24	supervision	supervision	NN
work_wgi74n6oavhdxhlomowlxqjdhu	114	25	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	114	26	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	114	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	115	1	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	115	2	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	3	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	115	4	maps	map	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	115	5	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	115	6	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	7	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	8	xj	xj	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	9	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	115	10	V	v	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	11	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	115	12	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	115	13	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	115	15	admissible	admissible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	115	16	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	115	17	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	115	18	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	115	19	xj	xj	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	115	20	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	115	21	⊆	⊆	.
work_wgi74n6oavhdxhlomowlxqjdhu	115	22	T	t	NN
work_wgi74n6oavhdxhlomowlxqjdhu	115	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	116	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	116	2	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	116	3	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	116	4	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	116	5	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	116	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	116	7	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	116	8	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	116	9	allow	allow	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	116	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	116	11	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	116	12	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	116	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	116	14	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	116	15	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	116	16	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	116	17	while	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	116	18	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	116	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	116	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	116	21	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	116	22	paper	paper	NN
work_wgi74n6oavhdxhlomowlxqjdhu	116	23	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	116	24	do	do	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	116	25	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	116	26	at-	at-	VB
work_wgi74n6oavhdxhlomowlxqjdhu	116	27	tempt	tempt	VB
work_wgi74n6oavhdxhlomowlxqjdhu	116	28	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	116	29	distinguish	distinguish	VB
work_wgi74n6oavhdxhlomowlxqjdhu	116	30	closed	closed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	116	31	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	116	32	class	class	NN
work_wgi74n6oavhdxhlomowlxqjdhu	116	33	versus	versus	IN
work_wgi74n6oavhdxhlomowlxqjdhu	116	34	open	open	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	116	35	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	116	36	class	class	NN
work_wgi74n6oavhdxhlomowlxqjdhu	116	37	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	116	38	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	116	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	117	1	When	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	117	2	provided	provide	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	117	3	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	117	4	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	117	5	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	117	6	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	117	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	117	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	117	9	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	117	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	117	11	admissible	admissible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	117	12	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	117	13	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	117	14	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	117	15	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	117	16	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	117	17	x	x	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	117	18	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	117	19	Y(x	Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	117	20	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	117	21	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	117	22	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	117	23	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	117	24	x1	x1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	117	25	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	117	26	×T	×T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	117	27	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	117	28	x2	x2	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	117	29	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	117	30	×	×	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	117	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	118	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	119	1	.×T	.×T	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	119	2	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	119	3	x|x|	x|x|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	119	4	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	119	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	120	1	When	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	120	2	no	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	120	3	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	120	4	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	120	5	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	120	6	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	120	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	120	8	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	120	9	simply	simply	RB
work_wgi74n6oavhdxhlomowlxqjdhu	120	10	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	120	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	120	12	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	120	13	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	120	14	Y(x	y(x	FW
work_wgi74n6oavhdxhlomowlxqjdhu	120	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	120	16	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	120	17	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	120	18	|x|	|x|	NN
work_wgi74n6oavhdxhlomowlxqjdhu	120	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	121	1	Let	let	VB
work_wgi74n6oavhdxhlomowlxqjdhu	121	2	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	121	3	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	121	4	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	121	5	ỹ1ỹ2	ỹ1ỹ2	NN
work_wgi74n6oavhdxhlomowlxqjdhu	121	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	122	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	123	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	124	1	ỹ|x|	ỹ|x|	LS
work_wgi74n6oavhdxhlomowlxqjdhu	124	2	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	124	3	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	124	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	124	5	projected	projected	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	124	6	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	124	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	124	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	124	9	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	124	10	x.	x.	.
work_wgi74n6oavhdxhlomowlxqjdhu	125	1	Note	note	VB
work_wgi74n6oavhdxhlomowlxqjdhu	125	2	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	125	3	{	{	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	125	4	ỹi	ỹi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	125	5	}	}	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	125	6	=	=	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	125	7	∅	∅	FW
work_wgi74n6oavhdxhlomowlxqjdhu	125	8	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	125	9	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	125	10	without	without	IN
work_wgi74n6oavhdxhlomowlxqjdhu	125	11	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	125	12	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	125	13	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	125	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	126	1	Next	next	RB
work_wgi74n6oavhdxhlomowlxqjdhu	126	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	126	3	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	126	4	define	define	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	126	5	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	126	6	piecewise	piecewise	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	7	operator	operator	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	8	_	_	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	126	9	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	126	10	couples	couple	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	126	11	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	126	12	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	126	13	Y(x	Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	126	14	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	126	15	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	126	16	respect	respect	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	17	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	126	18	every	every	DT
work_wgi74n6oavhdxhlomowlxqjdhu	126	19	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	20	index	index	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	21	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	126	22	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	126	23	results	result	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	126	24	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	126	25	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	126	26	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	27	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	126	28	type-	type-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	126	29	constrained	constrain	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	126	30	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	126	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	127	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	127	2	operator	operator	NN
work_wgi74n6oavhdxhlomowlxqjdhu	127	3	behaves	behave	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	127	4	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	127	5	follows	follow	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	127	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	127	7	coherent	coherent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	127	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	127	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	127	10	high	high	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	127	11	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	127	12	description	description	NN
work_wgi74n6oavhdxhlomowlxqjdhu	127	13	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	127	14	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	127	15	2.3	2.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	127	16	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	127	17	T̂	T̂	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	127	18	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	19	xi	xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	127	21	ỹi	ỹi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	22	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	23	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	127	24	ỹi	ỹi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	25	_	_	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	26	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	27	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	28	xi	xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	29	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	30	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	127	31	{	{	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	32	{	{	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	33	ỹi	ỹi	NN
work_wgi74n6oavhdxhlomowlxqjdhu	127	34	}	}	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	35	if	if	IN
work_wgi74n6oavhdxhlomowlxqjdhu	127	36	ỹi	ỹi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	37	∈T	∈t	FW
work_wgi74n6oavhdxhlomowlxqjdhu	127	38	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	39	xi	xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	40	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	41	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	42	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	43	xi	xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	127	44	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	127	45	otherwise	otherwise	RB
work_wgi74n6oavhdxhlomowlxqjdhu	127	46	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	128	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	128	2	denote	denote	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	128	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	128	4	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	128	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	128	6	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	128	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	128	8	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	128	9	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	128	10	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	128	11	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	128	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	128	13	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	128	14	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	128	15	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	128	16	T̂	T̂	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	128	17	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	128	18	x1	x1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	128	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	128	20	ỹ1)×T̂	ỹ1)×T̂	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	128	21	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	128	22	x2	x2	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	128	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	128	24	ỹ2)×.	ỹ2)×.	UH
work_wgi74n6oavhdxhlomowlxqjdhu	129	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	130	1	.×T̂	.×T̂	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	130	2	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	130	3	x|x|	x|x|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	130	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	130	5	ỹ|x|	ỹ|x|	NN
work_wgi74n6oavhdxhlomowlxqjdhu	130	6	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	130	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	131	1	Note	note	VB
work_wgi74n6oavhdxhlomowlxqjdhu	131	2	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	131	3	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	131	4	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	131	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	131	6	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	131	7	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	131	8	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	131	9	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	131	10	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	131	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	131	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	131	13	dictionary	dictionary	RB
work_wgi74n6oavhdxhlomowlxqjdhu	131	14	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	131	15	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	131	16	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	131	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	131	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	131	19	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	131	20	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	131	21	cou-	cou-	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	131	22	pled	plead	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	131	23	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	131	24	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	131	25	identical	identical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	131	26	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	131	27	that	that	RB
work_wgi74n6oavhdxhlomowlxqjdhu	131	28	is	is	RB
work_wgi74n6oavhdxhlomowlxqjdhu	131	29	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	131	30	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	131	31	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	131	32	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	131	33	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	131	34	Y(x	y(x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	131	35	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	131	36	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	132	1	3.1	3.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	132	2	HMMs	hmm	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	132	3	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	132	4	Coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	132	5	Constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	132	6	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	132	7	first	first	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	132	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	132	9	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	10	hidden	hide	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	132	11	Markov	Markov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	12	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	13	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	132	14	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	132	16	specifies	specify	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	132	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	132	18	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	132	19	distribution	distribution	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	132	21	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	132	22	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	23	x	x	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	132	24	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	25	X	x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	26	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	132	27	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	132	28	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	29	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	132	30	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	31	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	32	∈Y(x	∈Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	33	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	132	34	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	132	35	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	132	36	pβ(x	pβ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	37	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	132	38	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	39	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	132	40	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	132	41	|x|∏	|x|∏	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	132	42	i=1	i=1	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	132	43	pβ(xi	pβ(xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	44	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	45	yi)︸	yi)︸	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	46	︷︷	︷︷	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	47	︸	︸	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	132	48	emission	emission	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	49	pβ(yi	pβ(yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	50	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	51	yi−1)︸	yi−1)︸	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	132	52	︷︷	︷︷	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	53	︸	︸	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	132	54	transition	transition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	132	55	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	133	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	133	2	follow	follow	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	133	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	133	4	recent	recent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	133	5	trend	trend	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	133	7	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	133	8	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	133	9	log	log	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	133	11	linear	linear	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	12	parametrization	parametrization	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	133	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	133	15	emission	emission	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	16	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	133	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	133	18	transition	transition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	19	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	133	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	133	21	instead	instead	RB
work_wgi74n6oavhdxhlomowlxqjdhu	133	22	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	133	23	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	133	24	multinomial	multinomial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	133	25	parametriza-	parametriza-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	26	tion	tion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	133	27	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	133	28	Chen	Chen	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	133	29	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	133	30	2003	2003	CD
work_wgi74n6oavhdxhlomowlxqjdhu	133	31	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	133	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	134	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	134	2	allows	allow	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	134	3	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	134	4	parameters	parameter	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	134	5	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	134	6	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	134	7	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	134	8	shared	share	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	134	9	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	134	10	categorical	categorical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	134	11	events	event	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	134	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	134	13	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	134	14	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	134	15	been	be	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	134	16	shown	show	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	134	17	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	134	18	give	give	VB
work_wgi74n6oavhdxhlomowlxqjdhu	134	19	superior	superior	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	134	20	performance	performance	NN
work_wgi74n6oavhdxhlomowlxqjdhu	134	21	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	134	22	Berg-	Berg-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	134	23	Kirkpatrick	Kirkpatrick	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	134	24	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	134	25	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	134	26	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	134	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	134	28	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	134	29	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	134	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	135	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	135	2	categorical	categorical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	135	3	emission	emission	NN
work_wgi74n6oavhdxhlomowlxqjdhu	135	4	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	135	5	transition	transition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	135	6	events	event	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	135	7	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	135	8	represented	represent	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	135	9	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	135	10	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	135	11	vec-	vec-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	135	12	tors	tor	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	135	13	φ(xi	φ(xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	135	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	135	15	yi	yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	135	16	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	135	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	135	18	φ(yi	φ(yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	135	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	135	20	yi−1	yi−1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	135	21	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	135	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	136	1	Each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	136	2	element	element	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	136	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	136	5	parameter	parameter	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	6	vector	vector	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	7	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	8	corresponds	correspond	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	136	9	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	136	10	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	136	11	particular	particular	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	136	12	fea-	fea-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	13	ture	ture	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	14	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	136	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	136	16	component	component	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	17	log	log	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	18	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	136	19	linear	linear	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	20	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	136	21	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	136	22	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	136	23	pβ(xi	pβ(xi	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	136	24	|	|	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	136	25	yi	yi	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	136	26	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	27	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	136	28	exp	exp	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	29	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	30	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	31	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	136	32	φ(xi	φ(xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	33	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	136	34	yi	yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	35	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	36	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	37	∑	∑	.
work_wgi74n6oavhdxhlomowlxqjdhu	136	38	x′i∈V	x′i∈V	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	39	exp	exp	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	40	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	41	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	42	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	136	43	φ(x′i	φ(x′i	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	44	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	136	45	yi	yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	46	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	47	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	48	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	136	49	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	136	50	pβ(yi	pβ(yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	51	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	52	yi−1	yi−1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	53	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	54	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	136	55	exp	exp	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	56	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	57	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	58	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	136	59	φ(yi	φ(yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	60	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	136	61	yi−1	yi−1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	62	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	63	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	64	∑	∑	.
work_wgi74n6oavhdxhlomowlxqjdhu	136	65	y′i∈T	y′i∈T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	66	exp	exp	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	67	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	68	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	69	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	136	70	φ(y′i	φ(y′i	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	136	71	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	136	72	yi−1	yi−1	NN
work_wgi74n6oavhdxhlomowlxqjdhu	136	73	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	74	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	136	75	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	137	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	137	2	maximum	maximum	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	137	3	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	137	4	likelihood	likelihood	NN
work_wgi74n6oavhdxhlomowlxqjdhu	137	5	estimation	estimation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	137	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	137	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	137	8	parameters	parameter	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	137	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	137	10	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	137	11	seek	seek	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	137	12	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	137	13	maximize	maximize	VB
work_wgi74n6oavhdxhlomowlxqjdhu	137	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	137	15	likelihood	likelihood	NN
work_wgi74n6oavhdxhlomowlxqjdhu	137	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	137	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	137	18	observed	observed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	137	19	parts	part	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	137	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	137	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	137	22	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	137	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	138	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	138	2	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	138	3	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	138	4	need	need	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	138	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	138	6	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	138	7	marginal	marginal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	138	8	distribution	distribution	NN
work_wgi74n6oavhdxhlomowlxqjdhu	138	9	pβ(x	pβ(x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	138	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	11	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	13	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	14	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	138	17	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	138	18	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	138	19	x	x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	138	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	21	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	138	22	its	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	138	23	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	138	24	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	138	25	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	138	26	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	28	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	29	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	30	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	31	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	138	32	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	138	33	obtained	obtain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	138	34	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	138	35	marginalizing	marginalize	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	138	36	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	138	37	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	138	38	consistent	consistent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	138	39	outputs	output	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	138	40	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	138	41	pβ(x	pβ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	42	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	43	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	44	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	45	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	46	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	47	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	48	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	138	49	∑	∑	.
work_wgi74n6oavhdxhlomowlxqjdhu	138	50	y∈Ŷ(x	y∈ŷ(x	XX
work_wgi74n6oavhdxhlomowlxqjdhu	138	51	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	52	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	53	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	54	pβ(x	pβ(x	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	138	55	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	138	56	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	138	57	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	138	58	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	139	1	If	if	IN
work_wgi74n6oavhdxhlomowlxqjdhu	139	2	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	139	3	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	139	4	no	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	139	5	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	139	6	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	139	7	no	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	139	8	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	139	9	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	139	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	139	11	then	then	RB
work_wgi74n6oavhdxhlomowlxqjdhu	139	12	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	139	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	139	14	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	139	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	139	16	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	139	17	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	139	18	|x|	|x|	NN
work_wgi74n6oavhdxhlomowlxqjdhu	139	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	139	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	139	21	thus	thus	RB
work_wgi74n6oavhdxhlomowlxqjdhu	139	22	pβ(x	pβ(x	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	139	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	139	24	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	139	25	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	139	26	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	139	27	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	139	28	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	139	29	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	139	30	pβ(x	pβ(x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	139	31	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	139	32	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	139	33	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	139	34	reduces	reduce	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	139	35	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	139	36	fully	fully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	139	37	unsupervised	unsupervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	139	38	learning	learning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	139	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	140	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	140	2	`	`	''
work_wgi74n6oavhdxhlomowlxqjdhu	140	3	2-regularized	2-regularized	CD
work_wgi74n6oavhdxhlomowlxqjdhu	140	4	marginal	marginal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	140	5	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	140	6	log	log	NN
work_wgi74n6oavhdxhlomowlxqjdhu	140	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	140	8	likelihood	likelihood	NN
work_wgi74n6oavhdxhlomowlxqjdhu	140	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	140	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	140	11	constrained	constrained	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	140	12	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	140	13	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	140	14	D	d	NN
work_wgi74n6oavhdxhlomowlxqjdhu	140	15	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	140	16	{	{	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	140	17	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	140	18	x(i	x(i	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	140	19	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	140	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	140	21	ỹ(i))}ni=1	ỹ(i))}ni=1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	140	22	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	140	23	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	140	24	L(β;D	L(β;D	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	140	25	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	140	26	=	=	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	140	27	n∑	n∑	VB
work_wgi74n6oavhdxhlomowlxqjdhu	140	28	i=1	i=1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	140	29	log	log	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	140	30	pβ(x	pβ(x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	140	31	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	140	32	i),Ŷ(x(i	i),ŷ(x(i	NN
work_wgi74n6oavhdxhlomowlxqjdhu	140	33	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	140	34	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	140	35	ỹ(i)))−γ‖β‖22	ỹ(i)))−γ‖β‖22	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	140	36	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	141	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	141	2	1	1	LS
work_wgi74n6oavhdxhlomowlxqjdhu	141	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	141	4	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	141	5	follow	follow	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	141	6	Berg	Berg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	141	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	141	8	Kirkpatrick	Kirkpatrick	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	141	9	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	141	10	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	141	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	142	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	142	2	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	142	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	142	4	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	142	5	take	take	VB
work_wgi74n6oavhdxhlomowlxqjdhu	142	6	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	142	7	direct	direct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	142	8	gradient	gradient	NN
work_wgi74n6oavhdxhlomowlxqjdhu	142	9	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	142	10	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	142	11	optimizing	optimize	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	142	12	Eq	Eq	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	142	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	143	1	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	143	2	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	143	3	L	L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	143	4	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	143	5	BFGS	BFGS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	143	6	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	143	7	Liu	Liu	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	143	8	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	143	9	Nocedal	Nocedal	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	143	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	143	11	1989	1989	CD
work_wgi74n6oavhdxhlomowlxqjdhu	143	12	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	143	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	144	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	144	2	set	set	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	144	3	γ	γ	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	144	4	=	=	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	144	5	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	144	6	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	144	7	run	run	VB
work_wgi74n6oavhdxhlomowlxqjdhu	144	8	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	144	9	iterations	iteration	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	144	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	144	11	L	l	NN
work_wgi74n6oavhdxhlomowlxqjdhu	144	12	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	144	13	BFGS	BFGS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	144	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	145	1	One	one	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	145	2	could	could	MD
work_wgi74n6oavhdxhlomowlxqjdhu	145	3	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	145	4	em-	em-	VB
work_wgi74n6oavhdxhlomowlxqjdhu	145	5	ploy	ploy	VB
work_wgi74n6oavhdxhlomowlxqjdhu	145	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	7	Expectation	Expectation	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	145	9	Maximization	Maximization	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	10	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	145	11	EM	EM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	12	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	145	13	algorithm	algorithm	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	14	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	145	15	Dempster	Dempster	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	16	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	17	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	18	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	145	20	1977	1977	CD
work_wgi74n6oavhdxhlomowlxqjdhu	145	21	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	145	22	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	145	23	optimize	optimize	VB
work_wgi74n6oavhdxhlomowlxqjdhu	145	24	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	25	objective	objective	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	26	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	145	27	al-	al-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	145	28	though	though	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	29	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	30	relative	relative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	31	merits	merit	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	145	32	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	33	EM	EM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	34	versus	versus	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	35	direct	direct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	36	gradi-	gradi-	FW
work_wgi74n6oavhdxhlomowlxqjdhu	145	37	ent	ent	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	38	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	39	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	40	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	41	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	145	42	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	145	43	still	still	RB
work_wgi74n6oavhdxhlomowlxqjdhu	145	44	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	45	topic	topic	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	46	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	47	debate	debate	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	48	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	145	49	Berg	Berg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	50	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	145	51	Kirkpatrick	Kirkpatrick	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	52	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	145	53	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	54	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	55	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	145	56	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	145	57	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	145	58	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	59	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	60	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	61	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	62	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	145	63	2012).4	2012).4	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	64	Note	note	VB
work_wgi74n6oavhdxhlomowlxqjdhu	145	65	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	66	since	since	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	67	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	68	marginal	marginal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	69	likelihood	likelihood	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	70	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	145	71	non	non	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	72	-	-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	73	concave	concave	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	74	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	145	75	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	145	76	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	145	77	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	145	78	guaranteed	guarantee	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	145	79	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	145	80	find	find	VB
work_wgi74n6oavhdxhlomowlxqjdhu	145	81	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	145	82	local	local	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	145	83	maximum	maximum	NN
work_wgi74n6oavhdxhlomowlxqjdhu	145	84	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	145	85	Eq	Eq	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	145	86	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	146	1	1	1	LS
work_wgi74n6oavhdxhlomowlxqjdhu	146	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	147	1	After	after	IN
work_wgi74n6oavhdxhlomowlxqjdhu	147	2	estimating	estimate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	147	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	147	4	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	5	parameters	parameter	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	147	6	β	β	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	147	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	147	9	tag-	tag-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	10	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	11	y∗	y∗	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	147	12	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	13	Y(x	Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	14	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	147	15	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	147	16	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	147	17	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	18	x	x	LS
work_wgi74n6oavhdxhlomowlxqjdhu	147	19	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	20	X	x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	21	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	147	22	pre-	pre-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	147	23	dicted	dicte	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	147	24	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	147	25	choosing	choose	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	147	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	147	27	one	one	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	28	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	147	29	maximal	maximal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	147	30	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	147	31	prob-	prob-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	147	32	ability	ability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	147	33	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	147	34	y∗	y∗	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	35	←	←	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	36	arg	arg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	37	max	max	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	38	y∈Y(x	y∈Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	39	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	147	40	pβ(x	pβ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	41	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	147	42	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	147	43	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	147	44	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	148	1	4We	4we	CD
work_wgi74n6oavhdxhlomowlxqjdhu	148	2	trained	train	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	148	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	148	4	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	148	5	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	148	6	EM	EM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	148	7	as	as	RB
work_wgi74n6oavhdxhlomowlxqjdhu	148	8	well	well	RB
work_wgi74n6oavhdxhlomowlxqjdhu	148	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	148	10	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	148	11	achieved	achieve	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	148	12	better	well	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	148	13	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	148	14	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	148	15	direct	direct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	148	16	gradient	gradient	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	148	17	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	148	18	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	148	19	hence	hence	RB
work_wgi74n6oavhdxhlomowlxqjdhu	148	20	omit	omit	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	148	21	those	those	DT
work_wgi74n6oavhdxhlomowlxqjdhu	148	22	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	148	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	149	1	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	149	2	3.2	3.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	149	3	CRFs	crf	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	149	4	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	149	5	Coupled	Coupled	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	6	Constraints	Constraints	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	7	Whereas	whereas	IN
work_wgi74n6oavhdxhlomowlxqjdhu	149	8	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	9	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	10	models	model	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	149	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	12	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	149	13	probability	probability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	149	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	16	input	input	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	17	x	x	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	149	18	∈X	∈x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	19	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	149	20	output	output	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	21	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	22	∈Y(x	∈Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	23	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	149	25	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	149	26	locally	locally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	149	27	normalized	normalized	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	149	28	component	component	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	29	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	149	30	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	149	31	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	32	conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	149	33	random	random	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	149	34	field	field	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	35	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	36	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	37	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	38	instead	instead	RB
work_wgi74n6oavhdxhlomowlxqjdhu	149	39	models	model	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	149	40	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	41	probability	probability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	42	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	149	43	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	44	output	output	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	45	conditioned	condition	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	149	46	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	149	47	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	48	input	input	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	49	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	149	50	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	51	globally	globally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	149	52	nor-	nor-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	149	53	malized	malized	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	149	54	log	log	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	55	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	149	56	linear	linear	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	57	distribution	distribution	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	58	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	59	Lafferty	Lafferty	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	60	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	61	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	62	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	63	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	149	64	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	149	65	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	66	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	149	67	pθ(y	pθ(y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	68	|	|	POS
work_wgi74n6oavhdxhlomowlxqjdhu	149	69	x	x	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	149	70	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	71	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	149	72	exp	exp	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	73	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	74	θ	θ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	75	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	149	76	Φ(x	Φ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	77	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	149	78	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	79	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	80	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	81	∑	∑	.
work_wgi74n6oavhdxhlomowlxqjdhu	149	82	y′∈Y(x	y′∈Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	83	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	84	exp	exp	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	85	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	86	θ	θ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	87	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	149	88	Φ(x	Φ(x	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	149	89	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	149	90	y′	y′	LS
work_wgi74n6oavhdxhlomowlxqjdhu	149	91	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	92	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	149	93	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	149	94	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	149	95	θ	θ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	149	96	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	149	97	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	149	98	parameter	parameter	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	99	vector	vector	NN
work_wgi74n6oavhdxhlomowlxqjdhu	149	100	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	150	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	150	2	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	150	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	150	4	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	150	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	150	6	Y(x	Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	150	7	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	150	8	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	150	9	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	150	10	necessarily	necessarily	RB
work_wgi74n6oavhdxhlomowlxqjdhu	150	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	150	12	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	150	13	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	150	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	150	15	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	150	16	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	150	17	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	150	18	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	150	19	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	150	20	specifically	specifically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	150	21	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	150	22	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	150	23	us	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	150	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	150	25	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	150	26	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	150	27	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	150	28	dictionary-	dictionary-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	150	29	pruned	prune	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	150	30	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	150	31	without	without	IN
work_wgi74n6oavhdxhlomowlxqjdhu	150	32	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	150	33	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	150	34	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	150	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	151	1	With	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	151	2	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	151	3	first	first	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	151	4	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	151	5	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	151	6	Markov	Markov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	151	7	assumption	assumption	NN
work_wgi74n6oavhdxhlomowlxqjdhu	151	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	151	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	151	10	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	151	11	function	function	NN
work_wgi74n6oavhdxhlomowlxqjdhu	151	12	factors	factor	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	151	13	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	151	14	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	151	15	Φ(x	Φ(x	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	151	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	151	17	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	151	18	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	151	19	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	151	20	|x|∑	|x|∑	NN
work_wgi74n6oavhdxhlomowlxqjdhu	151	21	i=1	i=1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	151	22	φ(x	φ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	151	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	151	24	yi	yi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	151	25	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	151	26	yi−1	yi−1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	151	27	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	151	28	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	152	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	152	2	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	152	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	152	4	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	152	5	powerful	powerful	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	152	6	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	152	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	152	8	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	152	9	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	152	10	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	152	11	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	152	12	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	152	13	use	use	VB
work_wgi74n6oavhdxhlomowlxqjdhu	152	14	richer	rich	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	152	15	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	152	16	definitions	definition	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	152	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	152	18	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	152	19	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	152	20	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	152	21	in-	in-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	152	22	put	put	NN
work_wgi74n6oavhdxhlomowlxqjdhu	152	23	/	/	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	152	24	transition	transition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	152	25	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	152	26	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	152	27	features	feature	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	152	28	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	152	29	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	152	30	wider	wide	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	152	31	input	input	NN
work_wgi74n6oavhdxhlomowlxqjdhu	152	32	context	context	NN
work_wgi74n6oavhdxhlomowlxqjdhu	152	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	153	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	153	2	model	model	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	153	3	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	153	4	marginal	marginal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	153	5	conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	153	6	probabil-	probabil-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	153	7	ity	ity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	153	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	153	9	given	give	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	153	10	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	153	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	153	12	total	total	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	153	13	probability	probability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	153	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	153	15	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	153	16	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	153	17	sequences	sequence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	153	18	consistent	consistent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	153	19	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	153	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	153	21	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	153	22	Ŷ(x	Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	153	24	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	25	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	153	26	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	153	27	pθ(Ŷ(x	pθ(Ŷ(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	28	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	153	29	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	30	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	153	31	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	32	x	x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	33	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	153	34	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	153	35	∑	∑	.
work_wgi74n6oavhdxhlomowlxqjdhu	153	36	y∈Ŷ(x	y∈ŷ(x	XX
work_wgi74n6oavhdxhlomowlxqjdhu	153	37	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	153	38	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	39	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	153	40	pθ(y	pθ(y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	41	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	153	42	x	x	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	153	43	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	153	44	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	154	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	154	2	parameters	parameter	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	154	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	154	4	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	154	5	constrained	constrained	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	154	6	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	7	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	154	8	estimated	estimate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	154	9	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	154	10	maximizing	maximize	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	154	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	154	12	`	`	''
work_wgi74n6oavhdxhlomowlxqjdhu	154	13	2-regularized	2-regularized	CD
work_wgi74n6oavhdxhlomowlxqjdhu	154	14	marginal	marginal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	154	15	condi-	condi-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	16	tional	tional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	154	17	log	log	NN
work_wgi74n6oavhdxhlomowlxqjdhu	154	18	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	154	19	likelihood	likelihood	NN
work_wgi74n6oavhdxhlomowlxqjdhu	154	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	154	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	154	22	constrained	constrained	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	154	23	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	154	24	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	25	Riezler	Riezler	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	26	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	154	27	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	28	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	29	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	154	30	2002	2002	CD
work_wgi74n6oavhdxhlomowlxqjdhu	154	31	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	32	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	154	33	L(θ;D	l(θ;d	LS
work_wgi74n6oavhdxhlomowlxqjdhu	154	34	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	35	=	=	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	154	36	n∑	n∑	VB
work_wgi74n6oavhdxhlomowlxqjdhu	154	37	i=1	i=1	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	38	log	log	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	39	pθ(Ŷ(x(i	pθ(Ŷ(x(i	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	40	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	41	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	154	42	ỹ(i	ỹ(i	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	43	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	44	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	45	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	46	x(i	x(i	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	47	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	48	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	154	49	−γ‖θ‖22	−γ‖θ‖22	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	154	50	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	155	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	155	2	2	2	LS
work_wgi74n6oavhdxhlomowlxqjdhu	155	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	155	4	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	155	5	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	155	6	Eq	Eq	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	155	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	156	1	1	1	LS
work_wgi74n6oavhdxhlomowlxqjdhu	156	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	156	3	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	156	4	maximize	maximize	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	156	5	Eq	Eq	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	156	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	157	1	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	157	2	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	157	3	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	157	4	itera-	itera-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	157	5	tions	tion	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	157	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	157	7	L	l	NN
work_wgi74n6oavhdxhlomowlxqjdhu	157	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	157	9	BFGS	BFGS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	157	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	157	11	set	set	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	157	12	γ	γ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	157	13	=	=	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	157	14	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	157	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	158	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	158	2	contrast	contrast	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	3	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	158	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	158	5	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	158	7	after	after	IN
work_wgi74n6oavhdxhlomowlxqjdhu	158	8	estimating	estimate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	158	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	158	10	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	11	parameters	parameter	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	158	12	θ	θ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	158	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	158	15	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	158	17	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	18	y∗	y∗	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	19	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	20	Y(x	Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	21	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	158	22	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	158	23	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	158	24	sentence	sentence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	25	x	x	LS
work_wgi74n6oavhdxhlomowlxqjdhu	158	26	∈	∈	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	27	X	x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	28	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	158	29	chosen	choose	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	158	30	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	158	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	158	32	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	33	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	158	34	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	158	35	maximal	maximal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	158	36	conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	158	37	probability	probability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	158	38	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	158	39	y∗	y∗	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	40	←	←	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	41	arg	arg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	42	max	max	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	43	y∈Y(x	y∈Y(x	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	44	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	158	45	pθ(y	pθ(y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	46	|	|	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	158	47	x	x	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	158	48	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	158	49	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	159	1	4	4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	159	2	Empirical	Empirical	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	159	3	Study	Study	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	159	4	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	159	5	now	now	RB
work_wgi74n6oavhdxhlomowlxqjdhu	159	6	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	159	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	159	8	detailed	detailed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	159	9	empirical	empirical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	159	10	study	study	NN
work_wgi74n6oavhdxhlomowlxqjdhu	159	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	159	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	159	13	mod-	mod-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	159	14	els	els	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	159	15	proposed	propose	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	159	16	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	159	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	159	18	previous	previous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	159	19	sections	section	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	159	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	160	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	160	2	addition	addition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	160	3	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	160	4	comparing	compare	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	160	5	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	160	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	160	7	state	state	NN
work_wgi74n6oavhdxhlomowlxqjdhu	160	8	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	160	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	160	10	art	art	NN
work_wgi74n6oavhdxhlomowlxqjdhu	160	11	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	160	12	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	160	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	160	14	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	160	15	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	160	16	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	160	17	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	160	18	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	160	19	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	160	20	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	160	21	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	160	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	161	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	161	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	161	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	161	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	161	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	161	6	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	161	7	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	161	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	161	9	several	several	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	161	10	combinations	combination	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	161	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	161	12	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	161	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	161	14	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	161	15	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	161	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	161	17	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	161	18	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	161	19	incorporating	incorporate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	161	20	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	161	21	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	161	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	162	1	Both	both	DT
work_wgi74n6oavhdxhlomowlxqjdhu	162	2	generative	generative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	162	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	162	4	discriminative	discriminative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	162	5	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	162	6	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	162	7	explored	explore	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	162	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	163	1	4.1	4.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	163	2	Experimental	Experimental	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	163	3	Setup	Setup	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	163	4	Before	before	IN
work_wgi74n6oavhdxhlomowlxqjdhu	163	5	delving	delve	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	163	6	into	into	IN
work_wgi74n6oavhdxhlomowlxqjdhu	163	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	163	8	experimental	experimental	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	163	9	details	detail	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	163	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	163	11	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	163	12	present	present	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	163	13	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	163	14	setup	setup	NN
work_wgi74n6oavhdxhlomowlxqjdhu	163	15	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	163	16	datasets	dataset	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	163	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	164	1	Languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	164	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	165	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	165	2	evaluate	evaluate	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	165	3	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	165	4	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	165	5	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	165	6	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	165	7	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	165	8	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	165	9	previous	previous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	165	10	work	work	NN
work_wgi74n6oavhdxhlomowlxqjdhu	165	11	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	165	12	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	165	14	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	165	16	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	165	17	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	165	18	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	19	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	20	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	21	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	22	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	165	23	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	165	24	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	165	25	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	165	26	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	165	27	seven	seven	CD
work_wgi74n6oavhdxhlomowlxqjdhu	165	28	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	165	29	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	165	30	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	165	31	see	see	VB
work_wgi74n6oavhdxhlomowlxqjdhu	165	32	Ta-	Ta-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	165	33	ble	ble	VB
work_wgi74n6oavhdxhlomowlxqjdhu	165	34	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	165	35	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	165	36	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	166	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	166	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	166	3	former	former	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	166	4	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	166	5	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	166	6	all	all	RB
work_wgi74n6oavhdxhlomowlxqjdhu	166	7	belong	belong	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	166	8	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	166	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	166	10	Indo	indo	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	166	11	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	166	12	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	166	13	family	family	NN
work_wgi74n6oavhdxhlomowlxqjdhu	166	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	166	15	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	166	16	broaden	broaden	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	166	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	166	18	coverage	coverage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	166	19	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	166	20	language	language	VB
work_wgi74n6oavhdxhlomowlxqjdhu	166	21	families	family	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	166	22	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	166	23	distant	distant	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	166	24	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	166	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	166	26	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	166	27	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	166	28	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	166	29	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	166	30	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	166	31	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	166	32	Chinese	chinese	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	166	33	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	166	34	Japanese	japanese	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	166	35	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	166	36	Turk-	Turk-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	166	37	ish	ish	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	166	38	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	166	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	167	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	167	2	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	167	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	167	4	treebanks	treebank	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	167	5	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	167	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	167	7	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	8	shared	share	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	167	9	tasks	task	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	167	10	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	167	11	dependency	dependency	NN
work_wgi74n6oavhdxhlomowlxqjdhu	167	12	parsing	parse	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	167	13	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	167	14	Buchholz	Buchholz	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	15	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	167	16	Marsi	Marsi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	167	18	2006	2006	CD
work_wgi74n6oavhdxhlomowlxqjdhu	167	19	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	167	20	Nivre	Nivre	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	21	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	167	22	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	23	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	167	25	2007	2007	CD
work_wgi74n6oavhdxhlomowlxqjdhu	167	26	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	167	27	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	167	28	evaluation.5	evaluation.5	NN
work_wgi74n6oavhdxhlomowlxqjdhu	167	29	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	167	30	two-	two-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	167	31	letter	letter	NN
work_wgi74n6oavhdxhlomowlxqjdhu	167	32	abbreviations	abbreviation	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	167	33	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	167	34	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	167	35	ISO	ISO	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	167	36	639	639	CD
work_wgi74n6oavhdxhlomowlxqjdhu	167	37	-	-	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	167	38	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	167	39	standard	standard	NN
work_wgi74n6oavhdxhlomowlxqjdhu	167	40	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	167	41	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	167	42	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	167	43	referring	refer	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	167	44	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	167	45	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	167	46	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	167	47	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	167	48	tables	table	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	167	49	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	167	50	figures	figure	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	167	51	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	168	1	Tagset	tagset	NN
work_wgi74n6oavhdxhlomowlxqjdhu	168	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	169	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	169	2	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	169	3	cases	case	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	169	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	169	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	169	6	map	map	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	169	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	169	8	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	169	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	169	10	specific	specific	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	169	11	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	169	12	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	169	13	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	169	14	universal	universal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	169	15	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	169	16	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	169	17	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	169	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	169	19	mapping	mapping	NN
work_wgi74n6oavhdxhlomowlxqjdhu	169	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	169	21	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	169	22	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	169	23	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	169	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	170	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	170	2	2012).6	2012).6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	170	3	Since	since	IN
work_wgi74n6oavhdxhlomowlxqjdhu	170	4	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	170	5	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	170	6	indirect	indirect	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	170	7	super-	super-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	170	8	vision	vision	NN
work_wgi74n6oavhdxhlomowlxqjdhu	170	9	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	170	10	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	170	11	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	170	12	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	170	13	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	170	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	170	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	170	16	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	170	17	states	state	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	170	18	induced	induce	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	170	19	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	170	20	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	170	21	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	170	22	correspond	correspond	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	170	23	directly	directly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	170	24	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	170	25	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	170	26	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	170	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	170	28	enabling	enable	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	170	29	us	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	170	30	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	170	31	compute	compute	VB
work_wgi74n6oavhdxhlomowlxqjdhu	170	32	tagging	tag	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	170	33	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	170	34	without	without	IN
work_wgi74n6oavhdxhlomowlxqjdhu	170	35	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	170	36	greedy	greedy	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	170	37	1-to-1	1-to-1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	170	38	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	170	39	many	many	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	170	40	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	170	41	to-1	to-1	NN
work_wgi74n6oavhdxhlomowlxqjdhu	170	42	mapping	mapping	NN
work_wgi74n6oavhdxhlomowlxqjdhu	170	43	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	171	1	Bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	171	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	172	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	172	2	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	172	3	experiments	experiment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	172	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	172	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	172	6	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	172	7	English	English	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	172	8	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	172	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	172	10	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	172	11	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	172	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	173	1	Depending	depend	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	173	2	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	173	3	availability	availability	NN
work_wgi74n6oavhdxhlomowlxqjdhu	173	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	173	5	there	there	EX
work_wgi74n6oavhdxhlomowlxqjdhu	173	6	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	173	7	between	between	IN
work_wgi74n6oavhdxhlomowlxqjdhu	173	8	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	173	9	M	M	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	173	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	173	11	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	173	12	M	M	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	173	13	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	173	14	sentences	sentence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	173	15	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	173	16	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	173	17	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	173	18	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	174	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	174	2	majority	majority	NN
work_wgi74n6oavhdxhlomowlxqjdhu	174	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	174	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	174	5	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	174	6	data	data	NN
work_wgi74n6oavhdxhlomowlxqjdhu	174	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	174	8	gath-	gath-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	174	9	ered	ere	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	174	10	automatically	automatically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	174	11	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	174	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	174	13	web	web	NN
work_wgi74n6oavhdxhlomowlxqjdhu	174	14	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	174	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	174	16	method	method	NN
work_wgi74n6oavhdxhlomowlxqjdhu	174	17	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	174	18	Uszkoreit	Uszkoreit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	174	19	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	174	20	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	174	21	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	175	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	175	2	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	175	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	175	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	176	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	176	2	further	further	RB
work_wgi74n6oavhdxhlomowlxqjdhu	176	3	include	include	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	176	4	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	176	5	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	176	6	Europarl	Europarl	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	176	7	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	176	8	Koehn	Koehn	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	176	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	176	10	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	176	11	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	176	12	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	176	13	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	176	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	176	15	UN	UN	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	176	16	par-	par-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	176	17	allel	allel	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	176	18	corpus	corpus	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	176	19	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	176	20	UN	UN	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	176	21	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	176	22	2006	2006	CD
work_wgi74n6oavhdxhlomowlxqjdhu	176	23	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	176	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	176	25	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	176	26	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	176	27	covered	cover	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	176	28	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	176	29	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	176	30	corpora	corpora	NN
work_wgi74n6oavhdxhlomowlxqjdhu	176	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	177	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	177	2	English	english	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	177	3	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	177	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	177	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	177	6	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	177	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	177	8	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	9	tagged	tag	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	177	10	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	177	11	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	177	12	standard	standard	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	177	13	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	177	14	CRF	crf	NN
work_wgi74n6oavhdxhlomowlxqjdhu	177	15	tagger	tagger	NN
work_wgi74n6oavhdxhlomowlxqjdhu	177	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	177	17	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	177	18	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	177	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	177	20	Penn	Penn	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	21	Treebank	Treebank	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	22	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	177	23	Marcus	Marcus	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	24	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	25	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	26	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	177	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	177	28	1993	1993	CD
work_wgi74n6oavhdxhlomowlxqjdhu	177	29	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	177	30	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	177	31	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	177	32	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	177	33	mapped	map	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	177	34	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	177	35	universal	universal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	177	36	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	177	37	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	178	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	178	2	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	178	3	sen-	sen-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	178	4	5For	5for	CD
work_wgi74n6oavhdxhlomowlxqjdhu	178	5	French	French	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	178	6	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	178	7	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	178	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	178	9	treebank	treebank	NN
work_wgi74n6oavhdxhlomowlxqjdhu	178	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	178	11	Abeillé	Abeillé	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	178	12	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	178	13	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	178	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	179	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	179	2	2003	2003	CD
work_wgi74n6oavhdxhlomowlxqjdhu	179	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	179	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	180	1	6We	6we	CD
work_wgi74n6oavhdxhlomowlxqjdhu	180	2	use	use	NN
work_wgi74n6oavhdxhlomowlxqjdhu	180	3	version	version	NN
work_wgi74n6oavhdxhlomowlxqjdhu	180	4	1.03	1.03	CD
work_wgi74n6oavhdxhlomowlxqjdhu	180	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	180	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	180	7	mappings	mapping	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	180	8	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	180	9	at	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	180	10	http	http	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	180	11	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	180	12	//code.google.com	//code.google.com	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	180	13	/	/	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	180	14	p	p	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	180	15	/	/	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	180	16	universal	universal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	180	17	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	180	18	pos	pos	NN
work_wgi74n6oavhdxhlomowlxqjdhu	180	19	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	180	20	tags/.	tags/.	NN
work_wgi74n6oavhdxhlomowlxqjdhu	181	1	6	6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	181	2	tences	tence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	181	3	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	181	4	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	181	5	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	181	6	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	181	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	181	8	aligner	aligner	NN
work_wgi74n6oavhdxhlomowlxqjdhu	181	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	181	10	DeNero	DeNero	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	181	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	181	12	Macherey	Macherey	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	181	13	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	181	14	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	181	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	181	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	182	1	Intersected	intersect	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	182	2	high	high	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	182	3	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	182	4	confidence	confidence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	182	5	alignments	alignment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	182	6	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	182	7	confidence	confidence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	182	8	>	>	XX
work_wgi74n6oavhdxhlomowlxqjdhu	182	9	0.95	0.95	CD
work_wgi74n6oavhdxhlomowlxqjdhu	182	10	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	182	11	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	182	12	extracted	extract	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	182	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	182	14	ag-	ag-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	182	15	gregated	gregate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	182	16	into	into	IN
work_wgi74n6oavhdxhlomowlxqjdhu	182	17	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	182	18	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	182	19	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	182	20	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	182	21	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	182	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	183	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	183	2	purely	purely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	183	3	practical	practical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	183	4	reasons	reason	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	183	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	183	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	183	7	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	8	data	data	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	9	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	183	10	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	11	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	12	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	183	13	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	183	14	created	create	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	183	15	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	183	16	randomly	randomly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	183	17	sampling	sample	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	183	18	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	19	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	183	20	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	21	sentences	sentence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	183	22	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	183	23	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	183	24	total	total	NN
work_wgi74n6oavhdxhlomowlxqjdhu	183	25	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	183	26	500	500	CD
work_wgi74n6oavhdxhlomowlxqjdhu	183	27	K	K	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	183	28	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	183	29	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	184	1	Wiktionary	wiktionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	184	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	185	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	185	2	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	185	3	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	185	4	snapshot	snapshot	NN
work_wgi74n6oavhdxhlomowlxqjdhu	185	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	185	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	185	7	Wiktionary	wiktionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	185	8	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	185	9	definitions	definition	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	185	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	185	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	185	12	follow	follow	VB
work_wgi74n6oavhdxhlomowlxqjdhu	185	13	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	185	14	heuristics	heuristic	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	185	15	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	185	16	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	185	17	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	185	18	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	185	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	186	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	186	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	186	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	186	4	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	186	5	creating	create	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	186	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	186	7	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	186	8	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	186	9	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	186	10	mapping	map	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	186	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	186	12	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	186	13	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	186	14	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	186	15	universal	universal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	186	16	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	186	17	tags.7	tags.7	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	186	18	Features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	186	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	187	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	187	2	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	187	3	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	187	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	187	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	187	6	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	187	7	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	187	8	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	187	9	identity	identity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	187	10	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	187	11	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	187	12	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	187	13	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	187	14	pair	pair	NN
work_wgi74n6oavhdxhlomowlxqjdhu	187	15	transitions	transition	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	187	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	188	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	188	2	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	188	3	five	five	CD
work_wgi74n6oavhdxhlomowlxqjdhu	188	4	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	188	5	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	188	6	couple	couple	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	8	current	current	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	188	9	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	188	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	12	observed	observed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	188	13	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	14	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	188	15	analogous	analogous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	188	16	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	188	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	18	emission	emission	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	19	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	188	20	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	21	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	188	22	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	188	23	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	188	24	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	25	iden-	iden-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	188	26	tity	tity	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	188	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	188	28	suffixes	suffix	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	188	29	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	188	30	up	up	IN
work_wgi74n6oavhdxhlomowlxqjdhu	188	31	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	188	32	length	length	VB
work_wgi74n6oavhdxhlomowlxqjdhu	188	33	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	188	34	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	188	35	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	188	36	three	three	CD
work_wgi74n6oavhdxhlomowlxqjdhu	188	37	indicator	indicator	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	38	features	feature	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	188	39	that	that	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	40	fire	fire	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	41	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	188	42	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	43	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	44	starts	start	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	188	45	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	188	46	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	47	capital	capital	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	48	letter	letter	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	49	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	188	50	contains	contain	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	188	51	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	52	hyphen	hyphen	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	53	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	188	54	contains	contain	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	188	55	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	188	56	digit	digit	NN
work_wgi74n6oavhdxhlomowlxqjdhu	188	57	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	189	1	These	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	189	2	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	189	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	189	4	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	189	5	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	189	6	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	189	7	those	those	DT
work_wgi74n6oavhdxhlomowlxqjdhu	189	8	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	189	9	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	189	10	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	189	11	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	189	12	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	189	13	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	189	14	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	189	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	189	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	190	1	Finally	finally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	190	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	190	3	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	190	4	some	some	DT
work_wgi74n6oavhdxhlomowlxqjdhu	190	5	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	190	6	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	190	7	add	add	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	190	8	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	190	9	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	10	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	11	feature	feature	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	12	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	190	13	couples	couple	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	190	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	190	15	current	current	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	190	16	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	190	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	190	19	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	20	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	21	identity	identity	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	22	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	190	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	190	24	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	190	25	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	191	1	These	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	191	2	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	191	3	monolin-	monolin-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	191	4	gual	gual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	191	5	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	191	6	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	191	7	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	191	8	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	191	9	induced	induce	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	191	10	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	191	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	191	12	exchange	exchange	NN
work_wgi74n6oavhdxhlomowlxqjdhu	191	13	algorithm	algorithm	NN
work_wgi74n6oavhdxhlomowlxqjdhu	191	14	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	191	15	Uszkoreit	Uszkoreit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	191	16	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	191	17	Brants	Brants	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	191	18	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	191	19	2008	2008	CD
work_wgi74n6oavhdxhlomowlxqjdhu	191	20	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	191	21	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	192	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	192	2	set	set	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	192	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	192	4	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	192	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	192	6	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	192	7	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	192	8	256	256	CD
work_wgi74n6oavhdxhlomowlxqjdhu	192	9	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	192	10	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	192	11	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	192	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	192	13	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	192	14	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	192	15	has	have	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	192	16	previously	previously	RB
work_wgi74n6oavhdxhlomowlxqjdhu	192	17	been	be	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	192	18	shown	show	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	192	19	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	192	20	produce	produce	VB
work_wgi74n6oavhdxhlomowlxqjdhu	192	21	robust	robust	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	192	22	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	192	23	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	192	24	similar	similar	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	192	25	tasks	task	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	192	26	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	192	27	Turian	Turian	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	28	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	29	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	30	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	31	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	192	32	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	192	33	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	192	34	Täckström	Täckström	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	35	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	36	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	37	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	192	38	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	192	39	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	192	40	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	192	41	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	193	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	193	2	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	193	3	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	193	4	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	193	5	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	193	6	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	193	7	learned	learn	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	193	8	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	193	9	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	193	10	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	193	11	monolingual	monolingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	193	12	newswire	newswire	NN
work_wgi74n6oavhdxhlomowlxqjdhu	193	13	corpus	corpus	NN
work_wgi74n6oavhdxhlomowlxqjdhu	193	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	194	1	4.2	4.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	194	2	Models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	194	3	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	194	4	Type	Type	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	194	5	Constraints	Constraints	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	194	6	To	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	194	7	examine	examine	VB
work_wgi74n6oavhdxhlomowlxqjdhu	194	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	194	9	sole	sole	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	194	10	effect	effect	NN
work_wgi74n6oavhdxhlomowlxqjdhu	194	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	194	12	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	194	13	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	194	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	194	15	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	194	16	experiment	experiment	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	194	17	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	194	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	194	19	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	194	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	194	21	drawing	draw	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	194	22	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	194	23	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	194	24	three	three	CD
work_wgi74n6oavhdxhlomowlxqjdhu	194	25	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	194	26	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	194	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	195	1	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	195	2	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	195	3	compares	compare	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	195	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	195	5	per-	per-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	195	6	formance	formance	NN
work_wgi74n6oavhdxhlomowlxqjdhu	195	7	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	195	8	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	195	9	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	195	10	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	195	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	195	12	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	195	13	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	195	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	195	15	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	195	16	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	195	17	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	195	18	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	195	19	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	195	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	195	21	D&P	D&P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	195	22	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	195	23	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	195	24	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	195	25	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	195	26	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	195	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	196	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	196	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	196	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	196	4	LG&T	LG&T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	196	5	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	196	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	197	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	197	2	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	197	3	previous	previous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	197	4	work	work	NN
work_wgi74n6oavhdxhlomowlxqjdhu	197	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	197	6	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	197	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	197	8	done	do	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	197	9	exclusively	exclusively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	197	10	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	197	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	197	12	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	197	13	portion	portion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	197	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	197	15	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	197	16	treebank	treebank	NN
work_wgi74n6oavhdxhlomowlxqjdhu	197	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	197	18	stripped	strip	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	197	19	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	197	20	any	any	DT
work_wgi74n6oavhdxhlomowlxqjdhu	197	21	manual	manual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	197	22	linguistic	linguistic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	197	23	annotation	annotation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	197	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	198	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	198	2	first	first	RB
work_wgi74n6oavhdxhlomowlxqjdhu	198	3	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	198	4	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	198	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	198	6	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	198	7	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	198	8	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	198	9	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	198	10	generate	generate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	198	11	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	198	12	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	198	13	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	198	14	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	198	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	198	16	English	English	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	198	17	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	198	18	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	198	19	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	198	20	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	198	21	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	198	22	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	198	23	alignments	alignment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	198	24	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	198	25	aggregated	aggregate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	198	26	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	198	27	tag	tag	VB
work_wgi74n6oavhdxhlomowlxqjdhu	198	28	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	198	29	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	198	30	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	198	31	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	198	32	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	198	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	199	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	199	2	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	199	3	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	199	4	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	199	5	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	199	6	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	199	7	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	199	8	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	199	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	199	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	199	11	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	199	12	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	199	13	then	then	RB
work_wgi74n6oavhdxhlomowlxqjdhu	199	14	filtered	filter	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	199	15	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	199	16	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	199	17	threshold	threshold	NN
work_wgi74n6oavhdxhlomowlxqjdhu	199	18	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	199	19	0.2	0.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	199	20	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	199	21	remove	remove	VB
work_wgi74n6oavhdxhlomowlxqjdhu	199	22	noisy	noisy	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	199	23	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	199	24	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	199	25	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	199	26	create	create	VB
work_wgi74n6oavhdxhlomowlxqjdhu	199	27	7The	7the	CD
work_wgi74n6oavhdxhlomowlxqjdhu	199	28	definitions	definition	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	199	29	were	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	199	30	downloaded	download	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	199	31	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	199	32	August	August	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	199	33	31	31	CD
work_wgi74n6oavhdxhlomowlxqjdhu	199	34	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	199	35	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	199	36	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	199	37	http://toolserver.org/˜enwikt/definitions/.	http://toolserver.org/˜enwikt/definitions/.	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	200	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	200	2	snapshot	snapshot	NN
work_wgi74n6oavhdxhlomowlxqjdhu	200	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	200	4	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	200	5	recent	recent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	200	6	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	200	7	that	that	DT
work_wgi74n6oavhdxhlomowlxqjdhu	200	8	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	200	9	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	200	10	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	200	11	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	200	12	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	200	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	201	1	Prior	prior	RB
work_wgi74n6oavhdxhlomowlxqjdhu	201	2	work	work	VB
work_wgi74n6oavhdxhlomowlxqjdhu	201	3	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	201	4	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	201	5	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	201	6	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	201	7	Lang	Lang	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	201	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	202	1	D&P	D&P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	202	2	LG&T	LG&T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	202	3	YHMMproj	YHMMproj	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	202	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	203	1	YHMMwik	YHMMwik	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	203	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	204	1	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	2	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	3	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	204	4	C	c	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	5	bg	bg	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	6	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	7	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	8	84.2	84.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	9	68.1	68.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	10	87.2	87.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	11	87.9	87.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	12	cs	cs	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	13	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	14	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	15	75.4	75.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	16	70.2	70.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	17	75.4	75.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	18	79.2	79.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	19	da	da	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	20	83.2	83.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	21	83.3	83.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	22	87.7	87.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	23	82.0	82.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	24	78.4	78.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	25	89.5	89.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	26	de	de	IN
work_wgi74n6oavhdxhlomowlxqjdhu	204	27	82.8	82.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	28	85.8	85.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	29	86.6	86.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	30	85.1	85.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	31	80.0	80.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	32	88.3	88.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	33	el	el	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	34	82.5	82.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	35	79.2	79.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	36	83.3	83.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	37	83.8	83.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	38	86.0	86.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	39	83.2	83.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	40	es	es	XX
work_wgi74n6oavhdxhlomowlxqjdhu	204	41	84.2	84.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	42	86.4	86.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	43	83.9	83.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	44	83.7	83.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	45	88.3	88.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	46	87.3	87.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	47	fr	fr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	204	48	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	49	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	50	88.4	88.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	51	75.7	75.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	52	75.6	75.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	53	86.6	86.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	54	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	204	55	86.8	86.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	56	86.5	86.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	57	89.0	89.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	58	85.4	85.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	59	89.9	89.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	60	90.6	90.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	61	ja	ja	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	62	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	63	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	64	45.2	45.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	65	76.9	76.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	66	74.4	74.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	67	73.7	73.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	68	nl	nl	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	69	79.5	79.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	70	86.3	86.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	71	81.7	81.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	72	79.1	79.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	73	83.8	83.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	74	82.7	82.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	75	pt	pt	IN
work_wgi74n6oavhdxhlomowlxqjdhu	204	76	87.9	87.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	77	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	78	86.7	86.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	79	79.0	79.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	80	83.8	83.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	81	90.4	90.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	82	sl	sl	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	83	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	84	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	85	78.7	78.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	86	64.8	64.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	87	82.8	82.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	88	83.4	83.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	89	sv	sv	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	90	80.5	80.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	91	86.1	86.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	92	80.6	80.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	93	85.9	85.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	94	85.9	85.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	95	86.7	86.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	96	tr	tr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	204	97	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	98	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	99	66.2	66.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	100	44.1	44.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	101	65.1	65.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	102	65.7	65.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	103	zh	zh	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	104	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	105	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	106	59.2	59.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	107	73.9	73.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	108	63.2	63.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	109	73.0	73.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	110	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	111	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	204	112	8)	8)	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	113	83.4	83.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	114	84.8	84.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	115	84.9	84.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	116	83.0	83.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	117	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	118	87.3	87.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	119	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	120	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	121	–	–	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	122	78.5	78.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	123	75.9	75.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	124	80.0	80.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	125	83.2	83.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	126	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	127	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	204	128	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	204	129	Tagging	tag	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	204	130	accuracies	accuracy	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	204	131	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	204	132	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	204	133	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	204	134	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	204	135	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	204	136	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	204	137	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	205	1	D&P	D&P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	205	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	205	4	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	205	5	With	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	205	6	LP	LP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	7	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	205	8	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	205	9	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	205	10	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	205	11	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	205	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	205	13	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	14	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	205	15	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	16	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	205	17	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	205	18	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	205	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	205	20	while	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	205	21	LG&T	LG&T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	22	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	205	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	205	24	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	205	25	SHMM	SHMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	26	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	205	27	ME	ME	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	28	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	205	29	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	205	30	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	205	31	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	205	32	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	205	33	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	205	34	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	35	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	205	36	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	205	37	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	206	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	206	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	206	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	206	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	207	1	YHMMproj	YHMMproj	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	207	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	208	1	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	208	2	YHMMwik	YHMMwik	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	208	3	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	209	1	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	209	2	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	209	3	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	209	4	HMMs	hmm	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	209	5	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	209	6	solely	solely	RB
work_wgi74n6oavhdxhlomowlxqjdhu	209	7	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	209	8	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	209	9	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	209	10	derived	derive	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	209	11	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	209	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	209	13	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	209	14	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	209	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	209	16	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	209	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	209	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	209	19	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	209	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	209	21	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	209	22	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	209	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	209	24	respectively	respectively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	209	25	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	210	1	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	210	2	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	210	3	C	C	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	210	4	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	210	5	equivalent	equivalent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	210	6	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	210	7	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	210	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	210	9	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	210	10	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	210	11	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	210	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	211	1	All	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	211	2	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	211	3	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	211	4	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	211	5	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	211	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	211	7	treebank	treebank	NN
work_wgi74n6oavhdxhlomowlxqjdhu	211	8	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	211	9	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	211	10	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	211	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	211	12	stripped	strip	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	211	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	211	14	gold	gold	NN
work_wgi74n6oavhdxhlomowlxqjdhu	211	15	labels	label	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	211	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	212	1	Results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	212	2	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	212	3	averaged	average	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	212	4	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	212	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	212	6	8	8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	212	7	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	212	8	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	212	9	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	212	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	212	11	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	212	12	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	212	13	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	212	14	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	212	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	212	16	denoted	denote	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	212	17	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	212	18	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	212	19	8)	8)	CD
work_wgi74n6oavhdxhlomowlxqjdhu	212	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	212	21	as	as	RB
work_wgi74n6oavhdxhlomowlxqjdhu	212	22	well	well	RB
work_wgi74n6oavhdxhlomowlxqjdhu	212	23	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	212	24	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	212	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	212	26	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	212	27	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	212	28	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	212	29	15	15	CD
work_wgi74n6oavhdxhlomowlxqjdhu	212	30	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	212	31	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	212	32	denoted	denote	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	212	33	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	212	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	213	1	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	213	2	unweighted	unweighted	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	213	3	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	213	4	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	213	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	214	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	214	2	call	call	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	214	3	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	214	4	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	214	5	YHMMproj	yhmmproj	NN
work_wgi74n6oavhdxhlomowlxqjdhu	214	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	215	1	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	215	2	its	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	215	3	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	215	4	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	215	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	215	6	84.9	84.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	215	7	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	215	8	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	215	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	215	10	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	215	11	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	215	12	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	215	13	higher	high	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	215	14	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	215	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	215	16	83.4	83.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	215	17	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	215	18	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	215	19	D&P	D&P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	215	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	215	21	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	215	22	par	par	NN
work_wgi74n6oavhdxhlomowlxqjdhu	215	23	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	215	24	LG&T	LG&T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	215	25	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	215	26	84.8%).8	84.8%).8	VB
work_wgi74n6oavhdxhlomowlxqjdhu	215	27	Our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	215	28	next	next	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	215	29	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	215	30	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	215	31	YHMMwik	YHMMwik	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	215	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	215	33	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	216	1	simply	simply	RB
work_wgi74n6oavhdxhlomowlxqjdhu	216	2	draws	draw	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	216	3	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	216	4	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	216	5	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	216	6	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	216	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	217	1	It	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	217	2	slightly	slightly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	217	3	underperforms	underperform	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	217	4	LG&T	LG&T	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	217	5	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	217	6	83.0	83.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	217	7	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	217	8	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	217	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	217	10	presumably	presumably	RB
work_wgi74n6oavhdxhlomowlxqjdhu	217	11	because	because	IN
work_wgi74n6oavhdxhlomowlxqjdhu	217	12	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	217	13	used	use	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	217	14	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	217	15	second	second	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	217	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	217	17	order	order	NN
work_wgi74n6oavhdxhlomowlxqjdhu	217	18	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	217	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	218	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	218	2	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	218	3	simple	simple	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	218	4	extension	extension	NN
work_wgi74n6oavhdxhlomowlxqjdhu	218	5	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	218	6	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	218	7	two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	218	8	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	218	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	218	10	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	218	11	take	take	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	218	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	218	13	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	218	14	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	218	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	218	16	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	218	17	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	218	18	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	218	19	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	218	20	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	218	21	con-	con-	VB
work_wgi74n6oavhdxhlomowlxqjdhu	218	22	strain	strain	VB
work_wgi74n6oavhdxhlomowlxqjdhu	218	23	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	218	24	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	218	25	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	218	26	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	218	27	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	218	28	name	name	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	218	29	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	218	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	219	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	219	2	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	219	3	performs	perform	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	219	4	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	219	5	little	little	RB
work_wgi74n6oavhdxhlomowlxqjdhu	219	6	worse	bad	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	219	7	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	219	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	219	9	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	219	10	Indo	indo	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	219	11	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	219	12	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	219	13	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	219	14	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	219	15	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	219	16	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	219	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	219	18	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	219	19	gives	give	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	219	20	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	219	21	improvement	improvement	NN
work_wgi74n6oavhdxhlomowlxqjdhu	219	22	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	219	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	219	24	projected	projected	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	219	25	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	219	26	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	219	27	evaluated	evaluate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	219	28	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	219	29	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	219	30	15	15	CD
work_wgi74n6oavhdxhlomowlxqjdhu	219	31	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	219	32	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	219	33	80.0	80.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	219	34	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	219	35	vs.	vs.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	219	36	78.5	78.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	219	37	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	219	38	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	219	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	220	1	8Our	8our	CD
work_wgi74n6oavhdxhlomowlxqjdhu	220	2	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	220	3	corresponds	correspond	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	220	4	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	220	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	220	6	weaker	weak	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	220	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	220	8	“	"	``
work_wgi74n6oavhdxhlomowlxqjdhu	220	9	No	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	220	10	LP	LP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	220	11	”	"	''
work_wgi74n6oavhdxhlomowlxqjdhu	220	12	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	220	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	220	14	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	220	15	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	220	16	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	220	17	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	220	18	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	220	19	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	220	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	221	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	221	2	found	find	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	221	3	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	221	4	label	label	NN
work_wgi74n6oavhdxhlomowlxqjdhu	221	5	propagation	propagation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	221	6	was	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	221	7	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	221	8	beneficial	beneficial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	221	9	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	221	10	small	small	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	221	11	amounts	amount	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	221	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	221	13	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	221	14	were	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	221	15	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	221	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	222	1	7	7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	222	2	Token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	222	3	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	222	4	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	222	5	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	222	6	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	222	7	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	222	8	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	222	9	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	222	10	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	222	11	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	222	12	Lang	Lang	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	222	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	223	1	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	223	2	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	223	3	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	223	4	ỹHMM+C+L	ỹHMM+C+L	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	223	5	ỹCRF+C+L	ỹcrf+c+l	CD
work_wgi74n6oavhdxhlomowlxqjdhu	223	6	ŶHMMproj	ŷhmmproj	NN
work_wgi74n6oavhdxhlomowlxqjdhu	223	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	224	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	224	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	224	3	ŶHMMwik	ŶHMMwik	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	224	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	225	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	225	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	225	3	ŶHMMunion	ŷhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	225	4	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	225	5	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	225	6	ŶCRFproj	ŶCRFproj	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	225	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	226	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	226	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	226	3	ŶCRFwik	ŷcrfwik	NN
work_wgi74n6oavhdxhlomowlxqjdhu	226	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	227	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	3	ŶCRFunion+C+L	ŶCRFunion+C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	4	bg	bg	NN
work_wgi74n6oavhdxhlomowlxqjdhu	227	5	87.7	87.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	6	77.9	77.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	7	84.1	84.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	8	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	9	83.9	83.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	10	86.7	86.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	11	86.0	86.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	12	87.8	87.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	13	85.4	85.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	14	cs	cs	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	15	78.3	78.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	16	65.4	65.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	17	74.9	74.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	18	74.8	74.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	19	81.1	81.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	20	76.9	76.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	21	74.7	74.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	22	80.3	80.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	23	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	24	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	25	75.0	75.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	26	da	da	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	27	87.3	87.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	28	80.9	80.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	29	85.1	85.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	30	87.2	87.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	31	85.6	85.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	32	88.1	88.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	33	85.5	85.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	34	88.2	88.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	35	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	36	86.0	86.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	37	de	de	IN
work_wgi74n6oavhdxhlomowlxqjdhu	227	38	87.7	87.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	39	81.4	81.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	40	83.3	83.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	41	85.0	85.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	42	89.3	89.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	43	86.7	86.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	44	84.4	84.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	45	90.5	90.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	46	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	47	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	48	85.5	85.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	49	el	el	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	50	85.9	85.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	51	81.1	81.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	52	77.8	77.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	53	80.1	80.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	54	87.0	87.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	55	83.9	83.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	56	79.6	79.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	57	89.5	89.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	58	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	59	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	60	79.7	79.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	61	es	es	XX
work_wgi74n6oavhdxhlomowlxqjdhu	227	62	89.1	89.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	63	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	64	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	65	84.1	84.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	66	85.5	85.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	67	83.7	83.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	68	85.9	85.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	69	88.0	88.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	70	85.7	85.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	71	87.1	87.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	72	86.0	86.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	73	fr	fr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	227	74	88.4	88.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	75	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	76	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	77	83.5	83.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	78	84.7	84.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	79	85.9	85.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	80	86.4	86.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	81	87.4	87.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	82	84.9	84.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	83	87.2	87.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	84	85.6	85.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	85	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	227	86	89.6	89.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	87	85.2	85.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	88	88.5	88.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	89	88.7	88.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	90	87.6	87.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	91	89.8	89.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	92	88.3	88.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	93	89.3	89.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	94	89.4	89.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	95	ja	ja	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	96	72.8	72.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	97	47.6	47.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	98	54.2	54.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	99	43.2	43.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	100	76.1	76.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	101	70.5	70.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	102	44.9	44.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	103	81.0	81.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	104	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	105	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	106	68.0	68.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	107	nl	nl	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	108	83.1	83.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	109	78.4	78.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	110	82.4	82.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	111	82.3	82.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	112	84.2	84.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	113	83.2	83.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	114	83.1	83.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	115	85.9	85.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	116	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	117	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	118	83.2	83.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	119	pt	pt	IN
work_wgi74n6oavhdxhlomowlxqjdhu	227	120	89.1	89.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	121	84.7	84.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	122	87.0	87.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	123	86.6	86.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	124	88.7	88.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	125	88.0	88.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	126	87.9	87.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	127	91.0	91.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	128	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	129	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	130	88.3	88.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	131	sl	sl	NN
work_wgi74n6oavhdxhlomowlxqjdhu	227	132	82.4	82.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	133	69.8	69.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	134	78.2	78.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	135	78.5	78.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	136	81.8	81.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	137	80.1	80.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	138	79.7	79.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	139	82.3	82.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	140	80.0	80.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	141	sv	sv	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	142	86.1	86.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	143	80.1	80.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	144	84.2	84.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	145	82.3	82.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	146	87.9	87.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	147	86.9	86.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	148	84.4	84.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	149	88.9	88.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	150	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	151	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	152	85.5	85.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	153	tr	tr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	227	154	62.4	62.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	155	58.1	58.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	156	64.5	64.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	157	64.6	64.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	158	61.8	61.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	159	64.8	64.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	160	65.0	65.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	161	64.1	64.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	162	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	163	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	164	65.2	65.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	165	zh	zh	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	166	72.6	72.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	167	52.7	52.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	168	39.5	39.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	169	56.0	56.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	170	74.1	74.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	171	73.3	73.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	172	59.7	59.7	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	173	74.4	74.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	174	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	175	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	227	176	73.4	73.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	177	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	178	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	227	179	8)	8)	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	180	87.2	87.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	181	82.0	82.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	182	84.2	84.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	183	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	184	87.0	87.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	185	86.8	86.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	186	84.9	84.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	187	88.8	88.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	188	85.4	85.4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	189	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	227	190	82.8	82.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	191	74.1	74.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	192	76.9	76.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	193	77.6	77.6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	194	82.8	82.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	195	82.3	82.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	196	78.2	78.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	197	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	198	81.1	81.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	199	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	227	200	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	227	201	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	227	202	Tagging	tag	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	227	203	accuracies	accuracy	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	227	204	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	227	205	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	227	206	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	227	207	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	227	208	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	227	209	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	227	210	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	227	211	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	227	212	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	227	213	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	227	214	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	227	215	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	228	1	All	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	228	2	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	228	3	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	228	4	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	228	5	features	feature	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	228	6	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	228	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	229	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	230	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	231	1	+	+	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	231	2	C	c	NN
work_wgi74n6oavhdxhlomowlxqjdhu	231	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	231	4	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	231	5	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	231	6	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	231	7	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	231	8	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	231	9	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	231	10	sets	set	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	231	11	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	231	12	containing	contain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	231	13	500k	500k	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	231	14	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	231	15	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	231	16	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	231	17	partial	partial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	231	18	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	231	19	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	231	20	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	231	21	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	231	22	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	231	23	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	231	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	232	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	233	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	234	1	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	234	2	L	L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	234	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	234	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	235	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	235	2	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	235	3	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	235	4	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	235	5	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	235	6	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	235	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	235	8	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	235	9	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	235	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	235	11	larger	large	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	235	12	datasets	dataset	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	235	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	235	14	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	235	15	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	235	16	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	235	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	235	18	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	235	19	included	include	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	235	20	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	235	21	comparison	comparison	NN
work_wgi74n6oavhdxhlomowlxqjdhu	235	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	236	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	236	2	remaining	remain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	236	3	columns	column	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	236	4	correspond	correspond	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	236	5	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	236	6	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	236	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	236	8	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	236	9	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	236	10	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	236	11	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	236	12	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	236	13	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	236	14	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	236	15	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	236	16	ỹ	ỹ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	236	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	237	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	238	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	238	2	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	239	1	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	239	2	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	239	3	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	239	4	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	239	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	239	6	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	239	7	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	239	8	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	239	9	Ŷ	ŷ	NN
work_wgi74n6oavhdxhlomowlxqjdhu	239	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	240	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	241	1	.	.	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	241	2	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	241	3	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	242	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	242	2	latter	latter	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	242	3	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	242	4	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	242	5	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	242	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	242	7	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	242	8	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	242	9	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	242	10	·	·	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	242	11	proj	proj	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	242	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	243	1	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	243	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	243	3	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	243	4	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	243	5	·	·	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	243	6	wik	wik	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	243	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	243	8	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	244	1	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	244	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	244	3	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	244	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	244	5	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	244	6	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	244	7	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	244	8	·	·	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	244	9	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	244	10	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	244	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	244	12	respectively	respectively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	244	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	245	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	245	2	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	245	3	spaces	space	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	245	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	245	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	245	6	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	245	7	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	245	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	245	9	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	245	10	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	245	11	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	245	12	Ŷ	ŷ	NN
work_wgi74n6oavhdxhlomowlxqjdhu	245	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	246	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	247	1	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	247	2	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	248	1	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	248	2	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	248	3	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	248	4	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	248	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	248	6	respective	respective	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	248	7	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	248	8	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	248	9	used	use	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	248	10	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	248	11	derive	derive	VB
work_wgi74n6oavhdxhlomowlxqjdhu	248	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	248	13	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	248	14	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	248	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	249	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	249	2	observed	observed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	249	3	difference	difference	NN
work_wgi74n6oavhdxhlomowlxqjdhu	249	4	between	between	IN
work_wgi74n6oavhdxhlomowlxqjdhu	249	5	ŶCRFwik	ŷcrfwik	NN
work_wgi74n6oavhdxhlomowlxqjdhu	249	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	250	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	250	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	250	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	250	4	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	250	5	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	250	6	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	250	7	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	250	8	statistically	statistically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	250	9	significant	significant	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	250	10	at	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	250	11	p	p	NN
work_wgi74n6oavhdxhlomowlxqjdhu	250	12	<	<	XX
work_wgi74n6oavhdxhlomowlxqjdhu	250	13	0.01	0.01	CD
work_wgi74n6oavhdxhlomowlxqjdhu	250	14	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	250	15	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	250	16	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	250	17	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	250	18	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	250	19	p	p	NN
work_wgi74n6oavhdxhlomowlxqjdhu	250	20	<	<	XX
work_wgi74n6oavhdxhlomowlxqjdhu	250	21	0.015	0.015	CD
work_wgi74n6oavhdxhlomowlxqjdhu	250	22	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	250	23	*	*	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	250	24	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	250	25	according	accord	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	250	26	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	250	27	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	250	28	paired	pair	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	250	29	bootstrap	bootstrap	NN
work_wgi74n6oavhdxhlomowlxqjdhu	250	30	test	test	NN
work_wgi74n6oavhdxhlomowlxqjdhu	250	31	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	250	32	Efron	Efron	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	250	33	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	250	34	Tibshirani	Tibshirani	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	250	35	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	250	36	1993	1993	CD
work_wgi74n6oavhdxhlomowlxqjdhu	250	37	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	250	38	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	251	1	Significance	significance	NN
work_wgi74n6oavhdxhlomowlxqjdhu	251	2	was	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	251	3	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	251	4	assessed	assess	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	251	5	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	251	6	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	251	7	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	251	8	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	251	9	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	251	10	8)	8)	CD
work_wgi74n6oavhdxhlomowlxqjdhu	251	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	252	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	252	2	next	next	RB
work_wgi74n6oavhdxhlomowlxqjdhu	252	3	add	add	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	252	4	monolingual	monolingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	252	5	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	252	6	features	feature	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	252	7	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	252	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	252	9	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	252	10	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	252	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	252	12	union	union	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	252	13	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	252	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	253	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	2	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	253	4	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	5	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	253	6	C	C	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	253	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	253	8	significantly	significantly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	253	9	outperforms	outperform	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	253	10	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	11	other	other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	253	12	type-	type-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	13	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	253	14	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	253	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	253	16	demonstrating	demonstrate	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	253	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	18	utility	utility	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	19	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	20	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	21	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	253	22	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	23	features.9	features.9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	253	24	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	25	further	further	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	253	26	exploration	exploration	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	253	28	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	253	29	train	train	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	253	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	31	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	253	32	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	33	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	34	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	35	datasets	dataset	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	253	36	containing	contain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	253	37	500	500	CD
work_wgi74n6oavhdxhlomowlxqjdhu	253	38	K	K	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	253	39	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	253	40	sampled	sample	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	253	41	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	42	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	43	target	target	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	44	side	side	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	45	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	46	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	47	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	253	48	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	253	49	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	253	50	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	51	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	253	52	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	253	53	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	253	54	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	253	55	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	56	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	253	57	done	do	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	253	58	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	253	59	explore	explore	VB
work_wgi74n6oavhdxhlomowlxqjdhu	253	60	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	253	61	effects	effect	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	253	62	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	63	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	253	64	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	253	65	during	during	IN
work_wgi74n6oavhdxhlomowlxqjdhu	253	66	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	253	67	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	254	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	254	2	find	find	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	254	3	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	4	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	5	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	6	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	254	7	datasets	dataset	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	254	8	result	result	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	254	9	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	10	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	254	11	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	254	12	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	14	87.2	87.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	254	15	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	16	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	254	17	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	254	18	comparable	comparable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	254	19	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	254	21	87.3	87.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	254	22	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	23	reported	report	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	254	24	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	25	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	26	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	254	27	C	c	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	28	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	254	29	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	254	30	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	254	31	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	255	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	255	2	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	255	3	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	255	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	255	5	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	255	6	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	255	7	domain	domain	NN
work_wgi74n6oavhdxhlomowlxqjdhu	255	8	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	255	9	amount	amount	NN
work_wgi74n6oavhdxhlomowlxqjdhu	255	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	255	11	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	255	12	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	255	13	does	do	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	255	14	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	255	15	influence	influence	VB
work_wgi74n6oavhdxhlomowlxqjdhu	255	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	255	17	performance	performance	NN
work_wgi74n6oavhdxhlomowlxqjdhu	255	18	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	255	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	255	20	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	255	21	significantly	significantly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	255	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	256	1	Finally	finally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	256	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	256	3	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	256	4	train	train	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	256	5	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	256	6	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	256	7	where	where	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	256	8	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	256	9	treat	treat	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	256	10	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	256	11	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	256	12	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	256	13	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	256	14	partially	partially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	256	15	observed	observe	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	256	16	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	256	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	256	18	use	use	VB
work_wgi74n6oavhdxhlomowlxqjdhu	256	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	256	20	full	full	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	256	21	unpruned	unpruned	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	256	22	lattice	lattice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	256	23	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	256	24	computing	compute	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	256	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	256	26	partition	partition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	256	27	func-	func-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	256	28	9These	9These	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	256	29	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	256	30	monolingual	monolingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	256	31	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	256	32	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	257	1	Bilingual	bilingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	257	2	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	257	3	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	257	4	intro-	intro-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	257	5	duced	duce	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	257	6	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	257	7	Täckström	Täckström	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	257	8	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	257	9	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	257	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	258	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	258	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	258	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	258	4	might	may	MD
work_wgi74n6oavhdxhlomowlxqjdhu	258	5	bring	bring	VB
work_wgi74n6oavhdxhlomowlxqjdhu	258	6	additional	additional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	258	7	benefits	benefit	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	258	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	259	1	tion	tion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	259	2	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	259	3	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	259	4	3.2	3.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	259	5	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	259	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	260	1	Due	due	IN
work_wgi74n6oavhdxhlomowlxqjdhu	260	2	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	260	3	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	260	4	considerations	consideration	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	260	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	260	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	260	7	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	260	8	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	260	9	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	260	10	experiments	experiment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	260	11	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	260	12	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	260	13	shown	show	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	260	14	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	260	15	table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	260	16	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	260	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	261	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	261	2	ob-	ob-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	261	3	serve	serve	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	261	4	similar	similar	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	261	5	trends	trend	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	261	6	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	7	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	261	8	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	261	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	261	10	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	261	11	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	12	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	261	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	261	14	accuracies	accuracy	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	261	15	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	261	16	much	much	RB
work_wgi74n6oavhdxhlomowlxqjdhu	261	17	lower	low	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	261	18	compared	compare	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	261	19	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	261	21	type-	type-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	261	22	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	261	23	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	261	24	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	261	25	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	261	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	261	27	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	261	28	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	261	29	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	261	31	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	261	32	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	261	33	along	along	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	34	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	35	cluster	cluster	NN
work_wgi74n6oavhdxhlomowlxqjdhu	261	36	features	feature	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	261	37	achieves	achieve	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	261	38	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	261	39	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	261	40	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	261	41	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	42	79.3	79.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	261	43	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	261	44	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	261	45	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	261	46	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	261	47	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	261	48	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	261	49	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	262	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	262	2	result	result	NN
work_wgi74n6oavhdxhlomowlxqjdhu	262	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	262	4	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	262	5	unsurprising	unsurprise	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	262	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	263	1	First	first	RB
work_wgi74n6oavhdxhlomowlxqjdhu	263	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	263	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	263	4	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	263	5	’s	’s	NN
work_wgi74n6oavhdxhlomowlxqjdhu	263	6	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	263	7	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	263	8	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	263	9	fully	fully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	263	10	unconstrained	unconstrained	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	263	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	264	1	Second	second	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	264	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	264	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	264	4	dic-	dic-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	264	5	tionary	tionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	264	6	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	264	7	provides	provide	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	264	8	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	264	9	weak	weak	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	264	10	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	264	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	264	12	observation	observation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	264	13	con-	con-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	264	14	straints	straint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	264	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	264	16	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	264	17	do	do	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	264	18	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	264	19	provide	provide	VB
work_wgi74n6oavhdxhlomowlxqjdhu	264	20	sufficient	sufficient	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	264	21	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	264	22	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	264	23	successfully	successfully	RB
work_wgi74n6oavhdxhlomowlxqjdhu	264	24	train	train	VB
work_wgi74n6oavhdxhlomowlxqjdhu	264	25	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	264	26	discriminative	discriminative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	264	27	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	264	28	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	265	1	How-	How-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	265	2	ever	ever	RB
work_wgi74n6oavhdxhlomowlxqjdhu	265	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	265	4	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	265	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	265	6	will	will	MD
work_wgi74n6oavhdxhlomowlxqjdhu	265	7	observe	observe	VB
work_wgi74n6oavhdxhlomowlxqjdhu	265	8	next	next	RB
work_wgi74n6oavhdxhlomowlxqjdhu	265	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	265	10	coupling	couple	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	265	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	265	12	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	265	13	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	265	14	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	265	15	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	265	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	265	17	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	265	18	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	265	19	solves	solve	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	265	20	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	265	21	problem	problem	NN
work_wgi74n6oavhdxhlomowlxqjdhu	265	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	266	1	4.3	4.3	LS
work_wgi74n6oavhdxhlomowlxqjdhu	266	2	Models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	266	3	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	266	4	Token	Token	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	266	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	266	6	Type	Type	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	266	7	Constraints	Constraints	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	266	8	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	266	9	now	now	RB
work_wgi74n6oavhdxhlomowlxqjdhu	266	10	proceed	proceed	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	266	11	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	266	12	add	add	VB
work_wgi74n6oavhdxhlomowlxqjdhu	266	13	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	266	14	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	266	15	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	266	16	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	266	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	266	18	focusing	focus	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	266	19	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	266	20	particular	particular	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	266	21	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	266	22	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	266	23	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	266	24	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	266	25	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	266	26	8	8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	266	27	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	266	28	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	267	1	Since	since	IN
work_wgi74n6oavhdxhlomowlxqjdhu	267	2	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	267	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	267	4	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	267	5	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	267	6	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	267	7	generate	generate	VB
work_wgi74n6oavhdxhlomowlxqjdhu	267	8	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	267	9	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	267	10	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	267	11	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	267	12	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	267	13	monolingual	monolingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	267	14	treebanks	treebank	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	267	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	267	16	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	267	17	train	train	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	267	18	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	267	19	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	267	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	267	21	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	267	22	subsection	subsection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	267	23	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	267	24	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	267	25	500K	500k	CD
work_wgi74n6oavhdxhlomowlxqjdhu	267	26	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	267	27	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	267	28	datasets	dataset	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	267	29	sampled	sample	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	267	30	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	267	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	267	32	bi-	bi-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	267	33	text	text	NN
work_wgi74n6oavhdxhlomowlxqjdhu	267	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	268	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	268	2	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	268	3	baseline	baseline	NN
work_wgi74n6oavhdxhlomowlxqjdhu	268	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	268	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	268	6	first	first	RB
work_wgi74n6oavhdxhlomowlxqjdhu	268	7	train	train	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	268	8	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	268	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	268	10	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	268	11	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	268	12	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	268	13	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	268	14	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	268	15	projected	projected	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	268	16	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	268	17	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	268	18	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	268	19	ỹHMM+C+L	ỹHMM+C+L	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	268	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	268	21	ỹCRF+C+L	ỹcrf+c+l	CD
work_wgi74n6oavhdxhlomowlxqjdhu	268	22	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	268	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	269	1	As	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	269	2	shown	show	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	269	3	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	269	4	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	269	5	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	269	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	269	7	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	269	8	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	269	9	underperform	underperform	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	269	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	269	11	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	269	12	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	269	13	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	269	14	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	269	15	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	269	16	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	269	17	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	269	18	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	269	19	C+L),10	C+L),10	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	269	20	which	which	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	269	21	confirms	confirm	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	269	22	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	269	23	projected	project	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	269	24	to-	to-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	269	25	ken	ken	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	269	26	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	269	27	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	269	28	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	269	29	reliable	reliable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	269	30	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	269	31	their	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	269	32	own	own	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	269	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	270	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	270	2	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	270	3	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	270	4	line	line	NN
work_wgi74n6oavhdxhlomowlxqjdhu	270	5	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	270	6	similar	similar	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	270	7	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	270	8	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	270	9	previously	previously	RB
work_wgi74n6oavhdxhlomowlxqjdhu	270	10	examined	examine	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	270	11	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	270	12	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	270	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	270	14	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	270	15	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	270	16	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	270	17	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	270	18	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	271	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	271	2	then	then	RB
work_wgi74n6oavhdxhlomowlxqjdhu	271	3	study	study	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	271	4	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	271	5	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	271	6	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	271	7	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	271	8	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	271	9	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	271	10	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	271	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	272	1	These	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	272	2	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	272	3	use	use	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	272	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	272	5	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	272	6	three	three	CD
work_wgi74n6oavhdxhlomowlxqjdhu	272	7	dictio-	dictio-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	8	naries	narie	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	272	9	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	272	10	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	272	11	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	272	12	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	272	13	4.2	4.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	272	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	272	15	but	but	CC
work_wgi74n6oavhdxhlomowlxqjdhu	272	16	additionally	additionally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	272	17	couple	couple	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	272	19	derived	derive	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	272	20	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	21	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	272	22	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	272	23	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	272	24	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	272	25	con-	con-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	26	straints	straint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	272	27	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	272	28	see	see	VB
work_wgi74n6oavhdxhlomowlxqjdhu	272	29	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	272	30	caption	caption	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	31	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	272	32	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	33	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	272	34	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	272	35	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	272	36	list	list	NN
work_wgi74n6oavhdxhlomowlxqjdhu	272	37	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	272	38	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	272	39	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	272	40	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	273	1	Note	note	VB
work_wgi74n6oavhdxhlomowlxqjdhu	273	2	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	273	3	since	since	IN
work_wgi74n6oavhdxhlomowlxqjdhu	273	4	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	273	5	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	273	6	allow	allow	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	273	7	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	273	8	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	273	9	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	273	10	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	273	11	licensed	license	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	273	12	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	273	13	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	273	14	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	273	15	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	273	16	Step	step	NN
work_wgi74n6oavhdxhlomowlxqjdhu	273	17	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	273	18	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	273	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	273	20	trans-	trans-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	273	21	fer	fer	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	273	22	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	273	23	§	§	ADD
work_wgi74n6oavhdxhlomowlxqjdhu	273	24	2.3	2.3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	273	25	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	273	26	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	273	27	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	273	28	actual	actual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	273	29	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	273	30	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	273	31	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	273	32	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	273	33	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	273	34	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	273	35	vary	vary	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	273	36	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	273	37	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	273	38	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	273	39	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	273	40	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	274	1	From	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	274	2	Table	table	NN
work_wgi74n6oavhdxhlomowlxqjdhu	274	3	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	274	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	274	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	274	6	see	see	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	274	7	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	274	8	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	274	9	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	274	10	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	274	11	superior	superior	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	274	12	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	274	13	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	274	14	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	274	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	274	16	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	274	17	used	use	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	274	18	both	both	CC
work_wgi74n6oavhdxhlomowlxqjdhu	274	19	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	274	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	274	21	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	274	22	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	274	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	274	24	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	274	25	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	275	1	However	however	RB
work_wgi74n6oavhdxhlomowlxqjdhu	275	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	275	3	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	275	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	275	5	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	275	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	275	7	cou-	cou-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	275	8	pled	plead	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	275	9	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	275	10	do	do	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	275	11	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	275	12	provide	provide	VB
work_wgi74n6oavhdxhlomowlxqjdhu	275	13	any	any	DT
work_wgi74n6oavhdxhlomowlxqjdhu	275	14	benefit	benefit	NN
work_wgi74n6oavhdxhlomowlxqjdhu	275	15	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	275	16	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	275	17	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	275	18	alone	alone	RB
work_wgi74n6oavhdxhlomowlxqjdhu	275	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	275	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	275	21	particular	particular	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	275	22	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	275	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	275	24	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	275	25	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	275	26	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	275	27	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	275	28	union	union	NN
work_wgi74n6oavhdxhlomowlxqjdhu	275	29	dictionary	dictionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	275	30	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	275	31	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	275	32	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	275	33	derive	derive	VB
work_wgi74n6oavhdxhlomowlxqjdhu	275	34	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	275	35	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	275	36	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	275	37	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	275	38	ŶHMMproj	ŶHMMproj	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	275	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	276	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	276	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	276	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	276	4	ŶHMMunion	ŷhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	276	5	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	276	6	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	276	7	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	276	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	277	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	277	2	hypothesize	hypothesize	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	277	3	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	4	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	5	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	277	6	because	because	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	7	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	8	dictionar-	dictionar-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	277	9	ies	ies	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	277	10	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	277	11	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	12	particular	particular	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	277	13	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	14	former	former	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	277	15	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	277	16	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	277	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	18	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	277	19	bias	bias	NN
work_wgi74n6oavhdxhlomowlxqjdhu	277	20	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	22	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	277	23	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	277	24	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	277	25	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	277	26	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	277	27	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	277	28	so	so	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	29	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	31	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	277	32	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	277	33	unable	unable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	277	34	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	277	35	correct	correct	VB
work_wgi74n6oavhdxhlomowlxqjdhu	277	36	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	37	systematic	systematic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	277	38	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	277	39	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	277	40	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	277	41	pro-	pro-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	277	42	jections	jection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	277	43	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	277	44	see	see	VB
work_wgi74n6oavhdxhlomowlxqjdhu	277	45	§	§	CD
work_wgi74n6oavhdxhlomowlxqjdhu	277	46	2.1	2.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	277	47	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	277	48	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	278	1	Since	since	IN
work_wgi74n6oavhdxhlomowlxqjdhu	278	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	278	3	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	278	4	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	278	5	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	278	6	stronger	strong	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	278	7	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	278	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	278	9	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	278	10	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	278	11	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	278	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	278	13	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	278	14	mod-	mod-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	278	15	els	els	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	278	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	278	17	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	278	18	bias	bias	NN
work_wgi74n6oavhdxhlomowlxqjdhu	278	19	may	may	MD
work_wgi74n6oavhdxhlomowlxqjdhu	278	20	have	have	VB
work_wgi74n6oavhdxhlomowlxqjdhu	278	21	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	278	22	substantial	substantial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	278	23	impact	impact	NN
work_wgi74n6oavhdxhlomowlxqjdhu	278	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	279	1	With	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	279	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	279	3	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	279	4	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	279	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	279	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	279	7	difference	difference	NN
work_wgi74n6oavhdxhlomowlxqjdhu	279	8	between	between	IN
work_wgi74n6oavhdxhlomowlxqjdhu	279	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	279	10	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	279	11	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	279	12	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	279	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	279	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	279	15	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	279	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	279	17	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	279	18	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	279	19	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	279	20	negligible	negligible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	279	21	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	279	22	YHMMunion	yhmmunion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	279	23	+	+	CC
work_wgi74n6oavhdxhlomowlxqjdhu	279	24	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	279	25	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	279	26	ŶHMMwik	ŶHMMwik	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	279	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	280	1	+	+	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	280	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	280	3	both	both	DT
work_wgi74n6oavhdxhlomowlxqjdhu	280	4	av-	av-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	280	5	erage	erage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	280	6	at	at	IN
work_wgi74n6oavhdxhlomowlxqjdhu	280	7	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	280	8	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	280	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	280	10	82.8	82.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	280	11	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	280	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	281	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	281	2	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	281	3	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	281	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	281	5	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	281	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	281	7	other	other	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	281	8	hand	hand	NN
work_wgi74n6oavhdxhlomowlxqjdhu	281	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	281	10	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	281	11	able	able	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	281	12	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	281	13	take	take	VB
work_wgi74n6oavhdxhlomowlxqjdhu	281	14	advantage	advantage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	281	15	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	281	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	281	17	complementary	complementary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	281	18	information	information	NN
work_wgi74n6oavhdxhlomowlxqjdhu	281	19	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	281	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	281	21	coupled	couple	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	281	22	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	281	23	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	281	24	provided	provide	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	281	25	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	281	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	281	27	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	281	28	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	281	29	able	able	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	281	30	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	281	31	filter	filter	VB
work_wgi74n6oavhdxhlomowlxqjdhu	281	32	out	out	RP
work_wgi74n6oavhdxhlomowlxqjdhu	281	33	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	281	34	systematic	systematic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	281	35	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	281	36	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	281	37	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	281	38	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	281	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	282	1	With	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	282	2	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	282	3	dictionary	dictionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	282	4	derived	derive	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	282	5	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	282	6	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	282	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	282	8	pro-	pro-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	282	9	jected	jecte	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	282	10	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	282	11	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	282	12	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	282	13	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	282	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	282	15	ŶCRFwik	ŶCRFwik	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	282	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	283	1	+	+	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	283	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	283	3	performs	perform	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	283	4	10To	10To	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	283	5	make	make	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	283	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	283	7	comparison	comparison	NN
work_wgi74n6oavhdxhlomowlxqjdhu	283	8	fair	fair	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	283	9	vis	vis	FW
work_wgi74n6oavhdxhlomowlxqjdhu	283	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	283	11	a	a	FW
work_wgi74n6oavhdxhlomowlxqjdhu	283	12	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	283	13	vis	vis	FW
work_wgi74n6oavhdxhlomowlxqjdhu	283	14	potential	potential	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	283	15	divergences	divergence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	283	16	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	283	17	training	train	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	283	18	domains	domain	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	283	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	283	20	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	283	21	compare	compare	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	283	22	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	283	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	283	24	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	283	25	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	283	26	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	283	27	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	283	28	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	283	29	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	283	30	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	283	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	283	32	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	283	33	500	500	CD
work_wgi74n6oavhdxhlomowlxqjdhu	283	34	K	K	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	283	35	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	283	36	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	283	37	sets	set	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	283	38	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	284	1	0	0	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	284	2	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	3	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	4	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	5	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	6	25	25	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	7	50	50	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	8	75	75	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	9	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	10	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	11	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	12	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	13	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	14	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	15	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	16	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	17	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	18	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	19	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	20	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	21	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	22	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	23	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	24	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	25	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	26	Number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	27	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	28	token−level	token−level	.
work_wgi74n6oavhdxhlomowlxqjdhu	284	29	projections	projection	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	284	30	T	T	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	31	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	284	32	g	g	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	33	g	g	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	34	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	35	g	g	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	36	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	284	37	c	c	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	38	c	c	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	39	u	u	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	40	ra	ra	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	41	c	c	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	42	y	y	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	43	Number	Number	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	44	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	45	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	284	46	listed	list	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	284	47	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	48	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	49	Figure	Figure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	284	50	4	4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	284	51	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	284	52	Relative	relative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	284	53	influence	influence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	54	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	55	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	56	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	284	57	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	58	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	284	59	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	60	tagging	tag	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	284	61	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	62	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	284	63	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	284	64	ŶCRFwik	ŷcrfwik	NN
work_wgi74n6oavhdxhlomowlxqjdhu	284	65	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	285	1	+	+	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	285	2	C+L	C+L	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	285	3	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	285	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	286	1	Word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	2	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	286	3	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	286	4	categorized	categorize	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	286	5	according	accord	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	286	6	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	286	8	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	286	9	their	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	286	10	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	12	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	286	13	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	286	14	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	286	15	0,1,2	0,1,2	NFP
work_wgi74n6oavhdxhlomowlxqjdhu	286	16	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	286	17	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	18	+	+	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	286	19	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	286	20	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	286	21	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	22	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	23	representing	represent	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	286	24	no	no	DT
work_wgi74n6oavhdxhlomowlxqjdhu	286	25	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	286	26	entry	entry	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	27	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	286	28	top	top	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	29	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	286	30	axis	axis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	31	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	286	32	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	286	33	b	b	LS
work_wgi74n6oavhdxhlomowlxqjdhu	286	34	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	286	35	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	286	36	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	37	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	38	times	time	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	286	39	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	286	40	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	286	41	token-	token-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	42	constrained	constrain	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	286	43	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	44	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	286	45	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	46	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	47	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	286	48	divided	divide	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	286	49	into	into	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	50	buckets	bucket	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	286	51	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	286	52	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	53	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	286	54	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	55	-	-	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	286	56	9	9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	57	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	286	58	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	59	-	-	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	286	60	99	99	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	61	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	286	62	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	286	63	+	+	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	286	64	occurrences	occurrence	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	286	65	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	286	66	x	x	LS
work_wgi74n6oavhdxhlomowlxqjdhu	286	67	-	-	:
work_wgi74n6oavhdxhlomowlxqjdhu	286	68	axis	axis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	286	69	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	286	70	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	287	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	287	2	boxes	box	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	287	3	summarize	summarize	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	287	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	287	5	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	287	6	distributions	distribution	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	287	7	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	287	8	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	287	9	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	287	10	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	287	11	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	287	12	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	287	13	category	category	NN
work_wgi74n6oavhdxhlomowlxqjdhu	287	14	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	287	15	defined	define	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	287	16	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	287	17	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	287	18	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	287	19	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	287	20	b	b	NN
work_wgi74n6oavhdxhlomowlxqjdhu	287	21	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	287	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	288	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	2	horizontal	horizontal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	3	line	line	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	4	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	288	5	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	6	box	box	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	7	marks	mark	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	288	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	9	median	median	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	10	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	288	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	13	top	top	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	14	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	288	15	bottom	bottom	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	16	mark	mark	VB
work_wgi74n6oavhdxhlomowlxqjdhu	288	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	18	first	first	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	19	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	288	20	third	third	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	21	quantile	quantile	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	22	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	288	23	re-	re-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	288	24	spectively	spectively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	288	25	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	288	26	while	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	288	27	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	28	whiskers	whisker	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	288	29	mark	mark	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	288	30	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	31	minimum	minimum	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	32	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	288	33	maximum	maximum	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	288	34	values	value	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	288	35	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	288	36	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	288	37	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	38	distribution	distribution	NN
work_wgi74n6oavhdxhlomowlxqjdhu	288	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	289	1	better	well	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	289	2	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	289	3	all	all	PDT
work_wgi74n6oavhdxhlomowlxqjdhu	289	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	289	5	remaining	remain	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	289	6	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	289	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	289	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	289	9	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	289	10	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	289	11	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	289	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	289	13	88.8	88.8	CD
work_wgi74n6oavhdxhlomowlxqjdhu	289	14	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	289	15	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	289	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	289	17	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	289	18	Indo	indo	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	289	19	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	289	20	European	european	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	289	21	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	289	22	available	available	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	289	23	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	289	24	D&P	D&P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	289	25	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	289	26	LG&T.	LG&T.	.
work_wgi74n6oavhdxhlomowlxqjdhu	290	1	Averaged	average	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	290	2	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	290	3	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	290	4	15	15	CD
work_wgi74n6oavhdxhlomowlxqjdhu	290	5	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	290	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	290	7	its	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	290	8	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	290	9	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	290	10	84.5	84.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	290	11	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	290	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	291	1	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	291	2	Further	further	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	291	3	Analysis	Analysis	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	291	4	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	291	5	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	291	6	section	section	NN
work_wgi74n6oavhdxhlomowlxqjdhu	291	7	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	291	8	provide	provide	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	291	9	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	291	10	detailed	detailed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	291	11	analysis	analysis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	291	12	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	291	13	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	291	14	impact	impact	NN
work_wgi74n6oavhdxhlomowlxqjdhu	291	15	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	291	16	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	291	17	versus	versus	FW
work_wgi74n6oavhdxhlomowlxqjdhu	291	18	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	291	19	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	291	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	291	21	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	291	22	study	study	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	291	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	291	24	pruning	pruning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	291	25	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	291	26	filtering	filter	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	291	27	mistakes	mistake	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	291	28	resulting	result	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	291	29	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	291	30	in-	in-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	291	31	complete	complete	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	291	32	Wiktionary	wiktionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	291	33	entries	entry	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	291	34	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	291	35	detail	detail	NN
work_wgi74n6oavhdxhlomowlxqjdhu	291	36	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	292	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	292	2	analysis	analysis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	292	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	292	4	based	base	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	292	5	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	292	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	292	7	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	292	8	portion	portion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	292	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	292	10	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	292	11	treebank	treebank	NN
work_wgi74n6oavhdxhlomowlxqjdhu	292	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	293	1	5.1	5.1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	293	2	Influence	Influence	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	293	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	293	4	Token	Token	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	293	5	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	293	6	Type	Type	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	293	7	Constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	293	8	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	293	9	empirical	empirical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	293	10	success	success	NN
work_wgi74n6oavhdxhlomowlxqjdhu	293	11	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	293	12	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	293	13	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	293	14	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	293	15	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	293	16	cou-	cou-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	293	17	pled	plead	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	293	18	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	293	19	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	293	20	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	293	21	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	293	22	confirms	confirm	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	293	23	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	293	24	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	293	25	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	293	26	indeed	indeed	RB
work_wgi74n6oavhdxhlomowlxqjdhu	293	27	provide	provide	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	293	28	complementary	complementary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	293	29	signals	signal	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	293	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	294	1	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	294	2	4	4	CD
work_wgi74n6oavhdxhlomowlxqjdhu	294	3	provides	provide	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	294	4	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	294	5	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	294	6	detailed	detailed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	294	7	view	view	NN
work_wgi74n6oavhdxhlomowlxqjdhu	294	8	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	294	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	294	10	rela-	rela-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	294	11	tive	tive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	294	12	benefits	benefit	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	294	13	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	294	14	each	each	DT
work_wgi74n6oavhdxhlomowlxqjdhu	294	15	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	294	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	294	17	constraint	constraint	NN
work_wgi74n6oavhdxhlomowlxqjdhu	294	18	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	295	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	295	2	observe	observe	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	295	3	several	several	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	295	4	interesting	interesting	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	295	5	trends	trend	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	295	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	296	1	First	first	RB
work_wgi74n6oavhdxhlomowlxqjdhu	296	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	296	3	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	4	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	5	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	296	6	occur	occur	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	296	7	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	8	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	296	9	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	296	10	con-	con-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	11	straints	straint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	12	during	during	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	13	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	14	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	296	15	generally	generally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	296	16	tagged	tag	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	296	17	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	296	18	accurately	accurately	RB
work_wgi74n6oavhdxhlomowlxqjdhu	296	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	296	20	regardless	regardless	RB
work_wgi74n6oavhdxhlomowlxqjdhu	296	21	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	22	whether	whether	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	23	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	296	24	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	25	occur	occur	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	296	26	9	9	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	27	90.0	90.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	28	92.5	92.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	29	95.0	95.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	30	97.5	97.5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	31	100.0	100.0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	32	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	33	50	50	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	34	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	35	150	150	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	36	200	200	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	37	250	250	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	38	Number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	39	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	40	corrected	correct	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	296	41	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	42	entries	entrie	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	296	43	P	P	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	44	ru	ru	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	45	n	n	CC
work_wgi74n6oavhdxhlomowlxqjdhu	296	46	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	47	g	g	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	48	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	296	49	cc	cc	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	50	u	u	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	51	ra	ra	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	52	cy	cy	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	53	Figure	Figure	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	54	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	296	55	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	296	56	Average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	296	57	pruning	pruning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	58	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	59	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	296	60	line	line	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	61	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	296	62	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	63	lan-	lan-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	64	guages	guage	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	65	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	296	66	dots	dot	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	67	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	296	68	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	69	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	296	70	function	function	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	71	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	72	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	296	73	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	74	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	75	hypotheti-	hypotheti-	FW
work_wgi74n6oavhdxhlomowlxqjdhu	296	76	cally	cally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	296	77	corrected	correct	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	296	78	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	296	79	entries	entry	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	80	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	296	81	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	296	82	k	k	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	83	most	most	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	296	84	frequent	frequent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	296	85	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	296	86	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	296	87	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	297	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	297	2	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	297	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	297	4	position	position	NN
work_wgi74n6oavhdxhlomowlxqjdhu	297	5	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	297	6	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	297	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	297	8	x	x	NN
work_wgi74n6oavhdxhlomowlxqjdhu	297	9	-	-	:
work_wgi74n6oavhdxhlomowlxqjdhu	297	10	axis	axis	RB
work_wgi74n6oavhdxhlomowlxqjdhu	297	11	cor-	cor-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	297	12	responds	respond	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	297	13	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	297	14	manually	manually	RB
work_wgi74n6oavhdxhlomowlxqjdhu	297	15	correcting	correct	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	297	16	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	297	17	entries	entry	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	297	18	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	297	19	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	297	20	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	297	21	most	most	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	297	22	frequent	frequent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	297	23	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	297	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	297	25	while	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	297	26	position	position	NN
work_wgi74n6oavhdxhlomowlxqjdhu	297	27	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	297	28	corresponds	correspond	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	297	29	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	297	30	ex-	ex-	VB
work_wgi74n6oavhdxhlomowlxqjdhu	297	31	perimental	perimental	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	297	32	conditions	condition	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	297	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	298	1	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	298	2	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	298	3	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	299	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	299	2	most	most	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	299	3	common	common	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	299	4	scenario	scenario	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	5	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	299	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	299	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	299	8	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	9	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	10	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	299	11	have	have	VB
work_wgi74n6oavhdxhlomowlxqjdhu	299	12	exactly	exactly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	299	13	one	one	CD
work_wgi74n6oavhdxhlomowlxqjdhu	299	14	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	15	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	299	16	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	299	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	299	18	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	299	19	occur	occur	VB
work_wgi74n6oavhdxhlomowlxqjdhu	299	20	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	299	21	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	299	22	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	299	23	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	24	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	299	25	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	299	26	times	time	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	299	27	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	299	28	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	299	29	training	training	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	30	set	set	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	31	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	299	32	facet	facet	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	299	33	1	1	CD
work_wgi74n6oavhdxhlomowlxqjdhu	299	34	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	299	35	rightmost	rightmost	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	299	36	box	box	NN
work_wgi74n6oavhdxhlomowlxqjdhu	299	37	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	299	38	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	300	1	These	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	300	2	com-	com-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	300	3	mon	mon	NN
work_wgi74n6oavhdxhlomowlxqjdhu	300	4	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	300	5	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	300	6	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	300	7	typically	typically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	300	8	tagged	tag	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	300	9	very	very	RB
work_wgi74n6oavhdxhlomowlxqjdhu	300	10	accurately	accurately	RB
work_wgi74n6oavhdxhlomowlxqjdhu	300	11	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	300	12	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	300	13	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	300	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	301	1	Second	second	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	301	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	301	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	301	4	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	301	5	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	301	6	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	301	7	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	301	8	ambiguous	ambiguous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	301	9	accord-	accord-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	301	10	ing	ing	NN
work_wgi74n6oavhdxhlomowlxqjdhu	301	11	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	301	12	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	301	13	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	301	14	facets	facet	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	301	15	2	2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	301	16	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	301	17	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	301	18	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	301	19	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	301	20	predominantly	predominantly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	301	21	frequent	frequent	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	301	22	ones	one	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	301	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	302	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	302	2	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	302	3	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	302	4	typically	typically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	302	5	lower	low	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	302	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	302	7	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	302	8	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	302	9	compared	compare	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	302	10	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	302	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	302	12	unambiguous	unambiguous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	302	13	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	302	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	303	1	However	however	RB
work_wgi74n6oavhdxhlomowlxqjdhu	303	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	303	3	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	303	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	303	5	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	303	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	303	7	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	303	8	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	303	9	con-	con-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	303	10	straints	straint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	303	11	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	303	12	increased	increase	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	303	13	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	303	14	zero	zero	CD
work_wgi74n6oavhdxhlomowlxqjdhu	303	15	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	303	16	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	303	17	+	+	SYM
work_wgi74n6oavhdxhlomowlxqjdhu	303	18	observations	observation	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	303	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	303	20	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	303	21	ambiguous	ambiguous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	303	22	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	303	23	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	303	24	effectively	effectively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	303	25	disambiguated	disambiguate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	303	26	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	303	27	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	303	28	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	303	29	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	303	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	304	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	304	2	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	304	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	304	4	advantage	advantage	NN
work_wgi74n6oavhdxhlomowlxqjdhu	304	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	304	6	intersecting	intersect	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	304	7	token	token	NN
work_wgi74n6oavhdxhlomowlxqjdhu	304	8	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	304	9	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	304	10	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	304	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	305	1	Finally	finally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	305	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	305	3	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	305	4	generally	generally	RB
work_wgi74n6oavhdxhlomowlxqjdhu	305	5	helps	help	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	305	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	7	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	305	8	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	305	9	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	305	10	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	305	11	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	12	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	305	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	305	14	although	although	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	305	16	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	305	17	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	18	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	305	19	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	305	20	never	never	RB
work_wgi74n6oavhdxhlomowlxqjdhu	305	21	reach	reach	VB
work_wgi74n6oavhdxhlomowlxqjdhu	305	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	305	23	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	305	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	25	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	305	26	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	305	27	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	28	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	305	29	one	one	CD
work_wgi74n6oavhdxhlomowlxqjdhu	305	30	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	305	31	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	305	32	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	305	33	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	306	1	Interestingly	interestingly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	306	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	306	3	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	306	4	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	306	5	occur	occur	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	306	6	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	306	8	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	306	9	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	306	10	constraint	constraint	NN
work_wgi74n6oavhdxhlomowlxqjdhu	306	11	less	less	JJR
work_wgi74n6oavhdxhlomowlxqjdhu	306	12	than	than	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	13	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	306	14	times	time	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	306	15	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	306	16	tagged	tag	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	306	17	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	306	18	accurately	accurately	RB
work_wgi74n6oavhdxhlomowlxqjdhu	306	19	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	20	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	306	21	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	306	22	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	23	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	306	24	dictionary	dictionary	NN
work_wgi74n6oavhdxhlomowlxqjdhu	306	25	compared	compare	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	306	26	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	27	ambiguous	ambiguous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	306	28	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	306	29	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	306	30	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	306	32	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	306	33	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	306	34	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	306	35	projected	project	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	306	36	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	306	37	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	307	1	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	307	2	possible	possible	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	307	3	explanation	explanation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	307	4	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	307	5	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	307	6	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	307	7	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	307	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	307	9	ambiguous	ambiguous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	307	10	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	307	11	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	307	12	inherently	inherently	RB
work_wgi74n6oavhdxhlomowlxqjdhu	307	13	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	307	14	difficult	difficult	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	307	15	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	307	16	predict	predict	VB
work_wgi74n6oavhdxhlomowlxqjdhu	307	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	307	18	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	307	19	most	most	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	307	20	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	307	21	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	307	22	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	307	23	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	307	24	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	307	25	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	307	26	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	307	27	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	307	28	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	307	29	less	less	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	307	30	common	common	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	307	31	words	word	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	307	32	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	307	33	tend	tend	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	307	34	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	307	35	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	307	36	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	307	37	less	less	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	307	38	ambiguous	ambiguous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	307	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	308	1	zh	zh	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	2	tr	tr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	308	3	sv	sv	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	4	sl	sl	UH
work_wgi74n6oavhdxhlomowlxqjdhu	308	5	pt	pt	UH
work_wgi74n6oavhdxhlomowlxqjdhu	308	6	nl	nl	UH
work_wgi74n6oavhdxhlomowlxqjdhu	308	7	ja	ja	IN
work_wgi74n6oavhdxhlomowlxqjdhu	308	8	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	308	9	fr	fr	IN
work_wgi74n6oavhdxhlomowlxqjdhu	308	10	es	es	FW
work_wgi74n6oavhdxhlomowlxqjdhu	308	11	el	el	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	12	de	de	FW
work_wgi74n6oavhdxhlomowlxqjdhu	308	13	da	da	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	14	cs	cs	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	15	bg	bg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	16	avg	avg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	17	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	308	18	25	25	CD
work_wgi74n6oavhdxhlomowlxqjdhu	308	19	50	50	CD
work_wgi74n6oavhdxhlomowlxqjdhu	308	20	75	75	CD
work_wgi74n6oavhdxhlomowlxqjdhu	308	21	100	100	CD
work_wgi74n6oavhdxhlomowlxqjdhu	308	22	Proportion	proportion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	308	23	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	308	24	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	308	25	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	308	26	PRON	PRON	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	27	NOUN	NOUN	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	28	DET	DET	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	29	ADP	ADP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	30	PRT	PRT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	31	ADV	ADV	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	32	NUM	NUM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	33	CONJ	CONJ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	34	ADJ	ADJ	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	35	VERB	VERB	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	36	X	X	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	308	37	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	309	1	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	309	2	6	6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	309	3	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	309	4	Prevalence	Prevalence	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	309	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	309	6	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	309	7	mistakes	mistake	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	309	8	per	per	IN
work_wgi74n6oavhdxhlomowlxqjdhu	309	9	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	309	10	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	309	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	309	12	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	309	13	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	309	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	309	15	inference	inference	NN
work_wgi74n6oavhdxhlomowlxqjdhu	309	16	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	309	17	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	309	18	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	309	19	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	309	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	310	1	5.2	5.2	CD
work_wgi74n6oavhdxhlomowlxqjdhu	310	2	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	310	3	Pruning	Pruning	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	310	4	Mistakes	Mistakes	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	310	5	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	310	6	error	error	NN
work_wgi74n6oavhdxhlomowlxqjdhu	310	7	analysis	analysis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	310	8	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	310	9	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	310	10	et	et	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	310	11	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	310	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	311	1	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	311	2	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	311	3	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	311	4	showed	show	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	311	5	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	311	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	311	7	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	311	8	licensed	license	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	311	9	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	311	10	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	311	11	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	311	12	often	often	RB
work_wgi74n6oavhdxhlomowlxqjdhu	311	13	valid	valid	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	311	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	312	1	When	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	312	2	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	312	3	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	312	4	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	312	5	prune	prune	VB
work_wgi74n6oavhdxhlomowlxqjdhu	312	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	312	7	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	312	8	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	312	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	312	10	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	312	11	constrained	constrained	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	312	12	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	312	13	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	312	14	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	312	15	filter	filter	VB
work_wgi74n6oavhdxhlomowlxqjdhu	312	16	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	312	17	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	312	18	level	level	NN
work_wgi74n6oavhdxhlomowlxqjdhu	312	19	projec-	projec-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	312	20	tions	tion	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	312	21	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	312	22	it	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	312	23	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	312	24	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	312	25	important	important	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	312	26	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	312	27	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	312	28	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	312	29	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	312	30	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	312	31	mistakenly	mistakenly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	312	32	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	312	33	because	because	IN
work_wgi74n6oavhdxhlomowlxqjdhu	312	34	they	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	312	35	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	312	36	missing	miss	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	312	37	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	312	38	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	312	39	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	313	1	While	while	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	3	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	5	filtering	filtering	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	6	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	313	7	more	more	RBR
work_wgi74n6oavhdxhlomowlxqjdhu	313	8	difficult	difficult	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	313	9	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	313	10	study	study	VB
work_wgi74n6oavhdxhlomowlxqjdhu	313	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	313	12	due	due	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	13	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	14	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	15	lack	lack	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	17	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	18	gold	gold	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	313	19	standard	standard	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	313	20	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	21	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	23	bitext	bitext	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	24	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	313	25	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	26	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	313	27	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	313	28	position	position	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	29	0	0	CD
work_wgi74n6oavhdxhlomowlxqjdhu	313	30	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	32	x-	x-	XX
work_wgi74n6oavhdxhlomowlxqjdhu	313	33	axis	axis	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	34	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	313	35	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	313	36	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	37	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	38	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	39	pruning	pruning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	40	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	313	41	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	313	42	not	not	RB
work_wgi74n6oavhdxhlomowlxqjdhu	313	43	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	44	major	major	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	313	45	issue	issue	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	46	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	47	most	most	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	313	48	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	313	49	;	;	:
work_wgi74n6oavhdxhlomowlxqjdhu	313	50	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	313	51	average	average	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	313	52	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	313	53	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	313	54	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	55	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	313	56	almost	almost	RB
work_wgi74n6oavhdxhlomowlxqjdhu	313	57	95	95	CD
work_wgi74n6oavhdxhlomowlxqjdhu	313	58	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	313	59	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	314	1	However	however	RB
work_wgi74n6oavhdxhlomowlxqjdhu	314	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	314	3	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	314	4	some	some	DT
work_wgi74n6oavhdxhlomowlxqjdhu	314	5	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	314	6	such	such	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	314	7	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	314	8	Chinese	chinese	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	314	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	314	10	Czech	Czech	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	314	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	314	12	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	314	13	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	314	14	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	314	15	pruned	prune	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	314	16	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	314	17	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	314	18	search	search	NN
work_wgi74n6oavhdxhlomowlxqjdhu	314	19	space	space	NN
work_wgi74n6oavhdxhlomowlxqjdhu	314	20	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	314	21	nearly	nearly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	314	22	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	314	23	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	314	24	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	314	25	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	314	26	tokens	token	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	314	27	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	315	1	When	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	315	2	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	315	3	Wiktionary	Wiktionary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	315	4	as	as	IN
work_wgi74n6oavhdxhlomowlxqjdhu	315	5	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	315	6	pruner	pruner	NN
work_wgi74n6oavhdxhlomowlxqjdhu	315	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	315	8	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	315	9	upper	upper	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	315	10	bound	bind	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	315	11	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	315	12	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	315	13	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	315	14	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	315	15	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	315	16	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	315	17	therefore	therefore	RB
work_wgi74n6oavhdxhlomowlxqjdhu	315	18	only	only	RB
work_wgi74n6oavhdxhlomowlxqjdhu	315	19	around	around	RB
work_wgi74n6oavhdxhlomowlxqjdhu	315	20	90	90	CD
work_wgi74n6oavhdxhlomowlxqjdhu	315	21	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	315	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	316	1	However	however	RB
work_wgi74n6oavhdxhlomowlxqjdhu	316	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	316	3	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	316	4	5	5	CD
work_wgi74n6oavhdxhlomowlxqjdhu	316	5	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	316	6	shows	show	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	316	7	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	316	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	316	9	some	some	DT
work_wgi74n6oavhdxhlomowlxqjdhu	316	10	manual	manual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	316	11	effort	effort	NN
work_wgi74n6oavhdxhlomowlxqjdhu	316	12	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	316	13	might	may	MD
work_wgi74n6oavhdxhlomowlxqjdhu	316	14	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	316	15	able	able	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	316	16	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	316	17	remedy	remedy	VB
work_wgi74n6oavhdxhlomowlxqjdhu	316	18	many	many	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	316	19	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	316	20	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	316	21	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	316	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	317	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	2	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	317	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	317	4	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	5	adding	add	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	317	6	miss-	miss-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	317	7	ing	ing	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	317	8	valid	valid	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	317	9	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	317	10	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	317	12	250	250	CD
work_wgi74n6oavhdxhlomowlxqjdhu	317	13	most	most	RBS
work_wgi74n6oavhdxhlomowlxqjdhu	317	14	common	common	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	317	15	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	317	16	types	type	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	317	17	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	317	19	worst	bad	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	317	20	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	317	21	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	317	22	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	317	23	minimum	minimum	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	317	24	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	317	25	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	317	26	would	would	MD
work_wgi74n6oavhdxhlomowlxqjdhu	317	27	rise	rise	VB
work_wgi74n6oavhdxhlomowlxqjdhu	317	28	above	above	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	29	95	95	CD
work_wgi74n6oavhdxhlomowlxqjdhu	317	30	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	317	31	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	32	below	below	IN
work_wgi74n6oavhdxhlomowlxqjdhu	317	33	90	90	CD
work_wgi74n6oavhdxhlomowlxqjdhu	317	34	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	317	35	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	318	1	If	if	IN
work_wgi74n6oavhdxhlomowlxqjdhu	318	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	318	3	same	same	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	318	4	was	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	318	5	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	318	6	be	be	VB
work_wgi74n6oavhdxhlomowlxqjdhu	318	7	done	do	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	318	8	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	318	9	all	all	DT
work_wgi74n6oavhdxhlomowlxqjdhu	318	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	318	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	318	12	studied	study	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	318	13	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	318	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	318	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	318	16	mean	mean	NN
work_wgi74n6oavhdxhlomowlxqjdhu	318	17	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	318	18	accuracy	accuracy	NN
work_wgi74n6oavhdxhlomowlxqjdhu	318	19	would	would	MD
work_wgi74n6oavhdxhlomowlxqjdhu	318	20	reach	reach	VB
work_wgi74n6oavhdxhlomowlxqjdhu	318	21	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	318	22	97	97	CD
work_wgi74n6oavhdxhlomowlxqjdhu	318	23	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	318	24	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	319	1	Figure	figure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	319	2	6	6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	319	3	breaks	break	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	319	4	down	down	RP
work_wgi74n6oavhdxhlomowlxqjdhu	319	5	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	319	6	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	319	7	resulting	result	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	319	8	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	319	9	incorrect	incorrect	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	319	10	or	or	CC
work_wgi74n6oavhdxhlomowlxqjdhu	319	11	incomplete	incomplete	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	319	12	Wiktionary	wiktionary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	319	13	entries	entry	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	319	14	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	319	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	319	16	correct	correct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	319	17	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	319	18	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	319	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	320	1	From	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	320	2	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	320	3	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	320	4	observe	observe	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	320	5	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	320	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	320	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	320	8	many	many	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	320	9	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	320	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	320	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	320	12	pruning	pruning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	320	13	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	320	14	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	320	15	highly	highly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	320	16	skewed	skewed	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	320	17	towards	towards	IN
work_wgi74n6oavhdxhlomowlxqjdhu	320	18	specific	specific	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	320	19	tags	tag	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	320	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	321	1	For	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	321	2	example	example	NN
work_wgi74n6oavhdxhlomowlxqjdhu	321	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	321	4	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	321	5	Czech	Czech	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	321	6	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	321	7	80	80	CD
work_wgi74n6oavhdxhlomowlxqjdhu	321	8	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	321	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	321	10	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	321	11	pruning	prune	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	321	12	errors	error	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	321	13	are	be	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	321	14	caused	cause	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	321	15	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	321	16	mistak-	mistak-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	321	17	enly	enly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	321	18	pruned	pruned	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	321	19	pronouns	pronoun	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	321	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	322	1	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	322	2	6	6	CD
work_wgi74n6oavhdxhlomowlxqjdhu	322	3	Conclusions	Conclusions	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	322	4	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	322	5	considered	consider	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	322	6	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	322	7	problem	problem	NN
work_wgi74n6oavhdxhlomowlxqjdhu	322	8	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	322	9	constructing	construct	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	322	10	multilin-	multilin-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	322	11	gual	gual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	322	12	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	322	13	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	322	14	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	322	15	resource	resource	NN
work_wgi74n6oavhdxhlomowlxqjdhu	322	16	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	322	17	poor	poor	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	322	18	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	322	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	323	1	To	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	323	2	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	323	3	end	end	NN
work_wgi74n6oavhdxhlomowlxqjdhu	323	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	323	5	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	323	6	explored	explore	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	323	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	323	8	number	number	NN
work_wgi74n6oavhdxhlomowlxqjdhu	323	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	323	10	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	323	11	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	323	12	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	323	13	combine	combine	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	323	14	token	token	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	323	15	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	323	16	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	323	17	type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	323	18	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	323	19	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	323	20	different	different	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	323	21	sources	source	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	323	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	324	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	324	2	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	324	3	results	result	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	324	4	were	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	324	5	ob-	ob-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	324	6	tained	taine	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	324	7	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	324	8	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	324	9	partially	partially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	324	10	observed	observe	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	324	11	CRF	CRF	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	324	12	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	324	13	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	324	14	ef-	ef-	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	324	15	fectively	fectively	RB
work_wgi74n6oavhdxhlomowlxqjdhu	324	16	integrates	integrate	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	324	17	these	these	DT
work_wgi74n6oavhdxhlomowlxqjdhu	324	18	complementary	complementary	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	324	19	constraints	constraint	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	324	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	325	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	325	2	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	325	3	extensive	extensive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	325	4	empirical	empirical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	325	5	study	study	NN
work_wgi74n6oavhdxhlomowlxqjdhu	325	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	325	7	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	325	8	showed	show	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	325	9	that	that	IN
work_wgi74n6oavhdxhlomowlxqjdhu	325	10	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	325	11	approach	approach	NN
work_wgi74n6oavhdxhlomowlxqjdhu	325	12	substantially	substantially	RB
work_wgi74n6oavhdxhlomowlxqjdhu	325	13	improves	improve	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	325	14	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	325	15	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	325	16	state	state	NN
work_wgi74n6oavhdxhlomowlxqjdhu	325	17	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	325	18	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	325	19	art	art	NN
work_wgi74n6oavhdxhlomowlxqjdhu	325	20	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	325	21	this	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	325	22	context	context	NN
work_wgi74n6oavhdxhlomowlxqjdhu	325	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	326	1	Our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	326	2	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	326	3	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	326	4	significantly	significantly	RB
work_wgi74n6oavhdxhlomowlxqjdhu	326	5	out-	out-	RB
work_wgi74n6oavhdxhlomowlxqjdhu	326	6	performed	perform	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	326	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	326	8	second	second	RB
work_wgi74n6oavhdxhlomowlxqjdhu	326	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	326	10	best	good	JJS
work_wgi74n6oavhdxhlomowlxqjdhu	326	11	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	326	12	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	326	13	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	326	14	out	out	IN
work_wgi74n6oavhdxhlomowlxqjdhu	326	15	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	326	16	15	15	CD
work_wgi74n6oavhdxhlomowlxqjdhu	326	17	evaluated	evaluated	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	326	18	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	326	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	326	20	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	326	21	trained	train	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	326	22	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	326	23	identical	identical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	326	24	data	data	NN
work_wgi74n6oavhdxhlomowlxqjdhu	326	25	sets	set	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	326	26	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	326	27	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	326	28	an	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	326	29	insignificant	insignificant	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	326	30	difference	difference	NN
work_wgi74n6oavhdxhlomowlxqjdhu	326	31	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	326	32	3	3	CD
work_wgi74n6oavhdxhlomowlxqjdhu	326	33	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	326	34	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	327	1	Compared	compare	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	327	2	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	327	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	327	4	prior	prior	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	327	5	state	state	NN
work_wgi74n6oavhdxhlomowlxqjdhu	327	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	327	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	327	8	art	art	NN
work_wgi74n6oavhdxhlomowlxqjdhu	327	9	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	327	10	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	327	11	et	et	FW
work_wgi74n6oavhdxhlomowlxqjdhu	327	12	al	al	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	327	13	.	.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	327	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	327	15	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	327	16	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	327	17	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	327	18	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	327	19	observed	observe	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	327	20	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	327	21	relative	relative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	327	22	reduction	reduction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	327	23	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	327	24	error	error	NN
work_wgi74n6oavhdxhlomowlxqjdhu	327	25	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	327	26	25	25	CD
work_wgi74n6oavhdxhlomowlxqjdhu	327	27	%	%	NN
work_wgi74n6oavhdxhlomowlxqjdhu	327	28	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	327	29	averaged	average	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	327	30	over	over	IN
work_wgi74n6oavhdxhlomowlxqjdhu	327	31	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	327	32	eight	eight	CD
work_wgi74n6oavhdxhlomowlxqjdhu	327	33	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	327	34	common	common	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	327	35	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	327	36	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	327	37	studies	study	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	327	38	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	328	1	Acknowledgments	acknowledgment	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	328	2	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	328	3	thank	thank	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	328	4	Alexander	Alexander	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	328	5	Rush	Rush	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	328	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	328	7	help	help	NN
work_wgi74n6oavhdxhlomowlxqjdhu	328	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	328	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	328	10	hyper-	hyper-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	328	11	graph	graph	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	328	12	framework	framework	NN
work_wgi74n6oavhdxhlomowlxqjdhu	328	13	that	that	WDT
work_wgi74n6oavhdxhlomowlxqjdhu	328	14	was	be	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	328	15	used	use	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	328	16	to	to	TO
work_wgi74n6oavhdxhlomowlxqjdhu	328	17	implement	implement	VB
work_wgi74n6oavhdxhlomowlxqjdhu	328	18	our	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	328	19	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	328	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	328	21	Klaus	Klaus	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	328	22	Macherey	Macherey	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	328	23	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	328	24	help	help	NN
work_wgi74n6oavhdxhlomowlxqjdhu	328	25	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	328	26	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	328	27	bi-	bi-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	328	28	text	text	NN
work_wgi74n6oavhdxhlomowlxqjdhu	328	29	extraction	extraction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	328	30	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	329	1	This	this	DT
work_wgi74n6oavhdxhlomowlxqjdhu	329	2	work	work	NN
work_wgi74n6oavhdxhlomowlxqjdhu	329	3	benefited	benefit	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	329	4	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	329	5	many	many	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	329	6	dis-	dis-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	329	7	cussions	cussion	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	329	8	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	329	9	Yoav	Yoav	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	10	Goldberg	Goldberg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	11	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	329	12	Keith	Keith	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	13	Hall	Hall	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	329	15	Kuzman	Kuzman	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	16	Ganchev	Ganchev	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	17	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	329	18	Hao	Hao	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	19	Zhang	Zhang	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	329	20	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	330	1	We	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	330	2	also	also	RB
work_wgi74n6oavhdxhlomowlxqjdhu	330	3	thank	thank	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	330	4	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	330	5	editor	editor	NN
work_wgi74n6oavhdxhlomowlxqjdhu	330	6	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	330	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	330	8	three	three	CD
work_wgi74n6oavhdxhlomowlxqjdhu	330	9	anonymous	anonymous	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	330	10	reviewers	reviewer	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	330	11	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	330	12	their	-PRON-	PRP$
work_wgi74n6oavhdxhlomowlxqjdhu	330	13	valuable	valuable	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	330	14	feedback	feedback	NN
work_wgi74n6oavhdxhlomowlxqjdhu	330	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	331	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	331	2	first	first	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	331	3	author	author	NN
work_wgi74n6oavhdxhlomowlxqjdhu	331	4	is	be	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	331	5	grateful	grateful	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	331	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	331	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	331	8	financial	financial	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	331	9	support	support	NN
work_wgi74n6oavhdxhlomowlxqjdhu	331	10	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	331	11	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	331	12	Swedish	Swedish	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	13	National	National	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	14	Graduate	Graduate	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	15	School	School	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	16	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	331	17	Language	Language	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	18	Technology	Technology	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	19	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	331	20	GSLT	GSLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	331	21	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	331	22	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	332	1	References	References	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	2	Anne	Anne	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	3	Abeillé	Abeillé	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	332	5	Lionel	Lionel	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	6	Clément	Clément	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	332	8	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	332	9	François	François	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	10	Toussenel	Toussenel	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	332	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	333	1	2003	2003	CD
work_wgi74n6oavhdxhlomowlxqjdhu	333	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	334	1	Building	build	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	334	2	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	334	3	Treebank	Treebank	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	334	4	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	334	5	French	French	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	334	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	335	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	335	2	A.	a.	NN
work_wgi74n6oavhdxhlomowlxqjdhu	335	3	Abeillé	Abeillé	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	335	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	335	5	editor	editor	NN
work_wgi74n6oavhdxhlomowlxqjdhu	335	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	335	7	Treebanks	treebank	VBZ
work_wgi74n6oavhdxhlomowlxqjdhu	335	8	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	335	9	Building	build	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	335	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	335	11	Using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	335	12	Parsed	Parsed	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	335	13	Corpora	Corpora	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	335	14	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	335	15	chapter	chapter	NN
work_wgi74n6oavhdxhlomowlxqjdhu	335	16	10	10	CD
work_wgi74n6oavhdxhlomowlxqjdhu	335	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	336	1	Kluwer	Kluwer	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	336	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	337	1	Taylor	Taylor	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	2	Berg	Berg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	3	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	337	4	Kirkpatrick	Kirkpatrick	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	5	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	337	6	Alexandre	Alexandre	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	7	Bouchard	Bouchard	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	337	9	Côté	Côté	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	337	11	John	John	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	12	DeNero	DeNero	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	337	14	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	337	15	Dan	Dan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	16	Klein	Klein	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	337	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	338	1	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	338	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	339	1	Painless	painless	NN
work_wgi74n6oavhdxhlomowlxqjdhu	339	2	unsupervised	unsupervised	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	339	3	learning	learn	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	339	4	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	339	5	features	feature	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	339	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	340	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	340	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	340	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	340	4	NAACL	NAACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	340	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	340	6	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	340	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	341	1	Sabine	Sabine	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	341	2	Buchholz	Buchholz	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	341	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	341	4	Erwin	Erwin	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	341	5	Marsi	Marsi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	341	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	342	1	2006	2006	CD
work_wgi74n6oavhdxhlomowlxqjdhu	342	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	343	1	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	343	2	-	-	:
work_wgi74n6oavhdxhlomowlxqjdhu	343	3	X	X	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	343	4	shared	share	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	343	5	task	task	NN
work_wgi74n6oavhdxhlomowlxqjdhu	343	6	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	343	7	multilingual	multilingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	343	8	dependency	dependency	NN
work_wgi74n6oavhdxhlomowlxqjdhu	343	9	parsing	parsing	NN
work_wgi74n6oavhdxhlomowlxqjdhu	343	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	344	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	344	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	344	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	344	4	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	344	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	345	1	Stanley	Stanley	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	345	2	F	F	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	345	3	Chen	Chen	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	345	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	346	1	2003	2003	CD
work_wgi74n6oavhdxhlomowlxqjdhu	346	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	347	1	Conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	347	2	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	347	3	joint	joint	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	347	4	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	347	5	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	347	6	grapheme	grapheme	NN
work_wgi74n6oavhdxhlomowlxqjdhu	347	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	347	8	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	347	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	347	10	phoneme	phoneme	NN
work_wgi74n6oavhdxhlomowlxqjdhu	347	11	conversion	conversion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	347	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	348	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	348	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	348	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	348	4	Eurospeech	Eurospeech	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	348	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	349	1	Christos	Christos	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	349	2	Christodoulopoulos	Christodoulopoulos	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	349	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	349	4	Sharon	Sharon	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	349	5	Goldwater	Goldwater	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	349	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	349	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	349	8	Mark	Mark	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	349	9	Steedman	Steedman	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	349	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	350	1	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	350	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	351	1	Two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	351	2	decades	decade	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	351	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	351	4	unsupervised	unsupervised	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	351	5	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	351	6	induction	induction	NN
work_wgi74n6oavhdxhlomowlxqjdhu	351	7	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	351	8	How	how	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	351	9	far	far	RB
work_wgi74n6oavhdxhlomowlxqjdhu	351	10	have	have	VBP
work_wgi74n6oavhdxhlomowlxqjdhu	351	11	we	-PRON-	PRP
work_wgi74n6oavhdxhlomowlxqjdhu	351	12	come	come	VB
work_wgi74n6oavhdxhlomowlxqjdhu	351	13	?	?	.
work_wgi74n6oavhdxhlomowlxqjdhu	352	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	352	2	Proceed-	Proceed-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	352	3	ings	ing	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	352	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	352	5	EMNLP	EMNLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	352	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	353	1	Dipanjan	Dipanjan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	353	2	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	353	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	353	4	Slav	Slav	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	353	5	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	353	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	354	1	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	354	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	355	1	Unsupervised	unsupervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	355	2	part-	part-	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	355	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	355	4	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	355	5	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	355	6	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	355	7	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	355	8	bilingual	bilingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	355	9	graph	graph	NN
work_wgi74n6oavhdxhlomowlxqjdhu	355	10	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	355	11	based	base	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	355	12	projec-	projec-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	355	13	tions	tion	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	355	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	356	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	356	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	356	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	356	4	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	356	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	356	6	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	356	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	357	1	Arthur	Arthur	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	2	P.	P.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	3	Dempster	Dempster	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	357	5	Nan	Nan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	6	M.	M.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	7	Laird	Laird	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	357	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	357	10	Donald	Donald	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	11	B.	B.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	12	Rubin	Rubin	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	357	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	358	1	1977	1977	CD
work_wgi74n6oavhdxhlomowlxqjdhu	358	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	359	1	Maximum	maximum	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	359	2	likelihood	likelihood	NN
work_wgi74n6oavhdxhlomowlxqjdhu	359	3	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	359	4	incomplete	incomplete	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	359	5	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	359	6	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	359	7	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	359	8	EM	EM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	359	9	algorithm	algorithm	NN
work_wgi74n6oavhdxhlomowlxqjdhu	359	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	360	1	Journal	Journal	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	360	2	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	360	3	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	360	4	Royal	Royal	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	360	5	Statistical	Statistical	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	360	6	Society	Society	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	360	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	360	8	Series	Series	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	360	9	B	B	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	360	10	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	360	11	39	39	CD
work_wgi74n6oavhdxhlomowlxqjdhu	360	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	361	1	John	John	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	361	2	DeNero	DeNero	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	361	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	361	4	Klaus	Klaus	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	361	5	Macherey	Macherey	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	361	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	362	1	2011	2011	CD
work_wgi74n6oavhdxhlomowlxqjdhu	362	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	363	1	Model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	363	2	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	363	3	based	base	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	363	4	aligner	aligner	NN
work_wgi74n6oavhdxhlomowlxqjdhu	363	5	combination	combination	NN
work_wgi74n6oavhdxhlomowlxqjdhu	363	6	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	363	7	dual	dual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	363	8	decomposition	decomposition	NN
work_wgi74n6oavhdxhlomowlxqjdhu	363	9	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	364	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	364	2	Pro-	Pro-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	364	3	ceedings	ceeding	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	364	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	364	5	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	364	6	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	364	7	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	364	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	365	1	Brad	Brad	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	365	2	Efron	Efron	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	365	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	365	4	Robert	Robert	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	365	5	J.	J.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	365	6	Tibshirani	Tibshirani	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	365	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	366	1	1993	1993	CD
work_wgi74n6oavhdxhlomowlxqjdhu	366	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	367	1	An	an	DT
work_wgi74n6oavhdxhlomowlxqjdhu	367	2	Introduc-	Introduc-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	367	3	tion	tion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	367	4	to	to	IN
work_wgi74n6oavhdxhlomowlxqjdhu	367	5	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	367	6	Bootstrap	Bootstrap	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	367	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	368	1	Chapman	Chapman	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	368	2	&	&	CC
work_wgi74n6oavhdxhlomowlxqjdhu	368	3	Hall	Hall	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	368	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	368	5	New	New	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	368	6	York	York	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	368	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	368	8	NY	NY	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	368	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	368	10	USA	USA	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	368	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	369	1	Victoria	Victoria	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	369	2	Fossum	Fossum	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	369	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	369	4	Steven	Steven	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	369	5	Abney	Abney	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	369	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	370	1	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	370	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	371	1	Automatically	automatically	RB
work_wgi74n6oavhdxhlomowlxqjdhu	371	2	inducing	induce	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	371	3	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	371	4	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	371	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	371	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	371	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	371	8	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	371	9	tagger	tagger	NN
work_wgi74n6oavhdxhlomowlxqjdhu	371	10	by	by	IN
work_wgi74n6oavhdxhlomowlxqjdhu	371	11	projecting	project	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	371	12	from	from	IN
work_wgi74n6oavhdxhlomowlxqjdhu	371	13	multiple	multiple	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	371	14	source	source	NN
work_wgi74n6oavhdxhlomowlxqjdhu	371	15	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	371	16	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	371	17	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	371	18	corpora	corpora	NN
work_wgi74n6oavhdxhlomowlxqjdhu	371	19	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	372	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	372	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	372	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	372	4	IJCNLP	IJCNLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	372	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	373	1	Dan	Dan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	373	2	Garrette	Garrette	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	373	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	373	4	Jason	Jason	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	373	5	Baldridge	Baldridge	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	373	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	374	1	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	374	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	375	1	Type	type	NN
work_wgi74n6oavhdxhlomowlxqjdhu	375	2	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	375	3	supervised	supervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	375	4	hidden	hidden	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	375	5	markov	markov	NN
work_wgi74n6oavhdxhlomowlxqjdhu	375	6	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	375	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	375	8	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	375	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	375	10	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	375	11	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	375	12	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	375	13	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	375	14	with	with	IN
work_wgi74n6oavhdxhlomowlxqjdhu	375	15	incomplete	incomplete	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	375	16	tag	tag	NN
work_wgi74n6oavhdxhlomowlxqjdhu	375	17	dictionaries	dictionary	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	375	18	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	376	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	376	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	376	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	376	4	EMNLP-	EMNLP-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	376	5	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	376	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	377	1	Yoav	Yoav	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	377	2	Goldberg	Goldberg	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	377	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	377	4	Meni	Meni	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	377	5	Adler	Adler	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	377	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	377	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	377	8	Michael	Michael	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	377	9	Elhadad	Elhadad	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	377	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	378	1	2008	2008	CD
work_wgi74n6oavhdxhlomowlxqjdhu	378	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	379	1	EM	EM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	379	2	can	can	MD
work_wgi74n6oavhdxhlomowlxqjdhu	379	3	find	find	VB
work_wgi74n6oavhdxhlomowlxqjdhu	379	4	pretty	pretty	RB
work_wgi74n6oavhdxhlomowlxqjdhu	379	5	good	good	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	379	6	HMM	HMM	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	379	7	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	379	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	379	9	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	379	10	(	(	-LRB-
work_wgi74n6oavhdxhlomowlxqjdhu	379	11	when	when	WRB
work_wgi74n6oavhdxhlomowlxqjdhu	379	12	given	give	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	379	13	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	379	14	good	good	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	379	15	start	start	NN
work_wgi74n6oavhdxhlomowlxqjdhu	379	16	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	379	17	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	380	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	380	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	380	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	380	4	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	380	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	380	6	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	380	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	381	1	Philipp	Philipp	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	381	2	Koehn	Koehn	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	381	3	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	382	1	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	382	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	383	1	Europarl	europarl	NN
work_wgi74n6oavhdxhlomowlxqjdhu	383	2	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	383	3	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	383	4	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	383	5	corpus	corpus	NN
work_wgi74n6oavhdxhlomowlxqjdhu	383	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	383	7	statistical	statistical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	383	8	machine	machine	NN
work_wgi74n6oavhdxhlomowlxqjdhu	383	9	translation	translation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	383	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	384	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	384	2	MT	MT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	384	3	Summit	Summit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	384	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	385	1	John	John	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	2	D.	D.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	3	Lafferty	Lafferty	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	385	5	Andrew	Andrew	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	6	McCallum	McCallum	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	7	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	385	8	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	385	9	Fernando	Fernando	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	10	C.	C.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	11	N.	N.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	12	Pereira	Pereira	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	385	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	386	1	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	386	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	387	1	Conditional	conditional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	387	2	random	random	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	387	3	fields	field	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	387	4	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	387	5	Probabilistic	probabilistic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	387	6	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	387	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	387	8	segmenting	segmenting	NN
work_wgi74n6oavhdxhlomowlxqjdhu	387	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	387	10	labeling	label	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	387	11	sequence	sequence	NN
work_wgi74n6oavhdxhlomowlxqjdhu	387	12	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	387	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	388	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	388	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	388	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	388	4	ICML	ICML	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	388	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	389	1	Shen	Shen	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	389	2	Li	Li	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	389	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	389	4	João	João	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	389	5	Graça	Graça	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	389	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	389	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	389	8	Ben	Ben	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	389	9	Taskar	Taskar	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	389	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	390	1	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	390	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	391	1	Wiki	wiki	VB
work_wgi74n6oavhdxhlomowlxqjdhu	391	2	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	391	3	ly	ly	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	391	4	supervised	supervise	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	391	5	part	part	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	391	6	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	391	7	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	391	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	391	9	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	391	10	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	391	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	392	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	392	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	392	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	392	4	EMNLP	EMNLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	392	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	392	6	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	392	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	393	1	Dong	Dong	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	393	2	C.	C.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	393	3	Liu	Liu	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	393	4	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	393	5	Jorge	Jorge	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	393	6	Nocedal	Nocedal	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	393	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	394	1	1989	1989	CD
work_wgi74n6oavhdxhlomowlxqjdhu	394	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	395	1	On	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	395	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	395	3	limited	limited	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	395	4	memory	memory	NN
work_wgi74n6oavhdxhlomowlxqjdhu	395	5	BFGS	BFGS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	395	6	method	method	NN
work_wgi74n6oavhdxhlomowlxqjdhu	395	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	395	8	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	395	9	scale	scale	NN
work_wgi74n6oavhdxhlomowlxqjdhu	395	10	optimization	optimization	NN
work_wgi74n6oavhdxhlomowlxqjdhu	395	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	396	1	Mathematical	Mathematical	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	396	2	Programming	Programming	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	396	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	396	4	45	45	CD
work_wgi74n6oavhdxhlomowlxqjdhu	396	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	397	1	Mitchell	Mitchell	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	2	P.	P.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	3	Marcus	Marcus	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	397	5	Mary	Mary	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	6	Ann	Ann	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	7	Marcinkiewicz	Marcinkiewicz	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	397	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	397	10	Beat-	Beat-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	11	rice	rice	NN
work_wgi74n6oavhdxhlomowlxqjdhu	397	12	Santorini	Santorini	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	397	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	398	1	1993	1993	CD
work_wgi74n6oavhdxhlomowlxqjdhu	398	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	399	1	Building	build	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	399	2	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	399	3	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	399	4	annotated	annotate	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	399	5	corpus	corpus	NN
work_wgi74n6oavhdxhlomowlxqjdhu	399	6	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	399	7	English	English	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	399	8	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	399	9	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	399	10	Penn	Penn	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	399	11	treebank	treebank	NN
work_wgi74n6oavhdxhlomowlxqjdhu	399	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	400	1	Computational	Computational	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	400	2	Linguis-	Linguis-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	400	3	tics	tic	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	400	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	400	5	19(2	19(2	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	400	6	)	)	-RRB-
work_wgi74n6oavhdxhlomowlxqjdhu	400	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	401	1	Tahira	Tahira	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	2	Naseem	Naseem	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	401	4	Benjamin	Benjamin	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	5	Snyder	Snyder	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	401	7	Jacob	Jacob	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	8	Eisenstein	Eisenstein	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	401	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	401	11	Regina	Regina	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	12	Barzilay	Barzilay	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	401	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	402	1	2009	2009	CD
work_wgi74n6oavhdxhlomowlxqjdhu	402	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	403	1	Multilingual	multilingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	403	2	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	403	3	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	403	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	403	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	403	6	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	403	7	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	403	8	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	403	9	Two	two	CD
work_wgi74n6oavhdxhlomowlxqjdhu	403	10	unsupervised	unsupervise	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	403	11	approaches	approach	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	403	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	404	1	JAIR	JAIR	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	404	2	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	404	3	36	36	CD
work_wgi74n6oavhdxhlomowlxqjdhu	404	4	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	405	1	Joakim	Joakim	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	2	Nivre	Nivre	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	405	4	Johan	Johan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	5	Hall	Hall	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	405	7	Sandra	Sandra	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	8	Kübler	Kübler	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	405	10	Ryan	Ryan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	11	McDon-	McDon-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	12	ald	ald	NN
work_wgi74n6oavhdxhlomowlxqjdhu	405	13	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	405	14	Jens	Jens	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	15	Nilsson	Nilsson	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	16	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	405	17	Sebastian	Sebastian	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	18	Riedel	Riedel	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	405	20	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	405	21	Deniz	Deniz	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	22	Yuret	Yuret	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	405	23	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	406	1	2007	2007	CD
work_wgi74n6oavhdxhlomowlxqjdhu	406	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	407	1	The	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	407	2	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	407	3	2007	2007	CD
work_wgi74n6oavhdxhlomowlxqjdhu	407	4	shared	share	VBD
work_wgi74n6oavhdxhlomowlxqjdhu	407	5	task	task	NN
work_wgi74n6oavhdxhlomowlxqjdhu	407	6	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	407	7	dependency	dependency	NN
work_wgi74n6oavhdxhlomowlxqjdhu	407	8	parsing	parsing	NN
work_wgi74n6oavhdxhlomowlxqjdhu	407	9	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	408	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	408	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	408	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	408	4	EMNLP	EMNLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	408	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	408	6	CoNLL	CoNLL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	408	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	409	1	Slav	Slav	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	409	2	Petrov	Petrov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	409	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	409	4	Dipanjan	Dipanjan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	409	5	Das	Das	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	409	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	409	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	409	8	Ryan	Ryan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	409	9	McDonald	McDonald	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	409	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	410	1	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	410	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	411	1	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	411	2	universal	universal	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	411	3	part	part	NN
work_wgi74n6oavhdxhlomowlxqjdhu	411	4	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	411	5	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	411	6	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	411	7	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	411	8	tagset	tagset	NN
work_wgi74n6oavhdxhlomowlxqjdhu	411	9	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	412	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	412	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	412	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	412	4	LREC	LREC	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	412	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	413	1	Sujith	Sujith	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	413	2	Ravi	Ravi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	413	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	413	4	Kevin	Kevin	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	413	5	Knight	Knight	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	413	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	414	1	2009	2009	CD
work_wgi74n6oavhdxhlomowlxqjdhu	414	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	415	1	Minimized	minimize	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	415	2	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	415	3	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	415	4	unsupervised	unsupervised	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	415	5	part	part	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	415	6	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	415	7	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	415	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	415	9	speech	speech	NN
work_wgi74n6oavhdxhlomowlxqjdhu	415	10	tagging	tagging	NN
work_wgi74n6oavhdxhlomowlxqjdhu	415	11	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	416	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	416	2	Proceed-	Proceed-	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	416	3	ings	ing	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	416	4	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	416	5	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	416	6	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	416	7	IJCNLP	IJCNLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	416	8	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	417	1	11	11	CD
work_wgi74n6oavhdxhlomowlxqjdhu	417	2	Stefan	Stefan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	3	Riezler	Riezler	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	4	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	417	5	Tracy	Tracy	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	6	H.	H.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	7	King	King	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	417	9	Ronald	Ronald	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	10	M.	M.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	11	Kaplan	Kaplan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	12	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	417	13	Richard	Richard	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	14	Crouch	Crouch	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	15	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	417	16	John	John	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	17	T.	T.	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	18	Maxwell	Maxwell	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	19	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	417	20	III	III	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	21	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	417	22	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	417	23	Mark	Mark	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	24	Johnson	Johnson	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	417	25	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	418	1	2002	2002	CD
work_wgi74n6oavhdxhlomowlxqjdhu	418	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	419	1	Parsing	parse	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	419	2	the	the	DT
work_wgi74n6oavhdxhlomowlxqjdhu	419	3	wall	wall	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	419	4	street	street	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	419	5	journal	journal	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	419	6	using	use	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	419	7	a	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	419	8	lexical	lexical	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	419	9	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	419	10	functional	functional	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	419	11	grammar	grammar	NN
work_wgi74n6oavhdxhlomowlxqjdhu	419	12	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	419	13	discriminative	discriminative	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	419	14	estimation	estimation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	419	15	techniques	technique	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	419	16	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	420	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	420	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	420	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	420	4	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	420	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	421	1	Noah	Noah	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	421	2	Smith	Smith	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	421	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	421	4	Jason	Jason	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	421	5	Eisner	Eisner	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	421	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	422	1	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	422	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	423	1	Contrastive	contrastive	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	423	2	estima-	estima-	NN
work_wgi74n6oavhdxhlomowlxqjdhu	423	3	tion	tion	NN
work_wgi74n6oavhdxhlomowlxqjdhu	423	4	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	423	5	Training	train	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	423	6	log	log	NN
work_wgi74n6oavhdxhlomowlxqjdhu	423	7	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	423	8	linear	linear	NN
work_wgi74n6oavhdxhlomowlxqjdhu	423	9	models	model	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	423	10	on	on	IN
work_wgi74n6oavhdxhlomowlxqjdhu	423	11	unlabeled	unlabeled	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	423	12	data	datum	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	423	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	424	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	424	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	424	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	424	4	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	424	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	425	1	Oscar	Oscar	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	425	2	Täckström	Täckström	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	425	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	425	4	Ryan	Ryan	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	425	5	McDonald	McDonald	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	425	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	425	7	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	425	8	Jakob	Jakob	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	425	9	Uszkoreit	Uszkoreit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	425	10	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	426	1	2012	2012	CD
work_wgi74n6oavhdxhlomowlxqjdhu	426	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	427	1	Cross	cross	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	427	2	-	-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	427	3	lingual	lingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	427	4	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	427	5	clusters	cluster	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	427	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	427	7	direct	direct	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	427	8	transfer	transfer	NN
work_wgi74n6oavhdxhlomowlxqjdhu	427	9	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	427	10	linguistic	linguistic	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	427	11	structure	structure	NN
work_wgi74n6oavhdxhlomowlxqjdhu	427	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	428	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	428	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	428	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	428	4	NAACL	NAACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	428	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	428	6	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	428	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	429	1	Joseph	Joseph	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	2	Turian	Turian	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	429	4	Lev	Lev	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	429	6	Arie	Arie	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	7	Ratinov	Ratinov	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	8	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	429	9	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	429	10	Yoshua	Yoshua	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	11	Bengio	Bengio	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	429	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	430	1	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	430	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	431	1	Word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	431	2	representations	representation	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	431	3	:	:	:
work_wgi74n6oavhdxhlomowlxqjdhu	431	4	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	431	5	simple	simple	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	431	6	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	431	7	general	general	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	431	8	method	method	NN
work_wgi74n6oavhdxhlomowlxqjdhu	431	9	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	431	10	semi	semi	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	431	11	-	-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	431	12	supervised	supervised	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	431	13	learning	learning	NN
work_wgi74n6oavhdxhlomowlxqjdhu	431	14	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	432	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	432	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	432	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	432	4	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	432	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	433	1	UN	UN	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	433	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	434	1	2006	2006	CD
work_wgi74n6oavhdxhlomowlxqjdhu	434	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	435	1	ODS	ODS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	435	2	UN	UN	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	435	3	parallel	parallel	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	435	4	corpus	corpus	NN
work_wgi74n6oavhdxhlomowlxqjdhu	435	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	436	1	Jakob	Jakob	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	436	2	Uszkoreit	Uszkoreit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	436	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	436	4	Thorsten	Thorsten	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	436	5	Brants	Brants	NNPS
work_wgi74n6oavhdxhlomowlxqjdhu	436	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	437	1	2008	2008	CD
work_wgi74n6oavhdxhlomowlxqjdhu	437	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	438	1	Distributed	distribute	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	438	2	word	word	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	3	clustering	cluster	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	438	4	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	438	5	large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	438	6	scale	scale	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	7	class	class	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	8	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	438	9	based	base	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	438	10	language	language	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	11	modeling	modeling	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	12	in	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	438	13	machine	machine	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	14	translation	translation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	438	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	439	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	439	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	439	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	439	4	ACL	ACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	439	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	439	6	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	439	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	440	1	Jakob	Jakob	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	2	Uszkoreit	Uszkoreit	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	3	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	440	4	Jay	Jay	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	5	Ponte	Ponte	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	6	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	440	7	Ashok	Ashok	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	8	Popat	Popat	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	9	,	,	,
work_wgi74n6oavhdxhlomowlxqjdhu	440	10	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	440	11	Moshe	Moshe	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	12	Dubiner	Dubiner	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	440	13	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	441	1	2010	2010	CD
work_wgi74n6oavhdxhlomowlxqjdhu	441	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	442	1	Large	large	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	442	2	scale	scale	NN
work_wgi74n6oavhdxhlomowlxqjdhu	442	3	parallel	parallel	NN
work_wgi74n6oavhdxhlomowlxqjdhu	442	4	document	document	NN
work_wgi74n6oavhdxhlomowlxqjdhu	442	5	mining	mining	NN
work_wgi74n6oavhdxhlomowlxqjdhu	442	6	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	442	7	machine	machine	NN
work_wgi74n6oavhdxhlomowlxqjdhu	442	8	translation	translation	NN
work_wgi74n6oavhdxhlomowlxqjdhu	442	9	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	443	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	443	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	443	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	443	4	COLING	COLING	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	443	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	444	1	Chenhai	Chenhai	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	444	2	Xi	Xi	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	444	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	444	4	Rebecca	Rebecca	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	444	5	Hwa	Hwa	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	444	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	445	1	2005	2005	CD
work_wgi74n6oavhdxhlomowlxqjdhu	445	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	446	1	A	a	DT
work_wgi74n6oavhdxhlomowlxqjdhu	446	2	backoff	backoff	NN
work_wgi74n6oavhdxhlomowlxqjdhu	446	3	model	model	NN
work_wgi74n6oavhdxhlomowlxqjdhu	446	4	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	446	5	bootstrapping	bootstrappe	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	446	6	resources	resource	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	446	7	for	for	IN
work_wgi74n6oavhdxhlomowlxqjdhu	446	8	non	non	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	446	9	-	-	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	446	10	English	english	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	446	11	languages	language	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	446	12	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	447	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	447	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	447	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	447	4	HLT	HLT	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	447	5	-	-	HYPH
work_wgi74n6oavhdxhlomowlxqjdhu	447	6	EMNLP	EMNLP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	447	7	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	448	1	David	David	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	448	2	Yarowsky	Yarowsky	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	448	3	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	448	4	Grace	Grace	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	448	5	Ngai	Ngai	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	448	6	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	449	1	2001	2001	CD
work_wgi74n6oavhdxhlomowlxqjdhu	449	2	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	450	1	Inducing	induce	VBG
work_wgi74n6oavhdxhlomowlxqjdhu	450	2	mul-	mul-	RP
work_wgi74n6oavhdxhlomowlxqjdhu	450	3	tilingual	tilingual	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	450	4	POS	POS	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	450	5	taggers	tagger	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	450	6	and	and	CC
work_wgi74n6oavhdxhlomowlxqjdhu	450	7	NP	NP	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	450	8	bracketers	bracketer	NNS
work_wgi74n6oavhdxhlomowlxqjdhu	450	9	via	via	IN
work_wgi74n6oavhdxhlomowlxqjdhu	450	10	robust	robust	JJ
work_wgi74n6oavhdxhlomowlxqjdhu	450	11	projection	projection	NN
work_wgi74n6oavhdxhlomowlxqjdhu	450	12	across	across	IN
work_wgi74n6oavhdxhlomowlxqjdhu	450	13	aligned	align	VBN
work_wgi74n6oavhdxhlomowlxqjdhu	450	14	corpora	corpora	NN
work_wgi74n6oavhdxhlomowlxqjdhu	450	15	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	451	1	In	in	IN
work_wgi74n6oavhdxhlomowlxqjdhu	451	2	Proceedings	Proceedings	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	451	3	of	of	IN
work_wgi74n6oavhdxhlomowlxqjdhu	451	4	NAACL	NAACL	NNP
work_wgi74n6oavhdxhlomowlxqjdhu	451	5	.	.	.
work_wgi74n6oavhdxhlomowlxqjdhu	452	1	12	12	CD