id	sid	tid	token	lemma	pos
work_zhl5dw75nfhy7oputsllvjjcqq	1	1	Evaluating	evaluate	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	1	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	1	3	Stability	Stability	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	1	5	Embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	1	6	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	1	7	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	1	8	Word	Word	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	9	Similarities	Similarities	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	1	10	Maria	Maria	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	11	Antoniak	Antoniak	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	12	Cornell	Cornell	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	13	University	University	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	14	maa343@cornell.edu	maa343@cornell.edu	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	1	15	David	David	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	16	Mimno	Mimno	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	17	Cornell	Cornell	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	18	University	University	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	19	mimno@cornell.edu	mimno@cornell.edu	NN
work_zhl5dw75nfhy7oputsllvjjcqq	1	20	Abstract	Abstract	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	21	Word	Word	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	1	22	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	1	23	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	1	24	increasingly	increasingly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	1	25	being	be	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	1	26	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	1	27	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	1	28	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	1	29	tool	tool	NN
work_zhl5dw75nfhy7oputsllvjjcqq	1	30	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	1	31	study	study	VB
work_zhl5dw75nfhy7oputsllvjjcqq	1	32	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	1	33	associations	association	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	1	34	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	1	35	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	1	36	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	1	37	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	2	1	However	however	RB
work_zhl5dw75nfhy7oputsllvjjcqq	2	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	2	3	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	2	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	2	5	unclear	unclear	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	2	6	whether	whether	IN
work_zhl5dw75nfhy7oputsllvjjcqq	2	7	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	2	8	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	2	9	reflect	reflect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	2	10	enduring	endure	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	2	11	properties	property	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	2	12	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	2	13	lan-	lan-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	2	14	guage	guage	NN
work_zhl5dw75nfhy7oputsllvjjcqq	2	15	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	2	16	if	if	IN
work_zhl5dw75nfhy7oputsllvjjcqq	2	17	they	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	2	18	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	2	19	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	2	20	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	2	21	inconsequential	inconsequential	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	2	22	variations	variation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	2	23	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	2	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	2	25	source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	2	26	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	2	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	3	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	3	2	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	3	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	3	4	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	3	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	3	6	neighbor	neighbor	NN
work_zhl5dw75nfhy7oputsllvjjcqq	3	7	distances	distance	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	3	8	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	3	9	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	3	10	sen-	sen-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	3	11	sitive	sitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	3	12	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	3	13	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	3	14	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	3	15	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	3	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	3	17	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	3	18	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	3	19	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	3	20	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	3	21	variety	variety	NN
work_zhl5dw75nfhy7oputsllvjjcqq	3	22	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	3	23	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	3	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	4	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	4	2	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	4	3	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	4	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	4	5	including	include	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	4	6	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	4	7	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	4	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	4	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	4	10	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	4	11	set	set	NN
work_zhl5dw75nfhy7oputsllvjjcqq	4	12	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	4	13	result	result	VB
work_zhl5dw75nfhy7oputsllvjjcqq	4	14	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	4	15	substantial	substantial	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	4	16	variations	variation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	4	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	5	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	5	2	show	show	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	5	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	5	4	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	5	5	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	5	6	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	5	7	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	5	8	prominent	prominent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	5	9	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	5	10	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	5	11	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	5	12	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	5	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	6	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	6	2	recommend	recommend	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	6	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	6	4	users	user	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	6	5	never	never	RB
work_zhl5dw75nfhy7oputsllvjjcqq	6	6	rely	rely	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	6	7	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	6	8	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	6	9	embedding	embedding	NN
work_zhl5dw75nfhy7oputsllvjjcqq	6	10	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	6	11	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	6	12	distance	distance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	6	13	calculations	calculation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	6	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	6	15	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	6	16	rather	rather	RB
work_zhl5dw75nfhy7oputsllvjjcqq	6	17	average	average	VB
work_zhl5dw75nfhy7oputsllvjjcqq	6	18	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	6	19	multiple	multiple	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	6	20	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	6	21	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	6	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	6	23	especially	especially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	6	24	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	6	25	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	6	26	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	6	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	7	1	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	7	2	Introduction	introduction	NN
work_zhl5dw75nfhy7oputsllvjjcqq	7	3	Word	Word	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	7	4	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	7	5	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	7	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	7	7	popular	popular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	7	8	technique	technique	NN
work_zhl5dw75nfhy7oputsllvjjcqq	7	9	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	7	10	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	7	11	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	7	12	processing	processing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	7	13	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	7	14	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	7	15	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	7	16	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	7	17	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	7	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	7	19	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	7	20	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	7	21	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	7	22	vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	7	23	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	7	24	mapped	map	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	7	25	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	7	26	low	low	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	7	27	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	7	28	dimensional	dimensional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	7	29	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	7	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	8	1	Embedding	embedding	NN
work_zhl5dw75nfhy7oputsllvjjcqq	8	2	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	3	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	8	4	easily	easily	RB
work_zhl5dw75nfhy7oputsllvjjcqq	8	5	trained	train	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	8	6	—	—	:
work_zhl5dw75nfhy7oputsllvjjcqq	8	7	several	several	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	8	8	imple-	imple-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	8	9	mentations	mentation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	10	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	8	11	publicly	publicly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	8	12	available	available	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	8	13	—	—	:
work_zhl5dw75nfhy7oputsllvjjcqq	8	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	8	15	relationships	relationship	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	16	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	8	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	8	18	embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	8	19	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	8	21	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	8	22	measured	measure	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	8	23	via	via	IN
work_zhl5dw75nfhy7oputsllvjjcqq	8	24	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	8	25	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	8	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	8	27	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	8	28	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	8	29	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	8	30	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	8	31	reveal	reveal	VB
work_zhl5dw75nfhy7oputsllvjjcqq	8	32	latent	latent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	8	33	seman-	seman-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	8	34	tic	tic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	8	35	relationships	relationship	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	36	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	8	37	pairs	pair	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	38	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	8	39	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	8	40	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	9	1	Word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	9	2	em-	em-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	9	3	beddings	bedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	9	4	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	9	5	increasingly	increasingly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	9	6	being	be	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	9	7	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	9	8	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	9	9	researchers	researcher	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	9	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	9	11	unexpected	unexpected	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	9	12	ways	way	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	9	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	9	14	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	9	15	become	become	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	9	16	popular	popular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	9	17	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	9	18	fields	field	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	9	19	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	9	20	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	9	21	digital	digital	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	9	22	humanities	humanity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	9	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	9	24	computational	computational	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	9	25	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	9	26	science	science	NN
work_zhl5dw75nfhy7oputsllvjjcqq	9	27	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	9	28	Hamilton	Hamilton	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	29	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	30	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	31	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	9	33	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	9	34	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	9	35	Heuser	Heuser	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	36	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	9	37	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	9	38	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	9	39	Phillips	Phillips	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	40	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	41	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	42	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	9	43	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	9	44	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	9	45	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	9	46	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	10	1	Embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	10	2	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	10	3	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	10	4	analyses	analysis	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	10	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	10	6	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	10	7	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	10	8	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	10	9	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	10	10	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	10	11	robust	robust	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	10	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	10	13	valuable	valuable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	10	14	tool	tool	NN
work_zhl5dw75nfhy7oputsllvjjcqq	10	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	10	16	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	10	17	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	10	18	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	10	19	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	10	20	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	10	21	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	10	22	dramatically	dramatically	RB
work_zhl5dw75nfhy7oputsllvjjcqq	10	23	under	under	IN
work_zhl5dw75nfhy7oputsllvjjcqq	10	24	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	10	25	represent	represent	VB
work_zhl5dw75nfhy7oputsllvjjcqq	10	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	10	27	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	10	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	10	29	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	10	30	measurements	measurement	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	10	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	11	1	Embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	11	2	algo-	algo-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	11	3	rithms	rithms	NN
work_zhl5dw75nfhy7oputsllvjjcqq	11	4	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	11	5	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	11	6	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	11	7	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	11	8	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	9	they	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	11	10	appear	appear	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	11	11	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	12	factors	factor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	11	13	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	11	14	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	11	16	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	11	17	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	18	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	11	19	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	11	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	11	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	11	22	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	11	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	11	25	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	11	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	11	27	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	11	28	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	11	29	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	30	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	11	31	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	11	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	11	33	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	11	34	even	even	RB
work_zhl5dw75nfhy7oputsllvjjcqq	11	35	seeds	seed	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	11	36	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	11	37	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	11	38	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	11	39	generators	generator	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	11	40	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	12	1	If	if	IN
work_zhl5dw75nfhy7oputsllvjjcqq	12	2	users	user	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	12	3	do	do	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	12	4	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	12	5	account	account	VB
work_zhl5dw75nfhy7oputsllvjjcqq	12	6	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	12	7	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	12	8	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	12	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	12	10	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	12	11	conclusions	conclusion	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	12	12	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	12	13	likely	likely	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	12	14	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	12	15	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	12	16	invalid	invalid	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	12	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	13	1	Fortunately	fortunately	RB
work_zhl5dw75nfhy7oputsllvjjcqq	13	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	13	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	13	4	also	also	RB
work_zhl5dw75nfhy7oputsllvjjcqq	13	5	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	13	6	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	13	7	simply	simply	RB
work_zhl5dw75nfhy7oputsllvjjcqq	13	8	averaging	average	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	13	9	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	13	10	multiple	multiple	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	13	11	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	13	12	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	13	13	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	13	14	sufficient	sufficient	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	13	15	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	13	16	produce	produce	VB
work_zhl5dw75nfhy7oputsllvjjcqq	13	17	stable	stable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	13	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	13	19	reliable	reliable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	13	20	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	13	21	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	13	22	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	13	23	cases	case	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	13	24	tested	test	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	13	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	14	1	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	14	2	research	research	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	3	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	14	4	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	5	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	14	6	has	have	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	14	7	so	so	RB
work_zhl5dw75nfhy7oputsllvjjcqq	14	8	far	far	RB
work_zhl5dw75nfhy7oputsllvjjcqq	14	9	fo-	fo-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	14	10	cused	cuse	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	14	11	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	14	12	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	14	13	downstream	downstream	RB
work_zhl5dw75nfhy7oputsllvjjcqq	14	14	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	14	15	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	14	16	use	use	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	17	case	case	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	14	19	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	14	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	14	21	end	end	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	22	goal	goal	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	23	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	14	24	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	14	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	14	26	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	14	27	themselves	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	14	28	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	14	29	performance	performance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	30	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	14	31	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	14	32	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	14	33	complicated	complicated	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	14	34	task	task	NN
work_zhl5dw75nfhy7oputsllvjjcqq	14	35	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	15	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	15	2	exam-	exam-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	15	3	ple	ple	NN
work_zhl5dw75nfhy7oputsllvjjcqq	15	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	15	5	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	15	6	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	15	7	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	15	8	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	15	9	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	15	10	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	15	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	15	12	bottom	bottom	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	15	13	layer	layer	NN
work_zhl5dw75nfhy7oputsllvjjcqq	15	14	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	15	15	neural	neural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	15	16	network	network	NN
work_zhl5dw75nfhy7oputsllvjjcqq	15	17	architectures	architecture	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	15	18	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	15	19	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	20	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	15	21	Ben-	Ben-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	22	gio	gio	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	23	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	24	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	25	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	15	27	2003	2003	CD
work_zhl5dw75nfhy7oputsllvjjcqq	15	28	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	15	29	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	15	30	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	15	31	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	15	32	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	15	33	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	16	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	16	2	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	16	3	’	’	POS
work_zhl5dw75nfhy7oputsllvjjcqq	16	4	training	train	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	16	5	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	16	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	16	7	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	16	8	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	16	9	selected	select	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	16	10	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	16	11	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	16	12	as	as	RB
work_zhl5dw75nfhy7oputsllvjjcqq	16	13	large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	16	14	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	16	15	possible	possible	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	16	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	16	17	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	16	18	only	only	RB
work_zhl5dw75nfhy7oputsllvjjcqq	16	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	16	20	interest	interest	NN
work_zhl5dw75nfhy7oputsllvjjcqq	16	21	insofar	insofar	RB
work_zhl5dw75nfhy7oputsllvjjcqq	16	22	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	16	23	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	16	24	generalizes	generalize	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	16	25	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	16	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	16	27	downstream	downstream	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	16	28	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	16	29	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	16	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	17	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	17	2	contrast	contrast	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	17	4	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	17	5	researchers	researcher	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	17	6	take	take	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	17	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	17	8	corpus-	corpus-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	9	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	17	10	approach	approach	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	17	12	use	use	VB
work_zhl5dw75nfhy7oputsllvjjcqq	17	13	relationships	relationship	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	17	14	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	17	15	em-	em-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	17	16	beddings	bedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	17	17	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	17	18	direct	direct	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	17	19	evidence	evidence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	20	about	about	IN
work_zhl5dw75nfhy7oputsllvjjcqq	17	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	17	22	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	17	24	culture	culture	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	25	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	17	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	17	27	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	17	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	17	29	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	17	30	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	31	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	17	32	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	17	33	Bolukbasi	Bolukbasi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	34	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	17	35	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	36	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	37	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	17	38	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	17	39	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	17	40	Hamilton	Hamilton	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	41	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	42	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	43	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	44	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	17	45	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	17	46	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	17	47	Heuser	Heuser	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	17	48	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	17	49	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	17	50	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	17	51	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	18	1	Embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	18	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	18	3	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	18	4	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	18	5	if	if	IN
work_zhl5dw75nfhy7oputsllvjjcqq	18	6	they	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	18	7	were	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	18	8	simulations	simulation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	18	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	18	10	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	18	11	survey	survey	NN
work_zhl5dw75nfhy7oputsllvjjcqq	18	12	asking	ask	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	18	13	subjects	subject	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	18	14	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	18	15	free	free	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	18	16	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	18	17	associate	associate	NN
work_zhl5dw75nfhy7oputsllvjjcqq	18	18	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	18	19	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	18	20	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	18	21	terms	term	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	18	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	19	1	Unlike	unlike	IN
work_zhl5dw75nfhy7oputsllvjjcqq	19	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	19	3	downstream	downstream	RB
work_zhl5dw75nfhy7oputsllvjjcqq	19	4	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	19	5	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	19	6	approach	approach	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	19	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	19	9	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	10	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	19	11	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	19	12	approach	approach	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	13	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	19	14	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	19	15	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	19	16	direct	direct	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	19	17	human	human	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	19	18	analysis	analysis	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	19	20	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	19	21	neighbors	neighbor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	19	22	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	19	23	embed-	embed-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	19	24	ding	ding	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	19	25	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	19	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	19	27	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	19	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	19	29	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	30	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	19	31	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	19	32	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	19	33	simply	simply	RB
work_zhl5dw75nfhy7oputsllvjjcqq	19	34	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	19	35	off	off	IN
work_zhl5dw75nfhy7oputsllvjjcqq	19	36	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	19	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	19	38	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	19	39	shelf	shelf	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	40	convenience	convenience	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	41	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	19	42	rather	rather	RB
work_zhl5dw75nfhy7oputsllvjjcqq	19	43	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	19	44	central	central	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	19	45	object	object	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	46	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	19	47	study	study	NN
work_zhl5dw75nfhy7oputsllvjjcqq	19	48	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	20	1	107	107	CD
work_zhl5dw75nfhy7oputsllvjjcqq	20	2	Transactions	transaction	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	20	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	20	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	20	5	Association	Association	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	20	6	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	20	7	Computational	Computational	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	20	8	Linguistics	Linguistics	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	20	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	20	10	vol	vol	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	20	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	21	1	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	21	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	21	3	pp	pp	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	21	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	22	1	107–119	107–119	CD
work_zhl5dw75nfhy7oputsllvjjcqq	22	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	22	3	2018	2018	CD
work_zhl5dw75nfhy7oputsllvjjcqq	22	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	23	1	Action	Action	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	23	2	Editor	Editor	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	23	3	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	23	4	Ivan	Ivan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	23	5	Titov	Titov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	23	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	24	1	Submission	submission	NN
work_zhl5dw75nfhy7oputsllvjjcqq	24	2	batch	batch	NN
work_zhl5dw75nfhy7oputsllvjjcqq	24	3	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	24	4	6/2017	6/2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	24	5	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	24	6	Revision	revision	NN
work_zhl5dw75nfhy7oputsllvjjcqq	24	7	batch	batch	NN
work_zhl5dw75nfhy7oputsllvjjcqq	24	8	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	24	9	9/2017	9/2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	24	10	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	24	11	Published	publish	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	24	12	2/2018	2/2018	CD
work_zhl5dw75nfhy7oputsllvjjcqq	24	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	25	1	c	c	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	25	2	©	©	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	25	3	2018	2018	CD
work_zhl5dw75nfhy7oputsllvjjcqq	25	4	Association	Association	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	25	5	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	25	6	Computational	Computational	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	25	7	Linguistics	Linguistics	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	25	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	26	1	Distributed	distribute	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	26	2	under	under	IN
work_zhl5dw75nfhy7oputsllvjjcqq	26	3	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	26	4	CC	cc	NN
work_zhl5dw75nfhy7oputsllvjjcqq	26	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	26	6	BY	BY	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	26	7	4.0	4.0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	26	8	license	license	NN
work_zhl5dw75nfhy7oputsllvjjcqq	26	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	27	1	Downstream	downstream	RB
work_zhl5dw75nfhy7oputsllvjjcqq	27	2	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	27	3	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	4	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	27	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	27	6	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	7	Big	Big	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	27	8	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	9	Small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	10	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	27	12	difficult	difficult	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	13	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	27	14	impossi-	impossi-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	15	ble	ble	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	27	16	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	27	17	expand	expand	VB
work_zhl5dw75nfhy7oputsllvjjcqq	27	18	Source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	19	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	27	20	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	27	21	important	important	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	22	Source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	23	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	27	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	27	25	object	object	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	26	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	27	study	study	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	28	Only	only	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	29	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	30	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	27	31	important	important	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	32	Specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	33	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	27	34	fine	fine	RB
work_zhl5dw75nfhy7oputsllvjjcqq	27	35	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	27	36	grained	grain	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	37	comparisons	comparison	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	38	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	27	39	important	important	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	40	Embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	41	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	27	42	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	43	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	44	downstream	downstream	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	45	tasks	task	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	46	Embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	47	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	27	48	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	49	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	27	50	learn	learn	VB
work_zhl5dw75nfhy7oputsllvjjcqq	27	51	about	about	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	52	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	27	53	mental	mental	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	54	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	55	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	56	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	57	associa-	associa-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	27	58	tion	tion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	59	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	60	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	27	61	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	62	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	63	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	27	64	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	27	65	Table	Table	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	27	66	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	27	67	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	27	68	Comparison	comparison	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	69	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	27	70	downstream	downstream	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	71	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	27	72	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	73	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	27	74	corpus-	corpus-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	27	75	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	27	76	approaches	approach	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	77	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	27	78	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	27	79	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	27	80	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	28	1	While	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	28	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	28	3	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	28	4	may	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	28	5	appear	appear	VB
work_zhl5dw75nfhy7oputsllvjjcqq	28	6	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	28	7	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	28	8	properties	property	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	28	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	28	10	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	28	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	28	12	they	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	28	13	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	28	14	fact	fact	NN
work_zhl5dw75nfhy7oputsllvjjcqq	28	15	only	only	RB
work_zhl5dw75nfhy7oputsllvjjcqq	28	16	measure	measure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	28	17	properties	property	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	28	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	28	19	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	28	20	curated	curate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	28	21	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	28	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	28	23	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	28	24	could	could	MD
work_zhl5dw75nfhy7oputsllvjjcqq	28	25	suf-	suf-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	28	26	fer	fer	VB
work_zhl5dw75nfhy7oputsllvjjcqq	28	27	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	28	28	several	several	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	28	29	problems	problem	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	28	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	29	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	29	2	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	29	3	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	29	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	29	5	merely	merely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	29	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	29	7	sample	sample	NN
work_zhl5dw75nfhy7oputsllvjjcqq	29	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	29	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	29	10	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	29	11	’	’	POS
work_zhl5dw75nfhy7oputsllvjjcqq	29	12	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	29	13	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	29	14	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	29	15	Shazeer	Shazeer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	29	16	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	29	17	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	29	18	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	29	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	29	20	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	29	21	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	29	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	30	1	Sources	source	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	30	2	could	could	MD
work_zhl5dw75nfhy7oputsllvjjcqq	30	3	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	30	4	missing	miss	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	30	5	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	30	6	over	over	RB
work_zhl5dw75nfhy7oputsllvjjcqq	30	7	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	30	8	represented	represent	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	30	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	30	10	typos	typo	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	30	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	30	12	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	30	13	lexical	lexical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	30	14	variations	variation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	30	15	could	could	MD
work_zhl5dw75nfhy7oputsllvjjcqq	30	16	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	30	17	present	present	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	30	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	30	19	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	30	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	30	21	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	30	22	noted	note	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	30	23	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	30	24	Goodfellow	Goodfellow	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	30	25	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	30	26	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	30	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	31	1	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	31	2	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	31	3	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	31	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	31	5	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	31	6	Many	many	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	31	7	datasets	dataset	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	31	8	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	31	9	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	31	10	naturally	naturally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	31	11	arranged	arrange	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	31	12	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	31	13	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	31	14	way	way	NN
work_zhl5dw75nfhy7oputsllvjjcqq	31	15	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	31	16	successive	successive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	31	17	examples	example	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	31	18	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	31	19	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	31	20	cor-	cor-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	31	21	related	related	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	31	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	31	23	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	31	24	Furthermore	furthermore	RB
work_zhl5dw75nfhy7oputsllvjjcqq	31	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	31	26	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	31	27	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	31	28	vary	vary	VB
work_zhl5dw75nfhy7oputsllvjjcqq	31	29	consid-	consid-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	31	30	erably	erably	RB
work_zhl5dw75nfhy7oputsllvjjcqq	31	31	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	31	32	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	31	33	initializations	initialization	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	31	34	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	31	35	making	make	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	31	36	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	31	37	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	31	38	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	31	39	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	31	40	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	31	41	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	31	42	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	31	43	unstable	unstable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	31	44	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	32	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	32	2	hypothesize	hypothesize	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	32	3	that	that	DT
work_zhl5dw75nfhy7oputsllvjjcqq	32	4	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	32	5	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	32	6	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	32	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	32	8	poten-	poten-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	32	9	tially	tially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	32	10	idiosyncratic	idiosyncratic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	32	11	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	32	12	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	32	13	exacerbate	exacerbate	VB
work_zhl5dw75nfhy7oputsllvjjcqq	32	14	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	32	15	prob-	prob-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	32	16	lems	lem	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	32	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	32	18	lead	lead	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	32	19	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	32	20	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	32	21	variable	variable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	32	22	estimates	estimate	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	32	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	32	24	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	32	25	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	32	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	33	1	Such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	2	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	3	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	33	4	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	33	5	common	common	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	6	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	33	7	digital	digital	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	8	humanities	humanity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	33	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	33	10	computational	computational	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	11	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	12	science	science	NN
work_zhl5dw75nfhy7oputsllvjjcqq	33	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	33	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	33	15	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	33	16	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	33	17	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	33	18	impossible	impossible	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	33	19	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	33	20	mitigate	mitigate	VB
work_zhl5dw75nfhy7oputsllvjjcqq	33	21	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	33	22	problems	problem	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	33	23	simply	simply	RB
work_zhl5dw75nfhy7oputsllvjjcqq	33	24	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	33	25	expanding	expand	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	33	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	33	27	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	33	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	34	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	34	2	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	34	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	34	4	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	34	5	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	34	6	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	34	7	create	create	VB
work_zhl5dw75nfhy7oputsllvjjcqq	34	8	more	more	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	34	9	18th	18th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	34	10	Century	Century	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	34	11	English	english	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	34	12	books	book	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	34	13	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	34	14	change	change	VB
work_zhl5dw75nfhy7oputsllvjjcqq	34	15	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	34	16	topical	topical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	34	17	focus	focus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	34	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	35	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	35	2	explore	explore	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	35	3	causes	cause	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	35	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	5	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	35	6	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	35	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	35	8	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	35	9	range	range	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	35	10	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	35	12	fundamental	fundamental	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	35	13	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	35	14	nature	nature	NN
work_zhl5dw75nfhy7oputsllvjjcqq	35	15	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	16	certain	certain	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	35	17	al-	al-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	35	18	gorithms	gorithms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	35	19	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	20	more	more	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	35	21	troubling	troubling	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	35	22	sensitivities	sensitivity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	35	23	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	24	properties	property	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	35	25	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	35	27	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	35	28	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	35	29	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	35	30	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	31	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	35	32	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	35	33	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	35	34	absence	absence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	35	35	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	35	36	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	35	37	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	35	38	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	36	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	36	2	focus	focus	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	36	3	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	36	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	36	5	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	36	6	cor-	cor-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	36	7	pus	pus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	36	8	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	36	9	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	36	10	source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	36	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	36	12	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	36	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	36	14	viewing	view	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	36	15	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	36	16	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	36	17	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	36	18	fragile	fragile	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	36	19	artifact	artifact	NN
work_zhl5dw75nfhy7oputsllvjjcqq	36	20	curated	curate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	36	21	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	36	22	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	36	23	arbitrary	arbitrary	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	36	24	decisions	decision	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	36	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	37	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	37	2	ex-	ex-	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	37	3	amine	amine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	37	4	four	four	CD
work_zhl5dw75nfhy7oputsllvjjcqq	37	5	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	37	6	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	37	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	37	8	six	six	CD
work_zhl5dw75nfhy7oputsllvjjcqq	37	9	datasets	dataset	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	37	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	37	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	37	12	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	37	13	manipulate	manipulate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	37	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	37	15	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	37	16	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	37	17	shuffling	shuffle	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	37	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	37	19	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	37	20	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	37	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	37	22	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	37	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	37	24	taking	take	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	37	25	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	37	26	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	37	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	37	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	37	29	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	37	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	38	1	Finally	finally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	38	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	38	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	38	4	examine	examine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	38	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	38	6	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	38	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	38	8	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	38	9	manipulations	manipulation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	38	10	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	38	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	38	12	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	38	13	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	38	14	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	38	15	em-	em-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	38	16	beddings	bedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	38	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	39	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	39	2	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	39	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	39	4	there	there	EX
work_zhl5dw75nfhy7oputsllvjjcqq	39	5	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	39	6	considerable	considerable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	39	7	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	39	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	39	9	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	39	10	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	39	11	may	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	39	12	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	39	13	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	39	14	obvious	obvious	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	39	15	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	39	16	users	user	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	39	17	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	39	18	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	39	19	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	39	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	40	1	Rankings	ranking	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	40	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	40	3	most	most	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	40	4	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	40	5	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	40	6	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	40	7	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	40	8	reliable	reliable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	40	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	40	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	40	11	both	both	DT
work_zhl5dw75nfhy7oputsllvjjcqq	40	12	ordering	ordering	NN
work_zhl5dw75nfhy7oputsllvjjcqq	40	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	40	14	membership	membership	NN
work_zhl5dw75nfhy7oputsllvjjcqq	40	15	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	40	16	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	40	17	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	40	18	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	40	19	liable	liable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	40	20	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	40	21	change	change	VB
work_zhl5dw75nfhy7oputsllvjjcqq	40	22	significantly	significantly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	40	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	41	1	Some	some	DT
work_zhl5dw75nfhy7oputsllvjjcqq	41	2	uncer-	uncer-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	41	3	tainty	tainty	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	41	5	expected	expect	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	41	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	41	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	41	8	there	there	EX
work_zhl5dw75nfhy7oputsllvjjcqq	41	9	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	41	10	no	no	DT
work_zhl5dw75nfhy7oputsllvjjcqq	41	11	clear	clear	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	41	12	criterion	criterion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	13	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	41	14	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	41	15	acceptable	acceptable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	41	16	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	41	17	levels	level	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	41	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	41	19	variance	variance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	41	21	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	41	22	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	41	23	argue	argue	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	41	24	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	41	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	41	26	amount	amount	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	41	28	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	29	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	41	30	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	41	31	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	41	32	sufficient	sufficient	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	41	33	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	41	34	call	call	VB
work_zhl5dw75nfhy7oputsllvjjcqq	41	35	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	41	36	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	41	37	method	method	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	38	into	into	IN
work_zhl5dw75nfhy7oputsllvjjcqq	41	39	question	question	NN
work_zhl5dw75nfhy7oputsllvjjcqq	41	40	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	42	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	42	2	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	42	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	42	4	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	42	5	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	42	6	cases	case	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	42	7	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	42	8	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	42	9	there	there	EX
work_zhl5dw75nfhy7oputsllvjjcqq	42	10	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	42	11	zero	zero	CD
work_zhl5dw75nfhy7oputsllvjjcqq	42	12	set	set	NN
work_zhl5dw75nfhy7oputsllvjjcqq	42	13	overlap	overlap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	42	14	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	42	15	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	42	16	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	42	17	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	42	18	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	42	19	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	42	20	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	42	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	42	22	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	42	23	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	42	24	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	42	25	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	42	26	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	42	27	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	42	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	43	1	Smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	43	2	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	43	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	43	4	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	43	5	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	43	6	sizes	size	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	43	7	increase	increase	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	43	8	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	43	9	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	43	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	44	1	Our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	44	2	goal	goal	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	3	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	44	4	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	44	5	provide	provide	VB
work_zhl5dw75nfhy7oputsllvjjcqq	44	6	meth-	meth-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	44	7	ods	od	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	44	8	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	44	9	quantify	quantify	VB
work_zhl5dw75nfhy7oputsllvjjcqq	44	10	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	44	11	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	44	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	44	14	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	44	15	account	account	VB
work_zhl5dw75nfhy7oputsllvjjcqq	44	16	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	44	17	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	44	18	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	44	20	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	44	21	recommend	recommend	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	44	22	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	44	23	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	44	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	44	25	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	26	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	44	27	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	44	28	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	29	gets	get	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	44	30	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	44	31	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	44	32	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	33	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	44	34	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	44	35	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	44	36	averaged	average	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	44	37	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	44	38	many	many	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	44	39	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	44	40	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	44	41	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	45	1	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	45	2	Related	related	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	45	3	Work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	45	4	Word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	45	5	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	45	6	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	45	7	mappings	mapping	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	45	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	45	9	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	45	10	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	45	11	points	point	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	45	12	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	45	13	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	45	14	K	K	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	45	15	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	45	16	dimensional	dimensional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	45	17	continuous	continuous	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	45	18	space	space	NN
work_zhl5dw75nfhy7oputsllvjjcqq	45	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	45	20	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	45	21	K	K	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	45	22	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	45	23	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	45	24	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	45	25	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	45	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	45	27	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	45	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	45	29	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	45	30	vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	45	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	46	1	Re-	re-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	46	2	ducing	duce	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	46	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	46	4	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	46	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	46	6	dimensions	dimension	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	7	has	have	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	46	8	two	two	CD
work_zhl5dw75nfhy7oputsllvjjcqq	46	9	benefits	benefit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	10	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	46	11	first	first	RB
work_zhl5dw75nfhy7oputsllvjjcqq	46	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	46	13	large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	46	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	46	15	sparse	sparse	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	46	16	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	17	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	46	18	transformed	transform	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	46	19	into	into	IN
work_zhl5dw75nfhy7oputsllvjjcqq	46	20	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	46	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	46	22	dense	dense	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	46	23	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	24	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	46	25	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	46	26	second	second	RB
work_zhl5dw75nfhy7oputsllvjjcqq	46	27	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	46	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	46	29	conflation	conflation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	46	30	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	46	31	features	feature	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	32	uncovers	uncovers	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	46	33	latent	latent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	46	34	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	46	35	relationships	relationship	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	36	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	46	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	46	38	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	46	39	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	47	1	These	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	47	2	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	47	3	relationships	relationship	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	47	4	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	47	5	usually	usually	RB
work_zhl5dw75nfhy7oputsllvjjcqq	47	6	mea-	mea-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	47	7	sured	sure	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	47	8	via	via	IN
work_zhl5dw75nfhy7oputsllvjjcqq	47	9	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	47	10	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	47	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	47	12	though	though	IN
work_zhl5dw75nfhy7oputsllvjjcqq	47	13	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	47	14	metrics	metric	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	47	15	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	47	16	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	47	17	Euclidean	euclidean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	47	18	distance	distance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	47	19	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	47	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	47	21	Dice	Dice	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	47	22	coefficient	coefficient	NN
work_zhl5dw75nfhy7oputsllvjjcqq	47	23	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	47	24	possible	possible	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	47	25	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	47	26	Turney	Turney	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	47	27	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	47	28	Pantel	Pantel	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	47	29	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	47	30	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	47	31	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	47	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	48	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	48	2	focus	focus	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	48	3	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	48	4	four	four	CD
work_zhl5dw75nfhy7oputsllvjjcqq	48	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	48	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	48	7	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	48	8	popular	popular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	48	9	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	48	10	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	48	11	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	48	12	La-	la-	UH
work_zhl5dw75nfhy7oputsllvjjcqq	48	13	tent	tent	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	14	Semantic	Semantic	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	15	Analysis	Analysis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	16	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	17	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	18	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	19	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	20	Deerwester	Deerwester	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	21	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	48	22	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	23	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	24	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	25	1990	1990	CD
work_zhl5dw75nfhy7oputsllvjjcqq	48	26	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	27	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	28	Skip	Skip	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	29	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	48	30	Gram	Gram	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	31	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	48	32	Negative	Negative	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	33	Sampling	Sampling	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	34	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	35	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	36	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	37	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	38	Mikolov	Mikolov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	39	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	48	40	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	41	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	42	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	43	2013	2013	CD
work_zhl5dw75nfhy7oputsllvjjcqq	48	44	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	45	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	46	Global	Global	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	47	Vectors	Vectors	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	48	48	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	48	49	Word	Word	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	50	Rep-	Rep-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	51	resentation	resentation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	48	52	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	53	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	54	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	55	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	56	Pennington	Pennington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	57	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	48	58	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	59	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	60	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	61	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	48	62	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	63	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	64	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	48	65	Positive	Positive	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	66	Pointwise	Pointwise	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	67	Mutual	mutual	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	48	68	Information	information	NN
work_zhl5dw75nfhy7oputsllvjjcqq	48	69	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	70	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	71	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	72	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	73	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	74	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	48	75	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	48	76	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	48	77	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	48	78	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	79	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	80	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	48	81	Section	section	NN
work_zhl5dw75nfhy7oputsllvjjcqq	48	82	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	48	83	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	48	84	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	48	85	detailed	detailed	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	48	86	descriptions	description	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	48	87	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	48	88	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	48	89	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	48	90	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	48	91	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	49	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	49	2	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	49	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	49	4	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	49	5	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	49	6	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	49	7	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	49	8	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	49	9	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	49	10	fea-	fea-	DT
work_zhl5dw75nfhy7oputsllvjjcqq	49	11	tures	ture	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	49	12	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	49	13	downstream	downstream	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	49	14	tasks	task	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	49	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	50	1	Dependency	dependency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	2	parsing	parsing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	3	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	50	4	Chen	Chen	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	5	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	50	6	Manning	Manning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	50	8	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	50	9	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	50	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	50	11	named	name	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	50	12	entity	entity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	13	recogni-	recogni-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	50	14	tion	tion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	15	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	50	16	Turian	Turian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	17	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	18	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	19	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	50	21	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	50	22	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	50	23	Cherry	Cherry	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	24	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	50	25	Guo	Guo	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	50	27	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	50	28	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	50	29	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	50	30	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	50	31	bilingual	bilingual	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	50	32	lexicon	lexicon	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	33	induction	induction	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	34	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	50	35	Vulic	Vulic	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	36	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	50	37	Moens	Moens	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	50	38	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	50	39	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	50	40	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	50	41	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	50	42	just	just	RB
work_zhl5dw75nfhy7oputsllvjjcqq	50	43	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	50	44	few	few	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	50	45	examples	example	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	50	46	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	50	47	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	50	48	use	use	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	49	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	50	50	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	50	51	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	50	52	features	feature	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	50	53	has	have	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	50	54	increased	increase	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	50	55	performance	performance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	50	56	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	50	57	recent	recent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	50	58	years	year	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	50	59	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	51	1	Increasingly	increasingly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	51	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	51	3	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	51	4	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	51	5	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	51	6	been	be	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	51	7	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	51	8	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	51	9	evidence	evidence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	51	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	51	11	studies	study	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	51	12	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	51	13	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	51	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	51	15	culture	culture	NN
work_zhl5dw75nfhy7oputsllvjjcqq	51	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	52	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	52	2	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	52	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	52	4	Hamilton	Hamilton	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	52	5	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	52	6	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	52	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	53	1	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	53	2	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	53	3	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	53	4	train	train	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	5	separate	separate	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	53	6	em-	em-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	53	7	beddings	bedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	8	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	53	9	temporal	temporal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	53	10	segments	segment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	53	12	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	53	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	53	15	then	then	RB
work_zhl5dw75nfhy7oputsllvjjcqq	53	16	108	108	CD
work_zhl5dw75nfhy7oputsllvjjcqq	53	17	analyze	analyze	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	18	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	19	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	53	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	53	21	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	22	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	53	23	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	24	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	53	25	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	53	26	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	53	27	shifts	shift	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	28	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	53	29	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	53	30	Heuser	Heuser	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	53	31	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	53	32	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	53	33	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	53	34	uses	use	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	53	35	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	36	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	53	37	characterize	characterize	VB
work_zhl5dw75nfhy7oputsllvjjcqq	53	38	discourse	discourse	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	39	about	about	IN
work_zhl5dw75nfhy7oputsllvjjcqq	53	40	virtues	virtue	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	53	41	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	53	42	18th	18th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	53	43	Cen-	cen-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	44	tury	tury	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	45	English	English	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	53	46	text	text	NN
work_zhl5dw75nfhy7oputsllvjjcqq	53	47	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	54	1	Other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	54	2	studies	study	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	54	3	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	54	4	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	54	5	similar-	similar-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	54	6	ities	itie	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	54	7	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	54	8	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	54	9	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	54	10	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	54	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	54	12	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	54	13	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	54	14	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	54	15	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	54	16	geographical	geographical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	54	17	areas	area	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	54	18	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	54	19	Kulkarni	Kulkarni	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	20	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	21	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	22	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	54	24	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	54	25	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	54	26	Phillips	Phillips	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	27	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	28	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	29	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	30	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	54	31	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	54	32	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	54	33	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	54	34	time	time	NN
work_zhl5dw75nfhy7oputsllvjjcqq	54	35	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	54	36	Kim	Kim	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	37	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	54	38	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	39	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	54	40	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	54	41	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	54	42	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	54	43	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	55	1	Each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	55	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	55	3	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	55	4	studies	study	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	55	5	seeks	seek	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	55	6	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	55	7	reconstruct	reconstruct	VB
work_zhl5dw75nfhy7oputsllvjjcqq	55	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	55	9	mental	mental	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	55	10	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	55	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	55	12	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	55	13	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	55	14	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	55	15	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	55	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	56	1	An	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	56	2	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	3	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	56	4	highlights	highlight	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	56	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	56	6	contrast	contrast	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	7	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	56	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	56	9	downstream	downstream	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	10	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	56	11	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	56	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	56	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	14	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	56	15	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	56	16	per-	per-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	17	spectives	spective	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	56	18	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	56	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	56	20	exploration	exploration	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	56	22	implicit	implicit	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	56	23	bias	bias	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	24	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	56	25	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	56	26	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	56	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	57	1	Researchers	researcher	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	57	2	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	57	3	observed	observe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	57	4	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	57	5	embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	57	6	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	57	7	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	57	8	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	57	9	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	57	10	reflect	reflect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	57	11	cultural	cultural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	57	12	stereotypes	stereotype	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	57	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	57	14	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	57	15	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	57	16	associations	association	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	57	17	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	57	18	occupa-	occupa-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	57	19	tions	tion	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	57	20	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	57	21	genders	gender	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	57	22	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	57	23	Bolukbasi	Bolukbasi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	57	24	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	57	25	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	57	26	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	57	27	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	57	28	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	57	29	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	57	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	58	1	From	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	58	2	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	58	3	downstream	downstream	RB
work_zhl5dw75nfhy7oputsllvjjcqq	58	4	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	58	5	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	58	6	perspective	perspective	NN
work_zhl5dw75nfhy7oputsllvjjcqq	58	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	58	8	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	58	9	stereotypical	stereotypical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	58	10	associations	association	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	58	11	represent	represent	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	58	12	bias	bias	NN
work_zhl5dw75nfhy7oputsllvjjcqq	58	13	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	58	14	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	58	15	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	58	16	filtered	filter	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	58	17	out	out	RP
work_zhl5dw75nfhy7oputsllvjjcqq	58	18	before	before	IN
work_zhl5dw75nfhy7oputsllvjjcqq	58	19	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	58	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	58	21	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	58	22	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	58	23	features	feature	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	58	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	59	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	59	2	contrast	contrast	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	59	4	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	59	5	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	59	6	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	7	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	59	8	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	59	9	perspective	perspective	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	59	11	implicit	implicit	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	59	12	bias	bias	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	13	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	59	14	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	59	15	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	59	16	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	59	17	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	59	18	problem	problem	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	19	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	59	20	must	must	MD
work_zhl5dw75nfhy7oputsllvjjcqq	59	21	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	59	22	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	59	23	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	59	24	rather	rather	RB
work_zhl5dw75nfhy7oputsllvjjcqq	59	25	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	59	26	means	means	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	59	28	measurement	measurement	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	29	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	59	30	providing	provide	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	59	31	quantita-	quantita-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	59	32	tive	tive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	59	33	evidence	evidence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	34	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	59	35	bias	bias	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	36	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	59	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	59	38	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	59	39	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	59	40	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	60	1	Embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	60	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	60	3	usually	usually	RB
work_zhl5dw75nfhy7oputsllvjjcqq	60	4	evaluated	evaluate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	60	5	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	60	6	direct	direct	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	60	7	use	use	NN
work_zhl5dw75nfhy7oputsllvjjcqq	60	8	cases	case	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	60	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	60	10	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	60	11	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	60	12	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	60	13	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	60	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	60	15	analogy	analogy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	60	16	tasks	task	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	60	17	via	via	IN
work_zhl5dw75nfhy7oputsllvjjcqq	60	18	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	60	19	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	60	20	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	60	21	Mikolov	Mikolov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	22	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	60	23	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	24	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	60	26	2013	2013	CD
work_zhl5dw75nfhy7oputsllvjjcqq	60	27	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	60	28	Pennington	Pennington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	29	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	60	30	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	31	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	60	33	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	60	34	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	60	35	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	36	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	37	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	38	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	39	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	60	40	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	60	41	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	60	42	Shazeer	Shazeer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	43	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	44	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	45	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	60	46	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	60	47	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	60	48	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	60	49	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	61	1	Intrinsic	intrinsic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	61	2	evaluations	evaluation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	61	3	like	like	IN
work_zhl5dw75nfhy7oputsllvjjcqq	61	4	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	61	5	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	61	6	measure	measure	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	61	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	61	8	interpretability	interpretability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	61	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	61	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	61	11	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	61	12	rather	rather	RB
work_zhl5dw75nfhy7oputsllvjjcqq	61	13	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	61	14	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	61	15	downstream	downstream	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	61	16	task	task	NN
work_zhl5dw75nfhy7oputsllvjjcqq	61	17	performance	performance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	61	18	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	61	19	Gladkova	Gladkova	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	20	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	61	21	Drozd	Drozd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	61	23	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	61	24	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	61	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	61	26	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	61	27	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	61	28	some	some	DT
work_zhl5dw75nfhy7oputsllvjjcqq	61	29	research	research	NN
work_zhl5dw75nfhy7oputsllvjjcqq	61	30	does	do	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	61	31	evaluate	evaluate	VB
work_zhl5dw75nfhy7oputsllvjjcqq	61	32	embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	61	33	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	61	34	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	61	35	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	61	36	downstream	downstream	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	61	37	task	task	NN
work_zhl5dw75nfhy7oputsllvjjcqq	61	38	perfor-	perfor-	IN
work_zhl5dw75nfhy7oputsllvjjcqq	61	39	mance	mance	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	40	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	61	41	Pennington	Pennington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	42	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	61	43	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	44	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	45	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	61	46	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	61	47	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	61	48	Faruqui	Faruqui	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	49	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	50	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	51	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	61	52	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	61	53	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	61	54	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	61	55	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	61	56	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	61	57	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	61	58	benchmarks	benchmark	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	61	59	remain	remain	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	61	60	intrinsic	intrinsic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	61	61	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	62	1	There	there	EX
work_zhl5dw75nfhy7oputsllvjjcqq	62	2	has	have	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	62	3	been	be	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	62	4	some	some	DT
work_zhl5dw75nfhy7oputsllvjjcqq	62	5	recent	recent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	62	6	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	62	7	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	62	8	evaluating	evaluate	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	62	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	62	10	stability	stability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	62	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	62	12	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	62	13	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	62	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	63	1	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	63	2	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	63	3	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	63	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	64	1	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	64	2	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	64	3	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	64	4	focus	focus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	5	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	7	hyperparameter	hyperparameter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	8	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	64	9	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	10	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	11	algo-	algo-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	64	12	rithm	rithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	64	14	show	show	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	64	15	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	16	hyperparameters	hyperparameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	64	17	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	64	18	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	20	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	23	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	24	window	window	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	64	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	27	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	29	negative	negative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	64	30	sam-	sam-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	64	31	ples	ple	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	64	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	64	33	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	64	34	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	35	level	level	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	36	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	37	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	38	distribution	distribution	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	39	smoothing	smoothing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	40	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	64	41	affect	affect	VB
work_zhl5dw75nfhy7oputsllvjjcqq	64	42	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	64	43	performance	performance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	44	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	45	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	64	46	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	64	47	simi-	simi-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	64	48	larity	larity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	49	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	64	50	analogy	analogy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	64	51	tasks	task	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	64	52	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	65	1	Hellrich	Hellrich	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	2	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	65	3	Hahn	Hahn	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	4	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	65	5	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	65	6	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	65	7	examine	examine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	65	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	65	9	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	65	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	11	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	12	frequency	frequency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	65	14	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	15	am-	am-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	65	16	biguity	biguity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	65	18	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	65	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	65	20	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	22	training	train	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	65	23	epochs	epoch	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	65	24	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	65	26	reliability	reliability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	28	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	65	29	produced	produce	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	65	30	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	31	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	65	32	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	33	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	65	34	skip	skip	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	35	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	65	36	gram	gram	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	37	hierarchical	hierarchical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	65	38	softmax	softmax	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	39	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	65	40	SGHS	SGHS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	41	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	65	42	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	65	43	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	65	44	variant	variant	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	45	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	46	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	47	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	65	48	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	65	49	striving	strive	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	65	50	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	51	reproducibility	reproducibility	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	52	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	65	53	recommend-	recommend-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	54	ing	ing	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	55	against	against	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	56	sampling	sample	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	65	57	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	65	58	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	65	59	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	65	60	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	61	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	65	62	preserve	preserve	VB
work_zhl5dw75nfhy7oputsllvjjcqq	65	63	stability	stability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	65	64	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	66	1	Likewise	likewise	RB
work_zhl5dw75nfhy7oputsllvjjcqq	66	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	66	3	Tian	Tian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	66	4	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	66	5	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	66	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	67	1	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	67	2	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	67	3	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	67	4	explore	explore	VB
work_zhl5dw75nfhy7oputsllvjjcqq	67	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	67	6	ro-	ro-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	67	7	bustness	bustness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	67	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	67	9	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	67	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	67	11	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	67	12	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	67	13	trained	train	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	67	14	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	67	15	large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	67	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	67	17	generic	generic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	67	18	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	67	19	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	67	20	Wikipedia	Wikipedia	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	67	21	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	67	22	news	news	NN
work_zhl5dw75nfhy7oputsllvjjcqq	67	23	data	datum	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	67	24	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	67	25	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	67	26	propose	propose	VB
work_zhl5dw75nfhy7oputsllvjjcqq	67	27	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	67	28	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	67	29	align	align	VB
work_zhl5dw75nfhy7oputsllvjjcqq	67	30	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	67	31	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	67	32	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	67	33	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	67	34	iterations	iteration	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	67	35	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	68	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	68	2	contrast	contrast	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	68	4	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	68	5	goal	goal	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	6	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	68	7	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	68	8	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	68	9	produce	produce	VB
work_zhl5dw75nfhy7oputsllvjjcqq	68	10	artificially	artificially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	68	11	stable	stable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	68	12	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	68	13	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	68	14	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	68	15	identify	identify	VB
work_zhl5dw75nfhy7oputsllvjjcqq	68	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	68	17	factors	factor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	68	18	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	68	19	create	create	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	68	20	instability	instability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	21	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	68	22	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	68	23	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	68	24	statistical	statistical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	68	25	confi-	confi-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	26	dence	dence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	27	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	68	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	68	29	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	30	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	68	31	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	68	32	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	68	33	trained	train	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	68	34	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	68	35	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	68	36	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	68	37	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	68	38	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	68	39	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	69	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	69	2	focus	focus	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	69	3	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	69	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	69	5	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	69	6	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	69	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	69	8	fragile	fragile	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	69	9	artifact	artifact	NN
work_zhl5dw75nfhy7oputsllvjjcqq	69	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	69	11	source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	69	12	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	69	13	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	69	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	69	15	considering	consider	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	69	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	69	17	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	69	18	itself	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	69	19	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	69	20	merely	merely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	69	21	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	69	22	sample	sample	NN
work_zhl5dw75nfhy7oputsllvjjcqq	69	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	69	24	possible	possible	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	69	25	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	69	26	produced	produce	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	69	27	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	69	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	69	29	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	69	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	70	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	70	2	examine	examine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	70	3	whether	whether	IN
work_zhl5dw75nfhy7oputsllvjjcqq	70	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	70	5	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	70	6	accurately	accurately	RB
work_zhl5dw75nfhy7oputsllvjjcqq	70	7	model	model	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	70	8	those	those	DT
work_zhl5dw75nfhy7oputsllvjjcqq	70	9	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	70	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	70	11	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	70	12	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	70	13	sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	70	14	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	70	15	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	70	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	70	17	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	70	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	70	19	adding	add	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	70	20	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	70	21	removing	remove	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	70	22	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	70	23	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	70	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	70	25	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	70	26	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	70	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	71	1	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	71	2	Corpora	Corpora	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	71	3	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	71	4	collected	collect	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	71	5	two	two	CD
work_zhl5dw75nfhy7oputsllvjjcqq	71	6	sub	sub	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	7	-	-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	8	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	9	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	71	10	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	71	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	71	12	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	71	13	datasets	dataset	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	71	14	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	71	15	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	71	16	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	17	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	71	18	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	71	19	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	71	20	explore	explore	VB
work_zhl5dw75nfhy7oputsllvjjcqq	71	21	how	how	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	71	22	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	23	embed-	embed-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	71	24	dings	ding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	71	25	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	71	26	affected	affect	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	71	27	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	71	28	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	29	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	71	30	vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	31	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	71	32	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	71	33	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	71	34	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	71	35	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	71	36	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	71	37	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	71	38	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	71	39	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	72	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	72	2	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	3	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	72	4	bet-	bet-	VB
work_zhl5dw75nfhy7oputsllvjjcqq	72	5	ter	ter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	6	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	7	realistic	realistic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	72	8	examples	example	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	72	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	72	10	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	11	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	72	12	centered	center	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	72	13	re-	re-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	72	14	search	search	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	72	16	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	72	17	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	18	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	72	19	deliberately	deliberately	RB
work_zhl5dw75nfhy7oputsllvjjcqq	72	20	chosen	choose	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	72	21	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	72	22	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	72	23	publicly	publicly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	72	24	available	available	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	72	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	72	26	suggestive	suggestive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	72	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	72	28	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	72	29	research	research	NN
work_zhl5dw75nfhy7oputsllvjjcqq	72	30	ques-	ques-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	72	31	tions	tion	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	72	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	72	33	varied	vary	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	72	34	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	72	35	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	72	36	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	72	37	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	72	38	e.g.	e.g.	RB
work_zhl5dw75nfhy7oputsllvjjcqq	73	1	topic	topic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	73	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	73	3	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	73	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	73	5	vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	73	6	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	73	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	73	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	73	9	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	73	10	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	73	11	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	73	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	73	13	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	73	14	cor-	cor-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	73	15	pora	pora	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	73	16	typically	typically	RB
work_zhl5dw75nfhy7oputsllvjjcqq	73	17	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	73	18	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	73	19	training	train	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	73	20	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	73	21	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	73	22	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	73	23	e.g.	e.g.	RB
work_zhl5dw75nfhy7oputsllvjjcqq	74	1	Wikipedia	Wikipedia	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	74	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	74	3	Gigaword	Gigaword	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	74	4	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	74	5	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	75	1	Each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	75	2	dataset	dataset	NN
work_zhl5dw75nfhy7oputsllvjjcqq	75	3	was	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	75	4	created	create	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	75	5	or-	or-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	75	6	ganically	ganically	RB
work_zhl5dw75nfhy7oputsllvjjcqq	75	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	75	8	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	75	9	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	75	10	time	time	NN
work_zhl5dw75nfhy7oputsllvjjcqq	75	11	periods	period	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	75	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	75	13	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	75	14	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	75	15	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	75	16	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	75	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	75	18	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	75	19	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	75	20	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	75	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	76	1	Thus	thus	RB
work_zhl5dw75nfhy7oputsllvjjcqq	76	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	76	3	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	76	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	76	5	impossible	impossible	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	76	6	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	76	7	expand	expand	VB
work_zhl5dw75nfhy7oputsllvjjcqq	76	8	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	76	9	datasets	dataset	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	76	10	without	without	IN
work_zhl5dw75nfhy7oputsllvjjcqq	76	11	compromising	compromise	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	76	12	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	76	13	specificity	specificity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	76	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	77	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	77	2	process	process	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	77	3	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	77	4	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	77	5	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	77	6	lowercasing	lowercase	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	77	7	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	77	8	text	text	NN
work_zhl5dw75nfhy7oputsllvjjcqq	77	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	77	10	re-	re-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	77	11	moving	move	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	77	12	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	77	13	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	77	14	appear	appear	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	77	15	fewer	few	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	77	16	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	77	17	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	77	18	times	time	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	77	19	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	77	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	77	21	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	77	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	77	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	77	24	removing	remove	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	77	25	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	77	26	numbers	number	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	77	27	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	77	28	punctuation	punctuation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	77	29	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	78	1	Because	because	IN
work_zhl5dw75nfhy7oputsllvjjcqq	78	2	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	78	3	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	78	4	rely	rely	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	78	5	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	78	6	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	78	7	sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	78	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	78	9	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	78	10	Section	Section	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	78	11	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	78	12	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	78	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	78	14	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	78	15	operates	operate	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	78	16	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	78	17	removing	remove	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	78	18	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	78	19	multi-	multi-	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	78	20	plying	ply	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	78	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	78	22	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	78	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	78	24	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	78	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	78	26	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	78	27	also	also	RB
work_zhl5dw75nfhy7oputsllvjjcqq	78	28	remove	remove	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	78	29	duplicate	duplicate	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	78	30	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	78	31	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	78	32	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	78	33	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	78	34	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	79	1	U.S.	U.S.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	79	2	Federal	Federal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	79	3	Courts	Courts	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	79	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	79	5	Appeals	Appeals	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	79	6	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	79	7	U.S.	U.S.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	79	8	Federal	Federal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	79	9	courts	court	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	79	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	79	11	appeals	appeal	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	79	12	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	79	13	regional	regional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	79	14	courts	court	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	79	15	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	79	16	decide	decide	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	79	17	ap-	ap-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	79	18	peals	peal	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	79	19	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	79	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	79	21	district	district	NN
work_zhl5dw75nfhy7oputsllvjjcqq	79	22	courts	court	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	79	23	within	within	IN
work_zhl5dw75nfhy7oputsllvjjcqq	79	24	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	79	25	federal	federal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	79	26	ju-	ju-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	79	27	dicial	dicial	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	79	28	circuit	circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	79	29	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	80	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	80	2	examine	examine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	80	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	4	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	7	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	80	8	recent	recent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	9	five	five	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	10	years	year	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	13	4th	4th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	80	15	9th	9th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	16	circuits.1	circuits.1	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	80	17	1https://www.courtlistener.com/	1https://www.courtlistener.com/	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	18	109	109	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	19	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	20	Number	Number	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	22	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	23	Unique	unique	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	24	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	25	Vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	26	density	density	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	27	Words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	28	per	per	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	29	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	30	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	31	Sports	Sports	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	80	32	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	33	2000	2000	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	34	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	35	8,786	8,786	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	36	12,475	12,475	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	37	0.0020	0.0020	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	38	708	708	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	39	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	40	Music	Music	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	41	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	42	2000	2000	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	43	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	44	3,666	3,666	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	45	9,762	9,762	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	46	0.0037	0.0037	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	47	715	715	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	48	AskScience	AskScience	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	49	331,635	331,635	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	50	16,901	16,901	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	51	0.0012	0.0012	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	52	44	44	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	53	AskHistorians	AskHistorians	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	80	54	63,578	63,578	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	55	9,384	9,384	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	56	0.0022	0.0022	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	57	66	66	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	58	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	59	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	60	5,368	5,368	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	61	16,639	16,639	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	62	0.0014	0.0014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	63	2,281	2,281	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	64	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	65	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	66	9,729	9,729	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	67	22,146	22,146	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	68	0.0011	0.0011	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	69	2,108	2,108	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	70	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	71	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	72	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	80	73	Comparison	comparison	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	74	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	75	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	76	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	77	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	78	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	79	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	80	80	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	81	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	82	unique	unique	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	83	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	84	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	85	after	after	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	86	removing	remove	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	80	87	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	88	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	80	89	appear	appear	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	80	90	fewer	few	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	80	91	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	92	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	80	93	times	time	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	94	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	95	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	80	96	vocabulary	vocabulary	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	97	density	density	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	98	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	99	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	100	ratio	ratio	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	101	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	102	unique	unique	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	103	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	104	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	105	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	106	total	total	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	107	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	108	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	109	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	110	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	80	111	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	80	112	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	80	113	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	114	average	average	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	80	115	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	116	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	117	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	80	118	per	per	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	119	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	80	120	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	80	121	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	80	122	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	80	123	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	81	1	Setting	set	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	81	2	Method	Method	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	3	Tests	Tests	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	4	...	...	:
work_zhl5dw75nfhy7oputsllvjjcqq	81	5	Run	run	VB
work_zhl5dw75nfhy7oputsllvjjcqq	81	6	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	81	7	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	8	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	81	9	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	10	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	81	11	Fixed	fixed	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	81	12	Documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	81	13	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	14	consistent	consistent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	81	15	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	16	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	17	due	due	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	18	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	19	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	20	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	81	21	baseline	baseline	VB
work_zhl5dw75nfhy7oputsllvjjcqq	81	22	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	81	23	A	A	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	24	B	b	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	25	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	26	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	27	B	b	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	28	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	29	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	30	B	b	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	31	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	32	Shuffled	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	81	33	Documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	81	34	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	35	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	81	36	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	37	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	38	due	due	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	39	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	40	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	41	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	42	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	43	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	44	B	b	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	45	B	B	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	46	A	A	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	47	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	48	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	49	B	B	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	50	A	A	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	51	Bootstrap	Bootstrap	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	52	Documents	Documents	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	81	53	sampled	sample	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	81	54	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	55	replacement	replacement	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	56	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	57	due	due	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	58	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	59	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	60	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	61	B	B	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	62	A	A	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	63	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	64	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	65	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	66	B	b	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	67	B	b	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	68	B	B	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	69	B	B	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	81	70	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	71	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	81	72	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	81	73	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	81	74	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	81	75	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	81	76	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	81	77	manipulate	manipulate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	81	78	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	81	79	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	80	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	81	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	81	82	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	83	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	81	84	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	81	85	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	81	86	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	82	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	82	2	4th	4th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	82	3	circuit	circuit	NN
work_zhl5dw75nfhy7oputsllvjjcqq	82	4	contains	contain	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	82	5	Washington	Washington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	82	6	D.C.	D.C.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	82	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	82	8	sur-	sur-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	82	9	rounding	round	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	82	10	states	state	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	82	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	82	12	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	82	13	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	82	14	9th	9th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	82	15	circuit	circuit	NN
work_zhl5dw75nfhy7oputsllvjjcqq	82	16	contains	contain	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	82	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	82	18	entirety	entirety	NN
work_zhl5dw75nfhy7oputsllvjjcqq	82	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	82	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	82	21	west	west	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	82	22	coast	coast	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	82	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	83	1	Social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	83	2	science	science	NN
work_zhl5dw75nfhy7oputsllvjjcqq	83	3	research	research	NN
work_zhl5dw75nfhy7oputsllvjjcqq	83	4	questions	question	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	83	5	might	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	83	6	involve	involve	VB
work_zhl5dw75nfhy7oputsllvjjcqq	83	7	measuring	measure	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	83	8	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	83	9	widely	widely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	83	10	held	hold	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	83	11	belief	belief	NN
work_zhl5dw75nfhy7oputsllvjjcqq	83	12	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	83	13	certain	certain	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	83	14	courts	court	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	83	15	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	83	16	distinct	distinct	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	83	17	ideological	ideological	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	83	18	ten-	ten-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	83	19	dencies	dencie	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	83	20	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	83	21	Broscheid	Broscheid	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	83	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	83	23	2011	2011	CD
work_zhl5dw75nfhy7oputsllvjjcqq	83	24	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	83	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	84	1	Such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	84	2	bias	bias	NN
work_zhl5dw75nfhy7oputsllvjjcqq	84	3	may	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	84	4	result	result	VB
work_zhl5dw75nfhy7oputsllvjjcqq	84	5	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	84	6	measurable	measurable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	84	7	differences	difference	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	84	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	84	9	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	84	10	association	association	NN
work_zhl5dw75nfhy7oputsllvjjcqq	84	11	due	due	IN
work_zhl5dw75nfhy7oputsllvjjcqq	84	12	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	84	13	framing	framing	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	84	14	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	84	15	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	84	16	Card	Card	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	84	17	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	84	18	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	84	19	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	84	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	84	21	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	84	22	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	84	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	84	24	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	84	25	could	could	MD
work_zhl5dw75nfhy7oputsllvjjcqq	84	26	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	84	27	observable	observable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	84	28	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	84	29	comparing	compare	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	84	30	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	84	31	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	84	32	associated	associate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	84	33	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	84	34	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	84	35	given	give	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	84	36	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	84	37	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	84	38	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	85	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	85	2	treat	treat	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	85	3	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	85	4	opinion	opinion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	85	5	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	85	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	85	7	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	85	8	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	85	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	86	1	New	New	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	2	York	York	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	3	Times	Times	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	4	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	86	5	New	New	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	6	York	York	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	7	Times	Times	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	86	9	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	10	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	86	11	An-	an-	CC
work_zhl5dw75nfhy7oputsllvjjcqq	86	12	notated	notate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	86	13	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	14	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	86	15	Sandhaus	Sandhaus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	86	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	86	17	2008	2008	CD
work_zhl5dw75nfhy7oputsllvjjcqq	86	18	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	86	19	contains	contain	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	86	20	newspaper	newspaper	NN
work_zhl5dw75nfhy7oputsllvjjcqq	86	21	articles	article	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	86	22	tagged	tag	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	86	23	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	86	24	additional	additional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	86	25	metadata	metadata	NN
work_zhl5dw75nfhy7oputsllvjjcqq	86	26	reflecting	reflect	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	86	27	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	86	28	content	content	NN
work_zhl5dw75nfhy7oputsllvjjcqq	86	29	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	86	30	publication	publication	NN
work_zhl5dw75nfhy7oputsllvjjcqq	86	31	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	86	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	87	1	To	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	87	2	constrain	constrain	VB
work_zhl5dw75nfhy7oputsllvjjcqq	87	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	87	4	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	87	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	87	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	87	7	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	87	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	87	9	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	87	10	enhance	enhance	VB
work_zhl5dw75nfhy7oputsllvjjcqq	87	11	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	87	12	specificity	specificity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	87	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	87	14	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	87	15	extract	extract	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	87	16	data	datum	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	87	17	only	only	RB
work_zhl5dw75nfhy7oputsllvjjcqq	87	18	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	87	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	87	20	year	year	NN
work_zhl5dw75nfhy7oputsllvjjcqq	87	21	2000	2000	CD
work_zhl5dw75nfhy7oputsllvjjcqq	87	22	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	87	23	focus	focus	VB
work_zhl5dw75nfhy7oputsllvjjcqq	87	24	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	87	25	only	only	RB
work_zhl5dw75nfhy7oputsllvjjcqq	87	26	two	two	CD
work_zhl5dw75nfhy7oputsllvjjcqq	87	27	sections	section	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	87	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	87	29	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	87	30	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	87	31	dataset	dataset	NN
work_zhl5dw75nfhy7oputsllvjjcqq	87	32	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	87	33	sports	sport	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	87	34	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	87	35	music	music	NN
work_zhl5dw75nfhy7oputsllvjjcqq	87	36	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	88	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	88	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	88	3	resulting	result	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	88	4	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	88	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	88	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	88	7	sports	sport	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	88	8	section	section	NN
work_zhl5dw75nfhy7oputsllvjjcqq	88	9	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	88	10	substantially	substantially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	88	11	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	88	12	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	88	13	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	88	14	music	music	NN
work_zhl5dw75nfhy7oputsllvjjcqq	88	15	section	section	NN
work_zhl5dw75nfhy7oputsllvjjcqq	88	16	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	88	17	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	88	18	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	88	19	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	88	20	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	88	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	89	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	89	2	treat	treat	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	89	3	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	89	4	article	article	NN
work_zhl5dw75nfhy7oputsllvjjcqq	89	5	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	89	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	89	7	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	89	8	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	89	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	90	1	Reddit	Reddit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	90	2	Reddit2	Reddit2	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	90	3	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	90	4	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	90	5	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	90	6	website	website	NN
work_zhl5dw75nfhy7oputsllvjjcqq	90	7	containing	contain	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	90	8	thou-	thou-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	90	9	sands	sand	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	90	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	90	11	forums	forum	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	90	12	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	90	13	subreddits	subreddit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	90	14	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	90	15	organized	organize	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	90	16	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	90	17	topic	topic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	90	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	91	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	91	2	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	91	3	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	91	4	dataset	dataset	NN
work_zhl5dw75nfhy7oputsllvjjcqq	91	5	containing	contain	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	91	6	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	91	7	posts	post	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	91	8	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	91	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	91	10	years	year	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	91	11	2007-	2007-	CD
work_zhl5dw75nfhy7oputsllvjjcqq	91	12	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	91	13	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	91	14	two	two	CD
work_zhl5dw75nfhy7oputsllvjjcqq	91	15	subreddits	subreddit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	91	16	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	91	17	/r	/r	.
work_zhl5dw75nfhy7oputsllvjjcqq	91	18	/	/	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	91	19	AskScience	AskScience	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	91	20	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	91	21	/r	/r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	91	22	/	/	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	91	23	AskHistorians	AskHistorians	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	91	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	92	1	These	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	92	2	two	two	CD
work_zhl5dw75nfhy7oputsllvjjcqq	92	3	subreddits	subreddit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	92	4	allow	allow	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	92	5	users	user	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	92	6	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	92	7	post	post	VB
work_zhl5dw75nfhy7oputsllvjjcqq	92	8	any	any	DT
work_zhl5dw75nfhy7oputsllvjjcqq	92	9	question	question	NN
work_zhl5dw75nfhy7oputsllvjjcqq	92	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	92	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	92	12	topics	topic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	92	13	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	92	14	history	history	NN
work_zhl5dw75nfhy7oputsllvjjcqq	92	15	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	92	16	science	science	NN
work_zhl5dw75nfhy7oputsllvjjcqq	92	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	92	18	respectively	respectively	RB
work_zhl5dw75nfhy7oputsllvjjcqq	92	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	93	1	AskScience	AskScience	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	93	2	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	93	3	more	more	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	93	4	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	93	5	five	five	CD
work_zhl5dw75nfhy7oputsllvjjcqq	93	6	times	time	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	93	7	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	93	8	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	93	9	AskHistorians	AskHistorians	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	93	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	93	11	though	though	IN
work_zhl5dw75nfhy7oputsllvjjcqq	93	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	93	13	doc-	doc-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	93	14	2https://www.reddit.com/	2https://www.reddit.com/	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	93	15	ument	ument	NN
work_zhl5dw75nfhy7oputsllvjjcqq	93	16	length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	93	17	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	93	18	generally	generally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	93	19	longer	long	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	93	20	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	93	21	AskHistorians	AskHistorians	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	93	22	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	93	23	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	93	24	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	93	25	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	93	26	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	93	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	94	1	Reddit	Reddit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	94	2	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	94	3	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	94	4	popular	popular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	94	5	data	data	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	6	source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	7	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	94	8	computational	computational	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	94	9	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	94	10	science	science	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	11	research	research	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	12	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	94	13	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	94	14	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	94	16	subreddits	subreddit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	94	17	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	94	18	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	94	19	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	94	20	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	94	21	explore	explore	VB
work_zhl5dw75nfhy7oputsllvjjcqq	94	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	94	23	distinctiveness	distinctiveness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	24	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	94	25	dynamicity	dynamicity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	94	26	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	94	27	communities	community	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	94	28	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	94	29	Zhang	Zhang	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	94	30	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	94	31	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	94	32	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	94	33	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	94	34	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	94	35	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	94	36	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	95	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	95	2	treat	treat	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	95	3	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	95	4	original	original	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	95	5	post	post	NN
work_zhl5dw75nfhy7oputsllvjjcqq	95	6	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	95	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	95	8	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	95	9	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	95	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	96	1	4	4	LS
work_zhl5dw75nfhy7oputsllvjjcqq	96	2	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	96	3	Parameters	Parameters	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	96	4	Order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	96	5	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	96	6	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	96	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	96	8	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	96	9	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	96	10	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	96	11	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	96	12	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	96	13	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	96	14	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	96	15	sample	sample	VB
work_zhl5dw75nfhy7oputsllvjjcqq	96	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	96	17	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	96	18	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	96	19	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	96	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	96	21	SHUFFLED	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	96	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	96	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	96	24	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	96	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	97	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	97	2	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	97	3	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	97	4	includes	include	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	97	5	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	97	6	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	97	7	exactly	exactly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	97	8	once	once	RB
work_zhl5dw75nfhy7oputsllvjjcqq	97	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	97	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	97	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	97	12	doc-	doc-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	97	13	uments	ument	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	97	14	appear	appear	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	97	15	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	97	16	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	97	17	constant	constant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	97	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	97	19	chronological	chronological	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	97	20	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	97	21	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	97	22	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	97	23	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	97	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	98	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	98	2	purpose	purpose	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	98	4	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	98	5	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	6	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	98	7	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	98	8	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	98	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	98	10	baseline	baseline	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	11	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	12	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	98	13	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	98	14	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	98	16	independent	independent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	98	17	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	98	18	any	any	DT
work_zhl5dw75nfhy7oputsllvjjcqq	98	19	change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	20	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	98	21	input	input	NN
work_zhl5dw75nfhy7oputsllvjjcqq	98	22	data	datum	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	98	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	99	1	Algorith-	Algorith-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	99	2	mic	mic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	99	3	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	99	4	may	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	99	5	arise	arise	VB
work_zhl5dw75nfhy7oputsllvjjcqq	99	6	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	99	7	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	99	8	initializations	initialization	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	99	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	99	10	learned	learn	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	99	11	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	99	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	99	13	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	99	14	negative	negative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	99	15	sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	99	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	99	17	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	99	18	randomized	randomized	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	99	19	subsampling	subsampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	99	20	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	99	21	tokens	token	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	99	22	within	within	IN
work_zhl5dw75nfhy7oputsllvjjcqq	99	23	docu-	docu-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	99	24	ments	ment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	99	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	100	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	100	2	SHUFFLED	SHUFFLED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	100	3	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	100	4	includes	include	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	100	5	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	100	6	docu-	docu-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	100	7	ment	ment	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	100	8	exactly	exactly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	100	9	once	once	RB
work_zhl5dw75nfhy7oputsllvjjcqq	100	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	100	11	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	100	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	100	13	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	100	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	100	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	100	16	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	100	17	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	100	18	randomized	randomize	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	100	19	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	100	20	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	100	21	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	100	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	101	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	101	2	purpose	purpose	NN
work_zhl5dw75nfhy7oputsllvjjcqq	101	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	101	4	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	101	5	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	101	6	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	101	7	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	101	8	evaluate	evaluate	VB
work_zhl5dw75nfhy7oputsllvjjcqq	101	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	101	10	impact	impact	NN
work_zhl5dw75nfhy7oputsllvjjcqq	101	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	101	12	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	101	13	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	101	14	how	how	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	101	15	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	101	16	present	present	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	101	17	examples	example	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	101	18	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	101	19	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	101	20	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	101	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	102	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	102	2	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	102	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	102	4	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	102	5	could	could	MD
work_zhl5dw75nfhy7oputsllvjjcqq	102	6	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	102	7	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	102	8	important	important	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	102	9	factor	factor	NN
work_zhl5dw75nfhy7oputsllvjjcqq	102	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	102	11	algo-	algo-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	102	12	rithms	rithms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	102	13	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	102	14	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	102	15	online	online	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	102	16	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	102	17	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	102	18	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	102	19	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	102	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	103	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	103	2	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	103	3	setting	set	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	103	4	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	103	5	N	n	CD
work_zhl5dw75nfhy7oputsllvjjcqq	103	6	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	103	7	randomly	randomly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	103	8	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	103	9	replacement	replacement	NN
work_zhl5dw75nfhy7oputsllvjjcqq	103	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	103	11	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	103	12	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	103	13	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	103	14	equal	equal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	103	15	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	103	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	103	17	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	103	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	103	19	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	103	20	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	103	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	103	22	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	103	23	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	103	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	104	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	104	2	purpose	purpose	NN
work_zhl5dw75nfhy7oputsllvjjcqq	104	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	104	4	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	104	5	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	104	6	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	104	7	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	104	8	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	104	9	how	how	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	104	10	much	much	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	104	11	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	104	12	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	104	13	due	due	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	104	14	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	104	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	104	16	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	104	17	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	104	18	absence	absence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	104	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	104	20	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	104	21	sequences	sequence	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	104	22	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	104	23	110	110	CD
work_zhl5dw75nfhy7oputsllvjjcqq	104	24	tokens	token	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	104	25	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	104	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	104	27	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	104	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	105	1	See	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	105	2	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	105	3	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	105	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	105	5	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	105	6	comparison	comparison	NN
work_zhl5dw75nfhy7oputsllvjjcqq	105	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	105	8	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	105	9	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	105	10	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	105	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	106	1	Size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	106	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	106	3	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	106	4	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	106	5	expect	expect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	106	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	106	7	stability	stability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	106	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	106	9	embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	106	10	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	106	11	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	106	12	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	106	13	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	106	14	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	106	15	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	106	16	influenced	influence	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	106	17	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	106	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	106	19	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	106	20	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	106	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	106	22	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	106	23	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	106	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	107	1	As	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	107	2	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	107	3	add	add	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	107	4	more	more	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	107	5	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	107	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	107	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	107	8	impact	impact	NN
work_zhl5dw75nfhy7oputsllvjjcqq	107	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	107	10	any	any	DT
work_zhl5dw75nfhy7oputsllvjjcqq	107	11	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	107	12	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	107	13	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	107	14	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	107	15	less	less	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	107	16	significant	significant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	107	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	108	1	At	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	108	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	108	3	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	108	4	time	time	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	108	6	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	108	7	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	8	may	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	108	9	also	also	RB
work_zhl5dw75nfhy7oputsllvjjcqq	108	10	tend	tend	VB
work_zhl5dw75nfhy7oputsllvjjcqq	108	11	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	108	12	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	108	13	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	108	14	broad	broad	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	108	15	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	108	16	scope	scope	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	108	18	variable	variable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	108	19	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	108	20	style	style	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	21	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	108	22	topic	topic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	108	24	leading	lead	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	108	25	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	108	26	less	less	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	108	27	idiosyn-	idiosyn-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	108	28	cratic	cratic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	108	29	patterns	pattern	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	108	30	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	108	31	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	32	co	co	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	33	-	-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	34	occurrence	occurrence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	108	35	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	109	1	Therefore	therefore	RB
work_zhl5dw75nfhy7oputsllvjjcqq	109	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	109	3	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	109	4	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	109	5	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	109	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	109	7	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	109	8	curate	curate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	109	9	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	109	10	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	109	11	sub	sub	NN
work_zhl5dw75nfhy7oputsllvjjcqq	109	12	-	-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	109	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	109	14	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	109	15	contains	contain	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	109	16	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	109	17	%	%	NN
work_zhl5dw75nfhy7oputsllvjjcqq	109	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	109	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	109	20	total	total	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	109	21	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	109	22	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	109	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	110	1	These	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	110	2	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	110	3	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	110	4	selected	select	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	110	5	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	110	6	contiguous	contiguous	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	110	7	sequences	sequence	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	110	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	110	9	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	110	10	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	110	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	110	12	beginning	beginning	NN
work_zhl5dw75nfhy7oputsllvjjcqq	110	13	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	110	14	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	110	15	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	110	16	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	110	17	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	110	18	ensures	ensure	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	110	19	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	110	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	110	21	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	110	22	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	110	23	remains	remain	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	110	24	constant	constant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	110	25	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	110	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	111	1	Length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	111	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	111	3	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	111	4	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	111	5	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	111	6	two	two	CD
work_zhl5dw75nfhy7oputsllvjjcqq	111	7	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	111	8	seg-	seg-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	111	9	mentation	mentation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	111	10	strategies	strategy	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	111	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	112	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	112	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	112	3	first	first	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	112	4	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	112	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	112	6	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	112	7	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	112	8	instance	instance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	112	9	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	112	10	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	112	11	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	112	12	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	112	13	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	112	14	i.e.	i.e.	FW
work_zhl5dw75nfhy7oputsllvjjcqq	113	1	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	113	2	article	article	NN
work_zhl5dw75nfhy7oputsllvjjcqq	113	3	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	113	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	113	5	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	113	6	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	113	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	113	8	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	113	9	opinion	opinion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	113	10	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	113	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	113	12	Courts	Courts	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	113	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	113	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	113	15	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	113	16	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	113	17	post	post	NN
work_zhl5dw75nfhy7oputsllvjjcqq	113	18	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	113	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	113	20	Reddit	Reddit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	113	21	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	113	22	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	113	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	114	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	114	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	114	3	second	second	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	114	4	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	114	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	114	6	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	114	7	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	114	8	instance	instance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	114	9	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	114	10	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	114	11	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	114	12	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	114	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	115	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	115	2	ex-	ex-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	115	3	pect	pect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	115	4	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	115	5	choice	choice	NN
work_zhl5dw75nfhy7oputsllvjjcqq	115	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	115	7	segmentation	segmentation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	115	8	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	115	9	have	have	VB
work_zhl5dw75nfhy7oputsllvjjcqq	115	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	115	11	largest	large	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	115	12	impact	impact	NN
work_zhl5dw75nfhy7oputsllvjjcqq	115	13	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	115	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	115	15	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	115	16	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	115	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	116	1	Documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	116	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	116	3	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	116	4	characterized	characterize	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	116	5	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	116	6	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	116	7	bursty	bursty	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	116	8	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	116	9	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	116	10	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	116	11	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	116	12	locally	locally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	116	13	frequent	frequent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	116	14	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	116	15	globally	globally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	116	16	rare	rare	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	116	17	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	116	18	Madsen	Madsen	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	116	19	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	116	20	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	116	21	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	116	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	116	23	2005	2005	CD
work_zhl5dw75nfhy7oputsllvjjcqq	116	24	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	116	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	116	26	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	116	27	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	116	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	116	29	name	name	NN
work_zhl5dw75nfhy7oputsllvjjcqq	116	30	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	116	31	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	116	32	defendant	defendant	NN
work_zhl5dw75nfhy7oputsllvjjcqq	116	33	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	116	34	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	116	35	court	court	NN
work_zhl5dw75nfhy7oputsllvjjcqq	116	36	case	case	NN
work_zhl5dw75nfhy7oputsllvjjcqq	116	37	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	117	1	Sampling	sample	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	117	2	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	117	3	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	117	4	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	117	5	replacement	replacement	NN
work_zhl5dw75nfhy7oputsllvjjcqq	117	6	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	117	7	magnify	magnify	VB
work_zhl5dw75nfhy7oputsllvjjcqq	117	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	117	9	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	117	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	117	11	bursty	bursty	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	117	12	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	117	13	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	117	14	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	117	15	rare	rare	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	117	16	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	117	17	locally	locally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	117	18	frequent	frequent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	117	19	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	117	20	will	will	MD
work_zhl5dw75nfhy7oputsllvjjcqq	117	21	either	either	CC
work_zhl5dw75nfhy7oputsllvjjcqq	117	22	occur	occur	VB
work_zhl5dw75nfhy7oputsllvjjcqq	117	23	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	117	24	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	117	25	Bootstrap	Bootstrap	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	117	26	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	117	27	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	117	28	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	117	29	occur	occur	VB
work_zhl5dw75nfhy7oputsllvjjcqq	117	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	118	1	Sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	118	2	sentences	sentence	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	118	3	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	118	4	replacement	replacement	NN
work_zhl5dw75nfhy7oputsllvjjcqq	118	5	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	118	6	have	have	VB
work_zhl5dw75nfhy7oputsllvjjcqq	118	7	less	less	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	118	8	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	118	9	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	118	10	bursty	bursty	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	118	11	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	118	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	118	13	since	since	IN
work_zhl5dw75nfhy7oputsllvjjcqq	118	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	118	15	chance	chance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	118	16	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	118	17	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	118	18	entire	entire	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	118	19	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	118	20	will	will	MD
work_zhl5dw75nfhy7oputsllvjjcqq	118	21	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	118	22	removed	remove	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	118	23	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	118	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	118	25	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	118	26	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	118	27	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	118	28	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	118	29	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	119	1	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	119	2	Algorithms	Algorithms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	119	3	Evaluating	evaluate	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	119	4	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	119	5	current	current	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	119	6	embedding	embedding	NN
work_zhl5dw75nfhy7oputsllvjjcqq	119	7	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	119	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	119	9	im-	im-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	119	10	plementations	plementation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	119	11	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	119	12	beyond	beyond	IN
work_zhl5dw75nfhy7oputsllvjjcqq	119	13	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	119	14	scope	scope	NN
work_zhl5dw75nfhy7oputsllvjjcqq	119	15	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	119	16	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	119	17	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	119	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	119	19	so	so	RB
work_zhl5dw75nfhy7oputsllvjjcqq	119	20	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	119	21	select	select	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	119	22	four	four	CD
work_zhl5dw75nfhy7oputsllvjjcqq	119	23	categories	category	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	119	24	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	119	25	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	119	26	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	119	27	represent	represent	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	119	28	distinct	distinct	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	119	29	optimization	optimization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	119	30	strategies	strategy	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	119	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	120	1	Recall	recall	VB
work_zhl5dw75nfhy7oputsllvjjcqq	120	2	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	120	3	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	120	4	goal	goal	NN
work_zhl5dw75nfhy7oputsllvjjcqq	120	5	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	120	6	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	120	7	examine	examine	VB
work_zhl5dw75nfhy7oputsllvjjcqq	120	8	how	how	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	120	9	algorithms	algorithms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	120	10	respond	respond	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	120	11	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	120	12	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	120	13	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	120	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	120	15	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	120	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	120	17	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	120	18	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	120	19	maximize	maximize	VB
work_zhl5dw75nfhy7oputsllvjjcqq	120	20	performance	performance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	120	21	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	120	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	120	23	accuracy	accuracy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	120	24	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	120	25	effectiveness	effectiveness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	120	26	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	120	27	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	120	28	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	120	29	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	121	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	121	2	first	first	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	121	3	category	category	NN
work_zhl5dw75nfhy7oputsllvjjcqq	121	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	121	5	online	online	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	121	6	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	121	7	updates	update	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	121	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	121	9	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	121	10	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	121	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	121	12	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	121	13	updates	update	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	121	14	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	121	15	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	121	16	us-	us-	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	121	17	ing	e	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	121	18	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	121	19	gradients	gradient	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	121	20	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	121	21	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	121	22	proceeds	proceed	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	121	23	through	through	IN
work_zhl5dw75nfhy7oputsllvjjcqq	121	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	121	25	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	121	26	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	121	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	122	1	All	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	122	2	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	122	3	implemented	implement	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	122	4	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	122	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	122	6	word2vec	word2vec	NN
work_zhl5dw75nfhy7oputsllvjjcqq	122	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	122	8	fastText	fasttext	NN
work_zhl5dw75nfhy7oputsllvjjcqq	122	9	packages	package	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	122	10	follow	follow	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	122	11	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	122	12	format	format	NN
work_zhl5dw75nfhy7oputsllvjjcqq	122	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	122	14	including	include	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	122	15	skip	skip	NN
work_zhl5dw75nfhy7oputsllvjjcqq	122	16	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	122	17	gram	gram	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	122	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	122	19	CBOW	CBOW	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	122	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	122	21	negative	negative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	122	22	sam-	sam-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	122	23	pling	pling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	122	24	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	122	25	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	122	26	hierarchical	hierarchical	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	122	27	softmax	softmax	NN
work_zhl5dw75nfhy7oputsllvjjcqq	122	28	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	122	29	Mikolov	Mikolov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	122	30	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	122	31	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	122	32	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	122	33	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	122	34	2013	2013	CD
work_zhl5dw75nfhy7oputsllvjjcqq	122	35	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	122	36	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	123	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	123	2	focus	focus	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	123	3	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	123	4	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	123	5	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	123	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	123	7	popular	popular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	123	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	123	9	representative	representative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	123	10	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	123	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	124	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	124	2	second	second	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	124	3	category	category	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	124	5	batch	batch	VB
work_zhl5dw75nfhy7oputsllvjjcqq	124	6	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	124	7	updates	update	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	124	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	124	9	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	124	10	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	124	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	124	12	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	124	13	first	first	RB
work_zhl5dw75nfhy7oputsllvjjcqq	124	14	collects	collect	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	124	15	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	124	16	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	17	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	124	18	summary	summary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	19	statistics	statistic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	124	20	derived	derive	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	124	21	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	124	22	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	124	23	pass	pass	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	24	through	through	IN
work_zhl5dw75nfhy7oputsllvjjcqq	124	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	124	26	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	27	data	datum	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	124	28	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	124	29	takes	take	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	124	30	place	place	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	31	before	before	IN
work_zhl5dw75nfhy7oputsllvjjcqq	124	32	any	any	DT
work_zhl5dw75nfhy7oputsllvjjcqq	124	33	parame-	parame-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	34	ters	ter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	124	35	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	124	36	set	set	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	124	37	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	124	38	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	124	39	then	then	RB
work_zhl5dw75nfhy7oputsllvjjcqq	124	40	updates	update	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	124	41	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	42	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	124	43	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	124	44	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	124	45	optimization	optimization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	124	46	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	125	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	125	2	select	select	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	125	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	125	4	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	125	5	algo-	algo-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	125	6	rithm	rithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	125	7	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	125	8	Pennington	Pennington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	125	9	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	125	10	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	125	11	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	125	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	125	13	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	125	14	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	125	15	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	125	16	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	125	17	representative	representative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	125	18	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	125	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	126	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	126	2	third	third	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	126	3	category	category	NN
work_zhl5dw75nfhy7oputsllvjjcqq	126	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	126	5	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	126	6	factorization	factorization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	126	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	126	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	126	9	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	126	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	126	11	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	126	12	makes	make	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	126	13	deterministic	deterministic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	126	14	updates	update	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	126	15	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	126	16	model	model	VB
work_zhl5dw75nfhy7oputsllvjjcqq	126	17	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	126	18	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	126	19	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	126	20	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	126	21	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	126	22	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	126	23	summary	summary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	126	24	statistics	statistic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	126	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	127	1	As	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	127	2	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	127	3	representative	representative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	127	4	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	127	5	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	127	6	include	include	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	127	7	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	127	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	127	9	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	127	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	127	11	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	127	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	127	13	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	127	14	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	127	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	128	1	Finally	finally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	128	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	128	3	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	128	4	test	test	VB
work_zhl5dw75nfhy7oputsllvjjcqq	128	5	whether	whether	IN
work_zhl5dw75nfhy7oputsllvjjcqq	128	6	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	7	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	8	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	128	9	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	128	10	significant	significant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	128	11	factor	factor	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	12	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	128	13	include	include	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	128	14	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	128	15	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	16	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	128	17	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	128	18	embedding	embedding	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	19	method	method	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	20	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	128	21	uses	use	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	128	22	ma-	ma-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	128	23	trix	trix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	24	factorization	factorization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	128	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	128	26	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	128	27	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	128	28	Deerwester	Deerwester	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	128	29	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	128	30	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	128	31	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	128	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	128	33	1990	1990	CD
work_zhl5dw75nfhy7oputsllvjjcqq	128	34	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	128	35	Lan-	Lan-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	128	36	dauer	dauer	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	128	37	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	128	38	Dumais	Dumais	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	128	39	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	128	40	1997	1997	CD
work_zhl5dw75nfhy7oputsllvjjcqq	128	41	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	128	42	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	129	1	These	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	129	2	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	129	3	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	129	4	include	include	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	129	5	several	several	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	129	6	hyperparam-	hyperparam-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	129	7	eters	eter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	129	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	129	9	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	129	10	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	129	11	known	know	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	129	12	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	129	13	have	have	VB
work_zhl5dw75nfhy7oputsllvjjcqq	129	14	measurable	measurable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	129	15	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	129	16	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	129	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	129	18	resulting	result	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	129	19	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	129	20	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	129	21	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	129	22	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	129	23	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	129	24	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	129	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	129	26	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	129	27	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	129	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	130	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	130	2	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	130	3	attempted	attempt	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	130	4	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	130	5	choose	choose	VB
work_zhl5dw75nfhy7oputsllvjjcqq	130	6	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	130	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	8	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	130	9	parame-	parame-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	130	10	ters	ter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	130	11	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	130	12	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	130	13	commonly	commonly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	130	14	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	130	15	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	130	16	comparable	comparable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	130	17	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	18	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	130	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	130	20	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	130	21	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	130	22	emphasize	emphasize	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	130	23	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	24	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	130	25	full	full	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	130	26	evaluation	evaluation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	130	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	130	29	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	130	30	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	31	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	130	32	algorithmic	algorithmic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	130	33	parameter	parameter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	130	34	would	would	MD
work_zhl5dw75nfhy7oputsllvjjcqq	130	35	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	130	36	beyond	beyond	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	130	38	scope	scope	NN
work_zhl5dw75nfhy7oputsllvjjcqq	130	39	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	130	40	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	130	41	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	130	42	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	131	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	131	2	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	131	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	131	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	131	5	following	follow	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	131	6	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	131	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	131	8	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	131	9	set	set	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	131	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	131	11	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	131	12	window	window	NN
work_zhl5dw75nfhy7oputsllvjjcqq	131	13	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	131	14	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	131	15	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	131	16	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	131	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	131	18	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	131	19	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	131	20	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	131	21	100	100	CD
work_zhl5dw75nfhy7oputsllvjjcqq	131	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	132	1	Since	since	IN
work_zhl5dw75nfhy7oputsllvjjcqq	132	2	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	132	3	re-	re-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	132	4	move	move	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	132	5	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	132	6	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	132	7	occur	occur	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	132	8	fewer	few	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	132	9	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	132	10	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	132	11	times	time	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	132	12	during	during	IN
work_zhl5dw75nfhy7oputsllvjjcqq	132	13	preprocessing	preprocessing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	132	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	132	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	132	16	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	132	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	132	18	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	132	19	set	set	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	132	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	132	21	frequency	frequency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	132	22	threshold	threshold	NN
work_zhl5dw75nfhy7oputsllvjjcqq	132	23	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	132	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	132	25	following	follow	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	132	26	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	132	27	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	132	28	0	0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	132	29	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	133	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	133	2	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	133	3	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	133	4	hyperparameters	hyperparameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	133	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	133	6	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	133	7	follow	follow	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	133	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	133	9	de-	de-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	133	10	fault	fault	NN
work_zhl5dw75nfhy7oputsllvjjcqq	133	11	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	133	12	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	133	13	popular	popular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	133	14	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	133	15	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	133	16	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	133	17	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	133	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	133	19	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	133	20	described	describe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	133	21	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	133	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	133	23	following	following	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	133	24	sections	section	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	133	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	134	1	5.1	5.1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	134	2	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	3	Latent	Latent	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	4	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	134	5	analysis	analysis	NN
work_zhl5dw75nfhy7oputsllvjjcqq	134	6	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	134	7	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	8	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	134	9	factorizes	factorize	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	134	10	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	134	11	sparse	sparse	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	134	12	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	134	13	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	134	14	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	134	15	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	134	16	X	X	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	17	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	134	18	Deerwester	Deerwester	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	19	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	134	20	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	21	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	134	23	1990	1990	CD
work_zhl5dw75nfhy7oputsllvjjcqq	134	24	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	134	25	Landauer	Landauer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	26	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	134	27	Dumais	Dumais	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	134	28	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	134	29	1997	1997	CD
work_zhl5dw75nfhy7oputsllvjjcqq	134	30	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	134	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	135	1	X	X	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	2	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	135	3	factored	factor	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	135	4	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	135	5	singular	singular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	135	6	value	value	NN
work_zhl5dw75nfhy7oputsllvjjcqq	135	7	decomposition	decomposition	NN
work_zhl5dw75nfhy7oputsllvjjcqq	135	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	135	9	SVD	SVD	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	10	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	135	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	135	12	retaining	retain	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	135	13	K	K	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	14	singular	singular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	135	15	values	value	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	135	16	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	135	17	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	135	18	X	X	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	19	≈	≈	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	20	XK	XK	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	21	=	=	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	135	22	UK	UK	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	23	ΣKV	ΣKV	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	24	TK	TK	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	135	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	136	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	136	2	elements	element	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	136	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	136	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	136	5	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	6	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	136	7	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	8	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	9	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	136	10	weighted	weight	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	136	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	136	12	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	136	13	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	136	14	TF	TF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	136	15	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	136	16	IDF	IDF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	136	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	136	18	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	136	19	measures	measure	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	136	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	136	21	111	111	CD
work_zhl5dw75nfhy7oputsllvjjcqq	136	22	importance	importance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	136	24	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	136	25	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	26	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	136	27	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	136	28	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	29	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	136	30	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	136	31	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	136	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	137	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	137	2	dense	dense	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	137	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	137	4	low	low	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	137	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	137	6	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	7	approximation	approximation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	137	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	137	10	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	11	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	137	12	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	13	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	137	15	XK	XK	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	137	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	137	17	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	137	18	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	137	19	used	use	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	137	20	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	137	21	measure	measure	VB
work_zhl5dw75nfhy7oputsllvjjcqq	137	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	137	23	relatedness	relatedness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	24	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	137	25	terms	term	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	137	26	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	137	27	calculating	calculate	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	137	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	137	29	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	30	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	31	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	137	32	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	137	33	relevant	relevant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	137	34	rows	row	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	137	35	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	137	36	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	137	37	reduced	reduced	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	137	38	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	137	39	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	138	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	138	2	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	138	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	138	4	sci	sci	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	138	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	138	6	kit	kit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	138	7	learn3	learn3	NN
work_zhl5dw75nfhy7oputsllvjjcqq	138	8	package	package	NN
work_zhl5dw75nfhy7oputsllvjjcqq	138	9	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	138	10	train	train	VB
work_zhl5dw75nfhy7oputsllvjjcqq	138	11	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	138	12	LSA	lsa	NN
work_zhl5dw75nfhy7oputsllvjjcqq	138	13	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	138	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	139	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	139	2	create	create	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	139	3	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	139	4	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	139	6	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	7	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	8	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	139	9	TF	TF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	139	10	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	139	11	IDF	IDF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	139	12	weighting	weighting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	139	14	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	139	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	139	16	default	default	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	17	set-	set-	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	139	18	tings	ting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	139	19	except	except	IN
work_zhl5dw75nfhy7oputsllvjjcqq	139	20	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	139	21	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	139	22	add	add	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	139	23	L2	l2	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	139	24	normalization	normalization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	25	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	139	26	sub-	sub-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	139	27	linear	linear	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	139	28	TF	TF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	139	29	scaling	scaling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	30	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	139	31	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	139	32	scales	scale	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	139	33	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	139	34	importance	importance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	35	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	139	36	terms	term	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	139	37	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	139	38	high	high	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	139	39	frequency	frequency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	40	within	within	IN
work_zhl5dw75nfhy7oputsllvjjcqq	139	41	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	139	42	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	139	43	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	140	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	140	2	perform	perform	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	140	3	dimensionality	dimensionality	NN
work_zhl5dw75nfhy7oputsllvjjcqq	140	4	reduction	reduction	NN
work_zhl5dw75nfhy7oputsllvjjcqq	140	5	via	via	IN
work_zhl5dw75nfhy7oputsllvjjcqq	140	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	140	7	randomized	randomized	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	140	8	solver	solver	NN
work_zhl5dw75nfhy7oputsllvjjcqq	140	9	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	140	10	Halko	Halko	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	140	11	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	140	12	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	140	13	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	140	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	140	15	September	September	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	140	16	2009	2009	CD
work_zhl5dw75nfhy7oputsllvjjcqq	140	17	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	140	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	141	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	141	2	construction	construction	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	141	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	141	5	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	6	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	141	7	count	count	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	8	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	141	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	141	11	TF	TF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	141	12	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	141	13	IDF	IDF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	141	14	weighting	weighting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	15	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	141	16	introduce	introduce	VB
work_zhl5dw75nfhy7oputsllvjjcqq	141	17	no	no	DT
work_zhl5dw75nfhy7oputsllvjjcqq	141	18	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	19	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	141	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	141	21	final	final	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	141	22	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	141	23	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	141	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	142	1	However	however	RB
work_zhl5dw75nfhy7oputsllvjjcqq	142	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	142	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	142	4	expect	expect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	142	5	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	142	6	due	due	IN
work_zhl5dw75nfhy7oputsllvjjcqq	142	7	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	142	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	142	9	randomized	randomized	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	142	10	SVD	SVD	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	142	11	solver	solver	NN
work_zhl5dw75nfhy7oputsllvjjcqq	142	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	142	13	even	even	RB
work_zhl5dw75nfhy7oputsllvjjcqq	142	14	when	when	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	142	15	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	142	16	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	142	17	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	142	18	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	142	19	training	train	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	142	20	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	142	21	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	142	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	142	23	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	142	24	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	142	25	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	142	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	142	27	etc	etc	FW
work_zhl5dw75nfhy7oputsllvjjcqq	142	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	142	29	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	143	1	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	143	2	constant	constant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	143	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	144	1	5.2	5.2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	144	2	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	144	3	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	144	4	skip	skip	NN
work_zhl5dw75nfhy7oputsllvjjcqq	144	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	144	6	gram	gram	NN
work_zhl5dw75nfhy7oputsllvjjcqq	144	7	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	144	8	negative	negative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	144	9	sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	144	10	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	144	11	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	144	12	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	144	13	algo-	algo-	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	144	14	rithm	rithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	144	15	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	144	16	Mikolov	Mikolov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	144	17	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	144	18	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	144	19	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	144	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	144	21	2013	2013	CD
work_zhl5dw75nfhy7oputsllvjjcqq	144	22	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	144	23	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	144	24	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	144	25	online	online	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	144	26	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	144	27	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	144	28	uses	use	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	144	29	randomized	randomize	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	144	30	updates	update	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	144	31	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	144	32	predict	predict	VB
work_zhl5dw75nfhy7oputsllvjjcqq	144	33	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	144	34	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	144	35	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	144	36	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	144	37	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	144	38	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	145	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	145	2	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	145	3	iteration	iteration	NN
work_zhl5dw75nfhy7oputsllvjjcqq	145	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	145	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	145	6	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	145	7	pro-	pro-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	145	8	ceeds	ceed	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	145	9	through	through	IN
work_zhl5dw75nfhy7oputsllvjjcqq	145	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	145	11	original	original	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	145	12	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	145	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	145	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	145	15	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	145	16	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	145	17	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	145	18	token	token	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	145	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	145	20	updates	update	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	145	21	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	145	22	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	145	23	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	145	24	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	145	25	gra-	gra-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	145	26	dients	dient	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	145	27	calculated	calculate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	145	28	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	145	29	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	145	30	current	current	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	145	31	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	145	32	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	145	33	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	146	1	This	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	146	2	process	process	NN
work_zhl5dw75nfhy7oputsllvjjcqq	146	3	maximizes	maximize	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	146	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	146	5	likelihood	likelihood	NN
work_zhl5dw75nfhy7oputsllvjjcqq	146	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	146	7	observed	observe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	146	8	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	146	9	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	146	10	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	146	11	pairs	pair	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	146	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	146	13	minimizes	minimize	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	146	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	146	15	likelihood	likelihood	NN
work_zhl5dw75nfhy7oputsllvjjcqq	146	16	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	146	17	negative	negative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	146	18	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	146	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	147	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	147	2	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	147	3	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	147	4	implementation	implementation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	147	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	147	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	147	7	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	147	8	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	147	9	included	include	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	147	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	147	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	147	12	Python	Python	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	147	13	library	library	NN
work_zhl5dw75nfhy7oputsllvjjcqq	147	14	gensim4	gensim4	NN
work_zhl5dw75nfhy7oputsllvjjcqq	147	15	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	147	16	Řehůřek	Řehůřek	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	147	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	147	18	Sojka	sojka	RB
work_zhl5dw75nfhy7oputsllvjjcqq	147	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	147	20	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	147	21	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	147	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	148	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	148	2	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	148	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	148	4	default	default	NN
work_zhl5dw75nfhy7oputsllvjjcqq	148	5	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	148	6	pro-	pro-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	148	7	vided	vide	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	148	8	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	148	9	gensim	gensim	NN
work_zhl5dw75nfhy7oputsllvjjcqq	148	10	except	except	IN
work_zhl5dw75nfhy7oputsllvjjcqq	148	11	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	148	12	described	describe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	148	13	above	above	RB
work_zhl5dw75nfhy7oputsllvjjcqq	148	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	149	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	149	2	predict	predict	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	149	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	149	4	multiple	multiple	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	149	5	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	149	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	149	7	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	149	8	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	149	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	149	10	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	149	11	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	149	12	will	will	MD
work_zhl5dw75nfhy7oputsllvjjcqq	149	13	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	149	14	produce	produce	VB
work_zhl5dw75nfhy7oputsllvjjcqq	149	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	149	16	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	149	17	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	149	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	150	1	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	2	ran-	ran-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	150	3	domly	domly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	150	4	initializes	initialize	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	150	5	all	all	PDT
work_zhl5dw75nfhy7oputsllvjjcqq	150	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	150	7	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	150	8	before	before	IN
work_zhl5dw75nfhy7oputsllvjjcqq	150	9	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	150	10	begins	begin	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	150	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	150	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	150	13	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	150	14	relies	rely	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	150	15	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	150	16	negative	negative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	150	17	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	150	18	created	create	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	150	19	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	150	20	randomly	randomly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	150	21	selecting	select	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	150	22	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	150	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	150	24	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	150	25	pairs	pair	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	150	26	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	150	27	Mikolov	Mikolov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	28	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	150	29	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	30	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	31	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	150	32	2013	2013	CD
work_zhl5dw75nfhy7oputsllvjjcqq	150	33	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	150	34	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	35	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	36	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	37	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	150	38	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	150	39	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	150	40	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	150	41	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	151	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	151	2	also	also	RB
work_zhl5dw75nfhy7oputsllvjjcqq	151	3	expect	expect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	151	4	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	151	5	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	151	6	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	151	7	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	151	8	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	151	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	151	10	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	151	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	151	12	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	151	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	151	14	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	151	15	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	151	16	relies	rely	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	151	17	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	151	18	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	151	19	gradient	gradient	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	151	20	descent	descent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	151	21	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	151	22	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	151	23	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	151	24	biased	bias	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	151	25	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	151	26	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	151	27	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	151	28	influenced	influence	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	151	29	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	151	30	initial	initial	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	151	31	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	151	32	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	151	33	Bottou	Bottou	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	151	34	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	151	35	2012	2012	CD
work_zhl5dw75nfhy7oputsllvjjcqq	151	36	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	151	37	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	152	1	3http://scikit-learn.org/	3http://scikit-learn.org/	CD
work_zhl5dw75nfhy7oputsllvjjcqq	152	2	4https://radimrehurek.com/gensim/models/	4https://radimrehurek.com/gensim/models/	CD
work_zhl5dw75nfhy7oputsllvjjcqq	152	3	word2vec.html	word2vec.html	CD
work_zhl5dw75nfhy7oputsllvjjcqq	152	4	5.3	5.3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	152	5	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	6	Global	Global	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	7	Vectors	Vectors	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	152	8	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	152	9	Word	Word	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	10	Representation	Representation	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	11	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	152	12	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	13	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	152	14	uses	use	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	152	15	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	152	16	gradient	gradient	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	17	updates	updates	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	152	18	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	152	19	operates	operate	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	152	20	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	152	21	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	152	22	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	152	23	global	global	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	152	24	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	152	25	representation	representation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	26	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	152	27	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	28	co	co	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	29	-	-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	30	occurrence	occurrence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	31	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	152	32	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	152	33	calcu-	calcu-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	34	lated	late	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	152	35	once	once	RB
work_zhl5dw75nfhy7oputsllvjjcqq	152	36	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	152	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	152	38	beginning	beginning	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	39	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	152	40	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	152	41	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	42	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	152	43	Penning-	Penning-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	44	ton	ton	NN
work_zhl5dw75nfhy7oputsllvjjcqq	152	45	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	152	46	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	47	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	152	48	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	152	49	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	152	50	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	152	51	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	153	1	Words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	153	2	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	153	3	contexts	contexts	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	4	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	153	5	associated	associate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	153	6	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	153	7	bias	bias	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	8	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	153	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	153	10	bw	bw	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	153	12	bc	bc	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	153	14	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	153	15	w	w	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	16	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	153	17	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	153	18	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	19	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	153	20	c	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	21	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	153	22	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	153	23	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	24	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	153	25	learned	learn	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	153	26	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	153	27	minimizing	minimize	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	153	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	153	29	cost	cost	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	30	function	function	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	31	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	153	32	L	l	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	33	=	=	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	153	34	∑	∑	.
work_zhl5dw75nfhy7oputsllvjjcqq	153	35	w	w	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	36	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	153	37	c	c	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	38	f(xwc	f(xwc	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	39	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	153	40	~w	~w	NFP
work_zhl5dw75nfhy7oputsllvjjcqq	153	41	·	·	NFP
work_zhl5dw75nfhy7oputsllvjjcqq	153	42	~c	~c	NFP
work_zhl5dw75nfhy7oputsllvjjcqq	153	43	+	+	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	153	44	bw	bw	NN
work_zhl5dw75nfhy7oputsllvjjcqq	153	45	+	+	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	153	46	bc	bc	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	47	−	−	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	48	log(xwc	log(xwc	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	153	49	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	153	50	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	154	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	154	2	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	154	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	154	4	GloVe	glove	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	154	5	implementation	implementation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	154	6	provided	provide	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	154	7	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	154	8	Pennington	Pennington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	154	9	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	154	10	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	154	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	155	1	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	155	2	2014)5	2014)5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	155	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	156	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	156	2	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	156	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	156	4	default	default	NN
work_zhl5dw75nfhy7oputsllvjjcqq	156	5	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	156	6	provided	provide	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	156	7	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	156	8	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	156	9	except	except	IN
work_zhl5dw75nfhy7oputsllvjjcqq	156	10	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	156	11	described	describe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	156	12	above	above	RB
work_zhl5dw75nfhy7oputsllvjjcqq	156	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	157	1	Unlike	unlike	IN
work_zhl5dw75nfhy7oputsllvjjcqq	157	2	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	157	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	157	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	157	5	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	157	6	does	do	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	157	7	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	157	8	perform	perform	VB
work_zhl5dw75nfhy7oputsllvjjcqq	157	9	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	157	10	updates	update	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	157	11	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	157	12	examining	examine	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	157	13	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	157	14	original	original	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	157	15	docu-	docu-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	157	16	ments	ment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	157	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	158	1	As	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	158	2	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	158	3	result	result	NN
work_zhl5dw75nfhy7oputsllvjjcqq	158	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	158	5	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	158	6	expect	expect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	158	7	GloVe	GloVe	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	158	8	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	158	9	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	158	10	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	158	11	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	158	12	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	158	13	initializations	initialization	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	158	14	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	158	15	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	158	16	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	158	17	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	158	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	158	19	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	158	20	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	158	21	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	158	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	159	1	5.4	5.4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	159	2	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	3	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	159	4	positive	positive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	159	5	pointwise	pointwise	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	6	mutual	mutual	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	159	7	information	information	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	159	9	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	10	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	159	11	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	159	13	whose	whose	WP$
work_zhl5dw75nfhy7oputsllvjjcqq	159	14	cells	cell	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	159	15	represent	represent	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	159	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	159	17	PPMI	ppmi	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	159	19	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	159	20	pair	pair	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	159	22	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	159	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	159	24	contexts	contexts	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	159	26	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	159	27	factored	factor	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	159	28	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	159	29	sin-	sin-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	159	30	gular	gular	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	159	31	value	value	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	32	decomposition	decomposition	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	33	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	159	34	SVD	SVD	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	35	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	159	36	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	159	37	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	159	38	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	159	39	low-	low-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	159	40	dimensional	dimensional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	159	41	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	159	42	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	159	43	perform	perform	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	159	44	similarly	similarly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	159	45	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	159	46	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	47	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	159	48	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	49	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	159	50	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	51	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	159	52	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	159	53	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	159	54	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	159	55	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	159	56	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	160	1	PMI(w	pmi(w	NN
work_zhl5dw75nfhy7oputsllvjjcqq	160	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	160	3	c	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	160	4	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	5	=	=	NFP
work_zhl5dw75nfhy7oputsllvjjcqq	160	6	log	log	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	7	P	P	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	9	w	w	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	160	11	c	c	LS
work_zhl5dw75nfhy7oputsllvjjcqq	160	12	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	13	P	p	NN
work_zhl5dw75nfhy7oputsllvjjcqq	160	14	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	15	w)P	w)P	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	16	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	17	c	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	160	18	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	19	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	160	20	PPMI(w	PPMI(w	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	160	22	c	c	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	23	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	24	=	=	NFP
work_zhl5dw75nfhy7oputsllvjjcqq	160	25	max(PMI(w	max(PMI(w	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	160	26	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	160	27	c	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	160	28	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	29	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	160	30	0	0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	160	31	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	160	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	161	1	To	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	161	2	train	train	VB
work_zhl5dw75nfhy7oputsllvjjcqq	161	3	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	161	4	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	161	5	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	161	6	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	161	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	161	8	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	161	9	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	161	10	hyperwords,6	hyperwords,6	NN
work_zhl5dw75nfhy7oputsllvjjcqq	161	11	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	161	12	implementation	implementation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	161	13	provided	provide	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	161	14	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	161	15	part	part	NN
work_zhl5dw75nfhy7oputsllvjjcqq	161	16	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	161	17	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	161	18	et	et	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	161	19	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	161	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	162	1	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	2	2015).7	2015).7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	162	3	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	162	4	follow	follow	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	162	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	162	6	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	162	7	’	’	POS
work_zhl5dw75nfhy7oputsllvjjcqq	162	8	recom-	recom-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	162	9	mendations	mendation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	162	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	162	11	set	set	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	162	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	162	13	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	14	distributional	distributional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	162	15	smooth-	smooth-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	16	ing	ing	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	162	17	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	18	cds	cd	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	162	19	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	20	parameter	parameter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	21	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	162	22	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	162	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	162	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	162	25	eigenvalue	eigenvalue	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	26	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	27	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	28	eig	eig	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	29	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	30	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	162	31	0.5	0.5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	162	32	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	162	33	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	162	34	subsampling	subsampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	35	threshold	threshold	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	36	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	37	sub	sub	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	162	38	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	39	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	162	40	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	162	41	-	-	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	162	42	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	162	43	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	162	44	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	162	45	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	162	46	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	47	window	window	NN
work_zhl5dw75nfhy7oputsllvjjcqq	162	48	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	49	win	win	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	162	50	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	162	51	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	162	52	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	162	53	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	163	1	5http://nlp.stanford.edu/projects/glove/	5http://nlp.stanford.edu/projects/glove/	CD
work_zhl5dw75nfhy7oputsllvjjcqq	163	2	6https://bitbucket.org/omerlevy/	6https://bitbucket.org/omerlevy/	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	163	3	hyperwords	hyperwords	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	163	4	/	/	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	163	5	src	src	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	163	6	7We	7we	CD
work_zhl5dw75nfhy7oputsllvjjcqq	163	7	altered	alter	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	163	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	163	9	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	163	10	code	code	NN
work_zhl5dw75nfhy7oputsllvjjcqq	163	11	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	163	12	remove	remove	VB
work_zhl5dw75nfhy7oputsllvjjcqq	163	13	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	163	14	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	163	15	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	163	16	seed	seed	NN
work_zhl5dw75nfhy7oputsllvjjcqq	163	17	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	163	18	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	163	19	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	163	20	introduce	introduce	VB
work_zhl5dw75nfhy7oputsllvjjcqq	163	21	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	163	22	given	give	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	163	23	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	163	24	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	163	25	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	163	26	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	163	27	no	no	DT
work_zhl5dw75nfhy7oputsllvjjcqq	163	28	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	163	29	change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	163	30	was	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	163	31	made	make	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	163	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	164	1	112	112	CD
work_zhl5dw75nfhy7oputsllvjjcqq	164	2	Like	like	IN
work_zhl5dw75nfhy7oputsllvjjcqq	164	3	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	164	4	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	164	5	unlike	unlike	IN
work_zhl5dw75nfhy7oputsllvjjcqq	164	6	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	164	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	164	8	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	164	9	operates	operate	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	164	10	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	164	11	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	164	12	pre	pre	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	164	13	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	164	14	computed	computed	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	164	15	representation	representation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	164	16	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	164	17	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	164	18	co	co	NN
work_zhl5dw75nfhy7oputsllvjjcqq	164	19	-	-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	164	20	occurrence	occurrence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	164	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	164	22	so	so	RB
work_zhl5dw75nfhy7oputsllvjjcqq	164	23	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	164	24	do	do	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	164	25	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	164	26	expect	expect	VB
work_zhl5dw75nfhy7oputsllvjjcqq	164	27	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	164	28	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	164	29	vary	vary	VB
work_zhl5dw75nfhy7oputsllvjjcqq	164	30	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	164	31	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	164	32	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	164	33	or-	or-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	164	34	der	der	NN
work_zhl5dw75nfhy7oputsllvjjcqq	164	35	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	164	36	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	164	37	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	165	1	Unlike	unlike	IN
work_zhl5dw75nfhy7oputsllvjjcqq	165	2	both	both	CC
work_zhl5dw75nfhy7oputsllvjjcqq	165	3	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	165	4	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	165	5	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	165	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	165	7	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	165	8	uses	use	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	165	9	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	165	10	stable	stable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	165	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	165	12	non	non	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	165	13	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	165	14	stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	165	15	SVD	SVD	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	165	16	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	165	17	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	165	18	should	should	MD
work_zhl5dw75nfhy7oputsllvjjcqq	165	19	produce	produce	VB
work_zhl5dw75nfhy7oputsllvjjcqq	165	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	165	21	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	165	22	result	result	NN
work_zhl5dw75nfhy7oputsllvjjcqq	165	23	given	give	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	165	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	165	25	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	165	26	input	input	NN
work_zhl5dw75nfhy7oputsllvjjcqq	165	27	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	165	28	regardless	regardless	RB
work_zhl5dw75nfhy7oputsllvjjcqq	165	29	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	165	30	initialization	initialization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	165	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	166	1	However	however	RB
work_zhl5dw75nfhy7oputsllvjjcqq	166	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	166	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	166	4	ex-	ex-	RB
work_zhl5dw75nfhy7oputsllvjjcqq	166	5	pect	pect	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	166	6	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	166	7	due	due	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	166	8	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	166	9	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	166	10	’s	’s	POS
work_zhl5dw75nfhy7oputsllvjjcqq	166	11	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	166	12	subsampling	subsampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	166	13	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	166	14	frequent	frequent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	166	15	tokens	token	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	166	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	167	1	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	167	2	Methods	Methods	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	167	3	To	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	167	4	establish	establish	VB
work_zhl5dw75nfhy7oputsllvjjcqq	167	5	statistical	statistical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	167	6	significance	significance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	167	7	bounds	bound	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	8	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	167	9	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	167	10	observations	observation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	12	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	167	13	train	train	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	167	14	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	167	15	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	167	16	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	18	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	167	19	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	167	20	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	22	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	167	23	GloVe	glove	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	167	24	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	26	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	167	27	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	167	28	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	167	29	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	30	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	167	31	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	167	32	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	167	33	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	167	34	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	167	35	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	167	36	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	167	37	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	167	38	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	39	SHUFFLED	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	167	40	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	41	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	167	42	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	167	43	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	167	44	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	45	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	167	46	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	167	47	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	167	48	segmentation	segmentation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	167	49	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	167	50	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	167	51	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	167	52	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	167	53	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	167	54	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	168	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	168	2	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	168	3	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	168	5	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	168	6	select	select	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	168	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	168	8	set	set	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	168	10	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	168	11	relevant	relevant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	168	12	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	13	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	168	14	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	168	15	high	high	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	168	16	probability	probability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	17	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	168	18	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	168	19	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	168	20	LDA	LDA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	168	21	topic	topic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	22	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	23	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	168	24	Blei	Blei	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	168	25	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	168	26	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	168	27	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	168	28	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	168	29	2003	2003	CD
work_zhl5dw75nfhy7oputsllvjjcqq	168	30	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	168	31	trained	train	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	168	32	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	168	33	that	that	DT
work_zhl5dw75nfhy7oputsllvjjcqq	168	34	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	168	35	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	168	36	200	200	CD
work_zhl5dw75nfhy7oputsllvjjcqq	168	37	topics	topic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	168	38	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	169	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	169	2	calculate	calculate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	169	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	4	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	5	sim-	sim-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	169	6	ilarity	ilarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	169	8	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	9	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	10	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	11	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	169	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	13	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	169	14	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	169	15	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	169	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	17	vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	169	19	creating	create	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	169	20	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	21	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	22	ranking	ranking	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	169	24	all	all	PDT
work_zhl5dw75nfhy7oputsllvjjcqq	169	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	26	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	169	27	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	169	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	169	29	vocabulary	vocabulary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	169	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	170	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	170	2	calculate	calculate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	170	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	170	4	mean	mean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	170	5	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	170	6	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	170	7	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	170	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	170	10	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	11	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	170	12	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	170	13	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	170	14	pair	pair	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	15	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	170	16	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	17	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	18	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	170	19	vocabulary	vocabulary	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	170	20	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	21	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	170	22	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	170	23	set	set	NN
work_zhl5dw75nfhy7oputsllvjjcqq	170	24	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	170	25	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	170	26	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	170	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	171	1	From	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	171	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	171	3	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	171	5	queries	query	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	6	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	171	7	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	171	8	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	171	10	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	171	11	select	select	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	171	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	171	13	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	171	14	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	15	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	171	16	closely	closely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	171	17	related	related	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	171	18	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	171	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	171	20	set	set	NN
work_zhl5dw75nfhy7oputsllvjjcqq	171	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	171	22	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	171	23	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	24	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	171	25	compare	compare	VB
work_zhl5dw75nfhy7oputsllvjjcqq	171	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	171	27	mean	mean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	171	28	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	171	29	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	171	30	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	171	31	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	171	32	those	those	DT
work_zhl5dw75nfhy7oputsllvjjcqq	171	33	pairs	pair	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	34	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	171	35	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	171	36	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	172	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	172	2	calculate	calculate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	172	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	172	4	Jaccard	Jaccard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	172	5	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	172	6	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	172	7	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	172	8	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	172	9	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	172	10	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	172	11	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	172	12	com-	com-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	172	13	pare	pare	NN
work_zhl5dw75nfhy7oputsllvjjcqq	172	14	membership	membership	NN
work_zhl5dw75nfhy7oputsllvjjcqq	172	15	change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	172	16	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	172	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	172	18	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	172	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	172	20	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	172	21	closely	closely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	172	22	related	relate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	172	23	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	172	24	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	172	25	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	172	26	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	172	27	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	172	28	average	average	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	172	29	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	172	30	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	172	31	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	172	32	within	within	IN
work_zhl5dw75nfhy7oputsllvjjcqq	172	33	those	those	DT
work_zhl5dw75nfhy7oputsllvjjcqq	172	34	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	172	35	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	173	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	173	2	examine	examine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	173	3	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	173	4	metrics	metric	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	173	5	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	173	6	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	173	7	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	173	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	173	9	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	173	10	parameters	parameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	173	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	174	1	7	7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	174	2	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	174	3	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	174	4	begin	begin	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	174	5	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	174	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	174	7	case	case	NN
work_zhl5dw75nfhy7oputsllvjjcqq	174	8	study	study	NN
work_zhl5dw75nfhy7oputsllvjjcqq	174	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	174	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	174	11	framing	framing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	174	12	around	around	IN
work_zhl5dw75nfhy7oputsllvjjcqq	174	13	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	174	14	query	query	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	174	15	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	174	16	marijuana	marijuana	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	174	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	175	1	One	one	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	175	2	might	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	175	3	hypothesize	hypothesize	VB
work_zhl5dw75nfhy7oputsllvjjcqq	175	4	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	175	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	175	6	authors	author	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	175	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	175	8	various	various	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	175	9	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	175	10	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	175	11	e.g.	e.g.	RB
work_zhl5dw75nfhy7oputsllvjjcqq	176	1	judges	judge	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	176	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	176	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	176	4	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	176	5	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	176	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	176	7	journalists	journalist	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	176	8	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	176	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	176	10	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	176	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	176	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	176	13	users	user	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	176	14	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	176	15	Reddit	Reddit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	176	16	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	176	17	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	176	18	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	176	19	perceptions	perception	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	176	20	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	176	21	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	176	22	drug	drug	NN
work_zhl5dw75nfhy7oputsllvjjcqq	176	23	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	176	24	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	176	25	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	176	26	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	176	27	might	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	176	28	reflect	reflect	VB
work_zhl5dw75nfhy7oputsllvjjcqq	176	29	those	those	DT
work_zhl5dw75nfhy7oputsllvjjcqq	176	30	differences	difference	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	176	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	177	1	Indeed	indeed	RB
work_zhl5dw75nfhy7oputsllvjjcqq	177	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	177	3	after	after	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	4	qualitatively	qualitatively	RB
work_zhl5dw75nfhy7oputsllvjjcqq	177	5	examining	examine	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	177	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	7	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	9	most	most	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	177	10	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	177	11	terms	term	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	12	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	177	13	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	177	14	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	15	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	16	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	177	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	177	18	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	177	19	might	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	177	20	come	come	VB
work_zhl5dw75nfhy7oputsllvjjcqq	177	21	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	23	conclusion	conclusion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	24	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	26	allegedly	allegedly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	177	27	conservative	conservative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	177	28	4th	4th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	177	29	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	30	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	31	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	32	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	33	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	34	ALGORITHM	ALGORITHM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	35	0.00	0.00	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	36	0.01	0.01	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	37	0.02	0.02	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	38	0.03	0.03	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	39	0.04	0.04	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	40	0.05	0.05	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	41	0.06	0.06	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	42	0.07	0.07	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	43	S	s	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	44	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	45	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	46	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	47	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	48	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	49	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	50	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	51	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	52	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	53	V	v	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	54	IA	ia	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	55	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	56	IO	io	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	57	N	n	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	58	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	177	59	shuffled	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	177	60	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	61	Standard	Standard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	62	Deviation	Deviation	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	63	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	64	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	65	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	66	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	67	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	68	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	69	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	70	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	71	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	72	ALGORITHM	ALGORITHM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	73	0.00	0.00	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	74	0.02	0.02	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	75	0.04	0.04	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	76	0.06	0.06	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	77	0.08	0.08	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	78	0.10	0.10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	79	0.12	0.12	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	80	S	s	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	81	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	82	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	83	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	84	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	85	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	86	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	87	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	88	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	89	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	90	V	v	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	91	IA	ia	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	92	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	93	IO	io	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	94	N	n	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	95	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	177	96	shuffled	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	177	97	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	98	Standard	Standard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	99	Deviation	Deviation	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	100	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	101	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	102	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	103	Music	Music	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	104	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	105	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	106	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	107	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	177	108	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	109	mean	mean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	177	110	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	177	111	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	112	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	113	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	114	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	177	115	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	116	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	117	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	118	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	177	119	closest	close	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	177	120	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	121	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	122	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	123	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	124	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	125	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	177	126	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	127	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	128	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	129	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	177	130	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	131	Music	Music	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	177	132	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	177	133	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	177	134	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	177	135	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	177	136	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	177	137	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	178	1	Larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	178	2	variations	variation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	178	3	indicate	indicate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	178	4	less	less	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	178	5	stable	stable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	178	6	embed-	embed-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	178	7	dings	ding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	178	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	179	1	judges	judge	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	179	2	view	view	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	179	3	marijuana	marijuana	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	179	4	as	as	RB
work_zhl5dw75nfhy7oputsllvjjcqq	179	5	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	179	6	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	179	7	illegal	illegal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	179	8	drugs	drug	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	179	9	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	179	10	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	179	11	heroin	heroin	NN
work_zhl5dw75nfhy7oputsllvjjcqq	179	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	179	13	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	179	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	179	15	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	179	16	Reddit	Reddit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	179	17	users	user	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	179	18	view	view	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	179	19	mari-	mari-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	179	20	juana	juana	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	179	21	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	179	22	closer	close	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	179	23	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	179	24	legal	legal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	179	25	substances	substance	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	179	26	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	179	27	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	179	28	nicotine	nicotine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	179	29	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	179	30	alcohol	alcohol	NN
work_zhl5dw75nfhy7oputsllvjjcqq	179	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	180	1	However	however	RB
work_zhl5dw75nfhy7oputsllvjjcqq	180	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	180	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	180	4	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	180	5	patterns	pattern	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	180	6	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	180	7	cause	cause	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	180	8	us	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	180	9	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	180	10	lower	lower	VB
work_zhl5dw75nfhy7oputsllvjjcqq	180	11	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	180	12	confidence	confidence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	180	13	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	180	14	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	180	15	conclusions	conclusion	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	180	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	181	1	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	181	2	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	181	3	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	181	4	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	181	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	181	6	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	181	7	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	181	8	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	181	9	vary	vary	VB
work_zhl5dw75nfhy7oputsllvjjcqq	181	10	signifi-	signifi-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	181	11	cantly	cantly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	181	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	182	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	182	2	see	see	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	182	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	182	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	182	5	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	182	6	ranked	rank	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	182	7	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	182	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	182	9	chosen	choose	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	182	10	according	accord	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	182	11	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	182	12	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	182	13	mean	mean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	182	14	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	182	15	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	182	16	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	182	17	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	182	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	182	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	182	20	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	182	21	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	182	22	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	182	23	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	182	24	have	have	VB
work_zhl5dw75nfhy7oputsllvjjcqq	182	25	widely	widely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	182	26	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	182	27	mean	mean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	182	28	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	182	29	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	182	30	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	182	31	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	182	32	depending	depend	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	182	33	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	182	34	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	182	35	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	182	36	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	182	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	182	38	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	182	39	training	training	NN
work_zhl5dw75nfhy7oputsllvjjcqq	182	40	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	182	41	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	182	42	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	182	43	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	182	44	SHUFFLED	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	182	45	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	182	46	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	182	47	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	182	48	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	183	1	As	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	183	2	expected	expect	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	183	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	183	4	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	183	5	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	183	6	has	have	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	183	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	183	8	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	183	9	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	183	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	183	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	183	12	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	183	13	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	183	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	184	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	184	2	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	184	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	184	4	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	184	5	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	184	6	see	see	VB
work_zhl5dw75nfhy7oputsllvjjcqq	184	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	184	8	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	184	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	184	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	184	11	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	184	12	SVD	SVD	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	184	13	solver	solver	NN
work_zhl5dw75nfhy7oputsllvjjcqq	184	14	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	184	15	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	184	16	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	184	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	184	18	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	184	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	184	20	random	random	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	184	21	subsampling	subsampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	184	22	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	184	23	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	184	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	185	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	185	2	do	do	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	185	3	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	185	4	observe	observe	VB
work_zhl5dw75nfhy7oputsllvjjcqq	185	5	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	185	6	consistent	consistent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	185	7	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	185	8	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	185	9	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	185	10	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	185	11	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	185	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	185	13	SHUFFLED	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	185	14	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	185	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	186	1	Most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	186	2	importantly	importantly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	186	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	186	4	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	5	figures	figure	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	6	reveal	reveal	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	7	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	186	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	9	113	113	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	10	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	11	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	12	NYT	NYT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	13	Sports	Sports	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	186	14	Reddit	Reddit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	15	Ask	Ask	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	16	Science	Science	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	17	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	18	0.82	0.82	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	19	0.84	0.84	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	20	0.86	0.86	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	21	0.88	0.88	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	22	0.90	0.90	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	23	0.92	0.92	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	24	0.94	0.94	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	25	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	26	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	27	distribute	distribute	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	28	manufacture	manufacture	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	29	oxycodone	oxycodone	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	30	distributing	distribute	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	186	31	powder	powder	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	32	methamphetamine	methamphetamine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	33	crack	crack	VB
work_zhl5dw75nfhy7oputsllvjjcqq	186	34	distribution	distribution	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	35	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	36	heroin	heroin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	37	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	38	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	39	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	40	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	41	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	42	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	43	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	44	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	45	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	46	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	47	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	48	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	49	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	50	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	51	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	52	shuffled	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	53	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	54	0.45	0.45	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	55	0.50	0.50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	56	0.55	0.55	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	57	0.60	0.60	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	58	0.65	0.65	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	59	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	60	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	61	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	62	steroids	steroid	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	63	reserved	reserve	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	64	substance	substance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	65	involving	involve	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	186	66	violent	violent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	67	several	several	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	68	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	69	testing	testing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	70	tested	test	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	71	criticized	criticize	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	72	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	73	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	74	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	75	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	76	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	77	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	78	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	79	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	80	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	81	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	82	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	83	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	84	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	85	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	86	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	87	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	88	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	89	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	90	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	91	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	92	0.85	0.85	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	93	0.90	0.90	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	94	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	95	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	96	masturbation	masturbation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	97	medication	medication	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	98	tobacco	tobacco	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	99	stress	stress	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	100	nicotine	nicotine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	101	alcohol	alcohol	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	102	thc	thc	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	103	caffeine	caffeine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	104	smoking	smoking	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	105	cannabis	cannabis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	106	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	107	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	108	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	109	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	110	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	111	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	112	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	113	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	114	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	115	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	116	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	117	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	118	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	119	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	120	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	121	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	122	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	123	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	124	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	125	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	126	0.85	0.85	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	127	0.90	0.90	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	128	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	129	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	130	cigarettes	cigarette	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	131	powder	powder	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	132	narcotics	narcotic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	133	crack	crack	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	134	drugs	drug	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	135	pills	pill	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	136	methamphetamine	methamphetamine	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	137	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	138	oxycodone	oxycodone	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	139	heroin	heroin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	140	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	141	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	142	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	143	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	144	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	145	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	146	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	147	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	148	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	149	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	150	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	151	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	152	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	153	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	154	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	155	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	156	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	157	0.60	0.60	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	158	0.65	0.65	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	159	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	160	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	161	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	162	0.85	0.85	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	163	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	164	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	165	testing	testing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	166	abuse	abuse	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	167	substance	substance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	168	urine	urine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	169	alcohol	alcohol	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	170	counseling	counseling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	171	steroid	steroid	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	172	nandrolone	nandrolone	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	173	drug	drug	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	174	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	175	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	176	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	177	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	178	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	179	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	180	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	181	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	182	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	183	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	184	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	185	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	186	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	187	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	188	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	189	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	190	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	191	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	192	0.65	0.65	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	193	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	194	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	195	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	196	0.85	0.85	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	197	0.90	0.90	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	198	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	199	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	200	drug	drug	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	201	smoking	smoking	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	202	lsd	lsd	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	203	thc	thc	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	204	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	205	weed	weed	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	206	mdma	mdma	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	207	tobacco	tobacco	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	208	nicotine	nicotine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	209	cannabis	cannabis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	210	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	211	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	212	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	213	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	214	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	215	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	216	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	217	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	218	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	219	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	220	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	221	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	222	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	223	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	224	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	225	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	226	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	227	0.45	0.45	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	228	0.50	0.50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	229	0.55	0.55	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	230	0.60	0.60	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	231	0.65	0.65	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	232	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	233	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	234	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	235	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	236	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	237	possession	possession	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	238	growing	grow	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	186	239	smoked	smoke	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	240	grams	gram	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	241	drugs	drug	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	242	distribute	distribute	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	243	crack	crack	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	244	kilograms	kilograms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	245	heroin	heroin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	246	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	247	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	248	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	249	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	250	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	251	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	252	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	253	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	254	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	255	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	256	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	257	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	258	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	259	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	260	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	261	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	262	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	263	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	264	0.2	0.2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	265	0.3	0.3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	266	0.4	0.4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	267	0.5	0.5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	268	0.6	0.6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	269	0.7	0.7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	270	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	271	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	272	positive	positive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	273	suspensions	suspension	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	274	blaming	blame	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	186	275	steroid	steroid	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	276	purposes	purpose	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	277	addiction	addiction	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	278	testing	testing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	279	smoking	smoking	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	280	procedures	procedure	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	281	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	282	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	283	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	284	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	285	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	286	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	287	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	288	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	289	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	290	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	291	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	292	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	293	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	294	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	295	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	296	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	297	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	298	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	299	0.4	0.4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	300	0.5	0.5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	301	0.6	0.6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	302	0.7	0.7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	303	0.8	0.8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	304	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	305	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	306	effects	effect	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	307	drugs	drug	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	308	caffeine	caffeine	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	309	weed	weed	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	310	drug	drug	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	311	thc	thc	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	312	tobacco	tobacco	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	313	smoke	smoke	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	314	smoking	smoking	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	315	cannabis	cannabis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	316	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	317	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	318	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	319	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	320	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	321	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	322	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	323	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	324	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	325	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	326	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	327	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	328	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	329	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	330	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	331	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	332	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	333	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	334	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	335	0.85	0.85	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	336	0.90	0.90	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	337	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	338	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	339	methamphetamine	methamphetamine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	340	hydrochloride	hydrochloride	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	341	kilograms	kilograms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	342	paraphernalia	paraphernalia	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	343	kilogram	kilogram	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	344	grams	grams	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	345	crack	crack	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	346	powder	powder	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	347	heroin	heroin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	348	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	349	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	350	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	351	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	352	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	353	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	354	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	355	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	356	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	357	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	358	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	359	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	360	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	361	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	362	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	363	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	364	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	365	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	366	0.5	0.5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	367	0.6	0.6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	368	0.7	0.7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	369	0.8	0.8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	370	0.9	0.9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	371	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	372	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	373	steroid	steroid	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	374	testing	testing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	375	crack	crack	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	376	drugs	drug	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	377	substance	substance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	378	positive	positive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	379	tested	test	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	380	alcohol	alcohol	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	381	drug	drug	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	382	cocaine	cocaine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	383	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	384	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	385	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	386	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	387	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	388	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	389	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	390	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	391	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	392	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	393	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	394	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	395	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	396	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	397	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	398	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	186	399	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	400	0.70	0.70	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	401	0.75	0.75	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	402	0.80	0.80	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	403	0.85	0.85	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	404	0.90	0.90	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	405	0.95	0.95	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	406	Cosine	Cosine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	407	Similarity	Similarity	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	408	smoke	smoke	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	409	smokers	smoker	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	410	thc	thc	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	411	cigar	cigar	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	412	cigarettes	cigarette	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	413	smoking	smoke	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	186	414	weed	weed	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	415	tobacco	tobacco	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	416	nicotine	nicotine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	417	cannabis	cannabis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	418	M	M	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	419	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	420	st	st	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	421	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	422	i	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	186	423	m	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	186	424	il	il	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	425	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	426	r	r	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	427	W	W	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	428	o	o	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	429	rd	rd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	430	s	s	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	431	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	432	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	433	shuffled	shuffle	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	186	434	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	435	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	436	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	437	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	186	438	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	439	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	186	440	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	441	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	442	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	186	443	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	186	444	means	mean	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	445	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	186	446	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	447	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	448	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	186	449	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	450	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	451	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	452	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	186	453	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	454	query	query	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	455	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	456	marijuana	marijuana	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	457	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	186	458	its	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	186	459	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	186	460	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	186	461	neighbors	neighbor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	186	462	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	186	463	highest	high	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	186	464	mean	mean	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	186	465	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	466	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	467	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	186	468	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	186	469	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	186	470	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	186	471	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	187	1	Embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	187	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	187	3	learned	learn	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	187	4	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	187	5	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	187	6	segmented	segment	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	187	7	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	187	8	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	187	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	188	1	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	188	2	setting	set	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	188	3	causes	cause	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	188	4	large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	188	5	increases	increase	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	188	6	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	188	7	varia-	varia-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	188	8	tion	tion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	188	9	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	188	10	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	188	11	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	188	12	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	188	13	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	188	14	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	188	15	weaker	weak	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	188	16	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	188	17	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	188	18	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	188	19	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	188	20	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	188	21	corpora	corpora	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	188	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	188	23	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	188	24	large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	188	25	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	188	26	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	188	27	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	188	28	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	188	29	rankings	ranking	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	188	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	189	1	This	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	189	2	indicates	indicate	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	189	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	189	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	189	5	pres-	pres-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	189	6	ence	ence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	189	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	189	8	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	189	9	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	189	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	189	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	189	12	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	189	13	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	189	14	signifi-	signifi-	VB
work_zhl5dw75nfhy7oputsllvjjcqq	189	15	cantly	cantly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	189	16	affect	affect	VB
work_zhl5dw75nfhy7oputsllvjjcqq	189	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	189	18	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	189	19	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	189	20	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	189	21	embed-	embed-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	189	22	ding	ding	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	189	23	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	189	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	190	1	GloVe	GloVe	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	190	2	produced	produce	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	190	3	very	very	RB
work_zhl5dw75nfhy7oputsllvjjcqq	190	4	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	190	5	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	190	6	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	190	7	both	both	PDT
work_zhl5dw75nfhy7oputsllvjjcqq	190	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	190	9	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	190	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	190	11	SHUFFLED	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	190	12	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	190	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	190	14	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	190	15	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	190	16	means	mean	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	190	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	190	18	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	190	19	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	190	20	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	190	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	190	22	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	190	23	indi-	indi-	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	190	24	cates	cat	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	190	25	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	190	26	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	190	27	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	190	28	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	190	29	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	190	30	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	190	31	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	190	32	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	190	33	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	191	1	However	however	RB
work_zhl5dw75nfhy7oputsllvjjcqq	191	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	191	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	191	4	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	191	5	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	191	6	caused	cause	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	191	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	191	8	reduc-	reduc-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	191	9	tion	tion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	191	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	191	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	191	12	mean	mean	NN
work_zhl5dw75nfhy7oputsllvjjcqq	191	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	191	14	widened	widen	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	191	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	191	16	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	191	17	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	191	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	191	19	indicating	indicate	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	191	20	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	191	21	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	191	22	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	191	23	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	191	24	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	191	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	191	26	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	191	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	191	28	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	191	29	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	191	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	192	1	114	114	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	2	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	3	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	4	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	5	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	6	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	7	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	8	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	9	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	10	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	11	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	12	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	13	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	14	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	15	7	7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	16	viability	viability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	17	fetus	fetus	VB
work_zhl5dw75nfhy7oputsllvjjcqq	192	18	trimester	trimester	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	19	surgery	surgery	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	20	trimester	trimester	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	21	pregnancies	pregnancy	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	192	22	abdomen	abdomen	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	23	pregnancies	pregnancy	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	192	24	pregnancies	pregnancy	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	192	25	surgery	surgery	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	26	visit	visit	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	192	27	surgery	surgery	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	28	occupation	occupation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	29	tenure	tenure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	30	abortion	abortion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	31	gestation	gestation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	32	visit	visit	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	33	therapy	therapy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	34	incarceration	incarceration	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	35	viability	viability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	36	stepfather	stepfather	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	37	abortions	abortion	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	192	38	kindergarten	kindergarten	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	192	39	tenure	tenure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	40	pain	pain	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	41	visit	visit	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	42	abortion	abortion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	43	wife	wife	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	44	fetus	fetus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	45	viability	viability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	46	workday	workday	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	47	hospitalization	hospitalization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	48	arrival	arrival	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	49	tenure	tenure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	50	groin	groin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	51	gestation	gestation	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	52	headaches	headache	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	192	53	abortions	abortions	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	192	54	neck	neck	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	55	pain	pain	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	56	visit	visit	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	192	57	throat	throat	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	58	surgery	surgery	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	59	pregnant	pregnant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	192	60	hernia	hernia	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	61	headaches	headache	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	192	62	headaches	headaches	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	63	abortions	abortions	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	192	64	grandmother	grandmother	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	192	65	expiration	expiration	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	192	66	abortion	abortion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	67	summer	summer	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	68	trimester	trimester	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	69	birthday	birthday	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	70	pregnant	pregnant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	192	71	daughter	daughter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	72	sudden	sudden	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	192	73	pain	pain	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	74	suicide	suicide	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	75	experiencing	experience	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	192	76	neck	neck	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	77	birthday	birthday	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	78	panic	panic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	79	fetal	fetal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	192	80	bladder	bladder	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	81	abortion	abortion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	82	medications	medication	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	192	83	tenure	tenure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	84	fetus	fetus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	85	jaw	jaw	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	86	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	87	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	88	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	192	89	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	192	90	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	192	91	closest	close	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	192	92	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	192	93	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	192	94	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	192	95	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	96	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	97	pregnancy	pregnancy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	192	98	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	192	99	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	192	100	variable	variable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	192	101	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	193	1	None	none	NN
work_zhl5dw75nfhy7oputsllvjjcqq	193	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	193	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	193	4	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	193	5	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	193	6	appear	appear	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	193	7	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	193	8	every	every	DT
work_zhl5dw75nfhy7oputsllvjjcqq	193	9	run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	193	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	194	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	194	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	194	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	194	4	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	194	5	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	194	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	194	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	194	8	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	194	9	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	194	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	194	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	194	12	full	full	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	194	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	194	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	194	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	194	16	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	194	17	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	194	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	194	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	194	20	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	194	21	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	194	22	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	194	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	194	24	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	194	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	194	26	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	194	27	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	194	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	195	1	Run	run	VB
work_zhl5dw75nfhy7oputsllvjjcqq	195	2	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	3	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	4	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	5	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	6	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	7	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	8	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	9	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	10	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	11	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	12	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	13	Run	run	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	14	7	7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	15	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	16	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	17	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	18	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	19	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	20	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	21	selection	selection	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	22	genetics	genetic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	23	process	process	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	195	24	human	human	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	25	darwinian	darwinian	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	26	convergent	convergent	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	27	evolutionary	evolutionary	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	28	darwinian	darwinian	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	29	convergent	convergent	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	30	darwinian	darwinian	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	31	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	32	theory	theory	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	195	33	darwinian	darwinian	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	34	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	35	nature	nature	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	195	36	process	process	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	37	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	38	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	39	genetics	genetic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	40	evolutionary	evolutionary	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	41	species	species	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	195	42	evolutionary	evolutionary	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	43	darwinian	darwinian	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	44	convergent	convergent	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	45	genetics	genetics	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	46	human	human	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	47	genetics	genetics	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	195	48	convergent	convergent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	49	convergent	convergent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	50	abiogenesis	abiogenesis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	51	evolutionary	evolutionary	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	52	species	species	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	195	53	evolutionary	evolutionary	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	54	theory	theory	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	55	process	process	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	56	process	process	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	57	evolutionary	evolutionary	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	58	species	specie	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	59	did	do	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	195	60	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	61	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	62	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	63	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	64	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	65	human	human	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	66	convergent	convergent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	67	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	68	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	69	did	do	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	195	70	species	specie	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	71	nature	nature	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	72	natural	natural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	73	process	process	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	74	convergent	convergent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	75	process	process	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	76	human	human	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	77	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	78	species	specie	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	195	79	theory	theory	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	80	evolutionary	evolutionary	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	81	creationism	creationism	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	195	82	human	human	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	83	darwinian	darwinian	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	84	favor	favor	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	85	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	86	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	87	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	195	88	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	195	89	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	90	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	195	91	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	195	92	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	195	93	closest	close	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	195	94	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	195	95	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	195	96	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	195	97	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	98	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	99	evolution	evolution	NN
work_zhl5dw75nfhy7oputsllvjjcqq	195	100	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	195	101	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	195	102	variable	variable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	195	103	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	196	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	196	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	196	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	196	4	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	196	5	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	196	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	196	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	196	8	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	196	9	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	196	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	196	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	196	12	full	full	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	196	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	196	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	196	15	AskScience	AskScience	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	196	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	196	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	196	18	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	196	19	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	196	20	length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	196	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	196	22	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	196	23	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	196	24	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	196	25	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	196	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	197	1	These	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	197	2	patterns	pattern	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	197	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	197	4	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	197	5	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	197	6	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	197	7	variations	variation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	197	8	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	197	9	generalized	generalize	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	197	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	197	11	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	197	12	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	197	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	197	14	which	which	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	197	15	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	197	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	197	17	mean	mean	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	197	18	stan-	stan-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	197	19	dard	dard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	197	20	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	197	21	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	197	22	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	197	23	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	197	24	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	197	25	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	197	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	198	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	198	2	calculated	calculate	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	198	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	198	4	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	198	5	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	198	6	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	198	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	198	8	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	198	9	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	198	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	198	11	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	198	12	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	198	13	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	198	14	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	198	15	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	198	16	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	198	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	198	18	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	198	19	then	then	RB
work_zhl5dw75nfhy7oputsllvjjcqq	198	20	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	198	21	averaged	average	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	198	22	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	198	23	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	198	24	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	198	25	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	198	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	199	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	199	2	re-	re-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	199	3	sults	sult	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	199	4	show	show	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	199	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	199	6	average	average	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	199	7	levels	level	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	199	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	199	9	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	199	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	199	11	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	199	12	algorithm	algorithm	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	199	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	199	14	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	199	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	200	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	200	2	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	200	3	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	200	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	200	5	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	200	6	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	200	7	SHUFFLED	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	200	8	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	200	9	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	200	10	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	200	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	200	12	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	200	13	produce	produce	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	200	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	200	15	least	least	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	200	16	variable	variable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	200	17	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	200	18	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	200	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	200	20	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	200	21	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	200	22	pro-	pro-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	200	23	duces	duce	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	200	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	200	25	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	200	26	variable	variable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	200	27	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	200	28	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	200	29	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	200	30	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	200	31	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	200	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	201	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	201	2	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	201	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	201	4	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	201	5	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	201	6	has	have	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	201	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	201	8	significant	significant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	201	9	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	201	10	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	201	11	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	201	12	four	four	CD
work_zhl5dw75nfhy7oputsllvjjcqq	201	13	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	201	14	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	201	15	lesser	less	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	201	16	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	201	17	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	201	18	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	201	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	201	20	consistently	consistently	RB
work_zhl5dw75nfhy7oputsllvjjcqq	201	21	increasing	increase	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	201	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	201	23	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	201	24	devia-	devia-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	201	25	tions	tion	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	201	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	202	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	202	2	turn	turn	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	202	3	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	202	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	202	5	question	question	NN
work_zhl5dw75nfhy7oputsllvjjcqq	202	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	202	7	how	how	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	202	8	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	202	9	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	202	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	202	11	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	202	12	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	202	13	affects	affect	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	202	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	202	15	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	202	16	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	202	17	most	most	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	202	18	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	202	19	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	202	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	203	1	Are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	203	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	203	3	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	203	4	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	203	5	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	203	6	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	203	7	simply	simply	RB
work_zhl5dw75nfhy7oputsllvjjcqq	203	8	re	re	VB
work_zhl5dw75nfhy7oputsllvjjcqq	203	9	-	-	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	203	10	ordered	order	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	203	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	203	12	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	203	13	do	do	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	203	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	203	15	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	203	16	present	present	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	203	17	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	203	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	203	19	list	list	NN
work_zhl5dw75nfhy7oputsllvjjcqq	203	20	substantially	substantially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	203	21	change	change	VB
work_zhl5dw75nfhy7oputsllvjjcqq	203	22	?	?	.
work_zhl5dw75nfhy7oputsllvjjcqq	204	1	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	2	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	204	3	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	204	4	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	204	5	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	204	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	204	8	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	204	9	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	204	10	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	11	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	12	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	204	13	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	204	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	204	15	query	query	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	204	16	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	17	pregnancy	pregnancy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	18	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	204	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	204	20	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	204	21	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	204	22	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	204	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	205	1	Observing	Observing	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	2	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	3	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	205	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	205	5	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	205	6	might	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	205	7	believe	believe	VB
work_zhl5dw75nfhy7oputsllvjjcqq	205	8	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	9	judges	judge	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	205	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	205	12	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	13	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	14	associate	associate	NN
work_zhl5dw75nfhy7oputsllvjjcqq	205	15	pregnancy	pregnancy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	205	16	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	205	17	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	18	questions	question	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	205	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	20	viability	viability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	205	21	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	205	22	abortion	abortion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	205	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	205	24	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	25	observ-	observ-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	26	ing	ing	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	27	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	205	28	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	205	29	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	205	30	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	205	31	might	may	MD
work_zhl5dw75nfhy7oputsllvjjcqq	205	32	believe	believe	VB
work_zhl5dw75nfhy7oputsllvjjcqq	205	33	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	34	pregnancy	pregnancy	NN
work_zhl5dw75nfhy7oputsllvjjcqq	205	35	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	205	36	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	205	37	associated	associate	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	205	38	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	39	questions	question	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	205	40	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	205	41	prisons	prison	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	205	42	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	205	43	family	family	NN
work_zhl5dw75nfhy7oputsllvjjcqq	205	44	visits	visit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	205	45	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	206	1	Although	although	IN
work_zhl5dw75nfhy7oputsllvjjcqq	206	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	3	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	206	4	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	206	5	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	6	table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	206	7	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	206	8	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	9	produced	produce	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	206	10	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	206	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	12	same	same	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	206	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	206	14	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	206	15	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	206	16	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	206	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	206	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	19	membership	membership	NN
work_zhl5dw75nfhy7oputsllvjjcqq	206	20	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	206	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	22	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	206	23	changes	change	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	206	24	substantially	substantially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	206	25	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	206	26	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	206	27	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	206	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	206	29	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	206	30	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	206	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	207	1	As	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	207	2	another	another	DT
work_zhl5dw75nfhy7oputsllvjjcqq	207	3	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	207	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	207	5	Table	table	NN
work_zhl5dw75nfhy7oputsllvjjcqq	207	6	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	207	7	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	207	8	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	207	9	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	207	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	207	11	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	207	12	evolution	evolution	NN
work_zhl5dw75nfhy7oputsllvjjcqq	207	13	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	207	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	207	15	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	207	16	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	207	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	207	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	207	19	AskScience	AskScience	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	207	20	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	207	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	208	1	Although	although	IN
work_zhl5dw75nfhy7oputsllvjjcqq	208	2	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	208	3	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	208	4	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	208	5	less	less	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	208	6	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	208	7	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	208	8	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	208	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	208	10	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	208	11	still	still	RB
work_zhl5dw75nfhy7oputsllvjjcqq	208	12	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	208	13	cause	cause	NN
work_zhl5dw75nfhy7oputsllvjjcqq	208	14	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	208	15	concern	concern	NN
work_zhl5dw75nfhy7oputsllvjjcqq	208	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	209	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	209	2	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	209	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	209	4	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	209	5	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	209	6	ranks	rank	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	209	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	209	8	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	209	9	human	human	NN
work_zhl5dw75nfhy7oputsllvjjcqq	209	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	209	11	humans	human	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	209	12	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	209	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	209	14	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	209	15	Run	Run	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	209	16	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	209	17	includes	include	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	209	18	neither	neither	DT
work_zhl5dw75nfhy7oputsllvjjcqq	209	19	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	209	20	those	those	DT
work_zhl5dw75nfhy7oputsllvjjcqq	209	21	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	209	22	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	209	23	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	209	24	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	209	25	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	209	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	210	1	These	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	210	2	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	210	3	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	210	4	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	210	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	210	6	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	210	7	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	210	8	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	210	9	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	210	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	210	11	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	210	12	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	210	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	211	1	For	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	211	2	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	211	3	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	211	4	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	211	5	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	211	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	211	7	AskHistorians	AskHistorians	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	211	8	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	211	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	211	10	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	211	11	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	211	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	211	13	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	211	14	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	211	15	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	211	16	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	211	17	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	211	18	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	211	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	212	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	212	2	generate	generate	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	212	3	new	new	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	212	4	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	212	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	212	6	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	212	7	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	212	8	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	212	9	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	212	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	212	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	212	12	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	212	13	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	212	14	trained	train	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	212	15	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	212	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	212	17	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	212	18	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	212	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	212	20	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	212	21	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	212	22	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	212	23	Jac-	Jac-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	212	24	card	card	NN
work_zhl5dw75nfhy7oputsllvjjcqq	212	25	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	212	26	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	212	27	compare	compare	VB
work_zhl5dw75nfhy7oputsllvjjcqq	212	28	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	212	29	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	212	30	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	212	31	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	213	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	213	2	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	213	3	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	213	4	patterns	pattern	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	5	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	7	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	9	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	213	10	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	11	115	115	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	12	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	13	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	14	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	15	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	16	ALGORITHM	ALGORITHM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	17	0.0	0.0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	18	0.1	0.1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	19	0.2	0.2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	20	0.3	0.3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	21	0.4	0.4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	22	0.5	0.5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	23	0.6	0.6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	24	0.7	0.7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	25	0.8	0.8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	26	0.9	0.9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	27	JA	ja	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	28	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	29	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	30	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	31	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	32	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	33	S	s	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	34	IM	IM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	35	IL	IL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	36	A	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	37	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	38	IT	it	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	39	Y	y	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	40	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	213	41	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	213	42	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	43	Variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	44	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	45	Top	Top	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	46	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	47	Words	Words	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	48	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	49	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	50	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	51	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	52	ALGORITHM	ALGORITHM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	53	0.0	0.0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	54	0.1	0.1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	55	0.2	0.2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	56	0.3	0.3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	57	0.4	0.4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	58	0.5	0.5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	59	0.6	0.6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	60	0.7	0.7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	61	0.8	0.8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	62	0.9	0.9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	63	JA	ja	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	64	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	65	C	c	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	66	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	67	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	68	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	69	S	s	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	70	IM	IM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	71	IL	IL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	72	A	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	73	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	74	IT	it	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	75	Y	y	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	76	fixed	fix	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	213	77	shuffled	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	213	78	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	79	Variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	80	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	81	Top	Top	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	82	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	83	Words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	84	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	85	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	86	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	213	87	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	88	mean	mean	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	89	Jaccard	Jaccard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	213	90	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	91	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	92	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	93	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	213	94	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	95	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	96	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	97	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	213	98	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	99	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	213	100	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	213	101	closest	close	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	213	102	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	103	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	104	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	105	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	106	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	213	107	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	213	108	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	213	109	AskHistorians	AskHistorians	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	213	110	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	213	111	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	214	1	Larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	214	2	Jaccard	Jaccard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	214	3	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	214	4	indicates	indicate	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	214	5	more	more	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	214	6	consistency	consistency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	214	7	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	214	8	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	214	9	N	n	CD
work_zhl5dw75nfhy7oputsllvjjcqq	214	10	member-	member-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	214	11	ship	ship	NN
work_zhl5dw75nfhy7oputsllvjjcqq	214	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	215	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	215	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	215	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	215	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	215	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	215	6	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	215	7	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	215	8	length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	215	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	216	1	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	216	2	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	216	3	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	216	4	;	;	:
work_zhl5dw75nfhy7oputsllvjjcqq	216	5	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	216	6	displays	display	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	216	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	216	8	lowest	low	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	216	9	Jaccard	Jaccard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	216	10	simi-	simi-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	216	11	larity	larity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	216	12	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	216	13	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	216	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	216	15	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	216	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	216	17	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	216	18	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	216	19	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	216	20	higher	high	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	216	21	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	216	22	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	216	23	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	216	24	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	216	25	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	216	26	SHUFFLED	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	216	27	set-	set-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	216	28	tings	ting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	216	29	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	216	30	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	216	31	lower	low	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	216	32	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	216	33	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	216	34	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	216	35	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	216	36	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	216	37	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	217	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	217	2	display	display	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	217	3	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	217	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	217	5	both	both	DT
work_zhl5dw75nfhy7oputsllvjjcqq	217	6	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	217	7	=	=	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	217	8	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	217	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	217	10	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	217	11	=	=	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	217	12	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	217	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	217	14	emphasizing	emphasize	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	217	15	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	217	16	even	even	RB
work_zhl5dw75nfhy7oputsllvjjcqq	217	17	very	very	RB
work_zhl5dw75nfhy7oputsllvjjcqq	217	18	highly	highly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	217	19	ranked	rank	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	217	20	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	217	21	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	217	22	drop	drop	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	217	23	out	out	IN
work_zhl5dw75nfhy7oputsllvjjcqq	217	24	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	217	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	217	26	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	217	27	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	217	28	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	217	29	list	list	NN
work_zhl5dw75nfhy7oputsllvjjcqq	217	30	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	218	1	Even	even	RB
work_zhl5dw75nfhy7oputsllvjjcqq	218	2	when	when	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	218	3	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	218	4	do	do	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	218	5	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	218	6	drop	drop	VB
work_zhl5dw75nfhy7oputsllvjjcqq	218	7	out	out	IN
work_zhl5dw75nfhy7oputsllvjjcqq	218	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	218	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	218	10	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	218	11	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	218	12	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	218	13	list	list	NN
work_zhl5dw75nfhy7oputsllvjjcqq	218	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	218	15	they	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	218	16	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	218	17	change	change	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	218	18	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	218	19	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	218	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	218	21	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	218	22	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	218	23	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	218	24	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	218	25	Figure	figure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	218	26	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	218	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	219	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	219	2	show	show	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	219	3	both	both	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	4	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	5	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	219	6	example	example	NN
work_zhl5dw75nfhy7oputsllvjjcqq	219	7	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	219	8	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	9	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	219	10	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	219	11	men	man	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	219	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	219	13	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	14	aggregate	aggregate	NN
work_zhl5dw75nfhy7oputsllvjjcqq	219	15	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	219	16	all	all	PDT
work_zhl5dw75nfhy7oputsllvjjcqq	219	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	18	terms	term	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	219	19	whose	whose	WP$
work_zhl5dw75nfhy7oputsllvjjcqq	219	20	average	average	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	219	21	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	219	22	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	219	23	within	within	IN
work_zhl5dw75nfhy7oputsllvjjcqq	219	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	25	top-10	top-10	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	219	26	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	219	27	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	219	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	219	29	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	219	30	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	219	31	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	219	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	220	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	220	2	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	220	3	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	220	4	highlight	highlight	VB
work_zhl5dw75nfhy7oputsllvjjcqq	220	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	220	6	av-	av-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	220	7	erage	erage	NN
work_zhl5dw75nfhy7oputsllvjjcqq	220	8	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	220	9	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	220	10	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	220	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	220	12	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	220	13	do	do	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	220	14	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	220	15	show	show	VB
work_zhl5dw75nfhy7oputsllvjjcqq	220	16	outliers	outlier	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	220	17	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	220	18	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	220	19	figure	figure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	220	20	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	220	21	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	220	22	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	220	23	note	note	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	220	24	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	220	25	outliers	outlier	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	220	26	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	220	27	large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	220	28	falls	fall	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	220	29	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	220	30	jumps	jump	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	220	31	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	220	32	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	220	33	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	220	34	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	220	35	common	common	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	220	36	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	221	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	221	2	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	221	3	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	221	4	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	221	5	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	221	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	221	7	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	221	8	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	221	9	indicates	indicate	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	221	10	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	221	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	221	12	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	221	13	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	221	14	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	221	15	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	221	16	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	221	17	significantly	significantly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	221	18	affect	affect	VB
work_zhl5dw75nfhy7oputsllvjjcqq	221	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	221	20	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	221	21	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	221	22	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	221	23	rankings	ranking	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	221	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	222	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	222	2	also	also	RB
work_zhl5dw75nfhy7oputsllvjjcqq	222	3	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	222	4	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	5	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	6	segmentation	segmentation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	7	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	8	af-	af-	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	9	0	0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	10	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	11	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	12	15	15	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	13	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	14	25	25	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	15	30	30	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	16	35	35	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	17	RANK	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	18	children	child	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	19	women	woman	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	20	soldiers	soldier	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	21	boys	boy	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	222	22	girls	girl	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	23	horses	horse	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	24	officers	officer	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	25	people	people	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	26	peasants	peasant	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	222	27	bodies	body	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	28	W	w	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	29	O	o	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	30	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	31	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	32	Change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	33	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	34	Rank	Rank	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	222	35	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	222	36	"	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	222	37	men	man	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	38	"	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	222	39	0	0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	40	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	41	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	42	15	15	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	43	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	44	25	25	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	45	30	30	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	46	35	35	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	47	40	40	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	48	45	45	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	49	RANK	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	50	FOR	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	51	CURRENT	CURRENT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	222	52	ITERATION	iteration	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	53	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	54	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	55	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	56	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	57	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	58	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	59	7	7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	60	8	8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	61	9	9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	62	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	63	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	64	V	v	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	65	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	66	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	67	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	68	G	g	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	69	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	70	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	71	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	72	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	73	K	k	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	74	Change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	75	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	76	Rank	Rank	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	222	77	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	78	All	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	222	79	Queries	Queries	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	222	80	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	222	81	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	82	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	222	83	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	222	84	change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	85	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	86	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	222	87	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	88	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	89	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	90	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	222	91	BOOT-	BOOT-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	222	92	STRAP	STRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	222	93	setting	set	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	222	94	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	222	95	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	222	96	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	222	97	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	222	98	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	222	99	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	223	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	223	2	show	show	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	223	3	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	223	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	223	5	both	both	CC
work_zhl5dw75nfhy7oputsllvjjcqq	223	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	7	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	223	8	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	223	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	223	10	men	man	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	223	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	223	12	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	223	13	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	14	aggregate	aggregate	NN
work_zhl5dw75nfhy7oputsllvjjcqq	223	15	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	223	16	all	all	PDT
work_zhl5dw75nfhy7oputsllvjjcqq	223	17	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	18	queries	query	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	223	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	223	20	showing	show	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	223	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	22	change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	223	23	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	223	24	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	223	25	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	223	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	27	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	223	28	whose	whose	WP$
work_zhl5dw75nfhy7oputsllvjjcqq	223	29	average	average	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	223	30	ranking	ranking	NN
work_zhl5dw75nfhy7oputsllvjjcqq	223	31	falls	fall	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	223	32	within	within	IN
work_zhl5dw75nfhy7oputsllvjjcqq	223	33	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	34	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	223	35	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	223	36	neighbors	neighbor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	223	37	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	223	38	those	those	DT
work_zhl5dw75nfhy7oputsllvjjcqq	223	39	queries	query	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	223	40	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	224	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	224	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	224	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	224	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	224	5	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	224	6	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	224	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	224	8	AskHis-	askhis-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	224	9	torians	torians	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	224	10	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	224	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	224	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	224	13	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	224	14	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	224	15	length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	224	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	225	1	fects	fect	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	225	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	225	3	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	225	4	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	225	5	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	226	1	Figure	figure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	226	2	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	226	3	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	226	4	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	226	5	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	226	6	segmented	segment	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	226	7	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	226	8	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	226	9	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	226	10	fine	fine	RB
work_zhl5dw75nfhy7oputsllvjjcqq	226	11	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	226	12	grained	grain	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	226	13	level	level	NN
work_zhl5dw75nfhy7oputsllvjjcqq	226	14	produce	produce	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	226	15	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	226	16	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	226	17	less	less	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	226	18	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	226	19	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	226	20	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	226	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	226	22	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	226	23	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	226	24	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	226	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	227	1	Documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	227	2	segmented	segment	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	227	3	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	227	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	227	5	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	227	6	level	level	NN
work_zhl5dw75nfhy7oputsllvjjcqq	227	7	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	227	8	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	227	9	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	227	10	clustering	cluster	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	227	11	closer	close	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	227	12	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	227	13	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	227	14	median	median	NN
work_zhl5dw75nfhy7oputsllvjjcqq	227	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	227	16	while	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	227	17	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	227	18	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	227	19	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	227	20	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	227	21	deviations	deviation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	227	22	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	227	23	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	227	24	spread	spread	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	227	25	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	227	26	widely	widely	RB
work_zhl5dw75nfhy7oputsllvjjcqq	227	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	228	1	This	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	228	2	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	228	3	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	228	4	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	228	5	significant	significant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	228	6	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	228	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	228	8	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	228	9	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	228	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	228	11	9th	9th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	228	12	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	228	13	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	228	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	228	15	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	228	16	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	228	17	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	228	18	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	228	19	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	228	20	“	"	``
work_zhl5dw75nfhy7oputsllvjjcqq	228	21	docu-	docu-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	228	22	ments	ment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	228	23	”	"	''
work_zhl5dw75nfhy7oputsllvjjcqq	228	24	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	228	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	228	26	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	228	27	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	228	28	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	229	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	229	2	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	229	3	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	229	4	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	229	5	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	229	6	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	229	7	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	229	8	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	229	9	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	229	10	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	229	11	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	229	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	230	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	230	2	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	230	3	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	230	4	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	230	5	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	230	6	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	230	7	spread	spread	NN
work_zhl5dw75nfhy7oputsllvjjcqq	230	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	230	9	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	230	10	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	230	11	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	230	12	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	230	13	larger	large	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	230	14	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	230	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	230	16	indicating	indicate	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	230	17	greater	great	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	230	18	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	230	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	231	1	Finally	finally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	231	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	231	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	231	4	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	231	5	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	231	6	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	231	7	variance	variance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	231	8	usually	usually	RB
work_zhl5dw75nfhy7oputsllvjjcqq	231	9	stabilizes	stabilize	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	231	10	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	231	11	about	about	RB
work_zhl5dw75nfhy7oputsllvjjcqq	231	12	25	25	CD
work_zhl5dw75nfhy7oputsllvjjcqq	231	13	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	231	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	231	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	231	16	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	231	17	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	231	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	232	1	Figure	figure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	232	2	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	232	3	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	232	4	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	232	5	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	232	6	initially	initially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	232	7	increases	increase	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	232	8	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	232	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	232	10	number	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	232	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	232	12	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	232	13	trained	train	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	232	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	233	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	233	2	observe	observe	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	233	3	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	233	4	pattern	pattern	NN
work_zhl5dw75nfhy7oputsllvjjcqq	233	5	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	233	6	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	233	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	233	8	algorithms	algorithms	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	233	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	233	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	233	11	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	233	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	234	1	116	116	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	2	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	3	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	4	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	5	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	6	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	7	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	8	7	7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	9	8	8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	10	9	9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	11	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	12	RANK	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	13	0.05	0.05	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	14	0.00	0.00	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	15	0.05	0.05	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	16	0.10	0.10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	17	0.15	0.15	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	18	0.20	0.20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	19	S	s	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	20	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	21	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	22	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	23	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	24	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	25	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	26	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	27	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	28	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	29	V	v	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	30	IA	ia	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	31	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	32	IO	io	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	33	N	N	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	234	34	Document	Document	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	234	35	Size	Size	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	234	36	Comparison	Comparison	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	234	37	DOCUMENT	DOCUMENT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	234	38	SIZE	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	39	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	40	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	234	41	Figure	figure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	42	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	43	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	234	44	Standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	234	45	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	46	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	234	47	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	234	48	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	49	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	234	50	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	234	51	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	234	52	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	234	53	N	n	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	54	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	234	55	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	234	56	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	234	57	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	234	58	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	234	59	neighbors	neighbor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	234	60	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	235	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	235	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	235	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	235	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	235	5	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	235	6	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	235	7	sizes	size	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	235	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	235	9	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	235	10	vs	vs	IN
work_zhl5dw75nfhy7oputsllvjjcqq	235	11	whole	whole	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	235	12	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	235	13	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	235	14	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	235	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	235	16	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	235	17	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	235	18	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	235	19	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	235	20	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	235	21	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	235	22	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	235	23	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	235	24	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	235	25	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	236	1	1	1	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	2	2	2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	3	3	3	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	4	4	4	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	5	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	6	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	7	7	7	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	8	8	8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	9	9	9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	10	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	11	RANK	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	12	0.02	0.02	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	13	0.00	0.00	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	14	0.02	0.02	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	15	0.04	0.04	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	16	0.06	0.06	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	17	0.08	0.08	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	18	0.10	0.10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	19	0.12	0.12	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	20	0.14	0.14	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	21	S	s	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	22	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	23	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	24	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	25	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	26	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	27	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	28	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	29	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	30	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	31	V	v	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	32	IA	ia	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	33	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	34	IO	io	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	35	N	N	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	236	36	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	236	37	Size	Size	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	236	38	Comparison	Comparison	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	236	39	CORPUS	CORPUS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	236	40	SIZE	SIZE	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	236	41	0.2	0.2	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	42	1.0	1.0	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	43	Figure	figure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	44	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	45	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	236	46	Standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	236	47	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	48	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	236	49	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	236	50	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	51	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	236	52	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	236	53	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	236	54	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	236	55	N	n	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	56	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	236	57	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	236	58	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	236	59	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	236	60	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	236	61	neighbors	neighbor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	236	62	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	237	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	237	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	237	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	237	4	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	5	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	237	6	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	237	7	sizes	size	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	237	8	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	237	9	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	237	10	%	%	NN
work_zhl5dw75nfhy7oputsllvjjcqq	237	11	vs	vs	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	12	100	100	CD
work_zhl5dw75nfhy7oputsllvjjcqq	237	13	%	%	NN
work_zhl5dw75nfhy7oputsllvjjcqq	237	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	15	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	237	16	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	237	17	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	237	19	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	237	20	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	237	21	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	22	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	237	23	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	237	25	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	237	26	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	237	27	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	237	28	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	237	29	segmented	segment	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	237	30	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	237	31	sentence	sentence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	237	32	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	238	1	8	8	CD
work_zhl5dw75nfhy7oputsllvjjcqq	238	2	Discussion	discussion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	238	3	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	238	4	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	238	5	obvious	obvious	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	238	6	result	result	NN
work_zhl5dw75nfhy7oputsllvjjcqq	238	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	238	8	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	238	9	experiments	experiment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	238	10	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	238	11	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	238	12	emphasize	emphasize	VB
work_zhl5dw75nfhy7oputsllvjjcqq	238	13	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	238	14	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	238	15	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	238	16	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	238	17	even	even	RB
work_zhl5dw75nfhy7oputsllvjjcqq	238	18	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	238	19	single	single	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	238	20	objective	objective	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	238	21	view	view	NN
work_zhl5dw75nfhy7oputsllvjjcqq	238	22	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	238	23	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	238	24	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	238	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	238	26	much	much	RB
work_zhl5dw75nfhy7oputsllvjjcqq	238	27	less	less	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	238	28	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	238	29	objective	objective	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	238	30	view	view	NN
work_zhl5dw75nfhy7oputsllvjjcqq	238	31	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	238	32	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	238	33	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	239	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	239	2	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	3	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	239	4	itself	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	239	5	only	only	RB
work_zhl5dw75nfhy7oputsllvjjcqq	239	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	239	7	sample	sample	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	239	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	239	10	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	239	11	have	have	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	239	12	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	239	13	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	239	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	239	15	curation	curation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	16	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	239	17	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	239	18	sample	sample	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	19	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	239	20	its	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	239	21	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	239	23	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	24	length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	25	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	239	26	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	239	27	inclusion	inclusion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	28	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	239	29	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	239	30	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	239	31	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	239	32	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	239	33	cause	cause	VB
work_zhl5dw75nfhy7oputsllvjjcqq	239	34	significant	significant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	239	35	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	239	36	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	239	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	239	38	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	239	39	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	240	1	Happily	happily	RB
work_zhl5dw75nfhy7oputsllvjjcqq	240	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	240	3	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	240	4	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	240	5	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	240	6	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	240	7	quan-	quan-	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	240	8	tified	tifie	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	240	9	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	240	10	averaging	average	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	240	11	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	240	12	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	240	13	multiple	multiple	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	240	14	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	240	15	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	240	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	241	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	241	2	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	241	3	make	make	VB
work_zhl5dw75nfhy7oputsllvjjcqq	241	4	several	several	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	241	5	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	241	6	observations	observation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	241	7	about	about	IN
work_zhl5dw75nfhy7oputsllvjjcqq	241	8	al-	al-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	241	9	gorithm	gorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	241	10	sensitivities	sensitivity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	241	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	242	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	242	2	general	general	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	242	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	242	4	LSA	LSA	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	242	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	242	6	GloVe	GloVe	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	242	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	242	8	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	242	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	242	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	242	11	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	242	12	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	242	13	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	242	14	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	242	15	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	242	16	document	document	VB
work_zhl5dw75nfhy7oputsllvjjcqq	242	17	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	242	18	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	242	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	242	20	collections	collection	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	242	21	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	242	22	evaluated	evaluate	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	242	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	243	1	This	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	243	2	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	243	3	surprising	surprising	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	243	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	243	5	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	243	6	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	243	7	5	5	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	8	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	9	15	15	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	10	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	11	25	25	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	12	30	30	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	13	35	35	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	14	40	40	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	15	45	45	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	16	50	50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	17	NUMBER	number	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	18	OF	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	243	19	ITERATIONS	iterations	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	20	0.028	0.028	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	21	0.030	0.030	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	22	0.032	0.032	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	23	0.034	0.034	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	24	0.036	0.036	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	25	m	m	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	26	e	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	27	a	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	28	n	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	29	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	243	30	S	S	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	243	31	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	32	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	33	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	34	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	35	A	a	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	36	R	r	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	37	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	38	D	d	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	39	E	e	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	40	V	v	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	41	IA	ia	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	42	T	t	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	43	IO	io	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	44	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	45	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	243	46	Stability	stability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	47	over	over	IN
work_zhl5dw75nfhy7oputsllvjjcqq	243	48	Iterations	Iterations	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	243	49	Figure	Figure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	243	50	6	6	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	51	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	243	52	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	243	53	mean	mean	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	54	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	243	55	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	243	56	standard	standard	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	243	57	deviation	deviation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	58	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	243	59	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	243	60	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	61	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	243	62	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	243	63	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	243	64	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	65	term	term	NN
work_zhl5dw75nfhy7oputsllvjjcqq	243	66	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	243	67	its	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	243	68	20	20	CD
work_zhl5dw75nfhy7oputsllvjjcqq	243	69	nearest	near	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	243	70	neighbors	neighbor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	243	71	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	244	1	Results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	244	2	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	244	3	shown	show	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	244	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	244	5	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	244	6	numbers	number	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	244	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	244	8	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	244	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	244	10	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	244	11	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	244	12	setting	setting	NN
work_zhl5dw75nfhy7oputsllvjjcqq	244	13	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	244	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	244	15	4th	4th	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	244	16	Circuit	Circuit	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	244	17	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	244	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	245	1	had	have	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	245	2	expected	expect	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	245	3	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	245	4	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	245	5	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	245	6	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	245	7	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	245	8	document	document	VB
work_zhl5dw75nfhy7oputsllvjjcqq	245	9	order	order	NN
work_zhl5dw75nfhy7oputsllvjjcqq	245	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	245	11	anecdotally	anecdotally	RB
work_zhl5dw75nfhy7oputsllvjjcqq	245	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	245	13	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	245	14	had	have	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	245	15	observed	observe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	245	16	cases	case	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	245	17	where	where	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	245	18	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	245	19	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	245	20	were	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	245	21	affected	affect	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	245	22	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	245	23	groups	group	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	245	24	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	245	25	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	245	26	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	245	27	e.g.	e.g.	RB
work_zhl5dw75nfhy7oputsllvjjcqq	246	1	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	246	2	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	246	3	different	different	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	246	4	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	246	5	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	246	6	at	at	IN
work_zhl5dw75nfhy7oputsllvjjcqq	246	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	246	8	beginning	beginning	NN
work_zhl5dw75nfhy7oputsllvjjcqq	246	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	246	10	train-	train-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	246	11	ing	ing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	246	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	247	1	However	however	RB
work_zhl5dw75nfhy7oputsllvjjcqq	247	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	247	3	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	247	4	four	four	CD
work_zhl5dw75nfhy7oputsllvjjcqq	247	5	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	247	6	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	247	7	sensitive	sensitive	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	247	8	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	247	9	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	247	10	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	247	11	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	247	12	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	247	13	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	247	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	247	15	though	though	IN
work_zhl5dw75nfhy7oputsllvjjcqq	247	16	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	247	17	effect	effect	NN
work_zhl5dw75nfhy7oputsllvjjcqq	247	18	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	247	19	weaker	weak	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	247	20	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	247	21	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	247	22	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	248	1	Although	although	IN
work_zhl5dw75nfhy7oputsllvjjcqq	248	2	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	248	3	appears	appear	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	248	4	deterministic	deterministic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	248	5	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	248	6	due	due	IN
work_zhl5dw75nfhy7oputsllvjjcqq	248	7	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	248	8	its	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	248	9	pre	pre	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	248	10	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	248	11	computed	computed	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	248	12	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	13	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	248	14	context	context	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	15	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	16	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	248	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	248	18	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	248	19	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	248	20	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	248	21	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	248	22	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	23	produced	produce	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	248	24	results	result	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	248	25	under	under	IN
work_zhl5dw75nfhy7oputsllvjjcqq	248	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	248	27	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	248	28	ordering	ordering	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	29	whose	whose	WP$
work_zhl5dw75nfhy7oputsllvjjcqq	248	30	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	31	was	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	248	32	closest	close	JJS
work_zhl5dw75nfhy7oputsllvjjcqq	248	33	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	248	34	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	248	35	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	248	36	set-	set-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	248	37	ting	ting	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	248	38	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	249	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	249	2	attribute	attribute	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	249	3	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	249	4	intrinsic	intrinsic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	249	5	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	249	6	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	249	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	249	8	use	use	NN
work_zhl5dw75nfhy7oputsllvjjcqq	249	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	249	10	token	token	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	249	11	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	249	12	level	level	NN
work_zhl5dw75nfhy7oputsllvjjcqq	249	13	subsampling	subsampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	249	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	250	1	This	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	250	2	sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	3	method	method	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	4	introduces	introduce	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	250	5	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	6	into	into	IN
work_zhl5dw75nfhy7oputsllvjjcqq	250	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	250	8	source	source	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	9	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	10	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	250	11	ap-	ap-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	250	12	pears	pear	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	250	13	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	250	14	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	250	15	comparable	comparable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	250	16	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	250	17	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	250	18	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	19	resampling	resampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	20	method	method	NN
work_zhl5dw75nfhy7oputsllvjjcqq	250	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	251	1	Sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	251	2	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	251	3	PPMI	PPMI	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	251	4	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	251	5	inspired	inspire	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	251	6	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	251	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	251	8	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	251	9	method	method	NN
work_zhl5dw75nfhy7oputsllvjjcqq	251	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	251	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	251	12	word2vec	word2vec	NN
work_zhl5dw75nfhy7oputsllvjjcqq	251	13	implementation	implementation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	251	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	251	15	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	251	16	(	(	-LRB-
work_zhl5dw75nfhy7oputsllvjjcqq	251	17	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	251	18	et	et	FW
work_zhl5dw75nfhy7oputsllvjjcqq	251	19	al	al	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	251	20	.	.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	251	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	251	22	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	251	23	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	251	24	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	252	1	It	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	252	2	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	252	3	therefore	therefore	RB
work_zhl5dw75nfhy7oputsllvjjcqq	252	4	surprising	surprising	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	252	5	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	252	6	SGNS	SGNS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	252	7	shows	show	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	252	8	noticeable	noticeable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	252	9	differentiation	differentiation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	252	10	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	252	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	252	12	BOOTSTRAP	BOOTSTRAP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	252	13	setting	set	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	252	14	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	252	15	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	252	16	one	one	CD
work_zhl5dw75nfhy7oputsllvjjcqq	252	17	hand	hand	NN
work_zhl5dw75nfhy7oputsllvjjcqq	252	18	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	252	19	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	252	20	FIXED	FIXED	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	252	21	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	252	22	SHUFFLED	shuffle	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	252	23	settings	setting	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	252	24	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	252	25	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	252	26	other	other	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	252	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	253	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	253	2	use	use	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	4	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	5	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	6	sources	source	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	8	evidence	evidence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	9	needs	need	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	253	10	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	253	11	be	be	VB
work_zhl5dw75nfhy7oputsllvjjcqq	253	12	tempered	temper	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	253	13	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	253	15	understanding	understanding	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	16	that	that	WDT
work_zhl5dw75nfhy7oputsllvjjcqq	253	17	fine-	fine-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	253	18	grained	grained	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	253	19	distinctions	distinction	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	20	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	21	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	22	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	23	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	253	24	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	253	25	reliable	reliable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	253	26	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	253	27	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	28	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	253	29	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	30	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	253	31	longer	long	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	253	32	docu-	docu-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	33	ments	ment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	34	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	253	35	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	253	36	susceptible	susceptible	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	253	37	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	38	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	39	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	40	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	253	41	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	253	42	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	43	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	253	44	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	253	45	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	254	1	When	when	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	254	2	studying	study	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	254	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	254	4	top	top	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	254	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	254	6	N	n	NN
work_zhl5dw75nfhy7oputsllvjjcqq	254	7	most	most	RBS
work_zhl5dw75nfhy7oputsllvjjcqq	254	8	similar	similar	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	254	9	words	word	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	254	10	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	254	11	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	254	12	query	query	NN
work_zhl5dw75nfhy7oputsllvjjcqq	254	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	254	14	it	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	254	15	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	254	16	important	important	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	254	17	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	254	18	account	account	VB
work_zhl5dw75nfhy7oputsllvjjcqq	254	19	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	254	20	variation	variation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	254	21	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	254	22	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	254	23	lists	list	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	254	24	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	254	25	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	254	26	both	both	DT
work_zhl5dw75nfhy7oputsllvjjcqq	254	27	rank	rank	NN
work_zhl5dw75nfhy7oputsllvjjcqq	254	28	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	254	29	membership	membership	NN
work_zhl5dw75nfhy7oputsllvjjcqq	254	30	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	254	31	significantly	significantly	RB
work_zhl5dw75nfhy7oputsllvjjcqq	254	32	change	change	VB
work_zhl5dw75nfhy7oputsllvjjcqq	254	33	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	254	34	runs	run	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	254	35	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	255	1	Therefore	therefore	RB
work_zhl5dw75nfhy7oputsllvjjcqq	255	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	255	3	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	255	4	emphasize	emphasize	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	255	5	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	6	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	7	smaller	small	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	255	8	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	9	comes	come	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	255	10	greater	great	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	255	11	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	255	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	255	14	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	255	15	recommend	recommend	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	255	16	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	17	practitioners	practitioner	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	255	18	use	use	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	255	19	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	20	sampling	sampling	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	21	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	255	22	generate	generate	VB
work_zhl5dw75nfhy7oputsllvjjcqq	255	23	an	an	DT
work_zhl5dw75nfhy7oputsllvjjcqq	255	24	117	117	CD
work_zhl5dw75nfhy7oputsllvjjcqq	255	25	ensemble	ensemble	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	26	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	27	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	28	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	255	29	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	30	each	each	DT
work_zhl5dw75nfhy7oputsllvjjcqq	255	31	sub	sub	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	255	32	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	255	33	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	34	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	255	35	present	present	VB
work_zhl5dw75nfhy7oputsllvjjcqq	255	36	both	both	CC
work_zhl5dw75nfhy7oputsllvjjcqq	255	37	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	255	38	mean	mean	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	39	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	255	40	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	41	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	42	any	any	DT
work_zhl5dw75nfhy7oputsllvjjcqq	255	43	sum-	sum-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	255	44	mary	mary	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	255	45	statistics	statistic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	255	46	such	such	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	255	47	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	255	48	ordered	order	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	255	49	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	255	50	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	255	51	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	256	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	256	2	leave	leave	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	256	3	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	256	4	future	future	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	256	5	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	256	6	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	256	7	full	full	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	256	8	hyperparameter	hyperparameter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	256	9	sweep	sweep	NN
work_zhl5dw75nfhy7oputsllvjjcqq	256	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	256	11	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	256	12	three	three	CD
work_zhl5dw75nfhy7oputsllvjjcqq	256	13	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	256	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	257	1	While	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	257	2	these	these	DT
work_zhl5dw75nfhy7oputsllvjjcqq	257	3	hyperpa-	hyperpa-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	257	4	rameters	rameter	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	257	5	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	257	6	substantially	substantially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	257	7	impact	impact	VB
work_zhl5dw75nfhy7oputsllvjjcqq	257	8	performance	performance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	257	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	257	10	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	257	11	goal	goal	NN
work_zhl5dw75nfhy7oputsllvjjcqq	257	12	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	257	13	this	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	257	14	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	257	15	was	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	257	16	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	257	17	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	257	18	achieve	achieve	VB
work_zhl5dw75nfhy7oputsllvjjcqq	257	19	high	high	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	257	20	perfor-	perfor-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	257	21	mance	mance	NN
work_zhl5dw75nfhy7oputsllvjjcqq	257	22	but	but	CC
work_zhl5dw75nfhy7oputsllvjjcqq	257	23	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	257	24	examine	examine	VB
work_zhl5dw75nfhy7oputsllvjjcqq	257	25	how	how	WRB
work_zhl5dw75nfhy7oputsllvjjcqq	257	26	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	257	27	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	257	28	respond	respond	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	257	29	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	257	30	changes	change	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	257	31	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	257	32	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	257	33	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	257	34	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	258	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	258	2	make	make	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	258	3	no	no	DT
work_zhl5dw75nfhy7oputsllvjjcqq	258	4	claim	claim	NN
work_zhl5dw75nfhy7oputsllvjjcqq	258	5	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	258	6	one	one	CD
work_zhl5dw75nfhy7oputsllvjjcqq	258	7	algorithm	algorithm	NN
work_zhl5dw75nfhy7oputsllvjjcqq	258	8	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	258	9	better	well	JJR
work_zhl5dw75nfhy7oputsllvjjcqq	258	10	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	258	11	another	another	DT
work_zhl5dw75nfhy7oputsllvjjcqq	258	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	259	1	9	9	CD
work_zhl5dw75nfhy7oputsllvjjcqq	259	2	Conclusion	conclusion	NN
work_zhl5dw75nfhy7oputsllvjjcqq	259	3	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	259	4	find	find	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	259	5	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	259	6	there	there	EX
work_zhl5dw75nfhy7oputsllvjjcqq	259	7	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	259	8	several	several	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	259	9	sources	source	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	259	10	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	259	11	variability	variability	NN
work_zhl5dw75nfhy7oputsllvjjcqq	259	12	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	259	13	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	259	14	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	259	15	between	between	IN
work_zhl5dw75nfhy7oputsllvjjcqq	259	16	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	259	17	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	259	18	vec-	vec-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	259	19	tors	tor	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	259	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	260	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	260	2	size	size	NN
work_zhl5dw75nfhy7oputsllvjjcqq	260	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	260	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	260	5	corpus	corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	260	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	260	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	260	8	length	length	NN
work_zhl5dw75nfhy7oputsllvjjcqq	260	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	260	10	individual	individual	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	260	11	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	260	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	260	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	260	14	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	260	15	presence	presence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	260	16	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	260	17	absence	absence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	260	18	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	260	19	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	260	20	documents	document	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	260	21	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	260	22	all	all	DT
work_zhl5dw75nfhy7oputsllvjjcqq	260	23	affect	affect	VB
work_zhl5dw75nfhy7oputsllvjjcqq	260	24	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	260	25	resulting	result	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	260	26	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	260	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	261	1	While	while	IN
work_zhl5dw75nfhy7oputsllvjjcqq	261	2	differences	difference	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	261	3	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	261	4	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	261	5	association	association	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	261	6	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	261	7	measur-	measur-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	261	8	able	able	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	261	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	261	10	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	261	11	often	often	RB
work_zhl5dw75nfhy7oputsllvjjcqq	261	12	significant	significant	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	261	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	261	14	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	261	15	differences	difference	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	261	16	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	261	17	cosine	cosine	NN
work_zhl5dw75nfhy7oputsllvjjcqq	261	18	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	261	19	are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	261	20	not	not	RB
work_zhl5dw75nfhy7oputsllvjjcqq	261	21	reliable	reliable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	261	22	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	261	23	especially	especially	RB
work_zhl5dw75nfhy7oputsllvjjcqq	261	24	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	261	25	small	small	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	261	26	corpora	corpora	NN
work_zhl5dw75nfhy7oputsllvjjcqq	261	27	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	262	1	If	if	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	262	3	intention	intention	NN
work_zhl5dw75nfhy7oputsllvjjcqq	262	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	5	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	262	6	study	study	NN
work_zhl5dw75nfhy7oputsllvjjcqq	262	7	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	262	8	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	262	9	learn	learn	VB
work_zhl5dw75nfhy7oputsllvjjcqq	262	10	about	about	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	11	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	262	12	specific	specific	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	262	13	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	262	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	262	15	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	262	16	recommend	recommend	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	262	17	that	that	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	18	practitioners	practitioner	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	262	19	test	test	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	262	20	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	262	21	statistical	statistical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	262	22	confidence	confidence	NN
work_zhl5dw75nfhy7oputsllvjjcqq	262	23	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	24	similarities	similarity	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	262	25	based	base	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	262	26	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	27	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	262	28	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	262	29	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	30	training	train	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	262	31	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	262	32	multiple	multiple	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	262	33	bootstrap	bootstrap	NN
work_zhl5dw75nfhy7oputsllvjjcqq	262	34	samples	sample	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	262	35	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	263	1	10	10	CD
work_zhl5dw75nfhy7oputsllvjjcqq	263	2	Acknowledgements	acknowledgement	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	263	3	This	this	DT
work_zhl5dw75nfhy7oputsllvjjcqq	263	4	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	263	5	was	be	VBD
work_zhl5dw75nfhy7oputsllvjjcqq	263	6	supported	support	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	263	7	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	263	8	NSF	NSF	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	263	9	#	#	$
work_zhl5dw75nfhy7oputsllvjjcqq	263	10	1526155	1526155	CD
work_zhl5dw75nfhy7oputsllvjjcqq	263	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	263	12	#	#	NN
work_zhl5dw75nfhy7oputsllvjjcqq	263	13	1652536	1652536	CD
work_zhl5dw75nfhy7oputsllvjjcqq	263	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	263	15	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	263	16	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	263	17	Alfred	Alfred	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	263	18	P.	P.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	263	19	Sloan	Sloan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	263	20	Foundation	Foundation	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	263	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	264	1	We	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	264	2	would	would	MD
work_zhl5dw75nfhy7oputsllvjjcqq	264	3	like	like	VB
work_zhl5dw75nfhy7oputsllvjjcqq	264	4	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	264	5	thank	thank	VB
work_zhl5dw75nfhy7oputsllvjjcqq	264	6	Alexandra	Alexandra	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	7	Schofield	Schofield	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	264	9	Laure	Laure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	10	Thompson	Thompson	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	264	12	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	264	13	Action	Action	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	14	Editor	Editor	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	15	Ivan	Ivan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	16	Titov	Titov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	264	17	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	264	18	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	264	19	our	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	264	20	anonymous	anonymous	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	264	21	reviewers	reviewer	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	264	22	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	264	23	their	-PRON-	PRP$
work_zhl5dw75nfhy7oputsllvjjcqq	264	24	helpful	helpful	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	264	25	comments	comment	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	264	26	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	265	1	References	References	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	265	2	Yoshua	Yoshua	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	3	Bengio	Bengio	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	265	5	Réjean	Réjean	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	6	Ducharme	Ducharme	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	265	8	Pascal	Pascal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	9	Vincent	Vincent	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	265	11	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	265	12	Christian	Christian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	13	Jauvin	Jauvin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	265	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	266	1	2003	2003	CD
work_zhl5dw75nfhy7oputsllvjjcqq	266	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	267	1	A	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	267	2	neural	neural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	267	3	probabilistic	probabilistic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	267	4	lan-	lan-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	267	5	guage	guage	NN
work_zhl5dw75nfhy7oputsllvjjcqq	267	6	model	model	NN
work_zhl5dw75nfhy7oputsllvjjcqq	267	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	268	1	Journal	Journal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	268	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	268	3	Machine	Machine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	268	4	Learning	Learning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	268	5	Research	Research	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	268	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	268	7	3(Feb):1137–1155	3(feb):1137–1155	ADD
work_zhl5dw75nfhy7oputsllvjjcqq	268	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	269	1	David	David	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	2	M.	M.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	3	Blei	Blei	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	269	5	Andrew	Andrew	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	6	Y.	Y.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	7	Ng	Ng	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	269	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	269	10	Michael	Michael	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	11	I.	I.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	12	Jordan	Jordan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	269	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	270	1	2003	2003	CD
work_zhl5dw75nfhy7oputsllvjjcqq	270	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	271	1	Latent	latent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	271	2	dirichlet	dirichlet	NN
work_zhl5dw75nfhy7oputsllvjjcqq	271	3	allocation	allocation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	271	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	272	1	Journal	Journal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	272	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	272	3	Machine	Machine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	272	4	Learning	Learning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	272	5	research	research	NN
work_zhl5dw75nfhy7oputsllvjjcqq	272	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	272	7	3(Jan):993–1022	3(jan):993–1022	CD
work_zhl5dw75nfhy7oputsllvjjcqq	272	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	273	1	Tolga	Tolga	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	2	Bolukbasi	Bolukbasi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	273	4	Kai	Kai	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	273	6	Wei	Wei	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	7	Chang	Chang	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	273	9	James	James	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	10	Y.	Y.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	11	Zou	Zou	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	273	13	Venkatesh	Venkatesh	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	14	Saligrama	Saligrama	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	273	16	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	273	17	Adam	Adam	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	18	T.	T.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	19	Kalai	Kalai	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	273	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	274	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	274	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	275	1	Man	man	NN
work_zhl5dw75nfhy7oputsllvjjcqq	275	2	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	275	3	to	to	TO
work_zhl5dw75nfhy7oputsllvjjcqq	275	4	computer	computer	VB
work_zhl5dw75nfhy7oputsllvjjcqq	275	5	programmer	programmer	NN
work_zhl5dw75nfhy7oputsllvjjcqq	275	6	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	275	7	woman	woman	NN
work_zhl5dw75nfhy7oputsllvjjcqq	275	8	is	be	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	275	9	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	275	10	homemaker	homemaker	NN
work_zhl5dw75nfhy7oputsllvjjcqq	275	11	?	?	.
work_zhl5dw75nfhy7oputsllvjjcqq	276	1	Debiasing	debiasing	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	276	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	276	3	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	276	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	277	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	277	2	NIPS	NIPS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	277	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	277	4	pages	page	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	277	5	4349	4349	CD
work_zhl5dw75nfhy7oputsllvjjcqq	277	6	–	–	:
work_zhl5dw75nfhy7oputsllvjjcqq	277	7	4357	4357	CD
work_zhl5dw75nfhy7oputsllvjjcqq	277	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	278	1	Léon	léon	LS
work_zhl5dw75nfhy7oputsllvjjcqq	278	2	Bottou	Bottou	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	278	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	279	1	2012	2012	CD
work_zhl5dw75nfhy7oputsllvjjcqq	279	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	280	1	Stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	280	2	gradient	gradient	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	280	3	descent	descent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	280	4	tricks	trick	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	280	5	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	281	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	281	2	Neural	Neural	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	281	3	Networks	Networks	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	281	4	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	281	5	Tricks	trick	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	281	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	281	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	281	8	Trade	Trade	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	281	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	281	10	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	281	11	421–436	421–436	CD
work_zhl5dw75nfhy7oputsllvjjcqq	281	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	282	1	Springer	Springer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	282	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	283	1	Andreas	Andreas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	283	2	Broscheid	Broscheid	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	283	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	284	1	2011	2011	CD
work_zhl5dw75nfhy7oputsllvjjcqq	284	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	285	1	Comparing	compare	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	285	2	circuits	circuit	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	285	3	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	285	4	Are	be	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	285	5	some	some	DT
work_zhl5dw75nfhy7oputsllvjjcqq	285	6	U.S.	U.S.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	285	7	Courts	Courts	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	285	8	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	285	9	Appeals	Appeals	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	285	10	more	more	RBR
work_zhl5dw75nfhy7oputsllvjjcqq	285	11	liberal	liberal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	285	12	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	285	13	conservative	conservative	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	285	14	than	than	IN
work_zhl5dw75nfhy7oputsllvjjcqq	285	15	others	other	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	285	16	?	?	.
work_zhl5dw75nfhy7oputsllvjjcqq	286	1	Law	Law	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	286	2	&	&	CC
work_zhl5dw75nfhy7oputsllvjjcqq	286	3	Society	Society	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	286	4	Review	Review	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	286	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	286	6	45(1	45(1	LS
work_zhl5dw75nfhy7oputsllvjjcqq	286	7	)	)	-RRB-
work_zhl5dw75nfhy7oputsllvjjcqq	286	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	286	9	March	March	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	286	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	287	1	Dallas	Dallas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	2	Card	Card	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	287	4	Amber	Amber	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	5	E.	E.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	6	Boydstun	Boydstun	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	287	8	Justin	Justin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	9	H.	H.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	10	Gross	Gross	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	287	12	Philip	Philip	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	13	Resnik	Resnik	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	287	15	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	287	16	Noah	Noah	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	17	A.	A.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	18	Smith	Smith	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	287	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	288	1	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	288	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	289	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	289	2	media	media	NN
work_zhl5dw75nfhy7oputsllvjjcqq	289	3	frames	frame	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	289	4	corpus	corpus	NN
work_zhl5dw75nfhy7oputsllvjjcqq	289	5	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	289	6	Annotations	annotation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	289	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	289	8	frames	frame	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	289	9	across	across	IN
work_zhl5dw75nfhy7oputsllvjjcqq	289	10	issues	issue	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	289	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	290	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	290	2	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	290	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	291	1	Danqi	Danqi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	291	2	Chen	Chen	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	291	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	291	4	Christopher	Christopher	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	291	5	D.	D.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	291	6	Manning	Manning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	291	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	292	1	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	292	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	293	1	A	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	293	2	fast	fast	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	293	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	293	4	accurate	accurate	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	293	5	dependency	dependency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	293	6	parser	parser	NN
work_zhl5dw75nfhy7oputsllvjjcqq	293	7	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	293	8	neural	neural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	293	9	networks	network	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	293	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	294	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	294	2	EMNLP	EMNLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	294	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	294	4	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	294	5	740–750	740–750	CD
work_zhl5dw75nfhy7oputsllvjjcqq	294	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	295	1	Colin	Colin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	295	2	Cherry	Cherry	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	295	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	295	4	Hongyu	Hongyu	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	295	5	Guo	Guo	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	295	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	296	1	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	296	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	297	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	297	2	unreasonable	unreasonable	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	297	3	effectiveness	effectiveness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	297	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	297	5	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	297	6	representations	representation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	297	7	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	297	8	Twitter	Twitter	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	297	9	named	name	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	297	10	entity	entity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	297	11	recognition	recognition	NN
work_zhl5dw75nfhy7oputsllvjjcqq	297	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	298	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	298	2	HLT	HLT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	298	3	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	298	4	NAACL	NAACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	298	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	298	6	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	298	7	735–745	735–745	CD
work_zhl5dw75nfhy7oputsllvjjcqq	298	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	299	1	Scott	Scott	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	2	Deerwester	Deerwester	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	299	4	Susan	Susan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	5	T.	T.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	6	Dumais	Dumais	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	299	8	George	George	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	9	W.	W.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	10	Furnas	Furnas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	299	12	Thomas	Thomas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	13	K.	K.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	14	Landauer	Landauer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	299	16	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	299	17	Richard	Richard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	18	Harshman	Harshman	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	299	19	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	300	1	1990	1990	CD
work_zhl5dw75nfhy7oputsllvjjcqq	300	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	301	1	Indexing	indexing	NN
work_zhl5dw75nfhy7oputsllvjjcqq	301	2	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	301	3	latent	latent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	301	4	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	301	5	analysis	analysis	NN
work_zhl5dw75nfhy7oputsllvjjcqq	301	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	302	1	Journal	Journal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	302	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	302	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	302	4	American	American	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	302	5	Society	Society	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	302	6	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	302	7	Information	Information	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	302	8	Science	Science	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	302	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	302	10	41(6):391	41(6):391	CD
work_zhl5dw75nfhy7oputsllvjjcqq	302	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	303	1	Manaal	Manaal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	2	Faruqui	Faruqui	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	303	4	Jesse	Jesse	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	5	Dodge	Dodge	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	303	7	Sujay	Sujay	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	8	K.	K.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	9	Jauhar	Jauhar	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	10	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	303	11	Chris	Chris	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	12	Dyer	Dyer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	303	14	Eduard	Eduard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	15	Hovy	Hovy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	303	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	303	18	Noah	Noah	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	19	A.	A.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	20	Smith	Smith	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	303	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	304	1	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	304	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	305	1	Retrofitting	retrofit	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	305	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	305	3	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	305	4	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	305	5	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	305	6	lexicons	lexicon	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	305	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	306	1	HLT	HLT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	306	2	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	306	3	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	306	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	306	5	pages	page	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	306	6	1606–1615	1606–1615	CD
work_zhl5dw75nfhy7oputsllvjjcqq	306	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	307	1	Anna	Anna	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	307	2	Gladkova	Gladkova	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	307	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	307	4	Aleksandr	Aleksandr	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	307	5	Drozd	Drozd	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	307	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	308	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	308	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	309	1	Intrinsic	intrinsic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	309	2	evaluations	evaluation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	309	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	309	4	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	309	5	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	309	6	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	309	7	What	what	WP
work_zhl5dw75nfhy7oputsllvjjcqq	309	8	can	can	MD
work_zhl5dw75nfhy7oputsllvjjcqq	309	9	we	-PRON-	PRP
work_zhl5dw75nfhy7oputsllvjjcqq	309	10	do	do	VB
work_zhl5dw75nfhy7oputsllvjjcqq	309	11	bet-	bet-	NN
work_zhl5dw75nfhy7oputsllvjjcqq	309	12	ter	ter	NN
work_zhl5dw75nfhy7oputsllvjjcqq	309	13	?	?	.
work_zhl5dw75nfhy7oputsllvjjcqq	310	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	310	2	Proceedings	proceeding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	310	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	310	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	310	5	1st	1st	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	310	6	Workshop	Workshop	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	310	7	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	310	8	Evaluating	Evaluating	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	310	9	Vector	Vector	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	310	10	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	310	11	Space	Space	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	310	12	Representations	Representations	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	310	13	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	310	14	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	310	15	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	310	16	pages	page	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	310	17	36–42	36–42	CD
work_zhl5dw75nfhy7oputsllvjjcqq	310	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	311	1	Yoav	Yoav	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	311	2	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	311	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	312	1	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	312	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	313	1	Neural	Neural	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	313	2	Network	Network	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	313	3	Methods	Methods	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	313	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	313	5	Nat-	Nat-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	313	6	ural	ural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	313	7	Language	Language	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	313	8	Processing	Processing	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	313	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	314	1	Synthesis	Synthesis	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	314	2	Lectures	Lectures	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	314	3	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	314	4	Hu-	Hu-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	314	5	man	man	NN
work_zhl5dw75nfhy7oputsllvjjcqq	314	6	Language	Language	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	314	7	Technologies	Technologies	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	314	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	315	1	Morgan	Morgan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	315	2	&	&	CC
work_zhl5dw75nfhy7oputsllvjjcqq	315	3	Claypool	Claypool	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	315	4	Pub-	Pub-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	315	5	lishers	lisher	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	315	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	316	1	Ian	Ian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	316	2	Goodfellow	Goodfellow	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	316	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	316	4	Yoshua	Yoshua	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	316	5	Bengio	Bengio	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	316	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	316	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	316	8	Aaron	Aaron	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	316	9	Courville	Courville	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	316	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	317	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	317	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	318	1	Deep	Deep	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	318	2	Learning	Learning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	318	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	319	1	MIT	MIT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	319	2	Press	Press	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	319	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	320	1	Nathan	Nathan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	2	Halko	Halko	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	320	4	Per	Per	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	320	6	Gunnar	Gunnar	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	7	Martinsson	Martinsson	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	320	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	320	10	Joel	Joel	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	11	A.	A.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	12	Tropp	Tropp	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	320	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	321	1	September	September	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	321	2	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	321	3	2009	2009	CD
work_zhl5dw75nfhy7oputsllvjjcqq	321	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	322	1	Finding	find	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	322	2	structure	structure	NN
work_zhl5dw75nfhy7oputsllvjjcqq	322	3	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	322	4	randomness	randomness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	322	5	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	322	6	Stochastic	stochastic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	322	7	algorithms	algorithm	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	322	8	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	322	9	constructing	construct	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	322	10	approximate	approximate	RB
work_zhl5dw75nfhy7oputsllvjjcqq	322	11	ma-	ma-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	322	12	trix	trix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	322	13	decompositions	decomposition	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	322	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	323	1	Technical	Technical	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	323	2	Report	Report	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	323	3	No	no	UH
work_zhl5dw75nfhy7oputsllvjjcqq	323	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	324	1	2009	2009	CD
work_zhl5dw75nfhy7oputsllvjjcqq	324	2	-	-	SYM
work_zhl5dw75nfhy7oputsllvjjcqq	324	3	05	05	CD
work_zhl5dw75nfhy7oputsllvjjcqq	324	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	325	1	Applied	Applied	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	325	2	&	&	CC
work_zhl5dw75nfhy7oputsllvjjcqq	325	3	Computational	Computational	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	325	4	Mathematics	Mathematics	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	325	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	325	6	California	California	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	325	7	In-	In-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	325	8	stitute	stitute	NN
work_zhl5dw75nfhy7oputsllvjjcqq	325	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	325	10	Technology	Technology	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	325	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	326	1	William	William	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	2	L.	L.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	3	Hamilton	Hamilton	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	326	5	Jure	Jure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	6	Leskovec	Leskovec	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	326	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	326	9	Dan	Dan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	10	Jurafsky	Jurafsky	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	326	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	327	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	327	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	328	1	Diachronic	diachronic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	328	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	328	3	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	328	4	reveal	reveal	VBP
work_zhl5dw75nfhy7oputsllvjjcqq	328	5	statistical	statistical	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	328	6	laws	law	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	328	7	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	328	8	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	328	9	change	change	NN
work_zhl5dw75nfhy7oputsllvjjcqq	328	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	329	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	329	2	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	329	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	330	1	Johannes	Johannes	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	330	2	Hellrich	Hellrich	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	330	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	330	4	Udo	Udo	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	330	5	Hahn	Hahn	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	330	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	331	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	331	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	332	1	Bad	bad	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	332	2	company	company	NN
work_zhl5dw75nfhy7oputsllvjjcqq	332	3	–	–	:
work_zhl5dw75nfhy7oputsllvjjcqq	332	4	neighborhoods	neighborhood	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	332	5	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	332	6	neural	neural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	332	7	embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	332	8	spaces	space	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	332	9	considered	consider	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	332	10	harmful	harmful	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	332	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	333	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	333	2	Proceedings	Proceedings	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	333	4	COLING	COLING	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	5	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	333	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	333	7	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	333	8	26th	26th	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	333	9	In-	In-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	10	ternational	ternational	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	333	11	Conference	Conference	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	12	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	333	13	Computational	Computational	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	14	Linguistics	Linguistics	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	333	15	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	333	16	Technical	Technical	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	17	Papers	Papers	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	333	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	333	19	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	333	20	2785–2796	2785–2796	CD
work_zhl5dw75nfhy7oputsllvjjcqq	333	21	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	334	1	Ryan	Ryan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	334	2	Heuser	Heuser	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	334	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	335	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	335	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	336	1	Word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	336	2	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	336	3	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	336	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	336	5	eighteenth-	eighteenth-	CD
work_zhl5dw75nfhy7oputsllvjjcqq	336	6	century	century	NN
work_zhl5dw75nfhy7oputsllvjjcqq	336	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	337	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	337	2	IPAM	IPAM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	337	3	Workshop	Workshop	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	337	4	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	337	5	Cultural	Cultural	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	337	6	Analytics	Analytics	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	337	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	338	1	Yoon	Yoon	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	2	Kim	Kim	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	338	4	Yi	Yi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	338	6	I	I	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	7	Chiu	Chiu	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	338	9	Kentaro	Kentaro	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	10	Hanaki	Hanaki	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	11	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	338	12	Darshan	Darshan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	13	Hegde	Hegde	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	338	15	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	338	16	Slav	Slav	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	17	Petrov	Petrov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	338	18	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	339	1	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	339	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	340	1	Temporal	temporal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	340	2	analysis	analysis	NN
work_zhl5dw75nfhy7oputsllvjjcqq	340	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	340	4	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	340	5	through	through	IN
work_zhl5dw75nfhy7oputsllvjjcqq	340	6	neural	neural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	340	7	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	340	8	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	340	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	341	1	Proceedings	proceeding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	341	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	341	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	341	4	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	341	5	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	341	6	Workshop	Workshop	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	341	7	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	341	8	Language	Language	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	341	9	Technologies	Technologies	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	341	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	341	11	Computational	Computational	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	341	12	Social	Social	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	341	13	Science	Science	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	341	14	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	341	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	342	1	118	118	CD
work_zhl5dw75nfhy7oputsllvjjcqq	342	2	Vivek	Vivek	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	342	3	Kulkarni	Kulkarni	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	342	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	342	5	Bryan	Bryan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	342	6	Perozzi	Perozzi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	342	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	342	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	342	9	Steven	Steven	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	342	10	Skiena	Skiena	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	342	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	343	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	343	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	344	1	Freshman	freshman	NN
work_zhl5dw75nfhy7oputsllvjjcqq	344	2	or	or	CC
work_zhl5dw75nfhy7oputsllvjjcqq	344	3	fresher	fresher	NN
work_zhl5dw75nfhy7oputsllvjjcqq	344	4	?	?	.
work_zhl5dw75nfhy7oputsllvjjcqq	345	1	Quantifying	quantify	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	345	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	345	3	geographic	geographic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	345	4	vari-	vari-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	345	5	ation	ation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	345	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	345	7	language	language	NN
work_zhl5dw75nfhy7oputsllvjjcqq	345	8	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	345	9	online	online	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	345	10	social	social	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	345	11	media	medium	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	345	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	346	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	346	2	ICWSM	ICWSM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	346	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	346	4	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	346	5	615–618	615–618	CD
work_zhl5dw75nfhy7oputsllvjjcqq	346	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	347	1	Thomas	Thomas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	347	2	K.	K.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	347	3	Landauer	Landauer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	347	4	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	347	5	Susan	Susan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	347	6	T.	T.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	347	7	Dumais	Dumais	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	347	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	348	1	1997	1997	CD
work_zhl5dw75nfhy7oputsllvjjcqq	348	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	349	1	A	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	349	2	so-	so-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	349	3	lution	lution	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	4	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	349	5	Plato	Plato	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	349	6	’s	’s	POS
work_zhl5dw75nfhy7oputsllvjjcqq	349	7	problem	problem	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	8	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	349	9	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	349	10	latent	latent	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	349	11	semantic	semantic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	349	12	analysis	analysis	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	13	theory	theory	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	14	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	349	15	acquisition	acquisition	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	349	17	induction	induction	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	18	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	349	19	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	349	20	representation	representation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	21	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	349	22	knowledge	knowledge	NN
work_zhl5dw75nfhy7oputsllvjjcqq	349	23	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	350	1	Psychological	Psychological	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	350	2	Review	Review	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	350	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	350	4	104(2):211	104(2):211	CD
work_zhl5dw75nfhy7oputsllvjjcqq	350	5	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	351	1	Omer	Omer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	351	2	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	351	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	351	4	Yoav	Yoav	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	351	5	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	351	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	352	1	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	352	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	353	1	Neural	neural	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	353	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	353	3	embedding	embed	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	353	4	as	as	IN
work_zhl5dw75nfhy7oputsllvjjcqq	353	5	implicit	implicit	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	353	6	matrix	matrix	NN
work_zhl5dw75nfhy7oputsllvjjcqq	353	7	factorization	factorization	NN
work_zhl5dw75nfhy7oputsllvjjcqq	353	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	354	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	354	2	NIPS	NIPS	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	354	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	354	4	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	354	5	2177–2185	2177–2185	CD
work_zhl5dw75nfhy7oputsllvjjcqq	354	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	355	1	Omer	Omer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	355	2	Levy	Levy	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	355	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	355	4	Yoav	Yoav	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	355	5	Goldberg	Goldberg	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	355	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	355	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	355	8	Ido	Ido	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	355	9	Dagan	Dagan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	355	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	356	1	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	356	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	357	1	Im-	im-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	357	2	proving	prove	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	357	3	distributional	distributional	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	357	4	similarity	similarity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	357	5	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	357	6	lessons	lesson	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	357	7	learned	learn	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	357	8	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	357	9	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	357	10	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	357	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	358	1	Transactions	transaction	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	358	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	358	3	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	358	4	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	358	5	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	358	6	3:211–225	3:211–225	CD
work_zhl5dw75nfhy7oputsllvjjcqq	358	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	359	1	Rasmus	Rasmus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	2	E.	E.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	3	Madsen	Madsen	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	4	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	359	5	David	David	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	6	Kauchak	Kauchak	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	359	8	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	359	9	Charles	Charles	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	10	Elkan	Elkan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	359	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	360	1	2005	2005	CD
work_zhl5dw75nfhy7oputsllvjjcqq	360	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	361	1	Modeling	model	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	361	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	361	3	burstiness	burstiness	NN
work_zhl5dw75nfhy7oputsllvjjcqq	361	4	using	use	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	361	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	361	6	dirichlet	dirichlet	NN
work_zhl5dw75nfhy7oputsllvjjcqq	361	7	dis-	dis-	IN
work_zhl5dw75nfhy7oputsllvjjcqq	361	8	tribution	tribution	NN
work_zhl5dw75nfhy7oputsllvjjcqq	361	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	362	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	362	2	Proceedings	proceeding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	362	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	362	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	362	5	22nd	22nd	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	362	6	International	International	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	362	7	Con-	Con-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	362	8	ference	ference	NN
work_zhl5dw75nfhy7oputsllvjjcqq	362	9	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	362	10	Machine	Machine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	362	11	Learning	Learning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	362	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	362	13	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	362	14	545–552	545–552	CD
work_zhl5dw75nfhy7oputsllvjjcqq	362	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	363	1	ACM	ACM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	363	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	364	1	Tomas	Tomas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	2	Mikolov	Mikolov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	364	4	Wen	Wen	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	364	6	tau	tau	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	7	Yih	Yih	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	8	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	364	9	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	364	10	Geoffrey	Geoffrey	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	11	Zweig	Zweig	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	364	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	365	1	2013	2013	CD
work_zhl5dw75nfhy7oputsllvjjcqq	365	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	366	1	Linguistic	linguistic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	366	2	Regularities	Regularities	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	366	3	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	366	4	Continuous	continuous	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	366	5	Space	Space	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	366	6	Word	Word	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	366	7	Rep-	Rep-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	366	8	resentations	resentation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	366	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	367	1	HLT	HLT	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	367	2	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	367	3	NAACL	NAACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	367	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	368	1	Jeffrey	Jeffrey	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	2	Pennington	Pennington	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	368	4	Richard	Richard	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	5	Socher	Socher	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	368	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	368	8	Christopher	Christopher	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	9	D.	D.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	10	Manning	Manning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	368	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	369	1	2014	2014	CD
work_zhl5dw75nfhy7oputsllvjjcqq	369	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	370	1	GloVe	GloVe	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	370	2	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	370	3	Global	global	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	370	4	vectors	vector	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	370	5	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	370	6	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	370	7	representation	representation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	370	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	371	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	371	2	EMNLP	emnlp	NN
work_zhl5dw75nfhy7oputsllvjjcqq	371	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	371	4	volume	volume	NN
work_zhl5dw75nfhy7oputsllvjjcqq	371	5	14	14	CD
work_zhl5dw75nfhy7oputsllvjjcqq	371	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	371	7	pages	page	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	371	8	1532–43	1532–43	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	371	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	372	1	Lawrence	Lawrence	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	2	Phillips	Phillips	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	372	4	Kyle	Kyle	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	5	Shaffer	Shaffer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	372	7	Dustin	Dustin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	8	Arendt	Arendt	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	372	10	Nathan	Nathan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	11	Hodas	Hodas	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	372	13	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	372	14	Svitlana	Svitlana	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	15	Volkova	Volkova	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	372	16	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	373	1	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	373	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	374	1	Intrinsic	intrinsic	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	374	2	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	374	3	ex-	ex-	XX
work_zhl5dw75nfhy7oputsllvjjcqq	374	4	trinsic	trinsic	NN
work_zhl5dw75nfhy7oputsllvjjcqq	374	5	evaluation	evaluation	NN
work_zhl5dw75nfhy7oputsllvjjcqq	374	6	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	374	7	spatiotemporal	spatiotemporal	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	374	8	text	text	NN
work_zhl5dw75nfhy7oputsllvjjcqq	374	9	representations	representation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	374	10	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	374	11	Twitter	Twitter	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	374	12	streams	stream	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	374	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	375	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	375	2	Proceedings	proceeding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	375	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	375	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	375	5	2nd	2nd	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	375	6	Workshop	Workshop	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	375	7	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	375	8	Representation	Representation	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	375	9	Learning	Learning	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	375	10	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	375	11	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	375	12	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	375	13	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	375	14	201–210	201–210	CD
work_zhl5dw75nfhy7oputsllvjjcqq	375	15	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	376	1	Radim	Radim	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	376	2	Řehůřek	řehůřek	CD
work_zhl5dw75nfhy7oputsllvjjcqq	376	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	376	4	Petr	Petr	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	376	5	Sojka	sojka	RB
work_zhl5dw75nfhy7oputsllvjjcqq	376	6	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	377	1	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	377	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	378	1	Software	Software	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	378	2	Frame-	Frame-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	378	3	work	work	NN
work_zhl5dw75nfhy7oputsllvjjcqq	378	4	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	378	5	Topic	Topic	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	378	6	Modelling	Modelling	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	378	7	with	with	IN
work_zhl5dw75nfhy7oputsllvjjcqq	378	8	Large	large	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	378	9	Corpora	Corpora	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	378	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	379	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	379	2	Pro-	Pro-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	3	ceedings	ceeding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	379	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	379	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	379	6	LREC	LREC	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	7	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	379	8	Workshop	Workshop	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	9	on	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	379	10	New	New	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	11	Chal-	Chal-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	12	lenges	lenge	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	379	13	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	379	14	NLP	NLP	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	15	Frameworks	Frameworks	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	379	17	pages	page	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	379	18	45–50	45–50	CD
work_zhl5dw75nfhy7oputsllvjjcqq	379	19	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	379	20	Valletta	Valletta	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	21	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	379	22	Malta	Malta	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	379	23	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	379	24	May.	May	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	380	1	ELRA	ELRA	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	380	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	381	1	Evan	Evan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	381	2	Sandhaus	Sandhaus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	381	3	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	382	1	2008	2008	CD
work_zhl5dw75nfhy7oputsllvjjcqq	382	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	383	1	The	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	383	2	New	New	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	383	3	York	York	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	383	4	Times	Times	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	383	5	Annotated	Annotated	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	383	6	Corpus	Corpus	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	383	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	384	1	LDC2008T19	LDC2008T19	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	384	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	385	1	Linguistic	Linguistic	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	385	2	Data	Data	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	385	3	Consortium	Consortium	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	385	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	386	1	Noam	Noam	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	2	Shazeer	Shazeer	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	386	4	Ryan	Ryan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	5	Doherty	Doherty	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	386	7	Colin	Colin	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	8	Evans	Evans	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	386	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	386	11	Chris	Chris	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	12	Waterson	Waterson	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	386	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	387	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	387	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	388	1	Swivel	swivel	NN
work_zhl5dw75nfhy7oputsllvjjcqq	388	2	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	388	3	Improving	improve	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	388	4	Embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	388	5	by	by	IN
work_zhl5dw75nfhy7oputsllvjjcqq	388	6	Noticing	notice	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	388	7	What	what	WP
work_zhl5dw75nfhy7oputsllvjjcqq	388	8	’s	’	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	388	9	Missing	miss	VBG
work_zhl5dw75nfhy7oputsllvjjcqq	388	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	389	1	arXiv:1602.02215	arXiv:1602.02215	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	389	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	390	1	Yingtao	Yingtao	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	2	Tian	Tian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	390	4	Vivek	Vivek	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	5	Kulkarni	Kulkarni	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	390	7	Bryan	Bryan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	8	Perozzi	Perozzi	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	9	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	390	10	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	390	11	Steven	Steven	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	12	Skiena	Skiena	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	390	13	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	391	1	2016	2016	CD
work_zhl5dw75nfhy7oputsllvjjcqq	391	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	392	1	On	on	IN
work_zhl5dw75nfhy7oputsllvjjcqq	392	2	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	392	3	convergent	convergent	NN
work_zhl5dw75nfhy7oputsllvjjcqq	392	4	properties	property	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	392	5	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	392	6	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	392	7	embedding	embedding	NN
work_zhl5dw75nfhy7oputsllvjjcqq	392	8	methods	method	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	392	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	393	1	arXiv	arXiv	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	393	2	preprint	preprint	NN
work_zhl5dw75nfhy7oputsllvjjcqq	393	3	arXiv:1605.03956	arXiv:1605.03956	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	393	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	394	1	Joseph	Joseph	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	394	2	Turian	Turian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	394	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	394	4	Lev	Lev	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	394	5	Ratinov	Ratinov	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	394	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	394	7	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	394	8	Yoshua	Yoshua	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	394	9	Bengio	Bengio	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	394	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	395	1	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	395	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	396	1	Word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	396	2	representations	representation	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	396	3	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	396	4	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	396	5	simple	simple	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	396	6	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	396	7	general	general	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	396	8	method	method	NN
work_zhl5dw75nfhy7oputsllvjjcqq	396	9	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	396	10	semi	semi	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	396	11	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	396	12	supervised	supervised	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	396	13	learning	learning	NN
work_zhl5dw75nfhy7oputsllvjjcqq	396	14	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	397	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	397	2	Proceedings	Proceedings	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	397	3	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	397	4	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	397	5	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	397	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	397	7	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	397	8	384–394	384–394	CD
work_zhl5dw75nfhy7oputsllvjjcqq	397	9	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	398	1	Association	Association	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	398	2	for	for	IN
work_zhl5dw75nfhy7oputsllvjjcqq	398	3	Computational	Computational	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	398	4	Linguistics	Linguistics	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	398	5	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	399	1	Peter	Peter	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	399	2	D.	D.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	399	3	Turney	Turney	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	399	4	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	399	5	Patrick	Patrick	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	399	6	Pantel	Pantel	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	399	7	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	400	1	2010	2010	CD
work_zhl5dw75nfhy7oputsllvjjcqq	400	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	401	1	From	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	401	2	frequency	frequency	NN
work_zhl5dw75nfhy7oputsllvjjcqq	401	3	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	401	4	meaning	meaning	NN
work_zhl5dw75nfhy7oputsllvjjcqq	401	5	:	:	:
work_zhl5dw75nfhy7oputsllvjjcqq	401	6	Vector	vector	NN
work_zhl5dw75nfhy7oputsllvjjcqq	401	7	space	space	NN
work_zhl5dw75nfhy7oputsllvjjcqq	401	8	models	model	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	401	9	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	401	10	semantics	semantic	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	401	11	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	402	1	Journal	Journal	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	402	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	402	3	Artificial	Artificial	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	402	4	Intelligence	Intelligence	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	402	5	Research	Research	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	402	6	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	402	7	37:141–188	37:141–188	CD
work_zhl5dw75nfhy7oputsllvjjcqq	402	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	403	1	Ivan	Ivan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	403	2	Vulic	Vulic	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	403	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	403	4	Marie	Marie	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	403	5	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	403	6	Francine	Francine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	403	7	Moens	Moens	NNPS
work_zhl5dw75nfhy7oputsllvjjcqq	403	8	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	404	1	2015	2015	CD
work_zhl5dw75nfhy7oputsllvjjcqq	404	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	405	1	Bilingual	bilingual	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	405	2	word	word	NN
work_zhl5dw75nfhy7oputsllvjjcqq	405	3	embeddings	embedding	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	405	4	from	from	IN
work_zhl5dw75nfhy7oputsllvjjcqq	405	5	non	non	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	405	6	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	405	7	parallel	parallel	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	405	8	document	document	NN
work_zhl5dw75nfhy7oputsllvjjcqq	405	9	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	405	10	aligned	align	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	405	11	data	datum	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	405	12	applied	apply	VBN
work_zhl5dw75nfhy7oputsllvjjcqq	405	13	to	to	IN
work_zhl5dw75nfhy7oputsllvjjcqq	405	14	bilingual	bilingual	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	405	15	lexicon	lexicon	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	405	16	induction	induction	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	405	17	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	406	1	In	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	406	2	Proceed-	Proceed-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	406	3	ings	ing	NNS
work_zhl5dw75nfhy7oputsllvjjcqq	406	4	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	406	5	the	the	DT
work_zhl5dw75nfhy7oputsllvjjcqq	406	6	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	406	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	406	8	pages	page	VBZ
work_zhl5dw75nfhy7oputsllvjjcqq	406	9	719–725	719–725	CD
work_zhl5dw75nfhy7oputsllvjjcqq	406	10	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	407	1	ACL	ACL	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	407	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	408	1	Justine	Justine	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	2	Zhang	Zhang	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	3	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	408	4	William	William	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	5	L.	L.	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	6	Hamilton	Hamilton	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	7	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	408	8	Cristian	Cristian	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	9	Danescu-	Danescu-	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	10	Niculescu	Niculescu	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	11	-	-	HYPH
work_zhl5dw75nfhy7oputsllvjjcqq	408	12	Mizil	Mizil	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	13	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	408	14	Dan	Dan	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	15	Jurafsky	Jurafsky	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	16	,	,	,
work_zhl5dw75nfhy7oputsllvjjcqq	408	17	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	408	18	Jure	Jure	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	19	Leskovec	Leskovec	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	408	20	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	409	1	2017	2017	CD
work_zhl5dw75nfhy7oputsllvjjcqq	409	2	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	410	1	Community	community	NN
work_zhl5dw75nfhy7oputsllvjjcqq	410	2	identity	identity	NN
work_zhl5dw75nfhy7oputsllvjjcqq	410	3	and	and	CC
work_zhl5dw75nfhy7oputsllvjjcqq	410	4	user	user	NN
work_zhl5dw75nfhy7oputsllvjjcqq	410	5	engagement	engagement	NN
work_zhl5dw75nfhy7oputsllvjjcqq	410	6	in	in	IN
work_zhl5dw75nfhy7oputsllvjjcqq	410	7	a	a	DT
work_zhl5dw75nfhy7oputsllvjjcqq	410	8	multi	multi	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	410	9	-	-	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	410	10	community	community	JJ
work_zhl5dw75nfhy7oputsllvjjcqq	410	11	landscape	landscape	NN
work_zhl5dw75nfhy7oputsllvjjcqq	410	12	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	411	1	Proceedings	Proceedings	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	411	2	of	of	IN
work_zhl5dw75nfhy7oputsllvjjcqq	411	3	ICWSM	ICWSM	NNP
work_zhl5dw75nfhy7oputsllvjjcqq	411	4	.	.	.
work_zhl5dw75nfhy7oputsllvjjcqq	412	1	119	119	CD
work_zhl5dw75nfhy7oputsllvjjcqq	412	2	120	120	CD