id	sid	tid	token	lemma	pos
work_27d5tr326feydgcx32t4nhl424	1	1	Transactions	transaction	NNS
work_27d5tr326feydgcx32t4nhl424	1	2	of	of	IN
work_27d5tr326feydgcx32t4nhl424	1	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	1	4	Association	Association	NNP
work_27d5tr326feydgcx32t4nhl424	1	5	for	for	IN
work_27d5tr326feydgcx32t4nhl424	1	6	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	1	7	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	1	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	1	9	1	1	CD
work_27d5tr326feydgcx32t4nhl424	1	10	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	1	11	2013	2013	CD
work_27d5tr326feydgcx32t4nhl424	1	12	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	1	13	89–98	89–98	CD
work_27d5tr326feydgcx32t4nhl424	1	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	2	1	Action	Action	NNP
work_27d5tr326feydgcx32t4nhl424	2	2	Editor	Editor	NNP
work_27d5tr326feydgcx32t4nhl424	2	3	:	:	:
work_27d5tr326feydgcx32t4nhl424	2	4	Noah	Noah	NNP
work_27d5tr326feydgcx32t4nhl424	2	5	Smith	Smith	NNP
work_27d5tr326feydgcx32t4nhl424	2	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	3	1	Submitted	submit	VBN
work_27d5tr326feydgcx32t4nhl424	3	2	12/2012	12/2012	CD
work_27d5tr326feydgcx32t4nhl424	3	3	;	;	:
work_27d5tr326feydgcx32t4nhl424	3	4	Published	publish	VBN
work_27d5tr326feydgcx32t4nhl424	3	5	5/2013	5/2013	CD
work_27d5tr326feydgcx32t4nhl424	3	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	4	1	c	c	NNP
work_27d5tr326feydgcx32t4nhl424	4	2	©	©	NNP
work_27d5tr326feydgcx32t4nhl424	4	3	2013	2013	CD
work_27d5tr326feydgcx32t4nhl424	4	4	Association	Association	NNP
work_27d5tr326feydgcx32t4nhl424	4	5	for	for	IN
work_27d5tr326feydgcx32t4nhl424	4	6	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	4	7	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	4	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	5	1	A	a	DT
work_27d5tr326feydgcx32t4nhl424	5	2	Novel	Novel	NNP
work_27d5tr326feydgcx32t4nhl424	5	3	Feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	5	4	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	5	5	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	5	6	Bayesian	Bayesian	NNP
work_27d5tr326feydgcx32t4nhl424	5	7	Model	Model	NNP
work_27d5tr326feydgcx32t4nhl424	5	8	for	for	IN
work_27d5tr326feydgcx32t4nhl424	5	9	Query	Query	NNP
work_27d5tr326feydgcx32t4nhl424	5	10	Focused	Focused	NNP
work_27d5tr326feydgcx32t4nhl424	5	11	Multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	5	12	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	5	13	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	5	14	Summarization	Summarization	NNP
work_27d5tr326feydgcx32t4nhl424	5	15	Jiwei	Jiwei	NNP
work_27d5tr326feydgcx32t4nhl424	5	16	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	5	17	School	School	NNP
work_27d5tr326feydgcx32t4nhl424	5	18	of	of	IN
work_27d5tr326feydgcx32t4nhl424	5	19	Computer	Computer	NNP
work_27d5tr326feydgcx32t4nhl424	5	20	Science	Science	NNP
work_27d5tr326feydgcx32t4nhl424	5	21	Carnegie	Carnegie	NNP
work_27d5tr326feydgcx32t4nhl424	5	22	Mellon	Mellon	NNP
work_27d5tr326feydgcx32t4nhl424	5	23	University	University	NNP
work_27d5tr326feydgcx32t4nhl424	5	24	bdlijiwei@gmail.com	bdlijiwei@gmail.com	ADD
work_27d5tr326feydgcx32t4nhl424	5	25	Sujian	Sujian	NNP
work_27d5tr326feydgcx32t4nhl424	5	26	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	5	27	Laboratory	Laboratory	NNP
work_27d5tr326feydgcx32t4nhl424	5	28	of	of	IN
work_27d5tr326feydgcx32t4nhl424	5	29	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	5	30	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	5	31	Peking	Peking	NNP
work_27d5tr326feydgcx32t4nhl424	5	32	University	University	NNP
work_27d5tr326feydgcx32t4nhl424	5	33	lisujian@pku.edu.cn	lisujian@pku.edu.cn	NN
work_27d5tr326feydgcx32t4nhl424	5	34	Abstract	Abstract	NNP
work_27d5tr326feydgcx32t4nhl424	5	35	Supervised	supervise	VBD
work_27d5tr326feydgcx32t4nhl424	5	36	learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	5	37	methods	method	NNS
work_27d5tr326feydgcx32t4nhl424	5	38	and	and	CC
work_27d5tr326feydgcx32t4nhl424	5	39	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	5	40	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	5	41	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	5	42	model	model	NN
work_27d5tr326feydgcx32t4nhl424	5	43	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	5	44	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	5	45	successfully	successfully	RB
work_27d5tr326feydgcx32t4nhl424	5	46	applied	apply	VBN
work_27d5tr326feydgcx32t4nhl424	5	47	in	in	IN
work_27d5tr326feydgcx32t4nhl424	5	48	the	the	DT
work_27d5tr326feydgcx32t4nhl424	5	49	field	field	NN
work_27d5tr326feydgcx32t4nhl424	5	50	of	of	IN
work_27d5tr326feydgcx32t4nhl424	5	51	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	5	52	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	5	53	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	5	54	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	5	55	.	.	.
work_27d5tr326feydgcx32t4nhl424	6	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	6	2	this	this	DT
work_27d5tr326feydgcx32t4nhl424	6	3	paper	paper	NN
work_27d5tr326feydgcx32t4nhl424	6	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	6	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	6	6	propose	propose	VBP
work_27d5tr326feydgcx32t4nhl424	6	7	a	a	DT
work_27d5tr326feydgcx32t4nhl424	6	8	novel	novel	NN
work_27d5tr326feydgcx32t4nhl424	6	9	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	6	10	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	6	11	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	6	12	can	can	MD
work_27d5tr326feydgcx32t4nhl424	6	13	in-	in-	XX
work_27d5tr326feydgcx32t4nhl424	6	14	corporate	corporate	JJ
work_27d5tr326feydgcx32t4nhl424	6	15	rich	rich	JJ
work_27d5tr326feydgcx32t4nhl424	6	16	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	6	17	features	feature	VBZ
work_27d5tr326feydgcx32t4nhl424	6	18	into	into	IN
work_27d5tr326feydgcx32t4nhl424	6	19	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	6	20	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	6	21	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	6	22	in	in	IN
work_27d5tr326feydgcx32t4nhl424	6	23	a	a	DT
work_27d5tr326feydgcx32t4nhl424	6	24	principled	principle	VBN
work_27d5tr326feydgcx32t4nhl424	6	25	way	way	NN
work_27d5tr326feydgcx32t4nhl424	6	26	,	,	,
work_27d5tr326feydgcx32t4nhl424	6	27	thus	thus	RB
work_27d5tr326feydgcx32t4nhl424	6	28	taking	take	VBG
work_27d5tr326feydgcx32t4nhl424	6	29	advantages	advantage	NNS
work_27d5tr326feydgcx32t4nhl424	6	30	of	of	IN
work_27d5tr326feydgcx32t4nhl424	6	31	both	both	CC
work_27d5tr326feydgcx32t4nhl424	6	32	topic	topic	JJ
work_27d5tr326feydgcx32t4nhl424	6	33	model	model	NN
work_27d5tr326feydgcx32t4nhl424	6	34	and	and	CC
work_27d5tr326feydgcx32t4nhl424	6	35	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	6	36	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	6	37	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	6	38	learn-	learn-	JJ
work_27d5tr326feydgcx32t4nhl424	6	39	ing	ing	NN
work_27d5tr326feydgcx32t4nhl424	6	40	methods	method	NNS
work_27d5tr326feydgcx32t4nhl424	6	41	.	.	.
work_27d5tr326feydgcx32t4nhl424	7	1	Experimental	experimental	JJ
work_27d5tr326feydgcx32t4nhl424	7	2	results	result	NNS
work_27d5tr326feydgcx32t4nhl424	7	3	on	on	IN
work_27d5tr326feydgcx32t4nhl424	7	4	DUC2007	duc2007	NN
work_27d5tr326feydgcx32t4nhl424	7	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	7	6	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	7	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	7	8	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	7	9	demonstrate	demonstrate	VB
work_27d5tr326feydgcx32t4nhl424	7	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	7	11	effective-	effective-	JJ
work_27d5tr326feydgcx32t4nhl424	7	12	ness	ness	NN
work_27d5tr326feydgcx32t4nhl424	7	13	of	of	IN
work_27d5tr326feydgcx32t4nhl424	7	14	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	7	15	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	7	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	8	1	1	1	CD
work_27d5tr326feydgcx32t4nhl424	8	2	Introduction	introduction	NN
work_27d5tr326feydgcx32t4nhl424	8	3	Query	Query	NNP
work_27d5tr326feydgcx32t4nhl424	8	4	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	8	5	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	8	6	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	8	7	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	8	8	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	8	9	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	8	10	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	8	11	Nenkova	Nenkova	NNP
work_27d5tr326feydgcx32t4nhl424	8	12	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	8	13	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	8	14	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	8	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	8	16	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	8	17	;	;	:
work_27d5tr326feydgcx32t4nhl424	8	18	Wan	Wan	NNP
work_27d5tr326feydgcx32t4nhl424	8	19	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	8	20	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	8	21	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	8	22	,	,	,
work_27d5tr326feydgcx32t4nhl424	8	23	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	8	24	;	;	:
work_27d5tr326feydgcx32t4nhl424	8	25	Ouyang	Ouyang	NNP
work_27d5tr326feydgcx32t4nhl424	8	26	et	et	FW
work_27d5tr326feydgcx32t4nhl424	8	27	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	8	28	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	8	29	,	,	,
work_27d5tr326feydgcx32t4nhl424	8	30	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	8	31	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	8	32	can	can	MD
work_27d5tr326feydgcx32t4nhl424	8	33	facilitate	facilitate	VB
work_27d5tr326feydgcx32t4nhl424	8	34	users	user	NNS
work_27d5tr326feydgcx32t4nhl424	8	35	to	to	TO
work_27d5tr326feydgcx32t4nhl424	8	36	grasp	grasp	VB
work_27d5tr326feydgcx32t4nhl424	8	37	the	the	DT
work_27d5tr326feydgcx32t4nhl424	8	38	main	main	JJ
work_27d5tr326feydgcx32t4nhl424	8	39	idea	idea	NN
work_27d5tr326feydgcx32t4nhl424	8	40	of	of	IN
work_27d5tr326feydgcx32t4nhl424	8	41	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	8	42	.	.	.
work_27d5tr326feydgcx32t4nhl424	9	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	9	2	query	query	NN
work_27d5tr326feydgcx32t4nhl424	9	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	9	4	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	9	5	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	9	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	9	7	a	a	DT
work_27d5tr326feydgcx32t4nhl424	9	8	specific	specific	JJ
work_27d5tr326feydgcx32t4nhl424	9	9	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	9	10	description	description	NN
work_27d5tr326feydgcx32t4nhl424	9	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	9	12	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	9	13	as	as	IN
work_27d5tr326feydgcx32t4nhl424	9	14	a	a	DT
work_27d5tr326feydgcx32t4nhl424	9	15	query	query	NN
work_27d5tr326feydgcx32t4nhl424	9	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	9	17	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	9	18	expresses	express	VBZ
work_27d5tr326feydgcx32t4nhl424	9	19	the	the	DT
work_27d5tr326feydgcx32t4nhl424	9	20	most	most	RBS
work_27d5tr326feydgcx32t4nhl424	9	21	important	important	JJ
work_27d5tr326feydgcx32t4nhl424	9	22	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	9	23	information	information	NN
work_27d5tr326feydgcx32t4nhl424	9	24	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	9	25	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	9	26	before	before	IN
work_27d5tr326feydgcx32t4nhl424	9	27	the	the	DT
work_27d5tr326feydgcx32t4nhl424	9	28	document	document	NN
work_27d5tr326feydgcx32t4nhl424	9	29	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	9	30	,	,	,
work_27d5tr326feydgcx32t4nhl424	9	31	and	and	CC
work_27d5tr326feydgcx32t4nhl424	9	32	a	a	DT
work_27d5tr326feydgcx32t4nhl424	9	33	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	9	34	would	would	MD
work_27d5tr326feydgcx32t4nhl424	9	35	be	be	VB
work_27d5tr326feydgcx32t4nhl424	9	36	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	9	37	according	accord	VBG
work_27d5tr326feydgcx32t4nhl424	9	38	to	to	IN
work_27d5tr326feydgcx32t4nhl424	9	39	the	the	DT
work_27d5tr326feydgcx32t4nhl424	9	40	given	give	VBN
work_27d5tr326feydgcx32t4nhl424	9	41	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	9	42	.	.	.
work_27d5tr326feydgcx32t4nhl424	10	1	Supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	10	2	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	10	3	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	10	4	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	10	5	widely	widely	RB
work_27d5tr326feydgcx32t4nhl424	10	6	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	10	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	10	8	sum-	sum-	JJ
work_27d5tr326feydgcx32t4nhl424	10	9	marization	marization	NN
work_27d5tr326feydgcx32t4nhl424	10	10	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	10	11	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	10	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	10	13	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	10	14	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	10	15	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	10	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	10	17	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	10	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	10	19	Shen	Shen	NNP
work_27d5tr326feydgcx32t4nhl424	10	20	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	10	21	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	10	22	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	10	23	,	,	,
work_27d5tr326feydgcx32t4nhl424	10	24	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	10	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	10	26	Ouyang	Ouyang	NNP
work_27d5tr326feydgcx32t4nhl424	10	27	et	et	FW
work_27d5tr326feydgcx32t4nhl424	10	28	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	10	29	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	10	30	,	,	,
work_27d5tr326feydgcx32t4nhl424	10	31	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	10	32	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	10	33	.	.	.
work_27d5tr326feydgcx32t4nhl424	11	1	Supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	11	2	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	11	3	usually	usually	RB
work_27d5tr326feydgcx32t4nhl424	11	4	re-	re-	RB
work_27d5tr326feydgcx32t4nhl424	11	5	gard	gard	JJ
work_27d5tr326feydgcx32t4nhl424	11	6	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	11	7	as	as	IN
work_27d5tr326feydgcx32t4nhl424	11	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	11	9	classification	classification	NN
work_27d5tr326feydgcx32t4nhl424	11	10	or	or	CC
work_27d5tr326feydgcx32t4nhl424	11	11	regression	regression	NN
work_27d5tr326feydgcx32t4nhl424	11	12	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	11	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	11	14	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	11	15	various	various	JJ
work_27d5tr326feydgcx32t4nhl424	11	16	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	11	17	features	feature	VBZ
work_27d5tr326feydgcx32t4nhl424	11	18	to	to	TO
work_27d5tr326feydgcx32t4nhl424	11	19	build	build	VB
work_27d5tr326feydgcx32t4nhl424	11	20	a	a	DT
work_27d5tr326feydgcx32t4nhl424	11	21	classifier	classifier	NN
work_27d5tr326feydgcx32t4nhl424	11	22	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	11	23	on	on	IN
work_27d5tr326feydgcx32t4nhl424	11	24	labeled	label	VBN
work_27d5tr326feydgcx32t4nhl424	11	25	negative	negative	JJ
work_27d5tr326feydgcx32t4nhl424	11	26	or	or	CC
work_27d5tr326feydgcx32t4nhl424	11	27	positive	positive	JJ
work_27d5tr326feydgcx32t4nhl424	11	28	sam-	sam-	JJ
work_27d5tr326feydgcx32t4nhl424	11	29	ples	ple	NNS
work_27d5tr326feydgcx32t4nhl424	11	30	.	.	.
work_27d5tr326feydgcx32t4nhl424	12	1	However	however	RB
work_27d5tr326feydgcx32t4nhl424	12	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	12	3	existing	exist	VBG
work_27d5tr326feydgcx32t4nhl424	12	4	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	12	5	approaches	approaches	NNP
work_27d5tr326feydgcx32t4nhl424	12	6	sel-	sel-	NNP
work_27d5tr326feydgcx32t4nhl424	12	7	dom	dom	NNP
work_27d5tr326feydgcx32t4nhl424	12	8	exploit	exploit	VB
work_27d5tr326feydgcx32t4nhl424	12	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	12	10	intrinsic	intrinsic	JJ
work_27d5tr326feydgcx32t4nhl424	12	11	structure	structure	NN
work_27d5tr326feydgcx32t4nhl424	12	12	among	among	IN
work_27d5tr326feydgcx32t4nhl424	12	13	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	12	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	13	1	This	this	DT
work_27d5tr326feydgcx32t4nhl424	13	2	disadvantage	disadvantage	NN
work_27d5tr326feydgcx32t4nhl424	13	3	usually	usually	RB
work_27d5tr326feydgcx32t4nhl424	13	4	gives	give	VBZ
work_27d5tr326feydgcx32t4nhl424	13	5	rise	rise	NN
work_27d5tr326feydgcx32t4nhl424	13	6	to	to	IN
work_27d5tr326feydgcx32t4nhl424	13	7	serious	serious	JJ
work_27d5tr326feydgcx32t4nhl424	13	8	prob-	prob-	JJ
work_27d5tr326feydgcx32t4nhl424	13	9	lems	lem	NNS
work_27d5tr326feydgcx32t4nhl424	13	10	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	13	11	as	as	IN
work_27d5tr326feydgcx32t4nhl424	13	12	unbalance	unbalance	NN
work_27d5tr326feydgcx32t4nhl424	13	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	13	14	low	low	JJ
work_27d5tr326feydgcx32t4nhl424	13	15	recall	recall	NN
work_27d5tr326feydgcx32t4nhl424	13	16	in	in	IN
work_27d5tr326feydgcx32t4nhl424	13	17	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	13	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	14	1	Recently	recently	RB
work_27d5tr326feydgcx32t4nhl424	14	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	3	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	14	4	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	14	5	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	14	6	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	14	7	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	14	8	et	et	FW
work_27d5tr326feydgcx32t4nhl424	14	9	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	14	10	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	14	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	12	2003	2003	CD
work_27d5tr326feydgcx32t4nhl424	14	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	14	14	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	14	15	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	14	16	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	14	17	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	14	18	widely	widely	RB
work_27d5tr326feydgcx32t4nhl424	14	19	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	14	20	applied	apply	VBN
work_27d5tr326feydgcx32t4nhl424	14	21	in	in	IN
work_27d5tr326feydgcx32t4nhl424	14	22	multi-	multi-	FW
work_27d5tr326feydgcx32t4nhl424	14	23	document	document	NN
work_27d5tr326feydgcx32t4nhl424	14	24	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	14	25	in	in	IN
work_27d5tr326feydgcx32t4nhl424	14	26	that	that	DT
work_27d5tr326feydgcx32t4nhl424	14	27	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	14	28	ap-	ap-	NN
work_27d5tr326feydgcx32t4nhl424	14	29	proaches	proache	NNS
work_27d5tr326feydgcx32t4nhl424	14	30	can	can	MD
work_27d5tr326feydgcx32t4nhl424	14	31	offer	offer	VB
work_27d5tr326feydgcx32t4nhl424	14	32	clear	clear	JJ
work_27d5tr326feydgcx32t4nhl424	14	33	and	and	CC
work_27d5tr326feydgcx32t4nhl424	14	34	rigorous	rigorous	JJ
work_27d5tr326feydgcx32t4nhl424	14	35	probabilis-	probabilis-	NN
work_27d5tr326feydgcx32t4nhl424	14	36	tic	tic	JJ
work_27d5tr326feydgcx32t4nhl424	14	37	interpretations	interpretation	NNS
work_27d5tr326feydgcx32t4nhl424	14	38	for	for	IN
work_27d5tr326feydgcx32t4nhl424	14	39	summaries(Daume	summaries(daume	NN
work_27d5tr326feydgcx32t4nhl424	14	40	and	and	CC
work_27d5tr326feydgcx32t4nhl424	14	41	Marcu	Marcu	NNP
work_27d5tr326feydgcx32t4nhl424	14	42	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	43	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	14	44	;	;	:
work_27d5tr326feydgcx32t4nhl424	14	45	Haghighi	Haghighi	NNP
work_27d5tr326feydgcx32t4nhl424	14	46	and	and	CC
work_27d5tr326feydgcx32t4nhl424	14	47	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	14	48	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	49	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	14	50	;	;	:
work_27d5tr326feydgcx32t4nhl424	14	51	Jin	Jin	NNP
work_27d5tr326feydgcx32t4nhl424	14	52	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	14	53	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	14	54	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	14	55	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	56	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	14	57	;	;	:
work_27d5tr326feydgcx32t4nhl424	14	58	Mason	Mason	NNP
work_27d5tr326feydgcx32t4nhl424	14	59	and	and	CC
work_27d5tr326feydgcx32t4nhl424	14	60	Charniak	Charniak	NNP
work_27d5tr326feydgcx32t4nhl424	14	61	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	62	2011	2011	CD
work_27d5tr326feydgcx32t4nhl424	14	63	;	;	:
work_27d5tr326feydgcx32t4nhl424	14	64	Delort	Delort	NNP
work_27d5tr326feydgcx32t4nhl424	14	65	and	and	CC
work_27d5tr326feydgcx32t4nhl424	14	66	Alfon-	Alfon-	NNP
work_27d5tr326feydgcx32t4nhl424	14	67	seca	seca	NN
work_27d5tr326feydgcx32t4nhl424	14	68	,	,	,
work_27d5tr326feydgcx32t4nhl424	14	69	2012	2012	CD
work_27d5tr326feydgcx32t4nhl424	14	70	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	14	71	.	.	.
work_27d5tr326feydgcx32t4nhl424	15	1	Exiting	exit	VBG
work_27d5tr326feydgcx32t4nhl424	15	2	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	15	3	approaches	approach	VBZ
work_27d5tr326feydgcx32t4nhl424	15	4	label	label	NN
work_27d5tr326feydgcx32t4nhl424	15	5	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	15	6	tences	tence	NNS
work_27d5tr326feydgcx32t4nhl424	15	7	or	or	CC
work_27d5tr326feydgcx32t4nhl424	15	8	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	15	9	with	with	IN
work_27d5tr326feydgcx32t4nhl424	15	10	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	15	11	and	and	CC
work_27d5tr326feydgcx32t4nhl424	15	12	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	15	13	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	15	14	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	15	15	closely	closely	RB
work_27d5tr326feydgcx32t4nhl424	15	16	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	15	17	with	with	IN
work_27d5tr326feydgcx32t4nhl424	15	18	query	query	NN
work_27d5tr326feydgcx32t4nhl424	15	19	or	or	CC
work_27d5tr326feydgcx32t4nhl424	15	20	can	can	MD
work_27d5tr326feydgcx32t4nhl424	15	21	highly	highly	RB
work_27d5tr326feydgcx32t4nhl424	15	22	generalize	generalize	VB
work_27d5tr326feydgcx32t4nhl424	15	23	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	15	24	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	15	25	selected	select	VBN
work_27d5tr326feydgcx32t4nhl424	15	26	into	into	IN
work_27d5tr326feydgcx32t4nhl424	15	27	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	15	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	16	1	However	however	RB
work_27d5tr326feydgcx32t4nhl424	16	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	16	3	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	16	4	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	16	5	model	model	NN
work_27d5tr326feydgcx32t4nhl424	16	6	suffers	suffer	VBZ
work_27d5tr326feydgcx32t4nhl424	16	7	from	from	IN
work_27d5tr326feydgcx32t4nhl424	16	8	the	the	DT
work_27d5tr326feydgcx32t4nhl424	16	9	intrinsic	intrinsic	JJ
work_27d5tr326feydgcx32t4nhl424	16	10	disad-	disad-	JJ
work_27d5tr326feydgcx32t4nhl424	16	11	vantages	vantage	NNS
work_27d5tr326feydgcx32t4nhl424	16	12	that	that	IN
work_27d5tr326feydgcx32t4nhl424	16	13	it	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	16	14	only	only	RB
work_27d5tr326feydgcx32t4nhl424	16	15	uses	use	VBZ
work_27d5tr326feydgcx32t4nhl424	16	16	word	word	NN
work_27d5tr326feydgcx32t4nhl424	16	17	frequency	frequency	NN
work_27d5tr326feydgcx32t4nhl424	16	18	for	for	IN
work_27d5tr326feydgcx32t4nhl424	16	19	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	16	20	modeling	modeling	NN
work_27d5tr326feydgcx32t4nhl424	16	21	and	and	CC
work_27d5tr326feydgcx32t4nhl424	16	22	can	can	MD
work_27d5tr326feydgcx32t4nhl424	16	23	not	not	RB
work_27d5tr326feydgcx32t4nhl424	16	24	use	use	VB
work_27d5tr326feydgcx32t4nhl424	16	25	useful	useful	JJ
work_27d5tr326feydgcx32t4nhl424	16	26	text	text	NN
work_27d5tr326feydgcx32t4nhl424	16	27	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	16	28	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	16	29	as	as	IN
work_27d5tr326feydgcx32t4nhl424	16	30	position	position	NN
work_27d5tr326feydgcx32t4nhl424	16	31	,	,	,
work_27d5tr326feydgcx32t4nhl424	16	32	word	word	NN
work_27d5tr326feydgcx32t4nhl424	16	33	order	order	NN
work_27d5tr326feydgcx32t4nhl424	16	34	etc	etc	FW
work_27d5tr326feydgcx32t4nhl424	16	35	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	16	36	Zhu	Zhu	NNP
work_27d5tr326feydgcx32t4nhl424	16	37	and	and	CC
work_27d5tr326feydgcx32t4nhl424	16	38	Xing	Xing	NNP
work_27d5tr326feydgcx32t4nhl424	16	39	,	,	,
work_27d5tr326feydgcx32t4nhl424	16	40	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	16	41	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	16	42	.	.	.
work_27d5tr326feydgcx32t4nhl424	17	1	For	for	IN
work_27d5tr326feydgcx32t4nhl424	17	2	example	example	NN
work_27d5tr326feydgcx32t4nhl424	17	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	17	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	17	5	first	first	JJ
work_27d5tr326feydgcx32t4nhl424	17	6	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	17	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	17	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	17	9	document	document	NN
work_27d5tr326feydgcx32t4nhl424	17	10	may	may	MD
work_27d5tr326feydgcx32t4nhl424	17	11	be	be	VB
work_27d5tr326feydgcx32t4nhl424	17	12	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	17	13	important	important	JJ
work_27d5tr326feydgcx32t4nhl424	17	14	for	for	IN
work_27d5tr326feydgcx32t4nhl424	17	15	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	17	16	since	since	IN
work_27d5tr326feydgcx32t4nhl424	17	17	it	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	17	18	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	17	19	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	17	20	likely	likely	JJ
work_27d5tr326feydgcx32t4nhl424	17	21	to	to	TO
work_27d5tr326feydgcx32t4nhl424	17	22	give	give	VB
work_27d5tr326feydgcx32t4nhl424	17	23	a	a	DT
work_27d5tr326feydgcx32t4nhl424	17	24	global	global	JJ
work_27d5tr326feydgcx32t4nhl424	17	25	generalization	generalization	NN
work_27d5tr326feydgcx32t4nhl424	17	26	about	about	IN
work_27d5tr326feydgcx32t4nhl424	17	27	the	the	DT
work_27d5tr326feydgcx32t4nhl424	17	28	document	document	NN
work_27d5tr326feydgcx32t4nhl424	17	29	.	.	.
work_27d5tr326feydgcx32t4nhl424	18	1	It	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	18	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	18	3	hard	hard	JJ
work_27d5tr326feydgcx32t4nhl424	18	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	18	5	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	18	6	model	model	NN
work_27d5tr326feydgcx32t4nhl424	18	7	to	to	TO
work_27d5tr326feydgcx32t4nhl424	18	8	consider	consider	VB
work_27d5tr326feydgcx32t4nhl424	18	9	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	18	10	informa-	informa-	JJ
work_27d5tr326feydgcx32t4nhl424	18	11	tion	tion	NN
work_27d5tr326feydgcx32t4nhl424	18	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	18	13	making	make	VBG
work_27d5tr326feydgcx32t4nhl424	18	14	useful	useful	JJ
work_27d5tr326feydgcx32t4nhl424	18	15	information	information	NN
work_27d5tr326feydgcx32t4nhl424	18	16	lost	lose	VBN
work_27d5tr326feydgcx32t4nhl424	18	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	19	1	It	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	19	2	naturally	naturally	RB
work_27d5tr326feydgcx32t4nhl424	19	3	comes	come	VBZ
work_27d5tr326feydgcx32t4nhl424	19	4	to	to	IN
work_27d5tr326feydgcx32t4nhl424	19	5	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	19	6	minds	mind	NNS
work_27d5tr326feydgcx32t4nhl424	19	7	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	19	8	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	19	9	can	can	MD
work_27d5tr326feydgcx32t4nhl424	19	10	im-	im-	RB
work_27d5tr326feydgcx32t4nhl424	19	11	prove	prove	VB
work_27d5tr326feydgcx32t4nhl424	19	12	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	19	13	performance	performance	NN
work_27d5tr326feydgcx32t4nhl424	19	14	by	by	IN
work_27d5tr326feydgcx32t4nhl424	19	15	making	make	VBG
work_27d5tr326feydgcx32t4nhl424	19	16	full	full	JJ
work_27d5tr326feydgcx32t4nhl424	19	17	use	use	NN
work_27d5tr326feydgcx32t4nhl424	19	18	of	of	IN
work_27d5tr326feydgcx32t4nhl424	19	19	both	both	CC
work_27d5tr326feydgcx32t4nhl424	19	20	useful	useful	JJ
work_27d5tr326feydgcx32t4nhl424	19	21	text	text	NN
work_27d5tr326feydgcx32t4nhl424	19	22	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	19	23	and	and	CC
work_27d5tr326feydgcx32t4nhl424	19	24	the	the	DT
work_27d5tr326feydgcx32t4nhl424	19	25	latent	latent	NN
work_27d5tr326feydgcx32t4nhl424	19	26	seman-	seman-	NNP
work_27d5tr326feydgcx32t4nhl424	19	27	tic	tic	JJ
work_27d5tr326feydgcx32t4nhl424	19	28	structures	structure	NNS
work_27d5tr326feydgcx32t4nhl424	19	29	from	from	IN
work_27d5tr326feydgcx32t4nhl424	19	30	by	by	IN
work_27d5tr326feydgcx32t4nhl424	19	31	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	19	32	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	19	33	model	model	NN
work_27d5tr326feydgcx32t4nhl424	19	34	.	.	.
work_27d5tr326feydgcx32t4nhl424	20	1	One	one	CD
work_27d5tr326feydgcx32t4nhl424	20	2	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	20	3	work	work	NN
work_27d5tr326feydgcx32t4nhl424	20	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	20	5	from	from	IN
work_27d5tr326feydgcx32t4nhl424	20	6	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	20	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	20	8	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	20	9	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	20	10	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	20	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	20	12	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	20	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	20	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	21	1	They	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	21	2	built	build	VBD
work_27d5tr326feydgcx32t4nhl424	21	3	a	a	DT
work_27d5tr326feydgcx32t4nhl424	21	4	hierarchical	hierarchical	JJ
work_27d5tr326feydgcx32t4nhl424	21	5	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	21	6	model	model	NN
work_27d5tr326feydgcx32t4nhl424	21	7	called	call	VBD
work_27d5tr326feydgcx32t4nhl424	21	8	Hybh-	Hybh-	NNP
work_27d5tr326feydgcx32t4nhl424	21	9	sum	sum	NN
work_27d5tr326feydgcx32t4nhl424	21	10	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	21	11	on	on	IN
work_27d5tr326feydgcx32t4nhl424	21	12	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	21	13	for	for	IN
work_27d5tr326feydgcx32t4nhl424	21	14	topic	topic	JJ
work_27d5tr326feydgcx32t4nhl424	21	15	discovery	discovery	NN
work_27d5tr326feydgcx32t4nhl424	21	16	and	and	CC
work_27d5tr326feydgcx32t4nhl424	21	17	assumed	assume	VBD
work_27d5tr326feydgcx32t4nhl424	21	18	this	this	DT
work_27d5tr326feydgcx32t4nhl424	21	19	model	model	NN
work_27d5tr326feydgcx32t4nhl424	21	20	can	can	MD
work_27d5tr326feydgcx32t4nhl424	21	21	produce	produce	VB
work_27d5tr326feydgcx32t4nhl424	21	22	appropriate	appropriate	JJ
work_27d5tr326feydgcx32t4nhl424	21	23	scores	score	NNS
work_27d5tr326feydgcx32t4nhl424	21	24	for	for	IN
work_27d5tr326feydgcx32t4nhl424	21	25	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	21	26	tence	tence	NN
work_27d5tr326feydgcx32t4nhl424	21	27	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	21	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	22	1	Then	then	RB
work_27d5tr326feydgcx32t4nhl424	22	2	the	the	DT
work_27d5tr326feydgcx32t4nhl424	22	3	scores	score	NNS
work_27d5tr326feydgcx32t4nhl424	22	4	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	22	5	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	22	6	for	for	IN
work_27d5tr326feydgcx32t4nhl424	22	7	tun-	tun-	DT
work_27d5tr326feydgcx32t4nhl424	22	8	ing	e	VBG
work_27d5tr326feydgcx32t4nhl424	22	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	22	10	weights	weight	NNS
work_27d5tr326feydgcx32t4nhl424	22	11	of	of	IN
work_27d5tr326feydgcx32t4nhl424	22	12	various	various	JJ
work_27d5tr326feydgcx32t4nhl424	22	13	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	22	14	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	22	15	helpful	helpful	JJ
work_27d5tr326feydgcx32t4nhl424	22	16	for	for	IN
work_27d5tr326feydgcx32t4nhl424	22	17	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	22	18	generation	generation	NN
work_27d5tr326feydgcx32t4nhl424	22	19	.	.	.
work_27d5tr326feydgcx32t4nhl424	23	1	Their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	23	2	work	work	NN
work_27d5tr326feydgcx32t4nhl424	23	3	made	make	VBD
work_27d5tr326feydgcx32t4nhl424	23	4	a	a	DT
work_27d5tr326feydgcx32t4nhl424	23	5	good	good	JJ
work_27d5tr326feydgcx32t4nhl424	23	6	step	step	NN
work_27d5tr326feydgcx32t4nhl424	23	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	23	8	combining	combine	VBG
work_27d5tr326feydgcx32t4nhl424	23	9	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	23	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	23	11	with	with	IN
work_27d5tr326feydgcx32t4nhl424	23	12	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	23	13	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	23	14	super-	super-	NN
work_27d5tr326feydgcx32t4nhl424	23	15	vised	vise	VBN
work_27d5tr326feydgcx32t4nhl424	23	16	learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	23	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	24	1	However	however	RB
work_27d5tr326feydgcx32t4nhl424	24	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	24	3	what	what	WP
work_27d5tr326feydgcx32t4nhl424	24	4	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	24	5	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	24	6	con-	con-	NN
work_27d5tr326feydgcx32t4nhl424	24	7	fuses	fuse	VBZ
work_27d5tr326feydgcx32t4nhl424	24	8	us	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	24	9	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	24	10	that	that	IN
work_27d5tr326feydgcx32t4nhl424	24	11	whether	whether	IN
work_27d5tr326feydgcx32t4nhl424	24	12	a	a	DT
work_27d5tr326feydgcx32t4nhl424	24	13	topic	topic	JJ
work_27d5tr326feydgcx32t4nhl424	24	14	model	model	NN
work_27d5tr326feydgcx32t4nhl424	24	15	only	only	RB
work_27d5tr326feydgcx32t4nhl424	24	16	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	24	17	on	on	IN
work_27d5tr326feydgcx32t4nhl424	24	18	word	word	NN
work_27d5tr326feydgcx32t4nhl424	24	19	frequency	frequency	NN
work_27d5tr326feydgcx32t4nhl424	24	20	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	24	21	good	good	JJ
work_27d5tr326feydgcx32t4nhl424	24	22	enough	enough	RB
work_27d5tr326feydgcx32t4nhl424	24	23	to	to	TO
work_27d5tr326feydgcx32t4nhl424	24	24	generate	generate	VB
work_27d5tr326feydgcx32t4nhl424	24	25	an	an	DT
work_27d5tr326feydgcx32t4nhl424	24	26	appropriate	appropriate	JJ
work_27d5tr326feydgcx32t4nhl424	24	27	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	24	28	score	score	NN
work_27d5tr326feydgcx32t4nhl424	24	29	for	for	IN
work_27d5tr326feydgcx32t4nhl424	24	30	regression	regression	NN
work_27d5tr326feydgcx32t4nhl424	24	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	25	1	Actually	actually	RB
work_27d5tr326feydgcx32t4nhl424	25	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	25	3	how	how	WRB
work_27d5tr326feydgcx32t4nhl424	25	4	to	to	TO
work_27d5tr326feydgcx32t4nhl424	25	5	incorporate	incorporate	VB
work_27d5tr326feydgcx32t4nhl424	25	6	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	25	7	into	into	IN
work_27d5tr326feydgcx32t4nhl424	25	8	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	25	9	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	25	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	25	11	has	have	VBZ
work_27d5tr326feydgcx32t4nhl424	25	12	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	25	13	a	a	DT
work_27d5tr326feydgcx32t4nhl424	25	14	open	open	JJ
work_27d5tr326feydgcx32t4nhl424	25	15	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	25	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	26	1	Supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	26	2	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	26	3	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	26	4	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	26	5	as	as	IN
work_27d5tr326feydgcx32t4nhl424	26	6	sLDA(Blei	slda(blei	NN
work_27d5tr326feydgcx32t4nhl424	26	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	26	8	MacAuliffe	MacAuliffe	NNP
work_27d5tr326feydgcx32t4nhl424	26	9	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	26	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	26	11	give	give	VBP
work_27d5tr326feydgcx32t4nhl424	26	12	us	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	26	13	some	some	DT
work_27d5tr326feydgcx32t4nhl424	26	14	inspiration	inspiration	NN
work_27d5tr326feydgcx32t4nhl424	26	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	27	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	27	2	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	27	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	27	4	each	each	DT
work_27d5tr326feydgcx32t4nhl424	27	5	document	document	NN
work_27d5tr326feydgcx32t4nhl424	27	6	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	27	7	asso-	asso-	RB
work_27d5tr326feydgcx32t4nhl424	27	8	ciated	ciate	VBN
work_27d5tr326feydgcx32t4nhl424	27	9	with	with	IN
work_27d5tr326feydgcx32t4nhl424	27	10	a	a	DT
work_27d5tr326feydgcx32t4nhl424	27	11	labeled	label	VBN
work_27d5tr326feydgcx32t4nhl424	27	12	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	27	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	27	14	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	27	15	can	can	MD
work_27d5tr326feydgcx32t4nhl424	27	16	integrate	integrate	VB
work_27d5tr326feydgcx32t4nhl424	27	17	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	27	18	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	27	19	into	into	IN
work_27d5tr326feydgcx32t4nhl424	27	20	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	27	21	for	for	IN
work_27d5tr326feydgcx32t4nhl424	27	22	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	27	23	modeling	modeling	NN
work_27d5tr326feydgcx32t4nhl424	27	24	in	in	IN
work_27d5tr326feydgcx32t4nhl424	27	25	a	a	DT
work_27d5tr326feydgcx32t4nhl424	27	26	prin-	prin-	RB
work_27d5tr326feydgcx32t4nhl424	27	27	89	89	CD
work_27d5tr326feydgcx32t4nhl424	27	28	cipled	cipled	JJ
work_27d5tr326feydgcx32t4nhl424	27	29	way	way	NN
work_27d5tr326feydgcx32t4nhl424	27	30	.	.	.
work_27d5tr326feydgcx32t4nhl424	28	1	With	with	IN
work_27d5tr326feydgcx32t4nhl424	28	2	reference	reference	NN
work_27d5tr326feydgcx32t4nhl424	28	3	to	to	IN
work_27d5tr326feydgcx32t4nhl424	28	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	28	5	work	work	NN
work_27d5tr326feydgcx32t4nhl424	28	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	28	7	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	28	8	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	28	9	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	28	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	28	11	in	in	IN
work_27d5tr326feydgcx32t4nhl424	28	12	this	this	DT
work_27d5tr326feydgcx32t4nhl424	28	13	paper	paper	NN
work_27d5tr326feydgcx32t4nhl424	28	14	,	,	,
work_27d5tr326feydgcx32t4nhl424	28	15	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	28	16	propose	propose	VBP
work_27d5tr326feydgcx32t4nhl424	28	17	a	a	DT
work_27d5tr326feydgcx32t4nhl424	28	18	novel	novel	JJ
work_27d5tr326feydgcx32t4nhl424	28	19	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	28	20	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	28	21	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	28	22	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	28	23	model	model	NN
work_27d5tr326feydgcx32t4nhl424	28	24	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	28	25	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	28	26	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	28	27	for	for	IN
work_27d5tr326feydgcx32t4nhl424	28	28	multi-	multi-	FW
work_27d5tr326feydgcx32t4nhl424	28	29	document	document	NN
work_27d5tr326feydgcx32t4nhl424	28	30	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	28	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	29	1	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	29	2	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	29	3	can	can	MD
work_27d5tr326feydgcx32t4nhl424	29	4	natu-	natu-	VBG
work_27d5tr326feydgcx32t4nhl424	29	5	rally	rally	NN
work_27d5tr326feydgcx32t4nhl424	29	6	combine	combine	VB
work_27d5tr326feydgcx32t4nhl424	29	7	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	29	8	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	29	9	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	29	10	methods	method	NNS
work_27d5tr326feydgcx32t4nhl424	29	11	and	and	CC
work_27d5tr326feydgcx32t4nhl424	29	12	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	29	13	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	29	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	30	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	30	2	most	most	RBS
work_27d5tr326feydgcx32t4nhl424	30	3	important	important	JJ
work_27d5tr326feydgcx32t4nhl424	30	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	30	5	challeng-	challeng-	JJ
work_27d5tr326feydgcx32t4nhl424	30	6	ing	ing	NNP
work_27d5tr326feydgcx32t4nhl424	30	7	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	30	8	in	in	IN
work_27d5tr326feydgcx32t4nhl424	30	9	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	30	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	30	11	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	30	12	the	the	DT
work_27d5tr326feydgcx32t4nhl424	30	13	tuning	tuning	NN
work_27d5tr326feydgcx32t4nhl424	30	14	of	of	IN
work_27d5tr326feydgcx32t4nhl424	30	15	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	30	16	weights	weight	NNS
work_27d5tr326feydgcx32t4nhl424	30	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	31	1	To	to	TO
work_27d5tr326feydgcx32t4nhl424	31	2	solve	solve	VB
work_27d5tr326feydgcx32t4nhl424	31	3	this	this	DT
work_27d5tr326feydgcx32t4nhl424	31	4	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	31	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	31	6	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	31	7	transform	transform	VBP
work_27d5tr326feydgcx32t4nhl424	31	8	the	the	DT
work_27d5tr326feydgcx32t4nhl424	31	9	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	31	10	of	of	IN
work_27d5tr326feydgcx32t4nhl424	31	11	finding	find	VBG
work_27d5tr326feydgcx32t4nhl424	31	12	optimum	optimum	JJ
work_27d5tr326feydgcx32t4nhl424	31	13	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	31	14	weights	weight	NNS
work_27d5tr326feydgcx32t4nhl424	31	15	into	into	IN
work_27d5tr326feydgcx32t4nhl424	31	16	an	an	DT
work_27d5tr326feydgcx32t4nhl424	31	17	optimization	optimization	NN
work_27d5tr326feydgcx32t4nhl424	31	18	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	31	19	and	and	CC
work_27d5tr326feydgcx32t4nhl424	31	20	learn	learn	VB
work_27d5tr326feydgcx32t4nhl424	31	21	these	these	DT
work_27d5tr326feydgcx32t4nhl424	31	22	weights	weight	NNS
work_27d5tr326feydgcx32t4nhl424	31	23	in	in	IN
work_27d5tr326feydgcx32t4nhl424	31	24	a	a	DT
work_27d5tr326feydgcx32t4nhl424	31	25	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	31	26	way	way	NN
work_27d5tr326feydgcx32t4nhl424	31	27	.	.	.
work_27d5tr326feydgcx32t4nhl424	32	1	A	a	DT
work_27d5tr326feydgcx32t4nhl424	32	2	set	set	NN
work_27d5tr326feydgcx32t4nhl424	32	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	32	4	experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	32	5	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	32	6	con-	con-	NN
work_27d5tr326feydgcx32t4nhl424	32	7	ducted	ducte	VBD
work_27d5tr326feydgcx32t4nhl424	32	8	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	32	9	on	on	IN
work_27d5tr326feydgcx32t4nhl424	32	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	32	11	benchmark	benchmark	JJ
work_27d5tr326feydgcx32t4nhl424	32	12	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	32	13	of	of	IN
work_27d5tr326feydgcx32t4nhl424	32	14	DUC2007	DUC2007	NNP
work_27d5tr326feydgcx32t4nhl424	32	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	32	16	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	32	17	and	and	CC
work_27d5tr326feydgcx32t4nhl424	32	18	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	32	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	32	20	and	and	CC
work_27d5tr326feydgcx32t4nhl424	32	21	experimental	experimental	JJ
work_27d5tr326feydgcx32t4nhl424	32	22	results	result	NNS
work_27d5tr326feydgcx32t4nhl424	32	23	show	show	VBP
work_27d5tr326feydgcx32t4nhl424	32	24	the	the	DT
work_27d5tr326feydgcx32t4nhl424	32	25	effectiveness	effectiveness	NN
work_27d5tr326feydgcx32t4nhl424	32	26	of	of	IN
work_27d5tr326feydgcx32t4nhl424	32	27	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	32	28	model	model	NN
work_27d5tr326feydgcx32t4nhl424	32	29	.	.	.
work_27d5tr326feydgcx32t4nhl424	33	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	33	2	rest	rest	NN
work_27d5tr326feydgcx32t4nhl424	33	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	33	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	33	5	paper	paper	NN
work_27d5tr326feydgcx32t4nhl424	33	6	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	33	7	organized	organize	VBN
work_27d5tr326feydgcx32t4nhl424	33	8	as	as	IN
work_27d5tr326feydgcx32t4nhl424	33	9	follows	follow	VBZ
work_27d5tr326feydgcx32t4nhl424	33	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	34	1	Sec-	Sec-	NNP
work_27d5tr326feydgcx32t4nhl424	34	2	tion	tion	NN
work_27d5tr326feydgcx32t4nhl424	34	3	2	2	CD
work_27d5tr326feydgcx32t4nhl424	34	4	describes	describe	VBZ
work_27d5tr326feydgcx32t4nhl424	34	5	some	some	DT
work_27d5tr326feydgcx32t4nhl424	34	6	background	background	NN
work_27d5tr326feydgcx32t4nhl424	34	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	34	8	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	34	9	works	work	NNS
work_27d5tr326feydgcx32t4nhl424	34	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	35	1	Section	section	NN
work_27d5tr326feydgcx32t4nhl424	35	2	3	3	CD
work_27d5tr326feydgcx32t4nhl424	35	3	describes	describe	VBZ
work_27d5tr326feydgcx32t4nhl424	35	4	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	35	5	details	detail	NNS
work_27d5tr326feydgcx32t4nhl424	35	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	35	7	S	s	NN
work_27d5tr326feydgcx32t4nhl424	35	8	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	35	9	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	35	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	35	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	36	1	Section	section	NN
work_27d5tr326feydgcx32t4nhl424	36	2	4	4	CD
work_27d5tr326feydgcx32t4nhl424	36	3	demonstrates	demonstrate	VBZ
work_27d5tr326feydgcx32t4nhl424	36	4	details	detail	NNS
work_27d5tr326feydgcx32t4nhl424	36	5	of	of	IN
work_27d5tr326feydgcx32t4nhl424	36	6	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	36	7	approaches	approach	NNS
work_27d5tr326feydgcx32t4nhl424	36	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	36	9	including	include	VBG
work_27d5tr326feydgcx32t4nhl424	36	10	learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	36	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	36	12	inference	inference	NN
work_27d5tr326feydgcx32t4nhl424	36	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	36	14	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	36	15	gener-	gener-	JJ
work_27d5tr326feydgcx32t4nhl424	36	16	ation	ation	NN
work_27d5tr326feydgcx32t4nhl424	36	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	37	1	Section	section	NN
work_27d5tr326feydgcx32t4nhl424	37	2	5	5	CD
work_27d5tr326feydgcx32t4nhl424	37	3	provides	provide	VBZ
work_27d5tr326feydgcx32t4nhl424	37	4	experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	37	5	results	result	NNS
work_27d5tr326feydgcx32t4nhl424	37	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	37	7	Section	section	NN
work_27d5tr326feydgcx32t4nhl424	37	8	6	6	CD
work_27d5tr326feydgcx32t4nhl424	37	9	concludes	conclude	VBZ
work_27d5tr326feydgcx32t4nhl424	37	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	37	11	paper	paper	NN
work_27d5tr326feydgcx32t4nhl424	37	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	38	1	2	2	CD
work_27d5tr326feydgcx32t4nhl424	38	2	Related	related	JJ
work_27d5tr326feydgcx32t4nhl424	38	3	Work	work	NN
work_27d5tr326feydgcx32t4nhl424	38	4	A	a	DT
work_27d5tr326feydgcx32t4nhl424	38	5	variety	variety	NN
work_27d5tr326feydgcx32t4nhl424	38	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	38	7	approaches	approach	NNS
work_27d5tr326feydgcx32t4nhl424	38	8	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	38	9	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	38	10	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	38	11	for	for	IN
work_27d5tr326feydgcx32t4nhl424	38	12	query	query	NN
work_27d5tr326feydgcx32t4nhl424	38	13	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	38	14	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	38	15	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	38	16	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	38	17	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	38	18	summarizations	summarization	NNS
work_27d5tr326feydgcx32t4nhl424	38	19	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	38	20	as	as	IN
work_27d5tr326feydgcx32t4nhl424	38	21	unsupervised	unsupervise	VBN
work_27d5tr326feydgcx32t4nhl424	38	22	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	38	23	semi	semi	RB
work_27d5tr326feydgcx32t4nhl424	38	24	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	38	25	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	38	26	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	38	27	approaches	approach	NNS
work_27d5tr326feydgcx32t4nhl424	38	28	,	,	,
work_27d5tr326feydgcx32t4nhl424	38	29	supervised	supervise	VBD
work_27d5tr326feydgcx32t4nhl424	38	30	approaches	approach	NNS
work_27d5tr326feydgcx32t4nhl424	38	31	,	,	,
work_27d5tr326feydgcx32t4nhl424	38	32	and	and	CC
work_27d5tr326feydgcx32t4nhl424	38	33	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	38	34	approaches	approach	NNS
work_27d5tr326feydgcx32t4nhl424	38	35	.	.	.
work_27d5tr326feydgcx32t4nhl424	39	1	Unsupervised	unsupervise	VBN
work_27d5tr326feydgcx32t4nhl424	39	2	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	39	3	semi	semi	RB
work_27d5tr326feydgcx32t4nhl424	39	4	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	39	5	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	39	6	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	39	7	approaches	approach	NNS
work_27d5tr326feydgcx32t4nhl424	39	8	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	39	9	as	as	IN
work_27d5tr326feydgcx32t4nhl424	39	10	Lexrank	Lexrank	NNP
work_27d5tr326feydgcx32t4nhl424	39	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	39	12	Erkan	Erkan	NNP
work_27d5tr326feydgcx32t4nhl424	39	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	39	14	Radex	Radex	NNP
work_27d5tr326feydgcx32t4nhl424	39	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	39	16	2004	2004	CD
work_27d5tr326feydgcx32t4nhl424	39	17	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	39	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	39	19	manifold	manifold	NNP
work_27d5tr326feydgcx32t4nhl424	39	20	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	39	21	Wan	Wan	NNP
work_27d5tr326feydgcx32t4nhl424	39	22	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	39	23	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	39	24	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	39	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	39	26	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	39	27	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	39	28	treat	treat	VB
work_27d5tr326feydgcx32t4nhl424	39	29	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	39	30	as	as	IN
work_27d5tr326feydgcx32t4nhl424	39	31	a	a	DT
work_27d5tr326feydgcx32t4nhl424	39	32	graph-	graph-	NN
work_27d5tr326feydgcx32t4nhl424	39	33	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	39	34	ranking	rank	VBG
work_27d5tr326feydgcx32t4nhl424	39	35	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	39	36	.	.	.
work_27d5tr326feydgcx32t4nhl424	40	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	40	2	relatedness	relatedness	NN
work_27d5tr326feydgcx32t4nhl424	40	3	between	between	IN
work_27d5tr326feydgcx32t4nhl424	40	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	40	5	query	query	NN
work_27d5tr326feydgcx32t4nhl424	40	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	40	7	each	each	DT
work_27d5tr326feydgcx32t4nhl424	40	8	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	40	9	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	40	10	achieved	achieve	VBN
work_27d5tr326feydgcx32t4nhl424	40	11	by	by	IN
work_27d5tr326feydgcx32t4nhl424	40	12	impos-	impos-	JJ
work_27d5tr326feydgcx32t4nhl424	40	13	ing	ing	NNP
work_27d5tr326feydgcx32t4nhl424	40	14	querys	querys	NNP
work_27d5tr326feydgcx32t4nhl424	40	15	influence	influence	NN
work_27d5tr326feydgcx32t4nhl424	40	16	on	on	IN
work_27d5tr326feydgcx32t4nhl424	40	17	each	each	DT
work_27d5tr326feydgcx32t4nhl424	40	18	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	40	19	along	along	IN
work_27d5tr326feydgcx32t4nhl424	40	20	with	with	IN
work_27d5tr326feydgcx32t4nhl424	40	21	the	the	DT
work_27d5tr326feydgcx32t4nhl424	40	22	propagation	propagation	NN
work_27d5tr326feydgcx32t4nhl424	40	23	of	of	IN
work_27d5tr326feydgcx32t4nhl424	40	24	graph	graph	NN
work_27d5tr326feydgcx32t4nhl424	40	25	.	.	.
work_27d5tr326feydgcx32t4nhl424	41	1	Most	most	RBS
work_27d5tr326feydgcx32t4nhl424	41	2	supervised	supervise	VBD
work_27d5tr326feydgcx32t4nhl424	41	3	ap-	ap-	JJ
work_27d5tr326feydgcx32t4nhl424	41	4	proaches	proache	NNS
work_27d5tr326feydgcx32t4nhl424	41	5	regard	regard	NN
work_27d5tr326feydgcx32t4nhl424	41	6	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	41	7	task	task	NN
work_27d5tr326feydgcx32t4nhl424	41	8	as	as	IN
work_27d5tr326feydgcx32t4nhl424	41	9	a	a	DT
work_27d5tr326feydgcx32t4nhl424	41	10	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	41	11	level	level	NN
work_27d5tr326feydgcx32t4nhl424	41	12	two	two	CD
work_27d5tr326feydgcx32t4nhl424	41	13	class	class	NN
work_27d5tr326feydgcx32t4nhl424	41	14	classification	classification	NN
work_27d5tr326feydgcx32t4nhl424	41	15	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	41	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	42	1	Supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	42	2	machine	machine	NN
work_27d5tr326feydgcx32t4nhl424	42	3	learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	42	4	methods	method	NNS
work_27d5tr326feydgcx32t4nhl424	42	5	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	42	6	as	as	IN
work_27d5tr326feydgcx32t4nhl424	42	7	Support	Support	NNP
work_27d5tr326feydgcx32t4nhl424	42	8	Vector	Vector	NNP
work_27d5tr326feydgcx32t4nhl424	42	9	Machine(SVM	machine(svm	CD
work_27d5tr326feydgcx32t4nhl424	42	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	42	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	42	12	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	42	13	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	14	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	42	15	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	42	16	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	42	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	18	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	42	19	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	42	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	21	Maximum	Maximum	NNP
work_27d5tr326feydgcx32t4nhl424	42	22	En-	En-	NNP
work_27d5tr326feydgcx32t4nhl424	42	23	tropy	tropy	NN
work_27d5tr326feydgcx32t4nhl424	42	24	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	42	25	Osborne	Osborne	NNP
work_27d5tr326feydgcx32t4nhl424	42	26	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	27	2002	2002	CD
work_27d5tr326feydgcx32t4nhl424	42	28	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	42	29	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	30	Conditional	Conditional	NNP
work_27d5tr326feydgcx32t4nhl424	42	31	Random	Random	NNP
work_27d5tr326feydgcx32t4nhl424	42	32	Field	Field	NNP
work_27d5tr326feydgcx32t4nhl424	42	33	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	42	34	Shen	Shen	NNP
work_27d5tr326feydgcx32t4nhl424	42	35	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	42	36	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	42	37	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	42	38	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	39	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	42	40	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	42	41	and	and	CC
work_27d5tr326feydgcx32t4nhl424	42	42	regression	regression	NN
work_27d5tr326feydgcx32t4nhl424	42	43	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	42	44	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	42	45	Ouyang	Ouyang	NNP
work_27d5tr326feydgcx32t4nhl424	42	46	et	et	FW
work_27d5tr326feydgcx32t4nhl424	42	47	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	42	48	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	42	49	,	,	,
work_27d5tr326feydgcx32t4nhl424	42	50	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	42	51	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	42	52	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	42	53	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	42	54	adopted	adopt	VBN
work_27d5tr326feydgcx32t4nhl424	42	55	to	to	TO
work_27d5tr326feydgcx32t4nhl424	42	56	leverage	leverage	VB
work_27d5tr326feydgcx32t4nhl424	42	57	the	the	DT
work_27d5tr326feydgcx32t4nhl424	42	58	rich	rich	JJ
work_27d5tr326feydgcx32t4nhl424	42	59	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	42	60	features	feature	VBZ
work_27d5tr326feydgcx32t4nhl424	42	61	for	for	IN
work_27d5tr326feydgcx32t4nhl424	42	62	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	42	63	.	.	.
work_27d5tr326feydgcx32t4nhl424	43	1	Recently	recently	RB
work_27d5tr326feydgcx32t4nhl424	43	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	43	3	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	43	4	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	43	5	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	43	6	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	43	7	shown	show	VBN
work_27d5tr326feydgcx32t4nhl424	43	8	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	43	9	power	power	NN
work_27d5tr326feydgcx32t4nhl424	43	10	in	in	IN
work_27d5tr326feydgcx32t4nhl424	43	11	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	43	12	for	for	IN
work_27d5tr326feydgcx32t4nhl424	43	13	its	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	43	14	clear	clear	JJ
work_27d5tr326feydgcx32t4nhl424	43	15	probabilistic	probabilistic	JJ
work_27d5tr326feydgcx32t4nhl424	43	16	interpretation	interpretation	NN
work_27d5tr326feydgcx32t4nhl424	43	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	44	1	Daume	Daume	NNP
work_27d5tr326feydgcx32t4nhl424	44	2	and	and	CC
work_27d5tr326feydgcx32t4nhl424	44	3	Marcu	Marcu	NNP
work_27d5tr326feydgcx32t4nhl424	44	4	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	44	5	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	44	6	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	44	7	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	44	8	Bayesum	Bayesum	NNP
work_27d5tr326feydgcx32t4nhl424	44	9	model	model	NN
work_27d5tr326feydgcx32t4nhl424	44	10	for	for	IN
work_27d5tr326feydgcx32t4nhl424	44	11	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	44	12	extraction	extraction	NN
work_27d5tr326feydgcx32t4nhl424	44	13	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	44	14	on	on	IN
work_27d5tr326feydgcx32t4nhl424	44	15	query	query	NNP
work_27d5tr326feydgcx32t4nhl424	44	16	expansion	expansion	NN
work_27d5tr326feydgcx32t4nhl424	44	17	concept	concept	NN
work_27d5tr326feydgcx32t4nhl424	44	18	in	in	IN
work_27d5tr326feydgcx32t4nhl424	44	19	information	information	NN
work_27d5tr326feydgcx32t4nhl424	44	20	retrieval	retrieval	NN
work_27d5tr326feydgcx32t4nhl424	44	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	45	1	Haghighi	Haghighi	NNP
work_27d5tr326feydgcx32t4nhl424	45	2	and	and	CC
work_27d5tr326feydgcx32t4nhl424	45	3	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	45	4	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	45	5	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	45	6	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	45	7	proposed	propose	VBD
work_27d5tr326feydgcx32t4nhl424	45	8	topic-	topic-	NN
work_27d5tr326feydgcx32t4nhl424	45	9	sum	sum	NN
work_27d5tr326feydgcx32t4nhl424	45	10	and	and	CC
work_27d5tr326feydgcx32t4nhl424	45	11	hiersum	hiersum	NN
work_27d5tr326feydgcx32t4nhl424	45	12	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	45	13	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	45	14	a	a	DT
work_27d5tr326feydgcx32t4nhl424	45	15	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	45	16	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	45	17	like	like	JJ
work_27d5tr326feydgcx32t4nhl424	45	18	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	45	19	model	model	NN
work_27d5tr326feydgcx32t4nhl424	45	20	and	and	CC
work_27d5tr326feydgcx32t4nhl424	45	21	assign	assign	VB
work_27d5tr326feydgcx32t4nhl424	45	22	each	each	DT
work_27d5tr326feydgcx32t4nhl424	45	23	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	45	24	a	a	DT
work_27d5tr326feydgcx32t4nhl424	45	25	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	45	26	over	over	IN
work_27d5tr326feydgcx32t4nhl424	45	27	back-	back-	JJ
work_27d5tr326feydgcx32t4nhl424	45	28	ground	ground	NNP
work_27d5tr326feydgcx32t4nhl424	45	29	topic	topic	NNP
work_27d5tr326feydgcx32t4nhl424	45	30	,	,	,
work_27d5tr326feydgcx32t4nhl424	45	31	doc	doc	NN
work_27d5tr326feydgcx32t4nhl424	45	32	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	45	33	specific	specific	NN
work_27d5tr326feydgcx32t4nhl424	45	34	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	45	35	and	and	CC
work_27d5tr326feydgcx32t4nhl424	45	36	content	content	NN
work_27d5tr326feydgcx32t4nhl424	45	37	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	45	38	.	.	.
work_27d5tr326feydgcx32t4nhl424	46	1	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	46	2	and	and	CC
work_27d5tr326feydgcx32t4nhl424	46	3	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	46	4	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	46	5	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	46	6	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	46	7	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	46	8	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	46	9	made	make	VBD
work_27d5tr326feydgcx32t4nhl424	46	10	a	a	DT
work_27d5tr326feydgcx32t4nhl424	46	11	good	good	JJ
work_27d5tr326feydgcx32t4nhl424	46	12	step	step	NN
work_27d5tr326feydgcx32t4nhl424	46	13	in	in	IN
work_27d5tr326feydgcx32t4nhl424	46	14	combining	combine	VBG
work_27d5tr326feydgcx32t4nhl424	46	15	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	46	16	model	model	NN
work_27d5tr326feydgcx32t4nhl424	46	17	with	with	IN
work_27d5tr326feydgcx32t4nhl424	46	18	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	46	19	fea-	fea-	JJ
work_27d5tr326feydgcx32t4nhl424	46	20	ture	ture	NN
work_27d5tr326feydgcx32t4nhl424	46	21	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	46	22	regression	regression	NN
work_27d5tr326feydgcx32t4nhl424	46	23	for	for	IN
work_27d5tr326feydgcx32t4nhl424	46	24	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	46	25	scoring	score	VBG
work_27d5tr326feydgcx32t4nhl424	46	26	in	in	IN
work_27d5tr326feydgcx32t4nhl424	46	27	sum-	sum-	NNP
work_27d5tr326feydgcx32t4nhl424	46	28	marization	marization	NN
work_27d5tr326feydgcx32t4nhl424	46	29	.	.	.
work_27d5tr326feydgcx32t4nhl424	47	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	47	2	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	47	3	model	model	NN
work_27d5tr326feydgcx32t4nhl424	47	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	47	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	47	6	score	score	NN
work_27d5tr326feydgcx32t4nhl424	47	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	47	8	training	training	NN
work_27d5tr326feydgcx32t4nhl424	47	9	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	47	10	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	47	11	firstly	firstly	RB
work_27d5tr326feydgcx32t4nhl424	47	12	got	get	VBN
work_27d5tr326feydgcx32t4nhl424	47	13	through	through	IN
work_27d5tr326feydgcx32t4nhl424	47	14	a	a	DT
work_27d5tr326feydgcx32t4nhl424	47	15	novel	novel	NN
work_27d5tr326feydgcx32t4nhl424	47	16	hierarchi-	hierarchi-	NN
work_27d5tr326feydgcx32t4nhl424	47	17	cal	cal	NNP
work_27d5tr326feydgcx32t4nhl424	47	18	topic	topic	NNP
work_27d5tr326feydgcx32t4nhl424	47	19	model	model	NN
work_27d5tr326feydgcx32t4nhl424	47	20	.	.	.
work_27d5tr326feydgcx32t4nhl424	48	1	Then	then	RB
work_27d5tr326feydgcx32t4nhl424	48	2	a	a	DT
work_27d5tr326feydgcx32t4nhl424	48	3	featured	feature	VBN
work_27d5tr326feydgcx32t4nhl424	48	4	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	48	5	support	support	NN
work_27d5tr326feydgcx32t4nhl424	48	6	vec-	vec-	NN
work_27d5tr326feydgcx32t4nhl424	48	7	tor	tor	NN
work_27d5tr326feydgcx32t4nhl424	48	8	regression	regression	NN
work_27d5tr326feydgcx32t4nhl424	48	9	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	48	10	SVR	SVR	NNP
work_27d5tr326feydgcx32t4nhl424	48	11	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	48	12	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	48	13	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	48	14	for	for	IN
work_27d5tr326feydgcx32t4nhl424	48	15	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	48	16	score	score	NN
work_27d5tr326feydgcx32t4nhl424	48	17	pre-	pre-	XX
work_27d5tr326feydgcx32t4nhl424	48	18	diction	diction	NN
work_27d5tr326feydgcx32t4nhl424	48	19	.	.	.
work_27d5tr326feydgcx32t4nhl424	49	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	49	2	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	49	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	49	4	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	49	5	and	and	CC
work_27d5tr326feydgcx32t4nhl424	49	6	Hakkani-	Hakkani-	NNP
work_27d5tr326feydgcx32t4nhl424	49	7	Turs	Turs	NNP
work_27d5tr326feydgcx32t4nhl424	49	8	model	model	NN
work_27d5tr326feydgcx32t4nhl424	49	9	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	49	10	that	that	IN
work_27d5tr326feydgcx32t4nhl424	49	11	topic	topic	JJ
work_27d5tr326feydgcx32t4nhl424	49	12	model	model	NN
work_27d5tr326feydgcx32t4nhl424	49	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	49	14	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	49	15	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	49	16	re-	re-	JJ
work_27d5tr326feydgcx32t4nhl424	49	17	gression	gression	NN
work_27d5tr326feydgcx32t4nhl424	49	18	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	49	19	two	two	CD
work_27d5tr326feydgcx32t4nhl424	49	20	separate	separate	JJ
work_27d5tr326feydgcx32t4nhl424	49	21	processes	process	NNS
work_27d5tr326feydgcx32t4nhl424	49	22	and	and	CC
work_27d5tr326feydgcx32t4nhl424	49	23	the	the	DT
work_27d5tr326feydgcx32t4nhl424	49	24	score	score	NN
work_27d5tr326feydgcx32t4nhl424	49	25	of	of	IN
work_27d5tr326feydgcx32t4nhl424	49	26	training	training	NN
work_27d5tr326feydgcx32t4nhl424	49	27	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	49	28	may	may	MD
work_27d5tr326feydgcx32t4nhl424	49	29	be	be	VB
work_27d5tr326feydgcx32t4nhl424	49	30	biased	biased	JJ
work_27d5tr326feydgcx32t4nhl424	49	31	because	because	IN
work_27d5tr326feydgcx32t4nhl424	49	32	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	49	33	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	49	34	model	model	NN
work_27d5tr326feydgcx32t4nhl424	49	35	only	only	RB
work_27d5tr326feydgcx32t4nhl424	49	36	consider	consider	VB
work_27d5tr326feydgcx32t4nhl424	49	37	word	word	NN
work_27d5tr326feydgcx32t4nhl424	49	38	frequency	frequency	NN
work_27d5tr326feydgcx32t4nhl424	49	39	and	and	CC
work_27d5tr326feydgcx32t4nhl424	49	40	fail	fail	VB
work_27d5tr326feydgcx32t4nhl424	49	41	to	to	IN
work_27d5tr326feydgcx32t4nhl424	49	42	con-	con-	NNP
work_27d5tr326feydgcx32t4nhl424	49	43	sider	sider	VB
work_27d5tr326feydgcx32t4nhl424	49	44	other	other	JJ
work_27d5tr326feydgcx32t4nhl424	49	45	important	important	JJ
work_27d5tr326feydgcx32t4nhl424	49	46	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	49	47	.	.	.
work_27d5tr326feydgcx32t4nhl424	50	1	Supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	50	2	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	50	3	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	50	4	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	50	5	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	50	6	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	50	7	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	50	8	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	50	9	in	in	IN
work_27d5tr326feydgcx32t4nhl424	50	10	recent	recent	JJ
work_27d5tr326feydgcx32t4nhl424	50	11	years	year	NNS
work_27d5tr326feydgcx32t4nhl424	50	12	to	to	TO
work_27d5tr326feydgcx32t4nhl424	50	13	incorporate	incorporate	VB
work_27d5tr326feydgcx32t4nhl424	50	14	different	different	JJ
work_27d5tr326feydgcx32t4nhl424	50	15	kinds	kind	NNS
work_27d5tr326feydgcx32t4nhl424	50	16	of	of	IN
work_27d5tr326feydgcx32t4nhl424	50	17	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	50	18	into	into	IN
work_27d5tr326feydgcx32t4nhl424	50	19	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	50	20	model	model	NN
work_27d5tr326feydgcx32t4nhl424	50	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	51	1	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	51	2	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	51	3	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	51	4	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	51	5	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	51	6	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	51	7	for	for	IN
work_27d5tr326feydgcx32t4nhl424	51	8	doc-	doc-	NNP
work_27d5tr326feydgcx32t4nhl424	51	9	ument	ument	NNP
work_27d5tr326feydgcx32t4nhl424	51	10	response	response	NN
work_27d5tr326feydgcx32t4nhl424	51	11	pairs	pair	NNS
work_27d5tr326feydgcx32t4nhl424	51	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	51	13	Daniel	Daniel	NNP
work_27d5tr326feydgcx32t4nhl424	51	14	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	51	15	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	51	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	52	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	52	2	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	52	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	52	4	pro-	pro-	NN
work_27d5tr326feydgcx32t4nhl424	52	5	posed	pose	VBN
work_27d5tr326feydgcx32t4nhl424	52	6	Labeled	Labeled	NNP
work_27d5tr326feydgcx32t4nhl424	52	7	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	52	8	by	by	IN
work_27d5tr326feydgcx32t4nhl424	52	9	defining	define	VBG
work_27d5tr326feydgcx32t4nhl424	52	10	a	a	DT
work_27d5tr326feydgcx32t4nhl424	52	11	one	one	CD
work_27d5tr326feydgcx32t4nhl424	52	12	to	to	TO
work_27d5tr326feydgcx32t4nhl424	52	13	one	one	CD
work_27d5tr326feydgcx32t4nhl424	52	14	corre-	corre-	NN
work_27d5tr326feydgcx32t4nhl424	52	15	spondence	spondence	NN
work_27d5tr326feydgcx32t4nhl424	52	16	between	between	IN
work_27d5tr326feydgcx32t4nhl424	52	17	latent	latent	NN
work_27d5tr326feydgcx32t4nhl424	52	18	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	52	19	and	and	CC
work_27d5tr326feydgcx32t4nhl424	52	20	user	user	NN
work_27d5tr326feydgcx32t4nhl424	52	21	tags	tag	NNS
work_27d5tr326feydgcx32t4nhl424	52	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	53	1	Zhu	Zhu	NNP
work_27d5tr326feydgcx32t4nhl424	53	2	and	and	CC
work_27d5tr326feydgcx32t4nhl424	53	3	Xing	Xing	NNP
work_27d5tr326feydgcx32t4nhl424	53	4	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	53	5	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	53	6	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	53	7	proposed	propose	VBD
work_27d5tr326feydgcx32t4nhl424	53	8	conditional	conditional	JJ
work_27d5tr326feydgcx32t4nhl424	53	9	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	53	10	random	random	JJ
work_27d5tr326feydgcx32t4nhl424	53	11	field	field	NN
work_27d5tr326feydgcx32t4nhl424	53	12	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	53	13	CTRF	CTRF	NNP
work_27d5tr326feydgcx32t4nhl424	53	14	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	53	15	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	53	16	addresses	address	VBZ
work_27d5tr326feydgcx32t4nhl424	53	17	feature	feature	VBP
work_27d5tr326feydgcx32t4nhl424	53	18	and	and	CC
work_27d5tr326feydgcx32t4nhl424	53	19	indepen-	indepen-	VBP
work_27d5tr326feydgcx32t4nhl424	53	20	dent	dent	NN
work_27d5tr326feydgcx32t4nhl424	53	21	limitation	limitation	NN
work_27d5tr326feydgcx32t4nhl424	53	22	in	in	IN
work_27d5tr326feydgcx32t4nhl424	53	23	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	53	24	.	.	.
work_27d5tr326feydgcx32t4nhl424	54	1	3	3	CD
work_27d5tr326feydgcx32t4nhl424	54	2	Model	model	NN
work_27d5tr326feydgcx32t4nhl424	54	3	description	description	NN
work_27d5tr326feydgcx32t4nhl424	54	4	3.1	3.1	CD
work_27d5tr326feydgcx32t4nhl424	54	5	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	54	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	54	7	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	54	8	The	the	DT
work_27d5tr326feydgcx32t4nhl424	54	9	hierarchical	hierarchical	JJ
work_27d5tr326feydgcx32t4nhl424	54	10	Bayesian	Bayesian	NNP
work_27d5tr326feydgcx32t4nhl424	54	11	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	54	12	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	54	13	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	54	14	et	et	FW
work_27d5tr326feydgcx32t4nhl424	54	15	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	54	16	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	54	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	54	18	2003	2003	CD
work_27d5tr326feydgcx32t4nhl424	54	19	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	54	20	models	model	VBZ
work_27d5tr326feydgcx32t4nhl424	54	21	the	the	DT
work_27d5tr326feydgcx32t4nhl424	54	22	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	54	23	of	of	IN
work_27d5tr326feydgcx32t4nhl424	54	24	a	a	DT
work_27d5tr326feydgcx32t4nhl424	54	25	corpus	corpus	NN
work_27d5tr326feydgcx32t4nhl424	54	26	on	on	IN
work_27d5tr326feydgcx32t4nhl424	54	27	hidden	hidden	JJ
work_27d5tr326feydgcx32t4nhl424	54	28	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	54	29	as	as	IN
work_27d5tr326feydgcx32t4nhl424	54	30	shown	show	VBN
work_27d5tr326feydgcx32t4nhl424	54	31	in	in	IN
work_27d5tr326feydgcx32t4nhl424	54	32	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	54	33	1(a	1(a	CD
work_27d5tr326feydgcx32t4nhl424	54	34	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	54	35	.	.	.
work_27d5tr326feydgcx32t4nhl424	55	1	Let	let	VB
work_27d5tr326feydgcx32t4nhl424	55	2	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	55	3	be	be	VB
work_27d5tr326feydgcx32t4nhl424	55	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	55	5	number	number	NN
work_27d5tr326feydgcx32t4nhl424	55	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	55	7	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	55	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	55	9	M	m	NN
work_27d5tr326feydgcx32t4nhl424	55	10	be	be	VB
work_27d5tr326feydgcx32t4nhl424	55	11	the	the	DT
work_27d5tr326feydgcx32t4nhl424	55	12	number	number	NN
work_27d5tr326feydgcx32t4nhl424	55	13	of	of	IN
work_27d5tr326feydgcx32t4nhl424	55	14	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	55	15	in	in	IN
work_27d5tr326feydgcx32t4nhl424	55	16	the	the	DT
work_27d5tr326feydgcx32t4nhl424	55	17	cor-	cor-	NN
work_27d5tr326feydgcx32t4nhl424	55	18	pus	pus	NN
work_27d5tr326feydgcx32t4nhl424	55	19	and	and	CC
work_27d5tr326feydgcx32t4nhl424	55	20	V	v	NN
work_27d5tr326feydgcx32t4nhl424	55	21	be	be	VB
work_27d5tr326feydgcx32t4nhl424	55	22	vocabulary	vocabulary	NN
work_27d5tr326feydgcx32t4nhl424	55	23	size	size	NN
work_27d5tr326feydgcx32t4nhl424	55	24	.	.	.
work_27d5tr326feydgcx32t4nhl424	56	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	56	2	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	56	3	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	56	4	of	of	IN
work_27d5tr326feydgcx32t4nhl424	56	5	each	each	DT
work_27d5tr326feydgcx32t4nhl424	56	6	document	document	NN
work_27d5tr326feydgcx32t4nhl424	56	7	θm	θm	NNP
work_27d5tr326feydgcx32t4nhl424	56	8	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	56	9	drawn	draw	VBN
work_27d5tr326feydgcx32t4nhl424	56	10	from	from	IN
work_27d5tr326feydgcx32t4nhl424	56	11	a	a	DT
work_27d5tr326feydgcx32t4nhl424	56	12	prior	prior	JJ
work_27d5tr326feydgcx32t4nhl424	56	13	Dirichlet	Dirichlet	NNP
work_27d5tr326feydgcx32t4nhl424	56	14	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	56	15	Dir(α	dir(α	CD
work_27d5tr326feydgcx32t4nhl424	56	16	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	56	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	56	18	and	and	CC
work_27d5tr326feydgcx32t4nhl424	56	19	each	each	DT
work_27d5tr326feydgcx32t4nhl424	56	20	document	document	NN
work_27d5tr326feydgcx32t4nhl424	56	21	word	word	NN
work_27d5tr326feydgcx32t4nhl424	56	22	wmn	wmn	NNP
work_27d5tr326feydgcx32t4nhl424	56	23	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	56	24	sampled	sample	VBN
work_27d5tr326feydgcx32t4nhl424	56	25	from	from	IN
work_27d5tr326feydgcx32t4nhl424	56	26	a	a	DT
work_27d5tr326feydgcx32t4nhl424	56	27	topic	topic	JJ
work_27d5tr326feydgcx32t4nhl424	56	28	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	56	29	word	word	NN
work_27d5tr326feydgcx32t4nhl424	56	30	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	56	31	φz	φz	NNP
work_27d5tr326feydgcx32t4nhl424	56	32	spec-	spec-	NNP
work_27d5tr326feydgcx32t4nhl424	56	33	ified	ifie	VBN
work_27d5tr326feydgcx32t4nhl424	56	34	by	by	IN
work_27d5tr326feydgcx32t4nhl424	56	35	a	a	DT
work_27d5tr326feydgcx32t4nhl424	56	36	drawn	drawn	NN
work_27d5tr326feydgcx32t4nhl424	56	37	from	from	IN
work_27d5tr326feydgcx32t4nhl424	56	38	the	the	DT
work_27d5tr326feydgcx32t4nhl424	56	39	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	56	40	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	56	41	document	document	NN
work_27d5tr326feydgcx32t4nhl424	56	42	distribu-	distribu-	FW
work_27d5tr326feydgcx32t4nhl424	56	43	tion	tion	NN
work_27d5tr326feydgcx32t4nhl424	56	44	θm	θm	NNP
work_27d5tr326feydgcx32t4nhl424	56	45	.	.	.
work_27d5tr326feydgcx32t4nhl424	57	1	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	57	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	57	3	a	a	DT
work_27d5tr326feydgcx32t4nhl424	57	4	K×M	k×m	JJ
work_27d5tr326feydgcx32t4nhl424	57	5	dimensional	dimensional	JJ
work_27d5tr326feydgcx32t4nhl424	57	6	matrix	matrix	NN
work_27d5tr326feydgcx32t4nhl424	57	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	57	8	each	each	DT
work_27d5tr326feydgcx32t4nhl424	57	9	βk	βk	NN
work_27d5tr326feydgcx32t4nhl424	57	10	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	57	11	a	a	DT
work_27d5tr326feydgcx32t4nhl424	57	12	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	57	13	over	over	IN
work_27d5tr326feydgcx32t4nhl424	57	14	the	the	DT
work_27d5tr326feydgcx32t4nhl424	57	15	V	v	NN
work_27d5tr326feydgcx32t4nhl424	57	16	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	57	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	58	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	58	2	generat-	generat-	NN
work_27d5tr326feydgcx32t4nhl424	58	3	ing	ing	NN
work_27d5tr326feydgcx32t4nhl424	58	4	procedure	procedure	NN
work_27d5tr326feydgcx32t4nhl424	58	5	of	of	IN
work_27d5tr326feydgcx32t4nhl424	58	6	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	58	7	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	58	8	illustrated	illustrate	VBN
work_27d5tr326feydgcx32t4nhl424	58	9	in	in	IN
work_27d5tr326feydgcx32t4nhl424	58	10	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	58	11	2	2	CD
work_27d5tr326feydgcx32t4nhl424	58	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	58	13	θm	θm	NNP
work_27d5tr326feydgcx32t4nhl424	58	14	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	58	15	a	a	DT
work_27d5tr326feydgcx32t4nhl424	58	16	mixture	mixture	NN
work_27d5tr326feydgcx32t4nhl424	58	17	proportion	proportion	NN
work_27d5tr326feydgcx32t4nhl424	58	18	over	over	IN
work_27d5tr326feydgcx32t4nhl424	58	19	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	58	20	of	of	IN
work_27d5tr326feydgcx32t4nhl424	58	21	document	document	NN
work_27d5tr326feydgcx32t4nhl424	58	22	m	m	NNP
work_27d5tr326feydgcx32t4nhl424	58	23	and	and	CC
work_27d5tr326feydgcx32t4nhl424	58	24	zmn	zmn	NNP
work_27d5tr326feydgcx32t4nhl424	58	25	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	58	26	a	a	DT
work_27d5tr326feydgcx32t4nhl424	58	27	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	58	28	dimensional	dimensional	JJ
work_27d5tr326feydgcx32t4nhl424	58	29	variable	variable	NN
work_27d5tr326feydgcx32t4nhl424	58	30	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	58	31	presents	present	VBZ
work_27d5tr326feydgcx32t4nhl424	58	32	the	the	DT
work_27d5tr326feydgcx32t4nhl424	58	33	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	58	34	assignment	assignment	NN
work_27d5tr326feydgcx32t4nhl424	58	35	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	58	36	of	of	IN
work_27d5tr326feydgcx32t4nhl424	58	37	different	different	JJ
work_27d5tr326feydgcx32t4nhl424	58	38	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	58	39	.	.	.
work_27d5tr326feydgcx32t4nhl424	59	1	Supervised	Supervised	NNP
work_27d5tr326feydgcx32t4nhl424	59	2	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	59	3	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	59	4	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	59	5	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	59	6	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	59	7	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	59	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	59	9	McAuliffe	McAuliffe	NNP
work_27d5tr326feydgcx32t4nhl424	59	10	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	59	11	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	59	12	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	59	13	a	a	DT
work_27d5tr326feydgcx32t4nhl424	59	14	document	document	NN
work_27d5tr326feydgcx32t4nhl424	59	15	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	59	16	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	59	17	model	model	NN
work_27d5tr326feydgcx32t4nhl424	59	18	and	and	CC
work_27d5tr326feydgcx32t4nhl424	59	19	intro-	intro-	NN
work_27d5tr326feydgcx32t4nhl424	59	20	90	90	CD
work_27d5tr326feydgcx32t4nhl424	59	21	Figure	figure	NN
work_27d5tr326feydgcx32t4nhl424	59	22	1	1	CD
work_27d5tr326feydgcx32t4nhl424	59	23	:	:	:
work_27d5tr326feydgcx32t4nhl424	59	24	Graphical	graphical	JJ
work_27d5tr326feydgcx32t4nhl424	59	25	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	59	26	for	for	IN
work_27d5tr326feydgcx32t4nhl424	59	27	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	59	28	a	a	DT
work_27d5tr326feydgcx32t4nhl424	59	29	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	59	30	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	59	31	model	model	NN
work_27d5tr326feydgcx32t4nhl424	59	32	and	and	CC
work_27d5tr326feydgcx32t4nhl424	59	33	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	59	34	b	b	NN
work_27d5tr326feydgcx32t4nhl424	59	35	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	59	36	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	59	37	model	model	NN
work_27d5tr326feydgcx32t4nhl424	59	38	.	.	.
work_27d5tr326feydgcx32t4nhl424	60	1	1	1	LS
work_27d5tr326feydgcx32t4nhl424	60	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	61	1	Draw	draw	VB
work_27d5tr326feydgcx32t4nhl424	61	2	a	a	DT
work_27d5tr326feydgcx32t4nhl424	61	3	document	document	NN
work_27d5tr326feydgcx32t4nhl424	61	4	proportion	proportion	NN
work_27d5tr326feydgcx32t4nhl424	61	5	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	61	6	θm|α	θm|α	NNP
work_27d5tr326feydgcx32t4nhl424	61	7	∼	∼	NN
work_27d5tr326feydgcx32t4nhl424	61	8	Dir(α	Dir(α	NNP
work_27d5tr326feydgcx32t4nhl424	61	9	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	61	10	2	2	CD
work_27d5tr326feydgcx32t4nhl424	61	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	62	1	For	for	IN
work_27d5tr326feydgcx32t4nhl424	62	2	each	each	DT
work_27d5tr326feydgcx32t4nhl424	62	3	word	word	NN
work_27d5tr326feydgcx32t4nhl424	62	4	in	in	IN
work_27d5tr326feydgcx32t4nhl424	62	5	m	m	NNP
work_27d5tr326feydgcx32t4nhl424	62	6	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	62	7	a)draw	a)draw	CD
work_27d5tr326feydgcx32t4nhl424	62	8	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	62	9	assignment	assignment	NN
work_27d5tr326feydgcx32t4nhl424	62	10	zmn|θ	zmn|θ	NNP
work_27d5tr326feydgcx32t4nhl424	62	11	∼	∼	NNP
work_27d5tr326feydgcx32t4nhl424	62	12	Multi(θzmn	Multi(θzmn	NNP
work_27d5tr326feydgcx32t4nhl424	62	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	62	14	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	62	15	b)draw	b)draw	NN
work_27d5tr326feydgcx32t4nhl424	62	16	word	word	NN
work_27d5tr326feydgcx32t4nhl424	62	17	wmn|zmn	wmn|zmn	NNP
work_27d5tr326feydgcx32t4nhl424	62	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	62	19	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	62	20	∼	∼	NNP
work_27d5tr326feydgcx32t4nhl424	62	21	Multi(βzmn	Multi(βzmn	NNP
work_27d5tr326feydgcx32t4nhl424	62	22	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	62	23	Figure	figure	NN
work_27d5tr326feydgcx32t4nhl424	62	24	2	2	CD
work_27d5tr326feydgcx32t4nhl424	62	25	:	:	:
work_27d5tr326feydgcx32t4nhl424	62	26	Generation	generation	NN
work_27d5tr326feydgcx32t4nhl424	62	27	process	process	NN
work_27d5tr326feydgcx32t4nhl424	62	28	for	for	IN
work_27d5tr326feydgcx32t4nhl424	62	29	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	62	30	duces	duce	VBZ
work_27d5tr326feydgcx32t4nhl424	62	31	a	a	DT
work_27d5tr326feydgcx32t4nhl424	62	32	response	response	NN
work_27d5tr326feydgcx32t4nhl424	62	33	variable	variable	JJ
work_27d5tr326feydgcx32t4nhl424	62	34	to	to	IN
work_27d5tr326feydgcx32t4nhl424	62	35	each	each	DT
work_27d5tr326feydgcx32t4nhl424	62	36	document	document	NN
work_27d5tr326feydgcx32t4nhl424	62	37	for	for	IN
work_27d5tr326feydgcx32t4nhl424	62	38	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	62	39	discovering	discovering	NN
work_27d5tr326feydgcx32t4nhl424	62	40	,	,	,
work_27d5tr326feydgcx32t4nhl424	62	41	as	as	IN
work_27d5tr326feydgcx32t4nhl424	62	42	shown	show	VBN
work_27d5tr326feydgcx32t4nhl424	62	43	in	in	IN
work_27d5tr326feydgcx32t4nhl424	62	44	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	62	45	1(b	1(b	CD
work_27d5tr326feydgcx32t4nhl424	62	46	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	62	47	.	.	.
work_27d5tr326feydgcx32t4nhl424	63	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	63	2	the	the	DT
work_27d5tr326feydgcx32t4nhl424	63	3	gener-	gener-	JJ
work_27d5tr326feydgcx32t4nhl424	63	4	ative	ative	JJ
work_27d5tr326feydgcx32t4nhl424	63	5	procedure	procedure	NN
work_27d5tr326feydgcx32t4nhl424	63	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	63	7	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	63	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	63	10	document	document	NN
work_27d5tr326feydgcx32t4nhl424	63	11	pairwise	pairwise	NN
work_27d5tr326feydgcx32t4nhl424	63	12	la-	la-	NNP
work_27d5tr326feydgcx32t4nhl424	63	13	bel	bel	NNP
work_27d5tr326feydgcx32t4nhl424	63	14	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	63	15	draw	draw	NN
work_27d5tr326feydgcx32t4nhl424	63	16	from	from	IN
work_27d5tr326feydgcx32t4nhl424	63	17	y|−→zm	y|−→zm	NNP
work_27d5tr326feydgcx32t4nhl424	63	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	19	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	63	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	21	δ2	δ2	NNP
work_27d5tr326feydgcx32t4nhl424	63	22	∼	∼	NNP
work_27d5tr326feydgcx32t4nhl424	63	23	p(y|−→zm	p(y|−→zm	NNP
work_27d5tr326feydgcx32t4nhl424	63	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	25	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	63	26	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	27	δ2	δ2	NNP
work_27d5tr326feydgcx32t4nhl424	63	28	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	63	29	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	30	where−→zm	where−→zm	ADD
work_27d5tr326feydgcx32t4nhl424	63	31	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	63	32	1N	1N	NNP
work_27d5tr326feydgcx32t4nhl424	63	33	∑N	∑N	NNP
work_27d5tr326feydgcx32t4nhl424	63	34	n=1	n=1	NNP
work_27d5tr326feydgcx32t4nhl424	63	35	zm	zm	NNP
work_27d5tr326feydgcx32t4nhl424	63	36	,	,	,
work_27d5tr326feydgcx32t4nhl424	63	37	n	n	NNP
work_27d5tr326feydgcx32t4nhl424	63	38	.	.	.
work_27d5tr326feydgcx32t4nhl424	64	1	3.2	3.2	CD
work_27d5tr326feydgcx32t4nhl424	64	2	Problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	64	3	Formulation	Formulation	NNP
work_27d5tr326feydgcx32t4nhl424	64	4	Here	here	RB
work_27d5tr326feydgcx32t4nhl424	64	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	64	6	firstly	firstly	RB
work_27d5tr326feydgcx32t4nhl424	64	7	give	give	VBP
work_27d5tr326feydgcx32t4nhl424	64	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	64	9	standard	standard	JJ
work_27d5tr326feydgcx32t4nhl424	64	10	formulation	formulation	NN
work_27d5tr326feydgcx32t4nhl424	64	11	of	of	IN
work_27d5tr326feydgcx32t4nhl424	64	12	the	the	DT
work_27d5tr326feydgcx32t4nhl424	64	13	task	task	NN
work_27d5tr326feydgcx32t4nhl424	64	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	65	1	Let	let	VB
work_27d5tr326feydgcx32t4nhl424	65	2	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	65	3	be	be	VB
work_27d5tr326feydgcx32t4nhl424	65	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	65	5	number	number	NN
work_27d5tr326feydgcx32t4nhl424	65	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	65	7	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	65	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	65	9	V	v	NN
work_27d5tr326feydgcx32t4nhl424	65	10	be	be	VBP
work_27d5tr326feydgcx32t4nhl424	65	11	the	the	DT
work_27d5tr326feydgcx32t4nhl424	65	12	vo-	vo-	JJ
work_27d5tr326feydgcx32t4nhl424	65	13	cabulary	cabulary	JJ
work_27d5tr326feydgcx32t4nhl424	65	14	size	size	NN
work_27d5tr326feydgcx32t4nhl424	65	15	and	and	CC
work_27d5tr326feydgcx32t4nhl424	65	16	M	m	NN
work_27d5tr326feydgcx32t4nhl424	65	17	be	be	VB
work_27d5tr326feydgcx32t4nhl424	65	18	the	the	DT
work_27d5tr326feydgcx32t4nhl424	65	19	number	number	NN
work_27d5tr326feydgcx32t4nhl424	65	20	of	of	IN
work_27d5tr326feydgcx32t4nhl424	65	21	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	65	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	66	1	Each	each	DT
work_27d5tr326feydgcx32t4nhl424	66	2	document	document	NN
work_27d5tr326feydgcx32t4nhl424	66	3	Dm	Dm	NNP
work_27d5tr326feydgcx32t4nhl424	66	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	66	5	represented	represent	VBN
work_27d5tr326feydgcx32t4nhl424	66	6	with	with	IN
work_27d5tr326feydgcx32t4nhl424	66	7	a	a	DT
work_27d5tr326feydgcx32t4nhl424	66	8	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	66	9	of	of	IN
work_27d5tr326feydgcx32t4nhl424	66	10	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	66	11	Dm	Dm	NNP
work_27d5tr326feydgcx32t4nhl424	66	12	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	66	13	{	{	-LRB-
work_27d5tr326feydgcx32t4nhl424	66	14	Ss}s	Ss}s	NNP
work_27d5tr326feydgcx32t4nhl424	66	15	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	66	16	Nms=1	Nms=1	NNP
work_27d5tr326feydgcx32t4nhl424	66	17	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	66	18	Nm	Nm	NNP
work_27d5tr326feydgcx32t4nhl424	66	19	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	66	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	66	21	number	number	NN
work_27d5tr326feydgcx32t4nhl424	66	22	of	of	IN
work_27d5tr326feydgcx32t4nhl424	66	23	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	66	24	in	in	IN
work_27d5tr326feydgcx32t4nhl424	66	25	mth	mth	NNP
work_27d5tr326feydgcx32t4nhl424	66	26	document	document	NN
work_27d5tr326feydgcx32t4nhl424	66	27	.	.	.
work_27d5tr326feydgcx32t4nhl424	67	1	Each	each	DT
work_27d5tr326feydgcx32t4nhl424	67	2	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	67	3	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	67	4	represented	represent	VBN
work_27d5tr326feydgcx32t4nhl424	67	5	with	with	IN
work_27d5tr326feydgcx32t4nhl424	67	6	a	a	DT
work_27d5tr326feydgcx32t4nhl424	67	7	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	67	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	67	9	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	67	10	{	{	-LRB-
work_27d5tr326feydgcx32t4nhl424	67	11	wmsn}n	wmsn}n	NNP
work_27d5tr326feydgcx32t4nhl424	67	12	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	67	13	Nmsn=1	Nmsn=1	NNP
work_27d5tr326feydgcx32t4nhl424	67	14	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	67	15	Nms	Nms	NNP
work_27d5tr326feydgcx32t4nhl424	67	16	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	67	17	the	the	DT
work_27d5tr326feydgcx32t4nhl424	67	18	number	number	NN
work_27d5tr326feydgcx32t4nhl424	67	19	of	of	IN
work_27d5tr326feydgcx32t4nhl424	67	20	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	67	21	in	in	IN
work_27d5tr326feydgcx32t4nhl424	67	22	current	current	JJ
work_27d5tr326feydgcx32t4nhl424	67	23	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	67	24	.	.	.
work_27d5tr326feydgcx32t4nhl424	68	1	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	68	2	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	68	3	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	68	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	68	5	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	68	6	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	68	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	68	8	current	current	JJ
work_27d5tr326feydgcx32t4nhl424	68	9	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	68	10	and	and	CC
work_27d5tr326feydgcx32t4nhl424	68	11	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	68	12	assume	assume	VBP
work_27d5tr326feydgcx32t4nhl424	68	13	that	that	IN
work_27d5tr326feydgcx32t4nhl424	68	14	these	these	DT
work_27d5tr326feydgcx32t4nhl424	68	15	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	68	16	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	68	17	independent	independent	JJ
work_27d5tr326feydgcx32t4nhl424	68	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	69	1	3.3	3.3	CD
work_27d5tr326feydgcx32t4nhl424	69	2	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	69	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	69	4	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	69	5	zms	zms	NNP
work_27d5tr326feydgcx32t4nhl424	69	6	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	69	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	69	8	hidden	hidden	JJ
work_27d5tr326feydgcx32t4nhl424	69	9	variable	variable	NN
work_27d5tr326feydgcx32t4nhl424	69	10	indicating	indicate	VBG
work_27d5tr326feydgcx32t4nhl424	69	11	the	the	DT
work_27d5tr326feydgcx32t4nhl424	69	12	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	69	13	of	of	IN
work_27d5tr326feydgcx32t4nhl424	69	14	current	current	JJ
work_27d5tr326feydgcx32t4nhl424	69	15	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	69	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	70	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	70	2	S	s	NN
work_27d5tr326feydgcx32t4nhl424	70	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	70	4	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	70	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	70	6	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	70	7	make	make	VBP
work_27d5tr326feydgcx32t4nhl424	70	8	an	an	DT
work_27d5tr326feydgcx32t4nhl424	70	9	assump-	assump-	JJ
work_27d5tr326feydgcx32t4nhl424	70	10	tion	tion	NN
work_27d5tr326feydgcx32t4nhl424	70	11	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	70	12	words	word	VBZ
work_27d5tr326feydgcx32t4nhl424	70	13	in	in	IN
work_27d5tr326feydgcx32t4nhl424	70	14	the	the	DT
work_27d5tr326feydgcx32t4nhl424	70	15	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	70	16	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	70	17	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	70	18	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	70	19	from	from	IN
work_27d5tr326feydgcx32t4nhl424	70	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	70	21	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	70	22	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	70	23	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	70	24	was	be	VBD
work_27d5tr326feydgcx32t4nhl424	70	25	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	70	26	by	by	IN
work_27d5tr326feydgcx32t4nhl424	70	27	Gruber	Gruber	NNP
work_27d5tr326feydgcx32t4nhl424	70	28	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	70	29	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	70	30	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	70	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	71	1	zmsn	zmsn	NNP
work_27d5tr326feydgcx32t4nhl424	71	2	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	71	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	71	4	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	71	5	assignment	assignment	NN
work_27d5tr326feydgcx32t4nhl424	71	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	71	7	cur-	cur-	NN
work_27d5tr326feydgcx32t4nhl424	71	8	rent	rent	NN
work_27d5tr326feydgcx32t4nhl424	71	9	word	word	NN
work_27d5tr326feydgcx32t4nhl424	71	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	72	1	According	accord	VBG
work_27d5tr326feydgcx32t4nhl424	72	2	to	to	IN
work_27d5tr326feydgcx32t4nhl424	72	3	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	72	4	assumption	assumption	NN
work_27d5tr326feydgcx32t4nhl424	72	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	72	6	zmsn	zmsn	NN
work_27d5tr326feydgcx32t4nhl424	72	7	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	72	8	Figure	figure	NN
work_27d5tr326feydgcx32t4nhl424	72	9	3	3	CD
work_27d5tr326feydgcx32t4nhl424	72	10	:	:	:
work_27d5tr326feydgcx32t4nhl424	72	11	Graph	graph	NN
work_27d5tr326feydgcx32t4nhl424	72	12	model	model	NN
work_27d5tr326feydgcx32t4nhl424	72	13	for	for	IN
work_27d5tr326feydgcx32t4nhl424	72	14	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	72	15	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	72	16	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	72	17	model	model	NN
work_27d5tr326feydgcx32t4nhl424	72	18	1	1	CD
work_27d5tr326feydgcx32t4nhl424	72	19	.	.	.
work_27d5tr326feydgcx32t4nhl424	73	1	Draw	draw	VB
work_27d5tr326feydgcx32t4nhl424	73	2	a	a	DT
work_27d5tr326feydgcx32t4nhl424	73	3	document	document	NN
work_27d5tr326feydgcx32t4nhl424	73	4	proportion	proportion	NN
work_27d5tr326feydgcx32t4nhl424	73	5	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	73	6	θm|α	θm|α	NNP
work_27d5tr326feydgcx32t4nhl424	73	7	∼	∼	NN
work_27d5tr326feydgcx32t4nhl424	73	8	Dir(α	Dir(α	NNP
work_27d5tr326feydgcx32t4nhl424	73	9	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	73	10	2	2	CD
work_27d5tr326feydgcx32t4nhl424	73	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	74	1	For	for	IN
work_27d5tr326feydgcx32t4nhl424	74	2	each	each	DT
work_27d5tr326feydgcx32t4nhl424	74	3	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	74	4	in	in	IN
work_27d5tr326feydgcx32t4nhl424	74	5	m	m	NNP
work_27d5tr326feydgcx32t4nhl424	74	6	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	74	7	a)draw	a)draw	CD
work_27d5tr326feydgcx32t4nhl424	74	8	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	74	9	assignment	assignment	NN
work_27d5tr326feydgcx32t4nhl424	74	10	zms|θ	zms|θ	XX
work_27d5tr326feydgcx32t4nhl424	74	11	∼	∼	NN
work_27d5tr326feydgcx32t4nhl424	74	12	Multi(θzmn	Multi(θzmn	NNP
work_27d5tr326feydgcx32t4nhl424	74	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	74	14	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	74	15	b)draw	b)draw	NN
work_27d5tr326feydgcx32t4nhl424	74	16	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	74	17	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	74	18	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	74	19	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	74	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	74	21	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	74	22	∼	∼	NNP
work_27d5tr326feydgcx32t4nhl424	74	23	p	p	NNP
work_27d5tr326feydgcx32t4nhl424	74	24	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	74	25	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	74	26	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	74	27	,	,	,
work_27d5tr326feydgcx32t4nhl424	74	28	η	η	LS
work_27d5tr326feydgcx32t4nhl424	74	29	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	74	30	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	74	31	c)for	c)for	IN
work_27d5tr326feydgcx32t4nhl424	74	32	each	each	DT
work_27d5tr326feydgcx32t4nhl424	74	33	word	word	NN
work_27d5tr326feydgcx32t4nhl424	74	34	wmsn	wmsn	NN
work_27d5tr326feydgcx32t4nhl424	74	35	in	in	IN
work_27d5tr326feydgcx32t4nhl424	74	36	current	current	JJ
work_27d5tr326feydgcx32t4nhl424	74	37	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	74	38	draw	draw	VBP
work_27d5tr326feydgcx32t4nhl424	74	39	wmsn|zms	wmsn|zms	NNP
work_27d5tr326feydgcx32t4nhl424	74	40	,	,	,
work_27d5tr326feydgcx32t4nhl424	74	41	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	74	42	∼	∼	NN
work_27d5tr326feydgcx32t4nhl424	74	43	Multi(βzms	Multi(βzms	NNP
work_27d5tr326feydgcx32t4nhl424	74	44	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	74	45	Figure	figure	NN
work_27d5tr326feydgcx32t4nhl424	74	46	4	4	CD
work_27d5tr326feydgcx32t4nhl424	74	47	:	:	:
work_27d5tr326feydgcx32t4nhl424	74	48	generation	generation	NN
work_27d5tr326feydgcx32t4nhl424	74	49	process	process	NN
work_27d5tr326feydgcx32t4nhl424	74	50	for	for	IN
work_27d5tr326feydgcx32t4nhl424	74	51	S	s	NN
work_27d5tr326feydgcx32t4nhl424	74	52	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	74	53	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	74	54	zms	zms	NNP
work_27d5tr326feydgcx32t4nhl424	74	55	for	for	IN
work_27d5tr326feydgcx32t4nhl424	74	56	any	any	DT
work_27d5tr326feydgcx32t4nhl424	74	57	n	n	NN
work_27d5tr326feydgcx32t4nhl424	74	58	∈	∈	JJ
work_27d5tr326feydgcx32t4nhl424	74	59	[	[	-LRB-
work_27d5tr326feydgcx32t4nhl424	74	60	1,Nms	1,nms	NN
work_27d5tr326feydgcx32t4nhl424	74	61	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	74	62	.	.	.
work_27d5tr326feydgcx32t4nhl424	75	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	75	2	generative	generative	JJ
work_27d5tr326feydgcx32t4nhl424	75	3	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	75	4	of	of	IN
work_27d5tr326feydgcx32t4nhl424	75	5	S	s	NN
work_27d5tr326feydgcx32t4nhl424	75	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	75	7	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	75	8	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	75	9	shown	show	VBN
work_27d5tr326feydgcx32t4nhl424	75	10	in	in	IN
work_27d5tr326feydgcx32t4nhl424	75	11	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	75	12	3	3	CD
work_27d5tr326feydgcx32t4nhl424	75	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	75	14	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	75	15	4	4	CD
work_27d5tr326feydgcx32t4nhl424	75	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	76	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	76	2	can	can	MD
work_27d5tr326feydgcx32t4nhl424	76	3	see	see	VB
work_27d5tr326feydgcx32t4nhl424	76	4	that	that	IN
work_27d5tr326feydgcx32t4nhl424	76	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	76	6	generative	generative	JJ
work_27d5tr326feydgcx32t4nhl424	76	7	process	process	NN
work_27d5tr326feydgcx32t4nhl424	76	8	involves	involve	VBZ
work_27d5tr326feydgcx32t4nhl424	76	9	not	not	RB
work_27d5tr326feydgcx32t4nhl424	76	10	only	only	RB
work_27d5tr326feydgcx32t4nhl424	76	11	the	the	DT
work_27d5tr326feydgcx32t4nhl424	76	12	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	76	13	within	within	IN
work_27d5tr326feydgcx32t4nhl424	76	14	current	current	JJ
work_27d5tr326feydgcx32t4nhl424	76	15	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	76	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	76	17	but	but	CC
work_27d5tr326feydgcx32t4nhl424	76	18	also	also	RB
work_27d5tr326feydgcx32t4nhl424	76	19	a	a	DT
work_27d5tr326feydgcx32t4nhl424	76	20	series	series	NN
work_27d5tr326feydgcx32t4nhl424	76	21	of	of	IN
work_27d5tr326feydgcx32t4nhl424	76	22	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	76	23	features	feature	VBZ
work_27d5tr326feydgcx32t4nhl424	76	24	.	.	.
work_27d5tr326feydgcx32t4nhl424	77	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	77	2	mixture	mixture	NN
work_27d5tr326feydgcx32t4nhl424	77	3	weights	weight	VBZ
work_27d5tr326feydgcx32t4nhl424	77	4	over	over	IN
work_27d5tr326feydgcx32t4nhl424	77	5	fea-	fea-	JJ
work_27d5tr326feydgcx32t4nhl424	77	6	tures	ture	NNS
work_27d5tr326feydgcx32t4nhl424	77	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	77	8	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	77	9	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	77	10	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	77	11	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	77	12	defined	define	VBN
work_27d5tr326feydgcx32t4nhl424	77	13	with	with	IN
work_27d5tr326feydgcx32t4nhl424	77	14	a	a	DT
work_27d5tr326feydgcx32t4nhl424	77	15	generalized	generalize	VBN
work_27d5tr326feydgcx32t4nhl424	77	16	lin-	lin-	NN
work_27d5tr326feydgcx32t4nhl424	77	17	ear	ear	NN
work_27d5tr326feydgcx32t4nhl424	77	18	model	model	NN
work_27d5tr326feydgcx32t4nhl424	77	19	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	77	20	GLM	GLM	NNP
work_27d5tr326feydgcx32t4nhl424	77	21	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	77	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	78	1	p	p	NNP
work_27d5tr326feydgcx32t4nhl424	78	2	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	78	3	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	78	4	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	78	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	78	6	η	η	NN
work_27d5tr326feydgcx32t4nhl424	78	7	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	78	8	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	78	9	exp(zTmsη	exp(zTmsη	NNP
work_27d5tr326feydgcx32t4nhl424	78	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	78	11	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	78	12	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	78	13	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	78	14	zms	zms	NNP
work_27d5tr326feydgcx32t4nhl424	78	15	exp(zTmsη	exp(zTmsη	NNP
work_27d5tr326feydgcx32t4nhl424	78	16	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	78	17	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	78	18	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	78	19	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	78	20	1	1	CD
work_27d5tr326feydgcx32t4nhl424	78	21	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	78	22	Here	here	RB
work_27d5tr326feydgcx32t4nhl424	78	23	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	78	24	assume	assume	VBP
work_27d5tr326feydgcx32t4nhl424	78	25	that	that	IN
work_27d5tr326feydgcx32t4nhl424	78	26	each	each	DT
work_27d5tr326feydgcx32t4nhl424	78	27	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	78	28	has	have	VBZ
work_27d5tr326feydgcx32t4nhl424	78	29	T	T	NNP
work_27d5tr326feydgcx32t4nhl424	78	30	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	78	31	and	and	CC
work_27d5tr326feydgcx32t4nhl424	78	32	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	78	33	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	78	34	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	78	35	a	a	DT
work_27d5tr326feydgcx32t4nhl424	78	36	T	t	NN
work_27d5tr326feydgcx32t4nhl424	78	37	×	×	NN
work_27d5tr326feydgcx32t4nhl424	78	38	1	1	CD
work_27d5tr326feydgcx32t4nhl424	78	39	dimensional	dimensional	JJ
work_27d5tr326feydgcx32t4nhl424	78	40	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	78	41	.	.	.
work_27d5tr326feydgcx32t4nhl424	79	1	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	79	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	79	3	a	a	DT
work_27d5tr326feydgcx32t4nhl424	79	4	K	k	NN
work_27d5tr326feydgcx32t4nhl424	79	5	×	×	NN
work_27d5tr326feydgcx32t4nhl424	79	6	T	T	NNP
work_27d5tr326feydgcx32t4nhl424	79	7	weight	weight	NN
work_27d5tr326feydgcx32t4nhl424	79	8	matrix	matrix	NN
work_27d5tr326feydgcx32t4nhl424	79	9	of	of	IN
work_27d5tr326feydgcx32t4nhl424	79	10	each	each	DT
work_27d5tr326feydgcx32t4nhl424	79	11	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	79	12	upon	upon	IN
work_27d5tr326feydgcx32t4nhl424	79	13	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	79	14	,	,	,
work_27d5tr326feydgcx32t4nhl424	79	15	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	79	16	largely	largely	RB
work_27d5tr326feydgcx32t4nhl424	79	17	controls	control	VBZ
work_27d5tr326feydgcx32t4nhl424	79	18	the	the	DT
work_27d5tr326feydgcx32t4nhl424	79	19	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	79	20	generation	generation	NN
work_27d5tr326feydgcx32t4nhl424	79	21	proce-	proce-	XX
work_27d5tr326feydgcx32t4nhl424	79	22	dure	dure	NN
work_27d5tr326feydgcx32t4nhl424	79	23	.	.	.
work_27d5tr326feydgcx32t4nhl424	80	1	Unlike	unlike	IN
work_27d5tr326feydgcx32t4nhl424	80	2	s	s	NNP
work_27d5tr326feydgcx32t4nhl424	80	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	80	4	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	80	5	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	80	6	η	η	NN
work_27d5tr326feydgcx32t4nhl424	80	7	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	80	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	80	9	latent	latent	NN
work_27d5tr326feydgcx32t4nhl424	80	10	variable	variable	NN
work_27d5tr326feydgcx32t4nhl424	80	11	esti-	esti-	NN
work_27d5tr326feydgcx32t4nhl424	80	12	mated	mate	VBN
work_27d5tr326feydgcx32t4nhl424	80	13	from	from	IN
work_27d5tr326feydgcx32t4nhl424	80	14	the	the	DT
work_27d5tr326feydgcx32t4nhl424	80	15	maximum	maximum	JJ
work_27d5tr326feydgcx32t4nhl424	80	16	likelihood	likelihood	NN
work_27d5tr326feydgcx32t4nhl424	80	17	estimation	estimation	NN
work_27d5tr326feydgcx32t4nhl424	80	18	al-	al-	NNP
work_27d5tr326feydgcx32t4nhl424	80	19	gorithm	gorithm	NNP
work_27d5tr326feydgcx32t4nhl424	80	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	80	21	in	in	IN
work_27d5tr326feydgcx32t4nhl424	80	22	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	80	23	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	80	24	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	80	25	the	the	DT
work_27d5tr326feydgcx32t4nhl424	80	26	value	value	NN
work_27d5tr326feydgcx32t4nhl424	80	27	of	of	IN
work_27d5tr326feydgcx32t4nhl424	80	28	η	η	NN
work_27d5tr326feydgcx32t4nhl424	80	29	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	80	30	trained	train	VBN
work_27d5tr326feydgcx32t4nhl424	80	31	through	through	IN
work_27d5tr326feydgcx32t4nhl424	80	32	a	a	DT
work_27d5tr326feydgcx32t4nhl424	80	33	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	80	34	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	80	35	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	80	36	will	will	MD
work_27d5tr326feydgcx32t4nhl424	80	37	be	be	VB
work_27d5tr326feydgcx32t4nhl424	80	38	illustrated	illustrate	VBN
work_27d5tr326feydgcx32t4nhl424	80	39	in	in	IN
work_27d5tr326feydgcx32t4nhl424	80	40	detail	detail	NN
work_27d5tr326feydgcx32t4nhl424	80	41	in	in	IN
work_27d5tr326feydgcx32t4nhl424	80	42	Section	section	NN
work_27d5tr326feydgcx32t4nhl424	80	43	3	3	CD
work_27d5tr326feydgcx32t4nhl424	80	44	.	.	.
work_27d5tr326feydgcx32t4nhl424	81	1	3.4	3.4	CD
work_27d5tr326feydgcx32t4nhl424	81	2	Posterior	Posterior	NNP
work_27d5tr326feydgcx32t4nhl424	81	3	Inference	Inference	NNP
work_27d5tr326feydgcx32t4nhl424	81	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	81	5	Estimation	Estimation	NNP
work_27d5tr326feydgcx32t4nhl424	81	6	Given	give	VBN
work_27d5tr326feydgcx32t4nhl424	81	7	a	a	DT
work_27d5tr326feydgcx32t4nhl424	81	8	document	document	NN
work_27d5tr326feydgcx32t4nhl424	81	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	81	10	labels	label	NNS
work_27d5tr326feydgcx32t4nhl424	81	11	for	for	IN
work_27d5tr326feydgcx32t4nhl424	81	12	each	each	DT
work_27d5tr326feydgcx32t4nhl424	81	13	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	81	14	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	15	the	the	DT
work_27d5tr326feydgcx32t4nhl424	81	16	posterior	posterior	NN
work_27d5tr326feydgcx32t4nhl424	81	17	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	81	18	of	of	IN
work_27d5tr326feydgcx32t4nhl424	81	19	the	the	DT
work_27d5tr326feydgcx32t4nhl424	81	20	latent	latent	NN
work_27d5tr326feydgcx32t4nhl424	81	21	variables	variable	NNS
work_27d5tr326feydgcx32t4nhl424	81	22	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	81	23	:	:	:
work_27d5tr326feydgcx32t4nhl424	81	24	p(θ	p(θ	NNP
work_27d5tr326feydgcx32t4nhl424	81	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	26	z1	z1	NN
work_27d5tr326feydgcx32t4nhl424	81	27	:	:	:
work_27d5tr326feydgcx32t4nhl424	81	28	N|w1	N|w1	NNS
work_27d5tr326feydgcx32t4nhl424	81	29	:	:	:
work_27d5tr326feydgcx32t4nhl424	81	30	N	n	CD
work_27d5tr326feydgcx32t4nhl424	81	31	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	32	Y	Y	NNP
work_27d5tr326feydgcx32t4nhl424	81	33	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	34	α	α	NNP
work_27d5tr326feydgcx32t4nhl424	81	35	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	36	β1	β1	NNP
work_27d5tr326feydgcx32t4nhl424	81	37	:	:	:
work_27d5tr326feydgcx32t4nhl424	81	38	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	81	39	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	40	η	η	NN
work_27d5tr326feydgcx32t4nhl424	81	41	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	42	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	81	43	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	81	44	m	m	NNP
work_27d5tr326feydgcx32t4nhl424	81	45	p(θm|α	p(θm|α	NNP
work_27d5tr326feydgcx32t4nhl424	81	46	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	47	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	81	48	s[p(zms|θm)p	s[p(zms|θm)p	NFP
work_27d5tr326feydgcx32t4nhl424	81	49	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	81	50	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	81	51	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	81	52	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	53	η	η	LS
work_27d5tr326feydgcx32t4nhl424	81	54	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	55	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	81	56	n	n	NNP
work_27d5tr326feydgcx32t4nhl424	81	57	p(wmsn|zmsn	p(wmsn|zmsn	NNP
work_27d5tr326feydgcx32t4nhl424	81	58	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	59	βzmsn	βzmsn	NNP
work_27d5tr326feydgcx32t4nhl424	81	60	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	61	∫	∫	:
work_27d5tr326feydgcx32t4nhl424	81	62	dθp(θm|α	dθp(θm|α	NNP
work_27d5tr326feydgcx32t4nhl424	81	63	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	64	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	81	65	z	z	NN
work_27d5tr326feydgcx32t4nhl424	81	66	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	81	67	s[p(zms|θm)p	s[p(zms|θm)p	NFP
work_27d5tr326feydgcx32t4nhl424	81	68	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	81	69	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	81	70	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	81	71	,	,	,
work_27d5tr326feydgcx32t4nhl424	81	72	η	η	LS
work_27d5tr326feydgcx32t4nhl424	81	73	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	74	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	81	75	n	n	NNP
work_27d5tr326feydgcx32t4nhl424	81	76	p(wmsn|βzmsn	p(wmsn|βzmsn	NNP
work_27d5tr326feydgcx32t4nhl424	81	77	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	78	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	79	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	81	80	2	2	LS
work_27d5tr326feydgcx32t4nhl424	81	81	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	81	82	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	81	83	.	.	.
work_27d5tr326feydgcx32t4nhl424	82	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	82	2	2	2	LS
work_27d5tr326feydgcx32t4nhl424	82	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	82	4	can	can	MD
work_27d5tr326feydgcx32t4nhl424	82	5	not	not	RB
work_27d5tr326feydgcx32t4nhl424	82	6	be	be	VB
work_27d5tr326feydgcx32t4nhl424	82	7	efficiently	efficiently	RB
work_27d5tr326feydgcx32t4nhl424	82	8	computed	compute	VBN
work_27d5tr326feydgcx32t4nhl424	82	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	83	1	By	by	IN
work_27d5tr326feydgcx32t4nhl424	83	2	applying	apply	VBG
work_27d5tr326feydgcx32t4nhl424	83	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	83	4	Jensens	Jensens	NNP
work_27d5tr326feydgcx32t4nhl424	83	5	inequality	inequality	NN
work_27d5tr326feydgcx32t4nhl424	83	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	7	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	83	8	obtain	obtain	VBP
work_27d5tr326feydgcx32t4nhl424	83	9	a	a	DT
work_27d5tr326feydgcx32t4nhl424	83	10	lower	low	JJR
work_27d5tr326feydgcx32t4nhl424	83	11	bound	bound	NN
work_27d5tr326feydgcx32t4nhl424	83	12	of	of	IN
work_27d5tr326feydgcx32t4nhl424	83	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	83	14	log	log	NN
work_27d5tr326feydgcx32t4nhl424	83	15	likelihood	likelihood	NN
work_27d5tr326feydgcx32t4nhl424	83	16	of	of	IN
work_27d5tr326feydgcx32t4nhl424	83	17	document	document	NN
work_27d5tr326feydgcx32t4nhl424	83	18	p(θ	p(θ	NNP
work_27d5tr326feydgcx32t4nhl424	83	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	20	z1	z1	NNP
work_27d5tr326feydgcx32t4nhl424	83	21	:	:	:
work_27d5tr326feydgcx32t4nhl424	83	22	N|w1	N|w1	NNS
work_27d5tr326feydgcx32t4nhl424	83	23	:	:	:
work_27d5tr326feydgcx32t4nhl424	83	24	N	n	CD
work_27d5tr326feydgcx32t4nhl424	83	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	26	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	83	27	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	83	28	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	29	α	α	NNP
work_27d5tr326feydgcx32t4nhl424	83	30	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	31	β1	β1	NNP
work_27d5tr326feydgcx32t4nhl424	83	32	:	:	:
work_27d5tr326feydgcx32t4nhl424	83	33	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	83	34	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	35	η	η	LS
work_27d5tr326feydgcx32t4nhl424	83	36	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	37	≥	≥	NN
work_27d5tr326feydgcx32t4nhl424	83	38	L	L	NNP
work_27d5tr326feydgcx32t4nhl424	83	39	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	40	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	83	41	L	l	NN
work_27d5tr326feydgcx32t4nhl424	83	42	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	83	43	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	83	44	ms	ms	XX
work_27d5tr326feydgcx32t4nhl424	83	45	E[logP(zms|θ	E[logP(zms|θ	NNP
work_27d5tr326feydgcx32t4nhl424	83	46	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	47	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	48	+	+	CC
work_27d5tr326feydgcx32t4nhl424	83	49	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	83	50	ms	ms	NNP
work_27d5tr326feydgcx32t4nhl424	83	51	E[logP	E[logP	NNP
work_27d5tr326feydgcx32t4nhl424	83	52	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	83	53	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	83	54	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	83	55	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	56	η)]+	η)]+	NNP
work_27d5tr326feydgcx32t4nhl424	83	57	∑	∑	NNP
work_27d5tr326feydgcx32t4nhl424	83	58	m	m	NNP
work_27d5tr326feydgcx32t4nhl424	83	59	E[logP	E[logP	NNP
work_27d5tr326feydgcx32t4nhl424	83	60	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	83	61	θ|α	θ|α	CD
work_27d5tr326feydgcx32t4nhl424	83	62	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	63	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	64	+	+	$
work_27d5tr326feydgcx32t4nhl424	83	65	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	83	66	msn	msn	NNP
work_27d5tr326feydgcx32t4nhl424	83	67	E[logP(wmsn|zms	E[logP(wmsn|zms	NNP
work_27d5tr326feydgcx32t4nhl424	83	68	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	69	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	83	70	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	71	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	72	+	+	CC
work_27d5tr326feydgcx32t4nhl424	83	73	H(q	H(q	NNS
work_27d5tr326feydgcx32t4nhl424	83	74	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	75	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	83	76	3	3	CD
work_27d5tr326feydgcx32t4nhl424	83	77	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	78	91	91	CD
work_27d5tr326feydgcx32t4nhl424	83	79	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	83	80	H(q	h(q	NN
work_27d5tr326feydgcx32t4nhl424	83	81	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	82	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	83	83	−E[logq	−E[logq	NNP
work_27d5tr326feydgcx32t4nhl424	83	84	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	85	and	and	CC
work_27d5tr326feydgcx32t4nhl424	83	86	it	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	83	87	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	83	88	the	the	DT
work_27d5tr326feydgcx32t4nhl424	83	89	entropy	entropy	JJ
work_27d5tr326feydgcx32t4nhl424	83	90	of	of	IN
work_27d5tr326feydgcx32t4nhl424	83	91	variational	variational	JJ
work_27d5tr326feydgcx32t4nhl424	83	92	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	83	93	q	q	NNP
work_27d5tr326feydgcx32t4nhl424	83	94	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	83	95	defined	define	VBN
work_27d5tr326feydgcx32t4nhl424	83	96	as	as	IN
work_27d5tr326feydgcx32t4nhl424	83	97	q(θ	q(θ	NNP
work_27d5tr326feydgcx32t4nhl424	83	98	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	99	z|γ	z|γ	CD
work_27d5tr326feydgcx32t4nhl424	83	100	,	,	,
work_27d5tr326feydgcx32t4nhl424	83	101	φ	φ	NN
work_27d5tr326feydgcx32t4nhl424	83	102	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	103	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	83	104	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	83	105	mk	mk	NNP
work_27d5tr326feydgcx32t4nhl424	83	106	q(θm|γ	q(θm|γ	NNP
work_27d5tr326feydgcx32t4nhl424	83	107	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	108	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	83	109	sn	sn	NNP
work_27d5tr326feydgcx32t4nhl424	83	110	q(zmsn|φms	q(zmsn|φms	NNP
work_27d5tr326feydgcx32t4nhl424	83	111	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	112	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	83	113	4	4	LS
work_27d5tr326feydgcx32t4nhl424	83	114	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	83	115	here	here	RB
work_27d5tr326feydgcx32t4nhl424	83	116	γ	γ	PDT
work_27d5tr326feydgcx32t4nhl424	83	117	a	a	DT
work_27d5tr326feydgcx32t4nhl424	83	118	K	k	NN
work_27d5tr326feydgcx32t4nhl424	83	119	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	83	120	dimensional	dimensional	JJ
work_27d5tr326feydgcx32t4nhl424	83	121	Dirichlet	Dirichlet	NNP
work_27d5tr326feydgcx32t4nhl424	83	122	parameter	parameter	NN
work_27d5tr326feydgcx32t4nhl424	83	123	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	83	124	and	and	CC
work_27d5tr326feydgcx32t4nhl424	83	125	multinomial	multinomial	JJ
work_27d5tr326feydgcx32t4nhl424	83	126	parameters	parameter	NNS
work_27d5tr326feydgcx32t4nhl424	83	127	.	.	.
work_27d5tr326feydgcx32t4nhl424	84	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	84	2	first	first	JJ
work_27d5tr326feydgcx32t4nhl424	84	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	84	4	third	third	JJ
work_27d5tr326feydgcx32t4nhl424	84	5	and	and	CC
work_27d5tr326feydgcx32t4nhl424	84	6	forth	forth	RB
work_27d5tr326feydgcx32t4nhl424	84	7	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	84	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	84	9	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	84	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	85	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	85	2	3	3	LS
work_27d5tr326feydgcx32t4nhl424	85	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	85	4	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	85	5	identical	identical	JJ
work_27d5tr326feydgcx32t4nhl424	85	6	to	to	IN
work_27d5tr326feydgcx32t4nhl424	85	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	85	8	corre-	corre-	NN
work_27d5tr326feydgcx32t4nhl424	85	9	sponding	sponde	VBG
work_27d5tr326feydgcx32t4nhl424	85	10	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	85	11	for	for	IN
work_27d5tr326feydgcx32t4nhl424	85	12	unsupervised	unsupervised	JJ
work_27d5tr326feydgcx32t4nhl424	85	13	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	85	14	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	85	15	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	85	16	et	et	FW
work_27d5tr326feydgcx32t4nhl424	85	17	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	85	18	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	85	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	85	20	2003	2003	CD
work_27d5tr326feydgcx32t4nhl424	85	21	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	85	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	86	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	86	2	second	second	JJ
work_27d5tr326feydgcx32t4nhl424	86	3	term	term	NN
work_27d5tr326feydgcx32t4nhl424	86	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	86	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	86	6	expectation	expectation	NN
work_27d5tr326feydgcx32t4nhl424	86	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	86	8	log	log	NN
work_27d5tr326feydgcx32t4nhl424	86	9	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	86	10	of	of	IN
work_27d5tr326feydgcx32t4nhl424	86	11	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	86	12	given	give	VBN
work_27d5tr326feydgcx32t4nhl424	86	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	86	14	latent	latent	NN
work_27d5tr326feydgcx32t4nhl424	86	15	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	86	16	assign-	assign-	NN
work_27d5tr326feydgcx32t4nhl424	86	17	ments	ment	NNS
work_27d5tr326feydgcx32t4nhl424	86	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	87	1	E[logP	E[logP	NNP
work_27d5tr326feydgcx32t4nhl424	87	2	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	87	3	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	87	4	Yms|zms	Yms|zms	NNP
work_27d5tr326feydgcx32t4nhl424	87	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	87	6	η	η	LS
work_27d5tr326feydgcx32t4nhl424	87	7	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	87	8	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	87	9	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	87	10	E(zms	e(zms	ADD
work_27d5tr326feydgcx32t4nhl424	87	11	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	87	12	Tη	tη	NN
work_27d5tr326feydgcx32t4nhl424	87	13	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	87	14	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	87	15	−	−	NNP
work_27d5tr326feydgcx32t4nhl424	87	16	log	log	NN
work_27d5tr326feydgcx32t4nhl424	87	17	∑	∑	XX
work_27d5tr326feydgcx32t4nhl424	87	18	zms	zms	NNP
work_27d5tr326feydgcx32t4nhl424	87	19	exp(zTmsη	exp(zTmsη	NNP
work_27d5tr326feydgcx32t4nhl424	87	20	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	87	21	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	87	22	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	87	23	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	87	24	5	5	CD
work_27d5tr326feydgcx32t4nhl424	87	25	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	87	26	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	87	27	E(zms)T	E(zms)T	NNP
work_27d5tr326feydgcx32t4nhl424	87	28	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	87	29	a	a	DT
work_27d5tr326feydgcx32t4nhl424	87	30	1	1	CD
work_27d5tr326feydgcx32t4nhl424	87	31	×	×	NN
work_27d5tr326feydgcx32t4nhl424	87	32	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	87	33	dimensional	dimensional	JJ
work_27d5tr326feydgcx32t4nhl424	87	34	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	87	35	[	[	-LRB-
work_27d5tr326feydgcx32t4nhl424	87	36	φmsk	φmsk	NN
work_27d5tr326feydgcx32t4nhl424	87	37	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	87	38	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	87	39	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	87	40	K	K	NNP
work_27d5tr326feydgcx32t4nhl424	87	41	k=1	k=1	NNP
work_27d5tr326feydgcx32t4nhl424	87	42	.	.	.
work_27d5tr326feydgcx32t4nhl424	88	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	88	2	Bayes	Bayes	NNP
work_27d5tr326feydgcx32t4nhl424	88	3	estimation	estimation	NN
work_27d5tr326feydgcx32t4nhl424	88	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	88	5	S	s	NN
work_27d5tr326feydgcx32t4nhl424	88	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	88	7	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	88	8	model	model	NN
work_27d5tr326feydgcx32t4nhl424	88	9	can	can	MD
work_27d5tr326feydgcx32t4nhl424	88	10	be	be	VB
work_27d5tr326feydgcx32t4nhl424	88	11	got	get	VBN
work_27d5tr326feydgcx32t4nhl424	88	12	via	via	IN
work_27d5tr326feydgcx32t4nhl424	88	13	a	a	DT
work_27d5tr326feydgcx32t4nhl424	88	14	variational	variational	JJ
work_27d5tr326feydgcx32t4nhl424	88	15	EM	em	NN
work_27d5tr326feydgcx32t4nhl424	88	16	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	88	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	89	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	89	2	EM	EM	NNP
work_27d5tr326feydgcx32t4nhl424	89	3	procedure	procedure	NN
work_27d5tr326feydgcx32t4nhl424	89	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	89	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	89	6	lower	low	JJR
work_27d5tr326feydgcx32t4nhl424	89	7	bound	bind	VBN
work_27d5tr326feydgcx32t4nhl424	89	8	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	89	9	firstly	firstly	RB
work_27d5tr326feydgcx32t4nhl424	89	10	minimized	minimize	VBN
work_27d5tr326feydgcx32t4nhl424	89	11	with	with	IN
work_27d5tr326feydgcx32t4nhl424	89	12	respect	respect	NN
work_27d5tr326feydgcx32t4nhl424	89	13	to	to	IN
work_27d5tr326feydgcx32t4nhl424	89	14	γ	γ	NN
work_27d5tr326feydgcx32t4nhl424	89	15	and	and	CC
work_27d5tr326feydgcx32t4nhl424	89	16	φ	φ	NN
work_27d5tr326feydgcx32t4nhl424	89	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	89	18	and	and	CC
work_27d5tr326feydgcx32t4nhl424	89	19	then	then	RB
work_27d5tr326feydgcx32t4nhl424	89	20	minimized	minimize	VBN
work_27d5tr326feydgcx32t4nhl424	89	21	with	with	IN
work_27d5tr326feydgcx32t4nhl424	89	22	α	α	NNP
work_27d5tr326feydgcx32t4nhl424	89	23	and	and	CC
work_27d5tr326feydgcx32t4nhl424	89	24	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	89	25	by	by	IN
work_27d5tr326feydgcx32t4nhl424	89	26	fixing	fix	VBG
work_27d5tr326feydgcx32t4nhl424	89	27	γ	γ	NNP
work_27d5tr326feydgcx32t4nhl424	89	28	and	and	CC
work_27d5tr326feydgcx32t4nhl424	89	29	φ	φ	NN
work_27d5tr326feydgcx32t4nhl424	89	30	.	.	.
work_27d5tr326feydgcx32t4nhl424	89	31	E	e	NN
work_27d5tr326feydgcx32t4nhl424	89	32	-	-	NN
work_27d5tr326feydgcx32t4nhl424	89	33	step	step	NN
work_27d5tr326feydgcx32t4nhl424	89	34	:	:	:
work_27d5tr326feydgcx32t4nhl424	89	35	The	the	DT
work_27d5tr326feydgcx32t4nhl424	89	36	updating	updating	NN
work_27d5tr326feydgcx32t4nhl424	89	37	of	of	IN
work_27d5tr326feydgcx32t4nhl424	89	38	Dirichlet	Dirichlet	NNP
work_27d5tr326feydgcx32t4nhl424	89	39	parameter	parameter	NN
work_27d5tr326feydgcx32t4nhl424	89	40	γ	γ	NNP
work_27d5tr326feydgcx32t4nhl424	89	41	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	89	42	identical	identical	JJ
work_27d5tr326feydgcx32t4nhl424	89	43	to	to	IN
work_27d5tr326feydgcx32t4nhl424	89	44	that	that	DT
work_27d5tr326feydgcx32t4nhl424	89	45	of	of	IN
work_27d5tr326feydgcx32t4nhl424	89	46	unsupervised	unsupervised	JJ
work_27d5tr326feydgcx32t4nhl424	89	47	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	89	48	,	,	,
work_27d5tr326feydgcx32t4nhl424	89	49	and	and	CC
work_27d5tr326feydgcx32t4nhl424	89	50	does	do	VBZ
work_27d5tr326feydgcx32t4nhl424	89	51	not	not	RB
work_27d5tr326feydgcx32t4nhl424	89	52	involve	involve	VB
work_27d5tr326feydgcx32t4nhl424	89	53	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	89	54	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	89	55	−−→	−−→	NNP
work_27d5tr326feydgcx32t4nhl424	89	56	Yms	Yms	NNP
work_27d5tr326feydgcx32t4nhl424	89	57	.	.	.
work_27d5tr326feydgcx32t4nhl424	90	1	γnewm	γnewm	NNP
work_27d5tr326feydgcx32t4nhl424	90	2	←	←	NNP
work_27d5tr326feydgcx32t4nhl424	90	3	α	α	NNP
work_27d5tr326feydgcx32t4nhl424	90	4	+	+	CC
work_27d5tr326feydgcx32t4nhl424	90	5	∑	∑	CD
work_27d5tr326feydgcx32t4nhl424	90	6	s∈m	s∈m	NN
work_27d5tr326feydgcx32t4nhl424	90	7	φs	φs	NNP
work_27d5tr326feydgcx32t4nhl424	90	8	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	90	9	6	6	CD
work_27d5tr326feydgcx32t4nhl424	90	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	11	φnewsk	φnewsk	NN
work_27d5tr326feydgcx32t4nhl424	90	12	∝	∝	NN
work_27d5tr326feydgcx32t4nhl424	90	13	exp{E[logθm|γ	exp{E[logθm|γ	NNP
work_27d5tr326feydgcx32t4nhl424	90	14	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	15	+	+	CC
work_27d5tr326feydgcx32t4nhl424	90	16	Nms∑	Nms∑	NNP
work_27d5tr326feydgcx32t4nhl424	90	17	n=1	n=1	NNP
work_27d5tr326feydgcx32t4nhl424	90	18	E[log(wmsn|β1	e[log(wmsn|β1	NN
work_27d5tr326feydgcx32t4nhl424	90	19	:	:	:
work_27d5tr326feydgcx32t4nhl424	90	20	K)]+	K)]+	NNP
work_27d5tr326feydgcx32t4nhl424	90	21	T∑	T∑	NNP
work_27d5tr326feydgcx32t4nhl424	90	22	t=1	t=1	VBZ
work_27d5tr326feydgcx32t4nhl424	90	23	ηktYst	ηktyst	NN
work_27d5tr326feydgcx32t4nhl424	90	24	}	}	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	25	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	90	26	exp[Ψ(γmk	exp[ψ(γmk	XX
work_27d5tr326feydgcx32t4nhl424	90	27	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	28	−	−	NNP
work_27d5tr326feydgcx32t4nhl424	90	29	Ψ	Ψ	NNP
work_27d5tr326feydgcx32t4nhl424	90	30	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	90	31	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	90	32	k=1	k=1	VBD
work_27d5tr326feydgcx32t4nhl424	90	33	γmk	γmk	NNP
work_27d5tr326feydgcx32t4nhl424	90	34	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	35	+	+	NFP
work_27d5tr326feydgcx32t4nhl424	90	36	T∑	T∑	NNP
work_27d5tr326feydgcx32t4nhl424	90	37	t=1	t=1	VBZ
work_27d5tr326feydgcx32t4nhl424	90	38	ηktYst	ηktyst	NN
work_27d5tr326feydgcx32t4nhl424	90	39	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	40	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	90	41	7	7	LS
work_27d5tr326feydgcx32t4nhl424	90	42	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	43	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	90	44	Ψ	ψ	NN
work_27d5tr326feydgcx32t4nhl424	90	45	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	90	46	·	·	NFP
work_27d5tr326feydgcx32t4nhl424	90	47	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	90	48	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	90	49	the	the	DT
work_27d5tr326feydgcx32t4nhl424	90	50	log	log	NN
work_27d5tr326feydgcx32t4nhl424	90	51	Γ	γ	IN
work_27d5tr326feydgcx32t4nhl424	90	52	function	function	NN
work_27d5tr326feydgcx32t4nhl424	90	53	.	.	.
work_27d5tr326feydgcx32t4nhl424	91	1	ms	ms	NNP
work_27d5tr326feydgcx32t4nhl424	91	2	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	91	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	4	document	document	NN
work_27d5tr326feydgcx32t4nhl424	91	5	that	that	IN
work_27d5tr326feydgcx32t4nhl424	91	6	current	current	JJ
work_27d5tr326feydgcx32t4nhl424	91	7	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	91	8	comes	come	VBZ
work_27d5tr326feydgcx32t4nhl424	91	9	from	from	IN
work_27d5tr326feydgcx32t4nhl424	91	10	and	and	CC
work_27d5tr326feydgcx32t4nhl424	91	11	Yst	Yst	NNP
work_27d5tr326feydgcx32t4nhl424	91	12	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	91	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	14	tth	tth	NN
work_27d5tr326feydgcx32t4nhl424	91	15	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	91	16	of	of	IN
work_27d5tr326feydgcx32t4nhl424	91	17	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	91	18	s.	s.	NNP
work_27d5tr326feydgcx32t4nhl424	91	19	M	M	NNP
work_27d5tr326feydgcx32t4nhl424	91	20	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	91	21	step	step	NN
work_27d5tr326feydgcx32t4nhl424	91	22	:	:	:
work_27d5tr326feydgcx32t4nhl424	91	23	The	the	DT
work_27d5tr326feydgcx32t4nhl424	91	24	M	M	NNP
work_27d5tr326feydgcx32t4nhl424	91	25	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	91	26	step	step	NN
work_27d5tr326feydgcx32t4nhl424	91	27	for	for	IN
work_27d5tr326feydgcx32t4nhl424	91	28	updating	update	VBG
work_27d5tr326feydgcx32t4nhl424	91	29	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	91	30	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	91	31	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	32	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	91	33	as	as	IN
work_27d5tr326feydgcx32t4nhl424	91	34	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	35	pro-	pro-	NN
work_27d5tr326feydgcx32t4nhl424	91	36	cedure	cedure	NN
work_27d5tr326feydgcx32t4nhl424	91	37	in	in	IN
work_27d5tr326feydgcx32t4nhl424	91	38	unsupervised	unsupervised	JJ
work_27d5tr326feydgcx32t4nhl424	91	39	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	91	40	,	,	,
work_27d5tr326feydgcx32t4nhl424	91	41	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	91	42	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	43	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	91	44	of	of	IN
work_27d5tr326feydgcx32t4nhl424	91	45	a	a	DT
work_27d5tr326feydgcx32t4nhl424	91	46	word	word	NN
work_27d5tr326feydgcx32t4nhl424	91	47	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	91	48	from	from	IN
work_27d5tr326feydgcx32t4nhl424	91	49	a	a	DT
work_27d5tr326feydgcx32t4nhl424	91	50	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	91	51	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	91	52	proportional	proportional	JJ
work_27d5tr326feydgcx32t4nhl424	91	53	to	to	IN
work_27d5tr326feydgcx32t4nhl424	91	54	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	55	number	number	NN
work_27d5tr326feydgcx32t4nhl424	91	56	of	of	IN
work_27d5tr326feydgcx32t4nhl424	91	57	times	time	NNS
work_27d5tr326feydgcx32t4nhl424	91	58	this	this	DT
work_27d5tr326feydgcx32t4nhl424	91	59	word	word	NN
work_27d5tr326feydgcx32t4nhl424	91	60	assigned	assign	VBN
work_27d5tr326feydgcx32t4nhl424	91	61	to	to	IN
work_27d5tr326feydgcx32t4nhl424	91	62	the	the	DT
work_27d5tr326feydgcx32t4nhl424	91	63	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	91	64	.	.	.
work_27d5tr326feydgcx32t4nhl424	92	1	βnewkw	βnewkw	NNP
work_27d5tr326feydgcx32t4nhl424	92	2	=	=	``
work_27d5tr326feydgcx32t4nhl424	92	3	M∑	M∑	NNP
work_27d5tr326feydgcx32t4nhl424	92	4	m=1	m=1	NN
work_27d5tr326feydgcx32t4nhl424	92	5	Nm∑	nm∑	RB
work_27d5tr326feydgcx32t4nhl424	92	6	s=1	s=1	NN
work_27d5tr326feydgcx32t4nhl424	92	7	Nms∑	Nms∑	NNP
work_27d5tr326feydgcx32t4nhl424	92	8	n=1	n=1	NNP
work_27d5tr326feydgcx32t4nhl424	92	9	1(wmsn	1(wmsn	VBZ
work_27d5tr326feydgcx32t4nhl424	92	10	=	=	-RRB-
work_27d5tr326feydgcx32t4nhl424	92	11	w)φ	w)φ	NNP
work_27d5tr326feydgcx32t4nhl424	92	12	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	92	13	ms	ms	NN
work_27d5tr326feydgcx32t4nhl424	92	14	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	92	15	8)	8)	CD
work_27d5tr326feydgcx32t4nhl424	92	16	4	4	CD
work_27d5tr326feydgcx32t4nhl424	92	17	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	92	18	Approach	Approach	NNP
work_27d5tr326feydgcx32t4nhl424	92	19	4.1	4.1	CD
work_27d5tr326feydgcx32t4nhl424	92	20	Learning	learn	VBG
work_27d5tr326feydgcx32t4nhl424	92	21	In	in	IN
work_27d5tr326feydgcx32t4nhl424	92	22	this	this	DT
work_27d5tr326feydgcx32t4nhl424	92	23	subsection	subsection	NN
work_27d5tr326feydgcx32t4nhl424	92	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	92	25	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	92	26	describe	describe	VBP
work_27d5tr326feydgcx32t4nhl424	92	27	how	how	WRB
work_27d5tr326feydgcx32t4nhl424	92	28	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	92	29	learn	learn	VBP
work_27d5tr326feydgcx32t4nhl424	92	30	the	the	DT
work_27d5tr326feydgcx32t4nhl424	92	31	fea-	fea-	NNP
work_27d5tr326feydgcx32t4nhl424	92	32	ture	ture	NN
work_27d5tr326feydgcx32t4nhl424	92	33	weight	weight	NN
work_27d5tr326feydgcx32t4nhl424	92	34	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	92	35	in	in	IN
work_27d5tr326feydgcx32t4nhl424	92	36	a	a	DT
work_27d5tr326feydgcx32t4nhl424	92	37	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	92	38	way	way	NN
work_27d5tr326feydgcx32t4nhl424	92	39	.	.	.
work_27d5tr326feydgcx32t4nhl424	93	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	93	2	learning	learn	VBG
work_27d5tr326feydgcx32t4nhl424	93	3	pro-	pro-	NN
work_27d5tr326feydgcx32t4nhl424	93	4	cess	cess	NN
work_27d5tr326feydgcx32t4nhl424	93	5	of	of	IN
work_27d5tr326feydgcx32t4nhl424	93	6	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	93	7	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	93	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	93	9	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	93	10	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	93	11	combined	combine	VBN
work_27d5tr326feydgcx32t4nhl424	93	12	with	with	IN
work_27d5tr326feydgcx32t4nhl424	93	13	variational	variational	JJ
work_27d5tr326feydgcx32t4nhl424	93	14	inference	inference	NN
work_27d5tr326feydgcx32t4nhl424	93	15	of	of	IN
work_27d5tr326feydgcx32t4nhl424	93	16	S	s	NN
work_27d5tr326feydgcx32t4nhl424	93	17	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	93	18	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	93	19	.	.	.
work_27d5tr326feydgcx32t4nhl424	94	1	Given	give	VBN
work_27d5tr326feydgcx32t4nhl424	94	2	a	a	DT
work_27d5tr326feydgcx32t4nhl424	94	3	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	94	4	de-	de-	NN
work_27d5tr326feydgcx32t4nhl424	94	5	scription	scription	NN
work_27d5tr326feydgcx32t4nhl424	94	6	Q1	Q1	NNP
work_27d5tr326feydgcx32t4nhl424	94	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	94	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	94	9	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	94	10	of	of	IN
work_27d5tr326feydgcx32t4nhl424	94	11	training	training	NN
work_27d5tr326feydgcx32t4nhl424	94	12	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	94	13	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	94	14	from	from	IN
work_27d5tr326feydgcx32t4nhl424	94	15	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	94	16	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	94	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	94	18	human	human	JJ
work_27d5tr326feydgcx32t4nhl424	94	19	assessors	assessor	NNS
work_27d5tr326feydgcx32t4nhl424	94	20	assign	assign	VBP
work_27d5tr326feydgcx32t4nhl424	94	21	a	a	DT
work_27d5tr326feydgcx32t4nhl424	94	22	score	score	NN
work_27d5tr326feydgcx32t4nhl424	94	23	v(v	v(v	NNP
work_27d5tr326feydgcx32t4nhl424	94	24	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	94	25	−2,−1	−2,−1	NNP
work_27d5tr326feydgcx32t4nhl424	94	26	,	,	,
work_27d5tr326feydgcx32t4nhl424	94	27	0	0	CD
work_27d5tr326feydgcx32t4nhl424	94	28	,	,	,
work_27d5tr326feydgcx32t4nhl424	94	29	1	1	CD
work_27d5tr326feydgcx32t4nhl424	94	30	,	,	,
work_27d5tr326feydgcx32t4nhl424	94	31	1	1	CD
work_27d5tr326feydgcx32t4nhl424	94	32	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	94	33	to	to	IN
work_27d5tr326feydgcx32t4nhl424	94	34	each	each	DT
work_27d5tr326feydgcx32t4nhl424	94	35	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	94	36	in	in	IN
work_27d5tr326feydgcx32t4nhl424	94	37	S.	S.	NNP
work_27d5tr326feydgcx32t4nhl424	94	38	The	the	DT
work_27d5tr326feydgcx32t4nhl424	94	39	score	score	NN
work_27d5tr326feydgcx32t4nhl424	94	40	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	94	41	an	an	DT
work_27d5tr326feydgcx32t4nhl424	94	42	integer	integer	NN
work_27d5tr326feydgcx32t4nhl424	94	43	between	between	IN
work_27d5tr326feydgcx32t4nhl424	94	44	−2	−2	NNP
work_27d5tr326feydgcx32t4nhl424	94	45	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	94	46	the	the	DT
work_27d5tr326feydgcx32t4nhl424	94	47	least	least	RBS
work_27d5tr326feydgcx32t4nhl424	94	48	desired	desire	VBN
work_27d5tr326feydgcx32t4nhl424	94	49	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	94	50	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	94	51	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	94	52	and	and	CC
work_27d5tr326feydgcx32t4nhl424	94	53	+2	+2	NNP
work_27d5tr326feydgcx32t4nhl424	94	54	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	94	55	the	the	DT
work_27d5tr326feydgcx32t4nhl424	94	56	most	most	RBS
work_27d5tr326feydgcx32t4nhl424	94	57	desired	desire	VBN
work_27d5tr326feydgcx32t4nhl424	94	58	sum-	sum-	NNP
work_27d5tr326feydgcx32t4nhl424	94	59	mary	mary	NNP
work_27d5tr326feydgcx32t4nhl424	94	60	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	94	61	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	94	62	,	,	,
work_27d5tr326feydgcx32t4nhl424	94	63	and	and	CC
work_27d5tr326feydgcx32t4nhl424	94	64	score	score	NN
work_27d5tr326feydgcx32t4nhl424	94	65	0	0	CD
work_27d5tr326feydgcx32t4nhl424	94	66	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	94	67	neutral	neutral	JJ
work_27d5tr326feydgcx32t4nhl424	94	68	atti-	atti-	NN
work_27d5tr326feydgcx32t4nhl424	94	69	tude	tude	NN
work_27d5tr326feydgcx32t4nhl424	94	70	.	.	.
work_27d5tr326feydgcx32t4nhl424	95	1	Ov	ov	IN
work_27d5tr326feydgcx32t4nhl424	95	2	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	95	3	{	{	-LRB-
work_27d5tr326feydgcx32t4nhl424	95	4	ov1,ov2	ov1,ov2	NNP
work_27d5tr326feydgcx32t4nhl424	95	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	95	6	...	...	:
work_27d5tr326feydgcx32t4nhl424	95	7	,	,	,
work_27d5tr326feydgcx32t4nhl424	95	8	vvk}(v	vvk}(v	NNP
work_27d5tr326feydgcx32t4nhl424	95	9	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	95	10	−2,−1	−2,−1	NNP
work_27d5tr326feydgcx32t4nhl424	95	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	95	12	0	0	CD
work_27d5tr326feydgcx32t4nhl424	95	13	,	,	,
work_27d5tr326feydgcx32t4nhl424	95	14	1	1	CD
work_27d5tr326feydgcx32t4nhl424	95	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	95	16	2	2	CD
work_27d5tr326feydgcx32t4nhl424	95	17	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	95	18	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	95	19	the	the	DT
work_27d5tr326feydgcx32t4nhl424	95	20	set	set	NN
work_27d5tr326feydgcx32t4nhl424	95	21	containing	contain	VBG
work_27d5tr326feydgcx32t4nhl424	95	22	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	95	23	with	with	IN
work_27d5tr326feydgcx32t4nhl424	95	24	score	score	NN
work_27d5tr326feydgcx32t4nhl424	95	25	v.	v.	CC
work_27d5tr326feydgcx32t4nhl424	95	26	Let	let	VB
work_27d5tr326feydgcx32t4nhl424	95	27	φQk	φQk	NNS
work_27d5tr326feydgcx32t4nhl424	95	28	denote	denote	VB
work_27d5tr326feydgcx32t4nhl424	95	29	the	the	DT
work_27d5tr326feydgcx32t4nhl424	95	30	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	95	31	that	that	IN
work_27d5tr326feydgcx32t4nhl424	95	32	query	query	NN
work_27d5tr326feydgcx32t4nhl424	95	33	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	95	34	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	95	35	from	from	IN
work_27d5tr326feydgcx32t4nhl424	95	36	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	95	37	k.	k.	NN
work_27d5tr326feydgcx32t4nhl424	95	38	Since	since	IN
work_27d5tr326feydgcx32t4nhl424	95	39	query	query	NN
work_27d5tr326feydgcx32t4nhl424	95	40	does	do	VBZ
work_27d5tr326feydgcx32t4nhl424	95	41	not	not	RB
work_27d5tr326feydgcx32t4nhl424	95	42	belong	belong	VB
work_27d5tr326feydgcx32t4nhl424	95	43	to	to	IN
work_27d5tr326feydgcx32t4nhl424	95	44	any	any	DT
work_27d5tr326feydgcx32t4nhl424	95	45	docu-	docu-	JJ
work_27d5tr326feydgcx32t4nhl424	95	46	ment	ment	NN
work_27d5tr326feydgcx32t4nhl424	95	47	,	,	,
work_27d5tr326feydgcx32t4nhl424	95	48	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	95	49	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	95	50	the	the	DT
work_27d5tr326feydgcx32t4nhl424	95	51	following	follow	VBG
work_27d5tr326feydgcx32t4nhl424	95	52	strategy	strategy	NN
work_27d5tr326feydgcx32t4nhl424	95	53	to	to	IN
work_27d5tr326feydgcx32t4nhl424	95	54	leverage	leverage	NN
work_27d5tr326feydgcx32t4nhl424	95	55	φQk	φQk	NNS
work_27d5tr326feydgcx32t4nhl424	95	56	φQk	φQk	NNS
work_27d5tr326feydgcx32t4nhl424	95	57	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	95	58	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	95	59	w∈Q	w∈q	ADD
work_27d5tr326feydgcx32t4nhl424	95	60	βkw	βkw	FW
work_27d5tr326feydgcx32t4nhl424	95	61	·	·	NFP
work_27d5tr326feydgcx32t4nhl424	95	62	1	1	CD
work_27d5tr326feydgcx32t4nhl424	95	63	M	M	NNP
work_27d5tr326feydgcx32t4nhl424	95	64	M∑	M∑	NNP
work_27d5tr326feydgcx32t4nhl424	95	65	m=1	m=1	NN
work_27d5tr326feydgcx32t4nhl424	95	66	exp[Ψ(γmk)−Ψ	exp[ψ(γmk)−ψ	NN
work_27d5tr326feydgcx32t4nhl424	95	67	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	95	68	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	95	69	k=1	k=1	VBD
work_27d5tr326feydgcx32t4nhl424	95	70	γmk	γmk	NNP
work_27d5tr326feydgcx32t4nhl424	95	71	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	95	72	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	95	73	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	95	74	9	9	CD
work_27d5tr326feydgcx32t4nhl424	95	75	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	95	76	In	in	IN
work_27d5tr326feydgcx32t4nhl424	95	77	Equ	Equ	NNP
work_27d5tr326feydgcx32t4nhl424	95	78	.	.	.
work_27d5tr326feydgcx32t4nhl424	96	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	96	2	9	9	CD
work_27d5tr326feydgcx32t4nhl424	96	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	96	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	96	5	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	96	6	w∈Q	w∈Q	NNP
work_27d5tr326feydgcx32t4nhl424	96	7	βkw	βkw	FW
work_27d5tr326feydgcx32t4nhl424	96	8	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	96	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	96	10	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	96	11	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	96	12	all	all	DT
work_27d5tr326feydgcx32t4nhl424	96	13	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	96	14	in	in	IN
work_27d5tr326feydgcx32t4nhl424	96	15	query	query	NN
work_27d5tr326feydgcx32t4nhl424	96	16	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	96	17	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	96	18	from	from	IN
work_27d5tr326feydgcx32t4nhl424	96	19	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	96	20	k	k	NNS
work_27d5tr326feydgcx32t4nhl424	96	21	and	and	CC
work_27d5tr326feydgcx32t4nhl424	96	22	1	1	CD
work_27d5tr326feydgcx32t4nhl424	96	23	M	M	NNP
work_27d5tr326feydgcx32t4nhl424	96	24	∑M	∑M	NNP
work_27d5tr326feydgcx32t4nhl424	96	25	m=1	m=1	NNP
work_27d5tr326feydgcx32t4nhl424	96	26	exp[Ψ(γmk)−Ψ	exp[Ψ(γmk)−Ψ	NNP
work_27d5tr326feydgcx32t4nhl424	96	27	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	96	28	∑K	∑K	NNP
work_27d5tr326feydgcx32t4nhl424	96	29	k=1	k=1	NNP
work_27d5tr326feydgcx32t4nhl424	96	30	γmk	γmk	NNP
work_27d5tr326feydgcx32t4nhl424	96	31	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	96	32	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	96	33	can	can	MD
work_27d5tr326feydgcx32t4nhl424	96	34	be	be	VB
work_27d5tr326feydgcx32t4nhl424	96	35	seen	see	VBN
work_27d5tr326feydgcx32t4nhl424	96	36	as	as	IN
work_27d5tr326feydgcx32t4nhl424	96	37	the	the	DT
work_27d5tr326feydgcx32t4nhl424	96	38	average	average	JJ
work_27d5tr326feydgcx32t4nhl424	96	39	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	96	40	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	96	41	all	all	DT
work_27d5tr326feydgcx32t4nhl424	96	42	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	96	43	in	in	IN
work_27d5tr326feydgcx32t4nhl424	96	44	the	the	DT
work_27d5tr326feydgcx32t4nhl424	96	45	corpus	corpus	NNP
work_27d5tr326feydgcx32t4nhl424	96	46	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	96	47	talking	talk	VBG
work_27d5tr326feydgcx32t4nhl424	96	48	about	about	IN
work_27d5tr326feydgcx32t4nhl424	96	49	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	96	50	k.	k.	NN
work_27d5tr326feydgcx32t4nhl424	96	51	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	96	52	.	.	.
work_27d5tr326feydgcx32t4nhl424	97	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	97	2	9	9	CD
work_27d5tr326feydgcx32t4nhl424	97	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	97	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	97	5	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	97	6	on	on	IN
work_27d5tr326feydgcx32t4nhl424	97	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	97	8	assumption	assumption	NN
work_27d5tr326feydgcx32t4nhl424	97	9	that	that	IN
work_27d5tr326feydgcx32t4nhl424	97	10	query	query	NNP
work_27d5tr326feydgcx32t4nhl424	97	11	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	97	12	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	97	13	relevant	relevant	JJ
work_27d5tr326feydgcx32t4nhl424	97	14	to	to	IN
work_27d5tr326feydgcx32t4nhl424	97	15	the	the	DT
work_27d5tr326feydgcx32t4nhl424	97	16	main	main	JJ
work_27d5tr326feydgcx32t4nhl424	97	17	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	97	18	discussed	discuss	VBN
work_27d5tr326feydgcx32t4nhl424	97	19	by	by	IN
work_27d5tr326feydgcx32t4nhl424	97	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	97	21	document	document	NN
work_27d5tr326feydgcx32t4nhl424	97	22	corpus	corpus	NNP
work_27d5tr326feydgcx32t4nhl424	97	23	.	.	.
work_27d5tr326feydgcx32t4nhl424	98	1	This	this	DT
work_27d5tr326feydgcx32t4nhl424	98	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	98	3	a	a	DT
work_27d5tr326feydgcx32t4nhl424	98	4	reasonable	reasonable	JJ
work_27d5tr326feydgcx32t4nhl424	98	5	assumption	assumption	NN
work_27d5tr326feydgcx32t4nhl424	98	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	98	7	most	most	RBS
work_27d5tr326feydgcx32t4nhl424	98	8	previous	previous	JJ
work_27d5tr326feydgcx32t4nhl424	98	9	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	98	10	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	98	11	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	98	12	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	98	13	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	98	14	on	on	IN
work_27d5tr326feydgcx32t4nhl424	98	15	similar	similar	JJ
work_27d5tr326feydgcx32t4nhl424	98	16	as-	as-	NN
work_27d5tr326feydgcx32t4nhl424	98	17	sumptions	sumption	NNS
work_27d5tr326feydgcx32t4nhl424	98	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	99	1	Next	next	RB
work_27d5tr326feydgcx32t4nhl424	99	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	99	3	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	99	4	define	define	VBP
work_27d5tr326feydgcx32t4nhl424	99	5	φOv	φOv	NNS
work_27d5tr326feydgcx32t4nhl424	99	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	99	7	k	k	NN
work_27d5tr326feydgcx32t4nhl424	99	8	for	for	IN
work_27d5tr326feydgcx32t4nhl424	99	9	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	99	10	set	set	VBN
work_27d5tr326feydgcx32t4nhl424	99	11	Ov	ov	RP
work_27d5tr326feydgcx32t4nhl424	99	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	99	13	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	99	14	can	can	MD
work_27d5tr326feydgcx32t4nhl424	99	15	be	be	VB
work_27d5tr326feydgcx32t4nhl424	99	16	interpreted	interpret	VBN
work_27d5tr326feydgcx32t4nhl424	99	17	as	as	IN
work_27d5tr326feydgcx32t4nhl424	99	18	the	the	DT
work_27d5tr326feydgcx32t4nhl424	99	19	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	99	20	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	99	21	all	all	DT
work_27d5tr326feydgcx32t4nhl424	99	22	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	99	23	tences	tence	NNS
work_27d5tr326feydgcx32t4nhl424	99	24	in	in	IN
work_27d5tr326feydgcx32t4nhl424	99	25	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	99	26	Ov	Ov	NNS
work_27d5tr326feydgcx32t4nhl424	99	27	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	99	28	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	99	29	from	from	IN
work_27d5tr326feydgcx32t4nhl424	99	30	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	99	31	k.	k.	NNP
work_27d5tr326feydgcx32t4nhl424	99	32	φOv	φOv	NNP
work_27d5tr326feydgcx32t4nhl424	99	33	,	,	,
work_27d5tr326feydgcx32t4nhl424	99	34	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	99	35	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	99	36	1	1	CD
work_27d5tr326feydgcx32t4nhl424	99	37	|Ov|	|ov|	CD
work_27d5tr326feydgcx32t4nhl424	99	38	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	99	39	s∈Ov	s∈Ov	NNP
work_27d5tr326feydgcx32t4nhl424	99	40	φsk	φsk	NNP
work_27d5tr326feydgcx32t4nhl424	99	41	,	,	,
work_27d5tr326feydgcx32t4nhl424	99	42	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	99	43	∈	∈	NNP
work_27d5tr326feydgcx32t4nhl424	99	44	[	[	-LRB-
work_27d5tr326feydgcx32t4nhl424	99	45	1,K],v	1,k],v	CD
work_27d5tr326feydgcx32t4nhl424	99	46	∈	∈	NNP
work_27d5tr326feydgcx32t4nhl424	99	47	[	[	-LRB-
work_27d5tr326feydgcx32t4nhl424	99	48	−2	−2	NNP
work_27d5tr326feydgcx32t4nhl424	99	49	,	,	,
work_27d5tr326feydgcx32t4nhl424	99	50	2	2	CD
work_27d5tr326feydgcx32t4nhl424	99	51	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	99	52	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	99	53	10	10	CD
work_27d5tr326feydgcx32t4nhl424	99	54	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	99	55	|Ov|	|Ov|	NNP
work_27d5tr326feydgcx32t4nhl424	99	56	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	99	57	the	the	DT
work_27d5tr326feydgcx32t4nhl424	99	58	number	number	NN
work_27d5tr326feydgcx32t4nhl424	99	59	of	of	IN
work_27d5tr326feydgcx32t4nhl424	99	60	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	99	61	in	in	IN
work_27d5tr326feydgcx32t4nhl424	99	62	set	set	NN
work_27d5tr326feydgcx32t4nhl424	99	63	Ov	ov	RP
work_27d5tr326feydgcx32t4nhl424	99	64	.	.	.
work_27d5tr326feydgcx32t4nhl424	100	1	In-	In-	NNP
work_27d5tr326feydgcx32t4nhl424	100	2	spired	spire	VBN
work_27d5tr326feydgcx32t4nhl424	100	3	by	by	IN
work_27d5tr326feydgcx32t4nhl424	100	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	100	5	idea	idea	NN
work_27d5tr326feydgcx32t4nhl424	100	6	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	100	7	desired	desire	VBD
work_27d5tr326feydgcx32t4nhl424	100	8	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	100	9	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	100	10	would	would	MD
work_27d5tr326feydgcx32t4nhl424	100	11	be	be	VB
work_27d5tr326feydgcx32t4nhl424	100	12	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	100	13	semantically	semantically	RB
work_27d5tr326feydgcx32t4nhl424	100	14	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	100	15	with	with	IN
work_27d5tr326feydgcx32t4nhl424	100	16	the	the	DT
work_27d5tr326feydgcx32t4nhl424	100	17	query	query	NN
work_27d5tr326feydgcx32t4nhl424	100	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	100	19	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	100	20	transform	transform	VBP
work_27d5tr326feydgcx32t4nhl424	100	21	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	100	22	of	of	IN
work_27d5tr326feydgcx32t4nhl424	100	23	finding	find	VBG
work_27d5tr326feydgcx32t4nhl424	100	24	optimum	optimum	JJ
work_27d5tr326feydgcx32t4nhl424	100	25	η	η	NN
work_27d5tr326feydgcx32t4nhl424	100	26	to	to	IN
work_27d5tr326feydgcx32t4nhl424	100	27	the	the	DT
work_27d5tr326feydgcx32t4nhl424	100	28	following	follow	VBG
work_27d5tr326feydgcx32t4nhl424	100	29	optimization	optimization	NN
work_27d5tr326feydgcx32t4nhl424	100	30	problem	problem	NN
work_27d5tr326feydgcx32t4nhl424	100	31	:	:	:
work_27d5tr326feydgcx32t4nhl424	100	32	minηL(η	minηL(η	NNP
work_27d5tr326feydgcx32t4nhl424	100	33	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	100	34	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	100	35	v=2∑	v=2∑	NN
work_27d5tr326feydgcx32t4nhl424	100	36	v=−2	v=−2	IN
work_27d5tr326feydgcx32t4nhl424	100	37	v	v	LS
work_27d5tr326feydgcx32t4nhl424	100	38	·	·	NFP
work_27d5tr326feydgcx32t4nhl424	100	39	KL(Ov||Q	KL(Ov||Q	NNP
work_27d5tr326feydgcx32t4nhl424	100	40	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	100	41	;	;	:
work_27d5tr326feydgcx32t4nhl424	100	42	T∑	T∑	NNP
work_27d5tr326feydgcx32t4nhl424	100	43	t=1	t=1	VBZ
work_27d5tr326feydgcx32t4nhl424	100	44	ηkt	ηkt	NN
work_27d5tr326feydgcx32t4nhl424	100	45	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	100	46	1	1	CD
work_27d5tr326feydgcx32t4nhl424	100	47	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	100	48	11	11	CD
work_27d5tr326feydgcx32t4nhl424	100	49	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	100	50	1We	1we	CD
work_27d5tr326feydgcx32t4nhl424	100	51	select	select	VB
work_27d5tr326feydgcx32t4nhl424	100	52	multiple	multiple	JJ
work_27d5tr326feydgcx32t4nhl424	100	53	queries	query	NNS
work_27d5tr326feydgcx32t4nhl424	100	54	and	and	CC
work_27d5tr326feydgcx32t4nhl424	100	55	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	100	56	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	100	57	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	100	58	for	for	IN
work_27d5tr326feydgcx32t4nhl424	100	59	training	train	VBG
work_27d5tr326feydgcx32t4nhl424	100	60	92	92	CD
work_27d5tr326feydgcx32t4nhl424	100	61	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	100	62	KL(Ov||Q	KL(Ov||Q	NNP
work_27d5tr326feydgcx32t4nhl424	100	63	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	100	64	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	100	65	the	the	DT
work_27d5tr326feydgcx32t4nhl424	100	66	Kullback	Kullback	NNP
work_27d5tr326feydgcx32t4nhl424	100	67	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	100	68	Leibler	Leibler	NNP
work_27d5tr326feydgcx32t4nhl424	100	69	diver-	diver-	NN
work_27d5tr326feydgcx32t4nhl424	100	70	gence	gence	NN
work_27d5tr326feydgcx32t4nhl424	100	71	between	between	IN
work_27d5tr326feydgcx32t4nhl424	100	72	the	the	DT
work_27d5tr326feydgcx32t4nhl424	100	73	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	100	74	and	and	CC
work_27d5tr326feydgcx32t4nhl424	100	75	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	100	76	set	set	VBN
work_27d5tr326feydgcx32t4nhl424	100	77	Ov	ov	RP
work_27d5tr326feydgcx32t4nhl424	100	78	as	as	IN
work_27d5tr326feydgcx32t4nhl424	100	79	shown	show	VBN
work_27d5tr326feydgcx32t4nhl424	100	80	in	in	IN
work_27d5tr326feydgcx32t4nhl424	100	81	Eqn.(12	eqn.(12	NN
work_27d5tr326feydgcx32t4nhl424	100	82	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	100	83	.	.	.
work_27d5tr326feydgcx32t4nhl424	101	1	KL(Ov||Q	KL(Ov||Q	NNP
work_27d5tr326feydgcx32t4nhl424	101	2	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	101	3	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	101	4	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	101	5	k=1	k=1	VBD
work_27d5tr326feydgcx32t4nhl424	101	6	φOvklog	φOvklog	NNP
work_27d5tr326feydgcx32t4nhl424	101	7	φOvk	φOvk	NNP
work_27d5tr326feydgcx32t4nhl424	101	8	φQk	φQk	NNS
work_27d5tr326feydgcx32t4nhl424	101	9	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	101	10	12	12	CD
work_27d5tr326feydgcx32t4nhl424	101	11	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	101	12	In	in	IN
work_27d5tr326feydgcx32t4nhl424	101	13	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	101	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	102	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	102	2	11	11	CD
work_27d5tr326feydgcx32t4nhl424	102	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	102	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	102	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	102	6	can	can	MD
work_27d5tr326feydgcx32t4nhl424	102	7	see	see	VB
work_27d5tr326feydgcx32t4nhl424	102	8	that	that	IN
work_27d5tr326feydgcx32t4nhl424	102	9	O2	o2	NN
work_27d5tr326feydgcx32t4nhl424	102	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	102	11	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	102	12	contain	contain	VBP
work_27d5tr326feydgcx32t4nhl424	102	13	de-	de-	RB
work_27d5tr326feydgcx32t4nhl424	102	14	sirable	sirable	JJ
work_27d5tr326feydgcx32t4nhl424	102	15	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	102	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	102	17	would	would	MD
work_27d5tr326feydgcx32t4nhl424	102	18	be	be	VB
work_27d5tr326feydgcx32t4nhl424	102	19	given	give	VBN
work_27d5tr326feydgcx32t4nhl424	102	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	102	21	largest	large	JJS
work_27d5tr326feydgcx32t4nhl424	102	22	penalty	penalty	NN
work_27d5tr326feydgcx32t4nhl424	102	23	for	for	IN
work_27d5tr326feydgcx32t4nhl424	102	24	its	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	102	25	KL	KL	NNP
work_27d5tr326feydgcx32t4nhl424	102	26	divergence	divergence	NN
work_27d5tr326feydgcx32t4nhl424	102	27	from	from	IN
work_27d5tr326feydgcx32t4nhl424	102	28	Query	Query	NNP
work_27d5tr326feydgcx32t4nhl424	102	29	.	.	.
work_27d5tr326feydgcx32t4nhl424	103	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	103	2	case	case	NN
work_27d5tr326feydgcx32t4nhl424	103	3	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	103	4	just	just	RB
work_27d5tr326feydgcx32t4nhl424	103	5	opposite	opposite	JJ
work_27d5tr326feydgcx32t4nhl424	103	6	for	for	IN
work_27d5tr326feydgcx32t4nhl424	103	7	undesired	undesired	JJ
work_27d5tr326feydgcx32t4nhl424	103	8	set	set	NN
work_27d5tr326feydgcx32t4nhl424	103	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	104	1	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	104	2	idea	idea	NN
work_27d5tr326feydgcx32t4nhl424	104	3	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	104	4	to	to	TO
work_27d5tr326feydgcx32t4nhl424	104	5	incorporate	incorporate	VB
work_27d5tr326feydgcx32t4nhl424	104	6	the	the	DT
work_27d5tr326feydgcx32t4nhl424	104	7	minimization	minimization	NN
work_27d5tr326feydgcx32t4nhl424	104	8	pro-	pro-	NN
work_27d5tr326feydgcx32t4nhl424	104	9	cess	cess	NN
work_27d5tr326feydgcx32t4nhl424	104	10	of	of	IN
work_27d5tr326feydgcx32t4nhl424	104	11	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	104	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	105	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	105	2	11	11	CD
work_27d5tr326feydgcx32t4nhl424	105	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	105	4	into	into	IN
work_27d5tr326feydgcx32t4nhl424	105	5	variational	variational	JJ
work_27d5tr326feydgcx32t4nhl424	105	6	inference	inference	NN
work_27d5tr326feydgcx32t4nhl424	105	7	process	process	NN
work_27d5tr326feydgcx32t4nhl424	105	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	105	9	S	s	NN
work_27d5tr326feydgcx32t4nhl424	105	10	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	105	11	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	105	12	model	model	NN
work_27d5tr326feydgcx32t4nhl424	105	13	.	.	.
work_27d5tr326feydgcx32t4nhl424	106	1	Here	here	RB
work_27d5tr326feydgcx32t4nhl424	106	2	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	106	3	perform	perform	VBP
work_27d5tr326feydgcx32t4nhl424	106	4	gradient	gradient	NN
work_27d5tr326feydgcx32t4nhl424	106	5	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	106	6	optimization	optimization	NN
work_27d5tr326feydgcx32t4nhl424	106	7	method	method	NN
work_27d5tr326feydgcx32t4nhl424	106	8	to	to	TO
work_27d5tr326feydgcx32t4nhl424	106	9	minimize	minimize	VB
work_27d5tr326feydgcx32t4nhl424	106	10	Eqn.(11	Eqn.(11	NNP
work_27d5tr326feydgcx32t4nhl424	106	11	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	106	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	107	1	Firstly	firstly	RB
work_27d5tr326feydgcx32t4nhl424	107	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	107	3	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	107	4	derive	derive	VBP
work_27d5tr326feydgcx32t4nhl424	107	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	107	6	gradient	gradient	NN
work_27d5tr326feydgcx32t4nhl424	107	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	107	8	L(η	L(η	NNP
work_27d5tr326feydgcx32t4nhl424	107	9	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	107	10	with	with	IN
work_27d5tr326feydgcx32t4nhl424	107	11	respect	respect	NN
work_27d5tr326feydgcx32t4nhl424	107	12	to	to	IN
work_27d5tr326feydgcx32t4nhl424	107	13	η	η	NN
work_27d5tr326feydgcx32t4nhl424	107	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	108	1	∂L(η	∂L(η	NNP
work_27d5tr326feydgcx32t4nhl424	108	2	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	108	3	ηxy	ηxy	NNP
work_27d5tr326feydgcx32t4nhl424	108	4	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	108	5	v=2∑	v=2∑	NN
work_27d5tr326feydgcx32t4nhl424	108	6	v=−2	v=−2	IN
work_27d5tr326feydgcx32t4nhl424	108	7	v	v	LS
work_27d5tr326feydgcx32t4nhl424	108	8	·	·	NFP
work_27d5tr326feydgcx32t4nhl424	108	9	∂KL(Qv||Q	∂KL(Qv||Q	.
work_27d5tr326feydgcx32t4nhl424	108	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	108	11	∂ηxy	∂ηxy	NFP
work_27d5tr326feydgcx32t4nhl424	108	12	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	108	13	13	13	CD
work_27d5tr326feydgcx32t4nhl424	108	14	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	108	15	∂KL(Qv||Q	∂KL(Qv||Q	.
work_27d5tr326feydgcx32t4nhl424	108	16	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	108	17	∂ηxy	∂ηxy	NFP
work_27d5tr326feydgcx32t4nhl424	108	18	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	108	19	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	108	20	k=1	k=1	VBD
work_27d5tr326feydgcx32t4nhl424	108	21	1	1	CD
work_27d5tr326feydgcx32t4nhl424	108	22	|Qv|	|Qv|	NNS
work_27d5tr326feydgcx32t4nhl424	108	23	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	108	24	1	1	CD
work_27d5tr326feydgcx32t4nhl424	108	25	+	+	CC
work_27d5tr326feydgcx32t4nhl424	108	26	log	log	NN
work_27d5tr326feydgcx32t4nhl424	108	27	∑	∑	NNS
work_27d5tr326feydgcx32t4nhl424	108	28	s∈Qv	s∈Qv	.
work_27d5tr326feydgcx32t4nhl424	108	29	|Qv|	|Qv|	NNP
work_27d5tr326feydgcx32t4nhl424	108	30	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	108	31	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	108	32	s∈Qv	s∈Qv	NNP
work_27d5tr326feydgcx32t4nhl424	108	33	∂φsk	∂φsk	NNP
work_27d5tr326feydgcx32t4nhl424	108	34	∂ηxy	∂ηxy	:
work_27d5tr326feydgcx32t4nhl424	108	35	−	−	NNP
work_27d5tr326feydgcx32t4nhl424	108	36	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	108	37	k=1	k=1	VBD
work_27d5tr326feydgcx32t4nhl424	108	38	1	1	CD
work_27d5tr326feydgcx32t4nhl424	108	39	|Qv|	|Qv|	NNP
work_27d5tr326feydgcx32t4nhl424	108	40	∑	∑	CD
work_27d5tr326feydgcx32t4nhl424	108	41	s∈Qv	s∈Qv	NNS
work_27d5tr326feydgcx32t4nhl424	108	42	∂Qsk	∂qsk	CD
work_27d5tr326feydgcx32t4nhl424	108	43	ηxy	ηxy	NNP
work_27d5tr326feydgcx32t4nhl424	108	44	−	−	NNP
work_27d5tr326feydgcx32t4nhl424	108	45	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	108	46	k=1	k=1	VBD
work_27d5tr326feydgcx32t4nhl424	108	47	1	1	CD
work_27d5tr326feydgcx32t4nhl424	108	48	Qv	Qv	NNP
work_27d5tr326feydgcx32t4nhl424	108	49	∑	∑	XX
work_27d5tr326feydgcx32t4nhl424	108	50	s∈Qvφsk	s∈Qvφsk	NNP
work_27d5tr326feydgcx32t4nhl424	108	51	φQk	φQk	NNP
work_27d5tr326feydgcx32t4nhl424	108	52	∂φsk	∂φsk	NNP
work_27d5tr326feydgcx32t4nhl424	108	53	∂ηxy	∂ηxy	NNP
work_27d5tr326feydgcx32t4nhl424	108	54	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	108	55	14	14	CD
work_27d5tr326feydgcx32t4nhl424	108	56	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	108	57	For	for	IN
work_27d5tr326feydgcx32t4nhl424	108	58	simplification	simplification	NN
work_27d5tr326feydgcx32t4nhl424	108	59	,	,	,
work_27d5tr326feydgcx32t4nhl424	108	60	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	108	61	regard	regard	VBP
work_27d5tr326feydgcx32t4nhl424	108	62	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	108	63	and	and	CC
work_27d5tr326feydgcx32t4nhl424	108	64	γ	γ	NNP
work_27d5tr326feydgcx32t4nhl424	108	65	as	as	RB
work_27d5tr326feydgcx32t4nhl424	108	66	constant	constant	JJ
work_27d5tr326feydgcx32t4nhl424	108	67	during	during	IN
work_27d5tr326feydgcx32t4nhl424	108	68	updating	update	VBG
work_27d5tr326feydgcx32t4nhl424	108	69	process	process	NN
work_27d5tr326feydgcx32t4nhl424	108	70	of	of	IN
work_27d5tr326feydgcx32t4nhl424	108	71	η	η	NN
work_27d5tr326feydgcx32t4nhl424	108	72	,	,	,
work_27d5tr326feydgcx32t4nhl424	108	73	so	so	CC
work_27d5tr326feydgcx32t4nhl424	108	74	∂φQk	∂φQk	NNP
work_27d5tr326feydgcx32t4nhl424	108	75	∂ηxy	∂ηxy	CD
work_27d5tr326feydgcx32t4nhl424	108	76	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	108	77	0.2	0.2	CD
work_27d5tr326feydgcx32t4nhl424	108	78	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	108	79	can	can	MD
work_27d5tr326feydgcx32t4nhl424	108	80	further	further	RB
work_27d5tr326feydgcx32t4nhl424	108	81	get	get	VB
work_27d5tr326feydgcx32t4nhl424	108	82	first	first	JJ
work_27d5tr326feydgcx32t4nhl424	108	83	derivative	derivative	NN
work_27d5tr326feydgcx32t4nhl424	108	84	for	for	IN
work_27d5tr326feydgcx32t4nhl424	108	85	each	each	DT
work_27d5tr326feydgcx32t4nhl424	108	86	labeled	label	VBN
work_27d5tr326feydgcx32t4nhl424	108	87	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	108	88	.	.	.
work_27d5tr326feydgcx32t4nhl424	109	1	∂φsk	∂φsk	NFP
work_27d5tr326feydgcx32t4nhl424	109	2	ηxy	ηxy	NNP
work_27d5tr326feydgcx32t4nhl424	109	3	∝	∝	NNP
work_27d5tr326feydgcx32t4nhl424	109	4			NNP
work_27d5tr326feydgcx32t4nhl424	109	5			SYM
work_27d5tr326feydgcx32t4nhl424	109	6			NNP
work_27d5tr326feydgcx32t4nhl424	109	7	Ysyexp[Ψ(γmsi	ysyexp[ψ(γmsi	CD
work_27d5tr326feydgcx32t4nhl424	109	8	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	109	9	−	−	NNP
work_27d5tr326feydgcx32t4nhl424	109	10	Ψ	Ψ	NNP
work_27d5tr326feydgcx32t4nhl424	109	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	109	12	K∑	K∑	NNP
work_27d5tr326feydgcx32t4nhl424	109	13	k=1	k=1	JJ
work_27d5tr326feydgcx32t4nhl424	109	14	γmsk	γmsk	NN
work_27d5tr326feydgcx32t4nhl424	109	15	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	109	16	+	+	NFP
work_27d5tr326feydgcx32t4nhl424	109	17	T∑	T∑	NNP
work_27d5tr326feydgcx32t4nhl424	109	18	t=1	t=1	VBZ
work_27d5tr326feydgcx32t4nhl424	109	19	ηktYsy	ηktysy	NN
work_27d5tr326feydgcx32t4nhl424	109	20	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	109	21	×	×	ADD
work_27d5tr326feydgcx32t4nhl424	109	22	∏	∏	NNP
work_27d5tr326feydgcx32t4nhl424	109	23	w∈s	w∈s	ADD
work_27d5tr326feydgcx32t4nhl424	109	24	βkw	βkw	NN
work_27d5tr326feydgcx32t4nhl424	109	25	if	if	IN
work_27d5tr326feydgcx32t4nhl424	109	26	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	109	27	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	109	28	x	x	SYM
work_27d5tr326feydgcx32t4nhl424	109	29	0	0	CD
work_27d5tr326feydgcx32t4nhl424	109	30	if	if	IN
work_27d5tr326feydgcx32t4nhl424	109	31	k	k	NN
work_27d5tr326feydgcx32t4nhl424	109	32	6=	6=	CD
work_27d5tr326feydgcx32t4nhl424	109	33	x	x	SYM
work_27d5tr326feydgcx32t4nhl424	109	34	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	109	35	15	15	CD
work_27d5tr326feydgcx32t4nhl424	109	36	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	109	37	4.2	4.2	CD
work_27d5tr326feydgcx32t4nhl424	109	38	Feature	Feature	NNP
work_27d5tr326feydgcx32t4nhl424	109	39	Space	Space	NNP
work_27d5tr326feydgcx32t4nhl424	109	40	Lots	lot	NNS
work_27d5tr326feydgcx32t4nhl424	109	41	of	of	IN
work_27d5tr326feydgcx32t4nhl424	109	42	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	109	43	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	109	44	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	109	45	proven	prove	VBN
work_27d5tr326feydgcx32t4nhl424	109	46	to	to	TO
work_27d5tr326feydgcx32t4nhl424	109	47	be	be	VB
work_27d5tr326feydgcx32t4nhl424	109	48	useful	useful	JJ
work_27d5tr326feydgcx32t4nhl424	109	49	for	for	IN
work_27d5tr326feydgcx32t4nhl424	109	50	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	109	51	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	109	52	Louis	Louis	NNP
work_27d5tr326feydgcx32t4nhl424	109	53	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	109	54	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	109	55	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	109	56	,	,	,
work_27d5tr326feydgcx32t4nhl424	109	57	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	109	58	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	109	59	.	.	.
work_27d5tr326feydgcx32t4nhl424	110	1	Here	here	RB
work_27d5tr326feydgcx32t4nhl424	110	2	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	110	3	dis-	dis-	VBP
work_27d5tr326feydgcx32t4nhl424	110	4	cuss	cuss	NN
work_27d5tr326feydgcx32t4nhl424	110	5	several	several	JJ
work_27d5tr326feydgcx32t4nhl424	110	6	types	type	NNS
work_27d5tr326feydgcx32t4nhl424	110	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	110	8	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	110	9	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	110	10	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	110	11	adopted	adopt	VBN
work_27d5tr326feydgcx32t4nhl424	110	12	in	in	IN
work_27d5tr326feydgcx32t4nhl424	110	13	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	110	14	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	110	15	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	110	16	model	model	NN
work_27d5tr326feydgcx32t4nhl424	110	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	111	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	111	2	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	111	3	values	value	NNS
work_27d5tr326feydgcx32t4nhl424	111	4	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	111	5	either	either	CC
work_27d5tr326feydgcx32t4nhl424	111	6	binary	binary	JJ
work_27d5tr326feydgcx32t4nhl424	111	7	or	or	CC
work_27d5tr326feydgcx32t4nhl424	111	8	normalized	normalized	JJ
work_27d5tr326feydgcx32t4nhl424	111	9	to	to	IN
work_27d5tr326feydgcx32t4nhl424	111	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	111	11	interval	interval	NN
work_27d5tr326feydgcx32t4nhl424	111	12	[	[	-LRB-
work_27d5tr326feydgcx32t4nhl424	111	13	0,1	0,1	NFP
work_27d5tr326feydgcx32t4nhl424	111	14	]	]	-RRB-
work_27d5tr326feydgcx32t4nhl424	111	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	112	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	112	2	following	follow	VBG
work_27d5tr326feydgcx32t4nhl424	112	3	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	112	4	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	112	5	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	112	6	in	in	IN
work_27d5tr326feydgcx32t4nhl424	112	7	S	s	NN
work_27d5tr326feydgcx32t4nhl424	112	8	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	112	9	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	112	10	:	:	:
work_27d5tr326feydgcx32t4nhl424	112	11	Cosine	Cosine	NNP
work_27d5tr326feydgcx32t4nhl424	112	12	Similarity	Similarity	NNP
work_27d5tr326feydgcx32t4nhl424	112	13	with	with	IN
work_27d5tr326feydgcx32t4nhl424	112	14	query	query	NN
work_27d5tr326feydgcx32t4nhl424	112	15	:	:	:
work_27d5tr326feydgcx32t4nhl424	112	16	Cosine	cosine	NN
work_27d5tr326feydgcx32t4nhl424	112	17	similarity	similarity	NN
work_27d5tr326feydgcx32t4nhl424	112	18	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	112	19	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	112	20	on	on	IN
work_27d5tr326feydgcx32t4nhl424	112	21	the	the	DT
work_27d5tr326feydgcx32t4nhl424	112	22	tf	tf	NNP
work_27d5tr326feydgcx32t4nhl424	112	23	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	112	24	idf	idf	NNP
work_27d5tr326feydgcx32t4nhl424	112	25	value	value	NN
work_27d5tr326feydgcx32t4nhl424	112	26	of	of	IN
work_27d5tr326feydgcx32t4nhl424	112	27	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	112	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	113	1	2This	2this	CD
work_27d5tr326feydgcx32t4nhl424	113	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	113	3	reasonable	reasonable	JJ
work_27d5tr326feydgcx32t4nhl424	113	4	because	because	IN
work_27d5tr326feydgcx32t4nhl424	113	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	113	6	influence	influence	NN
work_27d5tr326feydgcx32t4nhl424	113	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	113	8	γ	γ	NNP
work_27d5tr326feydgcx32t4nhl424	113	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	113	10	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	113	11	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	113	12	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	113	13	embodied	embody	VBN
work_27d5tr326feydgcx32t4nhl424	113	14	in	in	IN
work_27d5tr326feydgcx32t4nhl424	113	15	φ	φ	NN
work_27d5tr326feydgcx32t4nhl424	113	16	during	during	IN
work_27d5tr326feydgcx32t4nhl424	113	17	each	each	DT
work_27d5tr326feydgcx32t4nhl424	113	18	iteration	iteration	NN
work_27d5tr326feydgcx32t4nhl424	113	19	.	.	.
work_27d5tr326feydgcx32t4nhl424	114	1	Local	Local	NNP
work_27d5tr326feydgcx32t4nhl424	114	2	Inner	inner	JJ
work_27d5tr326feydgcx32t4nhl424	114	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	114	4	document	document	NN
work_27d5tr326feydgcx32t4nhl424	114	5	Degree	degree	NN
work_27d5tr326feydgcx32t4nhl424	114	6	Order	order	NN
work_27d5tr326feydgcx32t4nhl424	114	7	:	:	:
work_27d5tr326feydgcx32t4nhl424	114	8	Local	local	JJ
work_27d5tr326feydgcx32t4nhl424	114	9	Inner	inner	JJ
work_27d5tr326feydgcx32t4nhl424	114	10	document	document	NN
work_27d5tr326feydgcx32t4nhl424	114	11	Degree	Degree	NNP
work_27d5tr326feydgcx32t4nhl424	114	12	Order	order	NN
work_27d5tr326feydgcx32t4nhl424	114	13	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	114	14	a	a	DT
work_27d5tr326feydgcx32t4nhl424	114	15	binary	binary	JJ
work_27d5tr326feydgcx32t4nhl424	114	16	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	114	17	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	114	18	indicates	indicate	VBZ
work_27d5tr326feydgcx32t4nhl424	114	19	whether	whether	IN
work_27d5tr326feydgcx32t4nhl424	114	20	Inner	inner	JJ
work_27d5tr326feydgcx32t4nhl424	114	21	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	114	22	document	document	NN
work_27d5tr326feydgcx32t4nhl424	114	23	Degree	Degree	NNP
work_27d5tr326feydgcx32t4nhl424	114	24	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	114	25	IDD	IDD	NNP
work_27d5tr326feydgcx32t4nhl424	114	26	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	114	27	of	of	IN
work_27d5tr326feydgcx32t4nhl424	114	28	sentence	sentence	NNP
work_27d5tr326feydgcx32t4nhl424	114	29	s	s	POS
work_27d5tr326feydgcx32t4nhl424	114	30	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	114	31	the	the	DT
work_27d5tr326feydgcx32t4nhl424	114	32	largest	large	JJS
work_27d5tr326feydgcx32t4nhl424	114	33	among	among	IN
work_27d5tr326feydgcx32t4nhl424	114	34	its	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	114	35	neighbors	neighbor	NNS
work_27d5tr326feydgcx32t4nhl424	114	36	.	.	.
work_27d5tr326feydgcx32t4nhl424	115	1	IDD	IDD	NNP
work_27d5tr326feydgcx32t4nhl424	115	2	means	mean	VBZ
work_27d5tr326feydgcx32t4nhl424	115	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	115	4	edge	edge	NN
work_27d5tr326feydgcx32t4nhl424	115	5	number	number	NN
work_27d5tr326feydgcx32t4nhl424	115	6	between	between	IN
work_27d5tr326feydgcx32t4nhl424	115	7	s	s	NN
work_27d5tr326feydgcx32t4nhl424	115	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	115	9	other	other	JJ
work_27d5tr326feydgcx32t4nhl424	115	10	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	115	11	tences	tence	NNS
work_27d5tr326feydgcx32t4nhl424	115	12	in	in	IN
work_27d5tr326feydgcx32t4nhl424	115	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	115	14	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	115	15	document	document	NN
work_27d5tr326feydgcx32t4nhl424	115	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	116	1	Document	Document	NNP
work_27d5tr326feydgcx32t4nhl424	116	2	Specific	Specific	NNP
work_27d5tr326feydgcx32t4nhl424	116	3	Word	Word	NNP
work_27d5tr326feydgcx32t4nhl424	116	4	:	:	:
work_27d5tr326feydgcx32t4nhl424	116	5	1	1	CD
work_27d5tr326feydgcx32t4nhl424	116	6	if	if	IN
work_27d5tr326feydgcx32t4nhl424	116	7	a	a	DT
work_27d5tr326feydgcx32t4nhl424	116	8	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	116	9	contains	contain	VBZ
work_27d5tr326feydgcx32t4nhl424	116	10	document	document	NN
work_27d5tr326feydgcx32t4nhl424	116	11	specific	specific	JJ
work_27d5tr326feydgcx32t4nhl424	116	12	word	word	NN
work_27d5tr326feydgcx32t4nhl424	116	13	,	,	,
work_27d5tr326feydgcx32t4nhl424	116	14	0	0	CD
work_27d5tr326feydgcx32t4nhl424	116	15	otherwise	otherwise	RB
work_27d5tr326feydgcx32t4nhl424	116	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	117	1	Average	average	JJ
work_27d5tr326feydgcx32t4nhl424	117	2	Unigram	Unigram	NNP
work_27d5tr326feydgcx32t4nhl424	117	3	Probability	Probability	NNP
work_27d5tr326feydgcx32t4nhl424	117	4	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	117	5	Nenkova	Nenkova	NNP
work_27d5tr326feydgcx32t4nhl424	117	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	117	7	Van-	Van-	NNP
work_27d5tr326feydgcx32t4nhl424	117	8	derwende	derwende	NNP
work_27d5tr326feydgcx32t4nhl424	117	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	117	10	2005	2005	CD
work_27d5tr326feydgcx32t4nhl424	117	11	;	;	:
work_27d5tr326feydgcx32t4nhl424	117	12	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	117	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	117	14	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	117	15	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	117	16	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	117	17	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	117	18	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	117	19	:	:	:
work_27d5tr326feydgcx32t4nhl424	117	20	As	as	IN
work_27d5tr326feydgcx32t4nhl424	117	21	for	for	IN
work_27d5tr326feydgcx32t4nhl424	117	22	sentence	sentence	NNP
work_27d5tr326feydgcx32t4nhl424	117	23	s	s	POS
work_27d5tr326feydgcx32t4nhl424	117	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	117	25	p(s	p(s	NNP
work_27d5tr326feydgcx32t4nhl424	117	26	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	117	27	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	117	28	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	117	29	w∈s	w∈s	NNP
work_27d5tr326feydgcx32t4nhl424	117	30	1	1	CD
work_27d5tr326feydgcx32t4nhl424	117	31	|s|pD(w	|s|pD(w	NNP
work_27d5tr326feydgcx32t4nhl424	117	32	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	117	33	,	,	,
work_27d5tr326feydgcx32t4nhl424	117	34	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	117	35	pD(w	pd(w	NN
work_27d5tr326feydgcx32t4nhl424	117	36	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	117	37	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	117	38	the	the	DT
work_27d5tr326feydgcx32t4nhl424	117	39	observed	observed	JJ
work_27d5tr326feydgcx32t4nhl424	117	40	unigram	unigram	JJ
work_27d5tr326feydgcx32t4nhl424	117	41	probability	probability	NN
work_27d5tr326feydgcx32t4nhl424	117	42	in	in	IN
work_27d5tr326feydgcx32t4nhl424	117	43	document	document	NN
work_27d5tr326feydgcx32t4nhl424	117	44	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	117	45	.	.	.
work_27d5tr326feydgcx32t4nhl424	118	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	118	2	addition	addition	NN
work_27d5tr326feydgcx32t4nhl424	118	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	118	4	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	118	5	also	also	RB
work_27d5tr326feydgcx32t4nhl424	118	6	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	118	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	118	8	commonly	commonly	RB
work_27d5tr326feydgcx32t4nhl424	118	9	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	118	10	fea-	fea-	JJ
work_27d5tr326feydgcx32t4nhl424	118	11	tures	ture	NNS
work_27d5tr326feydgcx32t4nhl424	118	12	including	include	VBG
work_27d5tr326feydgcx32t4nhl424	118	13	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	118	14	position	position	NN
work_27d5tr326feydgcx32t4nhl424	118	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	118	16	paragraph	paragraph	NN
work_27d5tr326feydgcx32t4nhl424	118	17	po-	po-	NN
work_27d5tr326feydgcx32t4nhl424	118	18	sition	sition	NN
work_27d5tr326feydgcx32t4nhl424	118	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	118	20	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	118	21	length	length	NN
work_27d5tr326feydgcx32t4nhl424	118	22	and	and	CC
work_27d5tr326feydgcx32t4nhl424	118	23	sentence	sentence	NNP
work_27d5tr326feydgcx32t4nhl424	118	24	bigram	bigram	NNP
work_27d5tr326feydgcx32t4nhl424	118	25	fre-	fre-	NNP
work_27d5tr326feydgcx32t4nhl424	118	26	quency	quency	NN
work_27d5tr326feydgcx32t4nhl424	118	27	.	.	.
work_27d5tr326feydgcx32t4nhl424	119	1	E	e	NN
work_27d5tr326feydgcx32t4nhl424	119	2	-	-	NN
work_27d5tr326feydgcx32t4nhl424	119	3	step	step	NN
work_27d5tr326feydgcx32t4nhl424	119	4	initialize	initialize	NN
work_27d5tr326feydgcx32t4nhl424	119	5	φ0sk	φ0sk	NNP
work_27d5tr326feydgcx32t4nhl424	119	6	:	:	:
work_27d5tr326feydgcx32t4nhl424	119	7	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	119	8	1	1	CD
work_27d5tr326feydgcx32t4nhl424	119	9	/	/	SYM
work_27d5tr326feydgcx32t4nhl424	119	10	K	k	NN
work_27d5tr326feydgcx32t4nhl424	119	11	for	for	IN
work_27d5tr326feydgcx32t4nhl424	119	12	all	all	DT
work_27d5tr326feydgcx32t4nhl424	119	13	i	i	PRP
work_27d5tr326feydgcx32t4nhl424	119	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	119	15	s.	s.	NNP
work_27d5tr326feydgcx32t4nhl424	119	16	initialize	initialize	VB
work_27d5tr326feydgcx32t4nhl424	119	17	γmi	γmi	NNP
work_27d5tr326feydgcx32t4nhl424	119	18	:	:	:
work_27d5tr326feydgcx32t4nhl424	119	19	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	119	20	αmi	αmi	JJ
work_27d5tr326feydgcx32t4nhl424	119	21	+	+	SYM
work_27d5tr326feydgcx32t4nhl424	119	22	N)m	N)m	NNP
work_27d5tr326feydgcx32t4nhl424	119	23	/	/	SYM
work_27d5tr326feydgcx32t4nhl424	119	24	K	k	NN
work_27d5tr326feydgcx32t4nhl424	119	25	for	for	IN
work_27d5tr326feydgcx32t4nhl424	119	26	all	all	DT
work_27d5tr326feydgcx32t4nhl424	119	27	i.	i.	NNP
work_27d5tr326feydgcx32t4nhl424	119	28	initialize	initialize	VB
work_27d5tr326feydgcx32t4nhl424	119	29	ηkt	ηkt	NN
work_27d5tr326feydgcx32t4nhl424	119	30	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	119	31	0	0	NFP
work_27d5tr326feydgcx32t4nhl424	119	32	for	for	IN
work_27d5tr326feydgcx32t4nhl424	119	33	all	all	DT
work_27d5tr326feydgcx32t4nhl424	119	34	k	k	NN
work_27d5tr326feydgcx32t4nhl424	119	35	and	and	CC
work_27d5tr326feydgcx32t4nhl424	119	36	t.	t.	NN
work_27d5tr326feydgcx32t4nhl424	119	37	while	while	IN
work_27d5tr326feydgcx32t4nhl424	119	38	not	not	RB
work_27d5tr326feydgcx32t4nhl424	119	39	convergence	convergence	NN
work_27d5tr326feydgcx32t4nhl424	119	40	for	for	IN
work_27d5tr326feydgcx32t4nhl424	119	41	m	m	NN
work_27d5tr326feydgcx32t4nhl424	119	42	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	119	43	1	1	CD
work_27d5tr326feydgcx32t4nhl424	119	44	:	:	:
work_27d5tr326feydgcx32t4nhl424	119	45	M	m	NN
work_27d5tr326feydgcx32t4nhl424	119	46	update	update	NN
work_27d5tr326feydgcx32t4nhl424	119	47	γt+1	γt+1	NNP
work_27d5tr326feydgcx32t4nhl424	119	48	m	m	NNP
work_27d5tr326feydgcx32t4nhl424	119	49	according	accord	VBG
work_27d5tr326feydgcx32t4nhl424	119	50	to	to	IN
work_27d5tr326feydgcx32t4nhl424	119	51	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	119	52	.	.	.
work_27d5tr326feydgcx32t4nhl424	120	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	120	2	6	6	CD
work_27d5tr326feydgcx32t4nhl424	120	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	120	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	120	5	s	s	NN
work_27d5tr326feydgcx32t4nhl424	120	6	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	120	7	1	1	CD
work_27d5tr326feydgcx32t4nhl424	120	8	:	:	:
work_27d5tr326feydgcx32t4nhl424	120	9	Nm	nm	IN
work_27d5tr326feydgcx32t4nhl424	120	10	for	for	IN
work_27d5tr326feydgcx32t4nhl424	120	11	k	k	NN
work_27d5tr326feydgcx32t4nhl424	120	12	=	=	SYM
work_27d5tr326feydgcx32t4nhl424	120	13	1	1	CD
work_27d5tr326feydgcx32t4nhl424	120	14	:	:	:
work_27d5tr326feydgcx32t4nhl424	120	15	K	k	NN
work_27d5tr326feydgcx32t4nhl424	120	16	update	update	NN
work_27d5tr326feydgcx32t4nhl424	120	17	φt+1sk	φt+1sk	VB
work_27d5tr326feydgcx32t4nhl424	120	18	according	accord	VBG
work_27d5tr326feydgcx32t4nhl424	120	19	to	to	IN
work_27d5tr326feydgcx32t4nhl424	120	20	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	120	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	121	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	121	2	7	7	LS
work_27d5tr326feydgcx32t4nhl424	121	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	121	4	normalize	normalize	VB
work_27d5tr326feydgcx32t4nhl424	121	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	121	6	sum	sum	NN
work_27d5tr326feydgcx32t4nhl424	121	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	121	8	φt+1sk	φt+1sk	NNP
work_27d5tr326feydgcx32t4nhl424	121	9	to	to	IN
work_27d5tr326feydgcx32t4nhl424	121	10	1	1	CD
work_27d5tr326feydgcx32t4nhl424	121	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	122	1	Minimize	Minimize	NNP
work_27d5tr326feydgcx32t4nhl424	122	2	L(η	L(η	NNP
work_27d5tr326feydgcx32t4nhl424	122	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	122	4	according	accord	VBG
work_27d5tr326feydgcx32t4nhl424	122	5	to	to	IN
work_27d5tr326feydgcx32t4nhl424	122	6	Eqn.(11)-(15	Eqn.(11)-(15	NNP
work_27d5tr326feydgcx32t4nhl424	122	7	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	122	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	123	1	M	m	NN
work_27d5tr326feydgcx32t4nhl424	123	2	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	123	3	step	step	NN
work_27d5tr326feydgcx32t4nhl424	123	4	:	:	:
work_27d5tr326feydgcx32t4nhl424	123	5	update	update	VB
work_27d5tr326feydgcx32t4nhl424	123	6	β	β	NNP
work_27d5tr326feydgcx32t4nhl424	123	7	according	accord	VBG
work_27d5tr326feydgcx32t4nhl424	123	8	to	to	IN
work_27d5tr326feydgcx32t4nhl424	123	9	Eqn	Eqn	NNP
work_27d5tr326feydgcx32t4nhl424	123	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	124	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	124	2	8)	8)	CD
work_27d5tr326feydgcx32t4nhl424	124	3	Figure	figure	NN
work_27d5tr326feydgcx32t4nhl424	124	4	5	5	CD
work_27d5tr326feydgcx32t4nhl424	124	5	:	:	:
work_27d5tr326feydgcx32t4nhl424	124	6	Learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	124	7	process	process	NN
work_27d5tr326feydgcx32t4nhl424	124	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	124	9	η	η	NN
work_27d5tr326feydgcx32t4nhl424	124	10	in	in	IN
work_27d5tr326feydgcx32t4nhl424	124	11	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	124	12	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	124	13	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	124	14	4.3	4.3	CD
work_27d5tr326feydgcx32t4nhl424	124	15	Sentence	Sentence	NNP
work_27d5tr326feydgcx32t4nhl424	124	16	Selection	Selection	NNP
work_27d5tr326feydgcx32t4nhl424	124	17	Strategy	Strategy	NNP
work_27d5tr326feydgcx32t4nhl424	124	18	Next	Next	NNP
work_27d5tr326feydgcx32t4nhl424	124	19	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	124	20	explain	explain	VBP
work_27d5tr326feydgcx32t4nhl424	124	21	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	124	22	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	124	23	selection	selection	NN
work_27d5tr326feydgcx32t4nhl424	124	24	strategy	strategy	NN
work_27d5tr326feydgcx32t4nhl424	124	25	.	.	.
work_27d5tr326feydgcx32t4nhl424	125	1	Ac-	ac-	JJ
work_27d5tr326feydgcx32t4nhl424	125	2	cording	cord	VBG
work_27d5tr326feydgcx32t4nhl424	125	3	to	to	IN
work_27d5tr326feydgcx32t4nhl424	125	4	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	125	5	intuition	intuition	NN
work_27d5tr326feydgcx32t4nhl424	125	6	that	that	IN
work_27d5tr326feydgcx32t4nhl424	125	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	125	8	desired	desire	VBN
work_27d5tr326feydgcx32t4nhl424	125	9	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	125	10	should	should	MD
work_27d5tr326feydgcx32t4nhl424	125	11	have	have	VB
work_27d5tr326feydgcx32t4nhl424	125	12	a	a	DT
work_27d5tr326feydgcx32t4nhl424	125	13	small	small	JJ
work_27d5tr326feydgcx32t4nhl424	125	14	KL	KL	NNP
work_27d5tr326feydgcx32t4nhl424	125	15	divergence	divergence	NN
work_27d5tr326feydgcx32t4nhl424	125	16	with	with	IN
work_27d5tr326feydgcx32t4nhl424	125	17	query	query	NN
work_27d5tr326feydgcx32t4nhl424	125	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	125	19	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	125	20	propose	propose	VBP
work_27d5tr326feydgcx32t4nhl424	125	21	a	a	DT
work_27d5tr326feydgcx32t4nhl424	125	22	function	function	NN
work_27d5tr326feydgcx32t4nhl424	125	23	to	to	TO
work_27d5tr326feydgcx32t4nhl424	125	24	score	score	VB
work_27d5tr326feydgcx32t4nhl424	125	25	a	a	DT
work_27d5tr326feydgcx32t4nhl424	125	26	set	set	NN
work_27d5tr326feydgcx32t4nhl424	125	27	of	of	IN
work_27d5tr326feydgcx32t4nhl424	125	28	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	125	29	Sum	Sum	NNP
work_27d5tr326feydgcx32t4nhl424	125	30	.	.	.
work_27d5tr326feydgcx32t4nhl424	126	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	126	2	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	126	3	a	a	DT
work_27d5tr326feydgcx32t4nhl424	126	4	decreasing	decrease	VBG
work_27d5tr326feydgcx32t4nhl424	126	5	logistic	logistic	JJ
work_27d5tr326feydgcx32t4nhl424	126	6	function	function	NN
work_27d5tr326feydgcx32t4nhl424	126	7	ζ(x	ζ(x	NNP
work_27d5tr326feydgcx32t4nhl424	126	8	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	126	9	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	126	10	1/(1	1/(1	CD
work_27d5tr326feydgcx32t4nhl424	126	11	+	+	SYM
work_27d5tr326feydgcx32t4nhl424	126	12	ex	ex	NN
work_27d5tr326feydgcx32t4nhl424	126	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	126	14	to	to	TO
work_27d5tr326feydgcx32t4nhl424	126	15	refine	refine	VB
work_27d5tr326feydgcx32t4nhl424	126	16	the	the	DT
work_27d5tr326feydgcx32t4nhl424	126	17	score	score	NN
work_27d5tr326feydgcx32t4nhl424	126	18	to	to	IN
work_27d5tr326feydgcx32t4nhl424	126	19	the	the	DT
work_27d5tr326feydgcx32t4nhl424	126	20	range	range	NN
work_27d5tr326feydgcx32t4nhl424	126	21	of	of	IN
work_27d5tr326feydgcx32t4nhl424	126	22	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	126	23	0,1	0,1	CD
work_27d5tr326feydgcx32t4nhl424	126	24	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	126	25	.	.	.
work_27d5tr326feydgcx32t4nhl424	127	1	Score(Sum	score(sum	LS
work_27d5tr326feydgcx32t4nhl424	127	2	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	127	3	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	127	4	ζ(KL(sum||Q	ζ(KL(sum||Q	NNP
work_27d5tr326feydgcx32t4nhl424	127	5	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	127	6	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	127	7	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	127	8	16	16	CD
work_27d5tr326feydgcx32t4nhl424	127	9	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	127	10	Let	let	VB
work_27d5tr326feydgcx32t4nhl424	127	11	Sum	Sum	NNP
work_27d5tr326feydgcx32t4nhl424	127	12	?	?	.
work_27d5tr326feydgcx32t4nhl424	128	1	denote	denote	VB
work_27d5tr326feydgcx32t4nhl424	128	2	the	the	DT
work_27d5tr326feydgcx32t4nhl424	128	3	optimum	optimum	JJ
work_27d5tr326feydgcx32t4nhl424	128	4	update	update	NN
work_27d5tr326feydgcx32t4nhl424	128	5	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	128	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	129	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	129	2	can	can	MD
work_27d5tr326feydgcx32t4nhl424	129	3	get	get	VB
work_27d5tr326feydgcx32t4nhl424	129	4	Sum	Sum	NNP
work_27d5tr326feydgcx32t4nhl424	129	5	?	?	.
work_27d5tr326feydgcx32t4nhl424	130	1	by	by	IN
work_27d5tr326feydgcx32t4nhl424	130	2	maximizing	maximize	VBG
work_27d5tr326feydgcx32t4nhl424	130	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	130	4	scoring	score	VBG
work_27d5tr326feydgcx32t4nhl424	130	5	function	function	NN
work_27d5tr326feydgcx32t4nhl424	130	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	131	1	Sum	Sum	NNP
work_27d5tr326feydgcx32t4nhl424	131	2	?	?	.
work_27d5tr326feydgcx32t4nhl424	132	1	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	132	2	arg	arg	NNP
work_27d5tr326feydgcx32t4nhl424	132	3	max	max	NNP
work_27d5tr326feydgcx32t4nhl424	132	4	Sum∈S&&words(Sum)≤L	Sum∈S&&words(Sum)≤L	VBD
work_27d5tr326feydgcx32t4nhl424	132	5	Score(Sum	score(sum	CD
work_27d5tr326feydgcx32t4nhl424	132	6	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	132	7	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	132	8	17	17	CD
work_27d5tr326feydgcx32t4nhl424	132	9	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	132	10	93	93	CD
work_27d5tr326feydgcx32t4nhl424	132	11	1	1	CD
work_27d5tr326feydgcx32t4nhl424	132	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	133	1	Learning	learn	VBG
work_27d5tr326feydgcx32t4nhl424	133	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	133	3	Given	give	VBN
work_27d5tr326feydgcx32t4nhl424	133	4	labeled	label	VBD
work_27d5tr326feydgcx32t4nhl424	133	5	set	set	NN
work_27d5tr326feydgcx32t4nhl424	133	6	Ov	ov	RP
work_27d5tr326feydgcx32t4nhl424	133	7	,	,	,
work_27d5tr326feydgcx32t4nhl424	133	8	learn	learn	VB
work_27d5tr326feydgcx32t4nhl424	133	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	133	10	feature	feature	NN
work_27d5tr326feydgcx32t4nhl424	133	11	weight	weight	NN
work_27d5tr326feydgcx32t4nhl424	133	12	vector	vector	NN
work_27d5tr326feydgcx32t4nhl424	133	13	η	η	NNP
work_27d5tr326feydgcx32t4nhl424	133	14	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	133	15	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	133	16	in	in	IN
work_27d5tr326feydgcx32t4nhl424	133	17	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	133	18	5	5	CD
work_27d5tr326feydgcx32t4nhl424	133	19	.	.	.
work_27d5tr326feydgcx32t4nhl424	134	1	2	2	LS
work_27d5tr326feydgcx32t4nhl424	134	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	135	1	Given	give	VBN
work_27d5tr326feydgcx32t4nhl424	135	2	new	new	JJ
work_27d5tr326feydgcx32t4nhl424	135	3	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	135	4	set	set	NN
work_27d5tr326feydgcx32t4nhl424	135	5	and	and	CC
work_27d5tr326feydgcx32t4nhl424	135	6	η	η	NN
work_27d5tr326feydgcx32t4nhl424	135	7	,	,	,
work_27d5tr326feydgcx32t4nhl424	135	8	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	135	9	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	135	10	in	in	IN
work_27d5tr326feydgcx32t4nhl424	135	11	section	section	NN
work_27d5tr326feydgcx32t4nhl424	135	12	3.3	3.3	CD
work_27d5tr326feydgcx32t4nhl424	135	13	for	for	IN
work_27d5tr326feydgcx32t4nhl424	135	14	inference	inference	NN
work_27d5tr326feydgcx32t4nhl424	135	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	136	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	136	2	The	the	DT
work_27d5tr326feydgcx32t4nhl424	136	3	only	only	JJ
work_27d5tr326feydgcx32t4nhl424	136	4	difference	difference	NN
work_27d5tr326feydgcx32t4nhl424	136	5	between	between	IN
work_27d5tr326feydgcx32t4nhl424	136	6	this	this	DT
work_27d5tr326feydgcx32t4nhl424	136	7	step	step	NN
work_27d5tr326feydgcx32t4nhl424	136	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	136	9	step	step	NN
work_27d5tr326feydgcx32t4nhl424	136	10	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	136	11	1	1	CD
work_27d5tr326feydgcx32t4nhl424	136	12	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	136	13	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	136	14	that	that	IN
work_27d5tr326feydgcx32t4nhl424	136	15	in	in	IN
work_27d5tr326feydgcx32t4nhl424	136	16	this	this	DT
work_27d5tr326feydgcx32t4nhl424	136	17	step	step	NN
work_27d5tr326feydgcx32t4nhl424	136	18	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	136	19	do	do	VBP
work_27d5tr326feydgcx32t4nhl424	136	20	not	not	RB
work_27d5tr326feydgcx32t4nhl424	136	21	need	need	VB
work_27d5tr326feydgcx32t4nhl424	136	22	minimize	minimize	NN
work_27d5tr326feydgcx32t4nhl424	136	23	L(η	L(η	NNP
work_27d5tr326feydgcx32t4nhl424	136	24	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	136	25	.	.	.
work_27d5tr326feydgcx32t4nhl424	137	1	3	3	LS
work_27d5tr326feydgcx32t4nhl424	137	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	138	1	Select	select	JJ
work_27d5tr326feydgcx32t4nhl424	138	2	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	138	3	for	for	IN
work_27d5tr326feydgcx32t4nhl424	138	4	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	138	5	from	from	IN
work_27d5tr326feydgcx32t4nhl424	138	6	algo-	algo-	JJ
work_27d5tr326feydgcx32t4nhl424	138	7	rithm	rithm	NN
work_27d5tr326feydgcx32t4nhl424	138	8	in	in	IN
work_27d5tr326feydgcx32t4nhl424	138	9	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	138	10	6	6	CD
work_27d5tr326feydgcx32t4nhl424	138	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	139	1	Figure	figure	NN
work_27d5tr326feydgcx32t4nhl424	139	2	6	6	CD
work_27d5tr326feydgcx32t4nhl424	139	3	:	:	:
work_27d5tr326feydgcx32t4nhl424	139	4	Summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	139	5	Generation	generation	NN
work_27d5tr326feydgcx32t4nhl424	139	6	by	by	IN
work_27d5tr326feydgcx32t4nhl424	139	7	S	s	NN
work_27d5tr326feydgcx32t4nhl424	139	8	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	139	9	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	139	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	140	1	A	a	DT
work_27d5tr326feydgcx32t4nhl424	140	2	greedy	greedy	JJ
work_27d5tr326feydgcx32t4nhl424	140	3	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	140	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	140	5	applied	apply	VBN
work_27d5tr326feydgcx32t4nhl424	140	6	by	by	IN
work_27d5tr326feydgcx32t4nhl424	140	7	adding	add	VBG
work_27d5tr326feydgcx32t4nhl424	140	8	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	140	9	one	one	CD
work_27d5tr326feydgcx32t4nhl424	140	10	by	by	IN
work_27d5tr326feydgcx32t4nhl424	140	11	one	one	PRP
work_27d5tr326feydgcx32t4nhl424	140	12	to	to	TO
work_27d5tr326feydgcx32t4nhl424	140	13	obtain	obtain	VB
work_27d5tr326feydgcx32t4nhl424	140	14	Sum	Sum	NNP
work_27d5tr326feydgcx32t4nhl424	140	15	?	?	.
work_27d5tr326feydgcx32t4nhl424	140	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	141	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	141	2	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	141	3	G	g	NN
work_27d5tr326feydgcx32t4nhl424	141	4	to	to	TO
work_27d5tr326feydgcx32t4nhl424	141	5	denote	denote	VB
work_27d5tr326feydgcx32t4nhl424	141	6	the	the	DT
work_27d5tr326feydgcx32t4nhl424	141	7	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	141	8	set	set	VBD
work_27d5tr326feydgcx32t4nhl424	141	9	containing	contain	VBG
work_27d5tr326feydgcx32t4nhl424	141	10	selected	select	VBN
work_27d5tr326feydgcx32t4nhl424	141	11	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	141	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	142	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	142	2	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	142	3	first	first	RB
work_27d5tr326feydgcx32t4nhl424	142	4	initializes	initialize	VBZ
work_27d5tr326feydgcx32t4nhl424	142	5	G	g	NN
work_27d5tr326feydgcx32t4nhl424	142	6	to	to	IN
work_27d5tr326feydgcx32t4nhl424	142	7	Φ	φ	NN
work_27d5tr326feydgcx32t4nhl424	142	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	142	9	X	X	NNP
work_27d5tr326feydgcx32t4nhl424	142	10	to	to	IN
work_27d5tr326feydgcx32t4nhl424	142	11	SU	SU	NNP
work_27d5tr326feydgcx32t4nhl424	142	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	143	1	Dur-	dur-	VB
work_27d5tr326feydgcx32t4nhl424	143	2	ing	e	VBG
work_27d5tr326feydgcx32t4nhl424	143	3	each	each	DT
work_27d5tr326feydgcx32t4nhl424	143	4	iteration	iteration	NN
work_27d5tr326feydgcx32t4nhl424	143	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	143	6	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	143	7	select	select	VBP
work_27d5tr326feydgcx32t4nhl424	143	8	one	one	CD
work_27d5tr326feydgcx32t4nhl424	143	9	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	143	10	from	from	IN
work_27d5tr326feydgcx32t4nhl424	143	11	X	X	NNS
work_27d5tr326feydgcx32t4nhl424	143	12	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	143	13	maximize	maximize	VBP
work_27d5tr326feydgcx32t4nhl424	143	14	Score(sm	Score(sm	.
work_27d5tr326feydgcx32t4nhl424	143	15	∪G	∪G	NNP
work_27d5tr326feydgcx32t4nhl424	143	16	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	143	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	144	1	To	to	TO
work_27d5tr326feydgcx32t4nhl424	144	2	avoid	avoid	VB
work_27d5tr326feydgcx32t4nhl424	144	3	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	144	4	re-	re-	JJ
work_27d5tr326feydgcx32t4nhl424	144	5	dundancy	dundancy	NN
work_27d5tr326feydgcx32t4nhl424	144	6	in	in	IN
work_27d5tr326feydgcx32t4nhl424	144	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	144	8	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	144	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	144	10	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	144	11	also	also	RB
work_27d5tr326feydgcx32t4nhl424	144	12	revise	revise	VBP
work_27d5tr326feydgcx32t4nhl424	144	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	144	14	MMR	MMR	NNP
work_27d5tr326feydgcx32t4nhl424	144	15	strategy	strategy	NN
work_27d5tr326feydgcx32t4nhl424	144	16	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	144	17	Goldstein	Goldstein	NNP
work_27d5tr326feydgcx32t4nhl424	144	18	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	144	19	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	144	20	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	144	21	,	,	,
work_27d5tr326feydgcx32t4nhl424	144	22	1999	1999	CD
work_27d5tr326feydgcx32t4nhl424	144	23	;	;	:
work_27d5tr326feydgcx32t4nhl424	144	24	Ouyang	Ouyang	NNP
work_27d5tr326feydgcx32t4nhl424	144	25	et	et	FW
work_27d5tr326feydgcx32t4nhl424	144	26	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	144	27	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	144	28	,	,	,
work_27d5tr326feydgcx32t4nhl424	144	29	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	144	30	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	144	31	in	in	IN
work_27d5tr326feydgcx32t4nhl424	144	32	the	the	DT
work_27d5tr326feydgcx32t4nhl424	144	33	process	process	NN
work_27d5tr326feydgcx32t4nhl424	144	34	of	of	IN
work_27d5tr326feydgcx32t4nhl424	144	35	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	144	36	selection	selection	NN
work_27d5tr326feydgcx32t4nhl424	144	37	.	.	.
work_27d5tr326feydgcx32t4nhl424	145	1	For	for	IN
work_27d5tr326feydgcx32t4nhl424	145	2	each	each	DT
work_27d5tr326feydgcx32t4nhl424	145	3	sm	sm	NN
work_27d5tr326feydgcx32t4nhl424	145	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	145	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	145	6	compute	compute	VBP
work_27d5tr326feydgcx32t4nhl424	145	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	145	8	semantic	semantic	JJ
work_27d5tr326feydgcx32t4nhl424	145	9	similarity	similarity	NN
work_27d5tr326feydgcx32t4nhl424	145	10	between	between	IN
work_27d5tr326feydgcx32t4nhl424	145	11	sm	sm	NN
work_27d5tr326feydgcx32t4nhl424	145	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	145	13	each	each	DT
work_27d5tr326feydgcx32t4nhl424	145	14	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	145	15	st	st	NNP
work_27d5tr326feydgcx32t4nhl424	145	16	in	in	IN
work_27d5tr326feydgcx32t4nhl424	145	17	set	set	VBN
work_27d5tr326feydgcx32t4nhl424	145	18	Y	Y	NNP
work_27d5tr326feydgcx32t4nhl424	145	19	in	in	IN
work_27d5tr326feydgcx32t4nhl424	145	20	Eqn.(18	Eqn.(18	NNP
work_27d5tr326feydgcx32t4nhl424	145	21	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	145	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	146	1	cos−sem(sm	cos−sem(sm	NNP
work_27d5tr326feydgcx32t4nhl424	146	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	146	3	st	st	NNP
work_27d5tr326feydgcx32t4nhl424	146	4	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	146	5	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	146	6	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	146	7	k	k	NN
work_27d5tr326feydgcx32t4nhl424	146	8	φsmkφstk√∑	φsmkφstk√∑	-LRB-
work_27d5tr326feydgcx32t4nhl424	146	9	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	146	10	φ	φ	NNP
work_27d5tr326feydgcx32t4nhl424	146	11	2	2	CD
work_27d5tr326feydgcx32t4nhl424	146	12	smk	smk	NNP
work_27d5tr326feydgcx32t4nhl424	146	13	√∑	√∑	NNP
work_27d5tr326feydgcx32t4nhl424	146	14	k	k	NNP
work_27d5tr326feydgcx32t4nhl424	146	15	φ	φ	NNP
work_27d5tr326feydgcx32t4nhl424	146	16	2	2	CD
work_27d5tr326feydgcx32t4nhl424	146	17	stk	stk	NN
work_27d5tr326feydgcx32t4nhl424	146	18	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	146	19	18	18	CD
work_27d5tr326feydgcx32t4nhl424	146	20	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	146	21	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	146	22	need	need	VBP
work_27d5tr326feydgcx32t4nhl424	146	23	to	to	TO
work_27d5tr326feydgcx32t4nhl424	146	24	assure	assure	VB
work_27d5tr326feydgcx32t4nhl424	146	25	that	that	IN
work_27d5tr326feydgcx32t4nhl424	146	26	the	the	DT
work_27d5tr326feydgcx32t4nhl424	146	27	value	value	NN
work_27d5tr326feydgcx32t4nhl424	146	28	of	of	IN
work_27d5tr326feydgcx32t4nhl424	146	29	semantic	semantic	JJ
work_27d5tr326feydgcx32t4nhl424	146	30	similar-	similar-	NN
work_27d5tr326feydgcx32t4nhl424	146	31	ity	ity	NN
work_27d5tr326feydgcx32t4nhl424	146	32	between	between	IN
work_27d5tr326feydgcx32t4nhl424	146	33	two	two	CD
work_27d5tr326feydgcx32t4nhl424	146	34	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	146	35	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	146	36	less	less	JJR
work_27d5tr326feydgcx32t4nhl424	146	37	than	than	IN
work_27d5tr326feydgcx32t4nhl424	146	38	Thsem	Thsem	NNP
work_27d5tr326feydgcx32t4nhl424	146	39	.	.	.
work_27d5tr326feydgcx32t4nhl424	147	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	147	2	whole	whole	JJ
work_27d5tr326feydgcx32t4nhl424	147	3	procedure	procedure	NN
work_27d5tr326feydgcx32t4nhl424	147	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	147	5	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	147	6	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	147	7	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	147	8	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	147	9	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	147	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	147	11	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	147	12	illustrated	illustrate	VBN
work_27d5tr326feydgcx32t4nhl424	147	13	in	in	IN
work_27d5tr326feydgcx32t4nhl424	147	14	Figure	Figure	NNP
work_27d5tr326feydgcx32t4nhl424	147	15	6	6	CD
work_27d5tr326feydgcx32t4nhl424	147	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	148	1	Thsem	Thsem	NNP
work_27d5tr326feydgcx32t4nhl424	148	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	148	3	set	set	VBN
work_27d5tr326feydgcx32t4nhl424	148	4	to	to	IN
work_27d5tr326feydgcx32t4nhl424	148	5	0.5	0.5	CD
work_27d5tr326feydgcx32t4nhl424	148	6	in	in	IN
work_27d5tr326feydgcx32t4nhl424	148	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	148	8	experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	148	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	149	1	5	5	CD
work_27d5tr326feydgcx32t4nhl424	149	2	Experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	149	3	5.1	5.1	CD
work_27d5tr326feydgcx32t4nhl424	149	4	Experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	149	5	Set	Set	VBN
work_27d5tr326feydgcx32t4nhl424	149	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	149	7	up	up	RP
work_27d5tr326feydgcx32t4nhl424	149	8	The	the	DT
work_27d5tr326feydgcx32t4nhl424	149	9	query	query	NN
work_27d5tr326feydgcx32t4nhl424	149	10	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	149	11	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	149	12	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	149	13	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	149	14	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	149	15	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	149	16	task	task	NN
work_27d5tr326feydgcx32t4nhl424	149	17	defined	define	VBN
work_27d5tr326feydgcx32t4nhl424	149	18	in	in	IN
work_27d5tr326feydgcx32t4nhl424	149	19	DUC3(Document	DUC3(Document	NNP
work_27d5tr326feydgcx32t4nhl424	149	20	Understanding	Understanding	NNP
work_27d5tr326feydgcx32t4nhl424	149	21	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	149	22	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	149	23	and	and	CC
work_27d5tr326feydgcx32t4nhl424	149	24	TAC4(Text	TAC4(Text	NNP
work_27d5tr326feydgcx32t4nhl424	149	25	Analysis	Analysis	NNP
work_27d5tr326feydgcx32t4nhl424	149	26	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	149	27	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	149	28	evaluations	evaluation	NNS
work_27d5tr326feydgcx32t4nhl424	149	29	requires	require	VBZ
work_27d5tr326feydgcx32t4nhl424	149	30	generating	generate	VBG
work_27d5tr326feydgcx32t4nhl424	149	31	a	a	DT
work_27d5tr326feydgcx32t4nhl424	149	32	concise	concise	NN
work_27d5tr326feydgcx32t4nhl424	149	33	and	and	CC
work_27d5tr326feydgcx32t4nhl424	149	34	well	well	RB
work_27d5tr326feydgcx32t4nhl424	149	35	organized	organized	JJ
work_27d5tr326feydgcx32t4nhl424	149	36	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	149	37	for	for	IN
work_27d5tr326feydgcx32t4nhl424	149	38	a	a	DT
work_27d5tr326feydgcx32t4nhl424	149	39	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	149	40	of	of	IN
work_27d5tr326feydgcx32t4nhl424	149	41	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	149	42	news	news	NN
work_27d5tr326feydgcx32t4nhl424	149	43	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	149	44	according	accord	VBG
work_27d5tr326feydgcx32t4nhl424	149	45	to	to	IN
work_27d5tr326feydgcx32t4nhl424	149	46	a	a	DT
work_27d5tr326feydgcx32t4nhl424	149	47	given	give	VBN
work_27d5tr326feydgcx32t4nhl424	149	48	query	query	NN
work_27d5tr326feydgcx32t4nhl424	149	49	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	149	50	de-	de-	RB
work_27d5tr326feydgcx32t4nhl424	149	51	scribes	scribe	VBZ
work_27d5tr326feydgcx32t4nhl424	149	52	the	the	DT
work_27d5tr326feydgcx32t4nhl424	149	53	users	user	NNS
work_27d5tr326feydgcx32t4nhl424	149	54	information	information	NN
work_27d5tr326feydgcx32t4nhl424	149	55	need	need	VBP
work_27d5tr326feydgcx32t4nhl424	149	56	.	.	.
work_27d5tr326feydgcx32t4nhl424	150	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	150	2	query	query	NN
work_27d5tr326feydgcx32t4nhl424	150	3	usually	usually	RB
work_27d5tr326feydgcx32t4nhl424	150	4	consists	consist	VBZ
work_27d5tr326feydgcx32t4nhl424	150	5	of	of	IN
work_27d5tr326feydgcx32t4nhl424	150	6	a	a	DT
work_27d5tr326feydgcx32t4nhl424	150	7	title	title	NN
work_27d5tr326feydgcx32t4nhl424	150	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	150	9	one	one	CD
work_27d5tr326feydgcx32t4nhl424	150	10	or	or	CC
work_27d5tr326feydgcx32t4nhl424	150	11	more	more	JJR
work_27d5tr326feydgcx32t4nhl424	150	12	narra-	narra-	JJ
work_27d5tr326feydgcx32t4nhl424	150	13	tive	tive	JJ
work_27d5tr326feydgcx32t4nhl424	150	14	/	/	SYM
work_27d5tr326feydgcx32t4nhl424	150	15	question	question	NN
work_27d5tr326feydgcx32t4nhl424	150	16	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	150	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	151	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	151	2	system	system	NN
work_27d5tr326feydgcx32t4nhl424	151	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	151	4	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	151	5	sum-	sum-	NN
work_27d5tr326feydgcx32t4nhl424	151	6	maries	mary	NNS
work_27d5tr326feydgcx32t4nhl424	151	7	for	for	IN
work_27d5tr326feydgcx32t4nhl424	151	8	DUC	DUC	NNP
work_27d5tr326feydgcx32t4nhl424	151	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	151	10	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	151	11	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	151	12	respectively	respectively	RB
work_27d5tr326feydgcx32t4nhl424	151	13	limited	limit	VBN
work_27d5tr326feydgcx32t4nhl424	151	14	to	to	IN
work_27d5tr326feydgcx32t4nhl424	151	15	3http://duc.nist.gov/.	3http://duc.nist.gov/.	CD
work_27d5tr326feydgcx32t4nhl424	152	1	4http://www.nist.gov/tac/.	4http://www.nist.gov/tac/.	LS
work_27d5tr326feydgcx32t4nhl424	153	1	250	250	CD
work_27d5tr326feydgcx32t4nhl424	153	2	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	153	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	153	4	100	100	CD
work_27d5tr326feydgcx32t4nhl424	153	5	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	153	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	154	1	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	154	2	experiment	experiment	NN
work_27d5tr326feydgcx32t4nhl424	154	3	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	154	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	154	5	composed	compose	VBN
work_27d5tr326feydgcx32t4nhl424	154	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	154	7	DUC	DUC	NNP
work_27d5tr326feydgcx32t4nhl424	154	8	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	154	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	154	10	TAC5	TAC5	NNP
work_27d5tr326feydgcx32t4nhl424	154	11	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	154	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	154	13	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	154	14	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	154	15	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	154	16	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	154	17	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	154	18	45	45	CD
work_27d5tr326feydgcx32t4nhl424	154	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	154	20	48	48	CD
work_27d5tr326feydgcx32t4nhl424	154	21	and	and	CC
work_27d5tr326feydgcx32t4nhl424	154	22	44	44	CD
work_27d5tr326feydgcx32t4nhl424	154	23	collections	collection	NNS
work_27d5tr326feydgcx32t4nhl424	154	24	respec-	respec-	VBG
work_27d5tr326feydgcx32t4nhl424	154	25	tively	tively	RB
work_27d5tr326feydgcx32t4nhl424	154	26	.	.	.
work_27d5tr326feydgcx32t4nhl424	155	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	155	2	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	155	3	experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	155	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	155	5	DUC	DUC	NNP
work_27d5tr326feydgcx32t4nhl424	155	6	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	155	7	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	155	8	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	155	9	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	155	10	as	as	IN
work_27d5tr326feydgcx32t4nhl424	155	11	training	training	NN
work_27d5tr326feydgcx32t4nhl424	155	12	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	155	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	155	14	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	155	15	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	155	16	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	155	17	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	155	18	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	155	19	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	155	20	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	155	21	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	155	22	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	155	23	as	as	IN
work_27d5tr326feydgcx32t4nhl424	155	24	the	the	DT
work_27d5tr326feydgcx32t4nhl424	155	25	test	test	NN
work_27d5tr326feydgcx32t4nhl424	155	26	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	155	27	.	.	.
work_27d5tr326feydgcx32t4nhl424	156	1	Stop	stop	VB
work_27d5tr326feydgcx32t4nhl424	156	2	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	156	3	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	156	4	in	in	IN
work_27d5tr326feydgcx32t4nhl424	156	5	both	both	DT
work_27d5tr326feydgcx32t4nhl424	156	6	documents	document	NNS
work_27d5tr326feydgcx32t4nhl424	156	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	156	8	queries	query	NNS
work_27d5tr326feydgcx32t4nhl424	156	9	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	156	10	removed	remove	VBN
work_27d5tr326feydgcx32t4nhl424	156	11	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	156	12	a	a	DT
work_27d5tr326feydgcx32t4nhl424	156	13	stop	stop	VB
work_27d5tr326feydgcx32t4nhl424	156	14	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	156	15	word	word	NN
work_27d5tr326feydgcx32t4nhl424	156	16	list	list	NN
work_27d5tr326feydgcx32t4nhl424	156	17	of	of	IN
work_27d5tr326feydgcx32t4nhl424	156	18	598	598	CD
work_27d5tr326feydgcx32t4nhl424	156	19	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	156	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	156	21	and	and	CC
work_27d5tr326feydgcx32t4nhl424	156	22	the	the	DT
work_27d5tr326feydgcx32t4nhl424	156	23	remaining	remain	VBG
work_27d5tr326feydgcx32t4nhl424	156	24	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	156	25	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	156	26	stemmed	stem	VBN
work_27d5tr326feydgcx32t4nhl424	156	27	by	by	IN
work_27d5tr326feydgcx32t4nhl424	156	28	Porter	Porter	NNP
work_27d5tr326feydgcx32t4nhl424	156	29	Stem-	Stem-	NNP
work_27d5tr326feydgcx32t4nhl424	156	30	mer6	mer6	NNP
work_27d5tr326feydgcx32t4nhl424	156	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	157	1	As	as	IN
work_27d5tr326feydgcx32t4nhl424	157	2	for	for	IN
work_27d5tr326feydgcx32t4nhl424	157	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	157	4	automatic	automatic	JJ
work_27d5tr326feydgcx32t4nhl424	157	5	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	157	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	157	7	summa-	summa-	JJ
work_27d5tr326feydgcx32t4nhl424	157	8	rization	rization	NN
work_27d5tr326feydgcx32t4nhl424	157	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	157	10	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	157	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	157	12	Recall	recall	NN
work_27d5tr326feydgcx32t4nhl424	157	13	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	157	14	Oriented	orient	VBN
work_27d5tr326feydgcx32t4nhl424	157	15	Understudy	Understudy	NNP
work_27d5tr326feydgcx32t4nhl424	157	16	for	for	IN
work_27d5tr326feydgcx32t4nhl424	157	17	Gisting	Gisting	NNP
work_27d5tr326feydgcx32t4nhl424	157	18	Evaluation	Evaluation	NNP
work_27d5tr326feydgcx32t4nhl424	157	19	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	157	20	measures	measure	NNS
work_27d5tr326feydgcx32t4nhl424	157	21	,	,	,
work_27d5tr326feydgcx32t4nhl424	157	22	including	include	VBG
work_27d5tr326feydgcx32t4nhl424	157	23	ROUGE-	ROUGE-	NNP
work_27d5tr326feydgcx32t4nhl424	157	24	1	1	CD
work_27d5tr326feydgcx32t4nhl424	157	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	157	26	ROUGE-2	rouge-2	NN
work_27d5tr326feydgcx32t4nhl424	157	27	,	,	,
work_27d5tr326feydgcx32t4nhl424	157	28	and	and	CC
work_27d5tr326feydgcx32t4nhl424	157	29	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	157	30	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	157	31	SU47	SU47	NNP
work_27d5tr326feydgcx32t4nhl424	157	32	and	and	CC
work_27d5tr326feydgcx32t4nhl424	157	33	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	157	34	corre-	corre-	NN
work_27d5tr326feydgcx32t4nhl424	157	35	sponding	sponde	VBG
work_27d5tr326feydgcx32t4nhl424	157	36	95	95	CD
work_27d5tr326feydgcx32t4nhl424	157	37	%	%	NN
work_27d5tr326feydgcx32t4nhl424	157	38	confidence	confidence	NN
work_27d5tr326feydgcx32t4nhl424	157	39	intervals	interval	NNS
work_27d5tr326feydgcx32t4nhl424	157	40	,	,	,
work_27d5tr326feydgcx32t4nhl424	157	41	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	157	42	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	157	43	to	to	TO
work_27d5tr326feydgcx32t4nhl424	157	44	eval-	eval-	VB
work_27d5tr326feydgcx32t4nhl424	157	45	uate	uate	VB
work_27d5tr326feydgcx32t4nhl424	157	46	the	the	DT
work_27d5tr326feydgcx32t4nhl424	157	47	performance	performance	NN
work_27d5tr326feydgcx32t4nhl424	157	48	of	of	IN
work_27d5tr326feydgcx32t4nhl424	157	49	the	the	DT
work_27d5tr326feydgcx32t4nhl424	157	50	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	157	51	.	.	.
work_27d5tr326feydgcx32t4nhl424	158	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	158	2	order	order	NN
work_27d5tr326feydgcx32t4nhl424	158	3	to	to	TO
work_27d5tr326feydgcx32t4nhl424	158	4	obtain	obtain	VB
work_27d5tr326feydgcx32t4nhl424	158	5	a	a	DT
work_27d5tr326feydgcx32t4nhl424	158	6	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	158	7	comprehensive	comprehensive	JJ
work_27d5tr326feydgcx32t4nhl424	158	8	measure	measure	NN
work_27d5tr326feydgcx32t4nhl424	158	9	of	of	IN
work_27d5tr326feydgcx32t4nhl424	158	10	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	158	11	quality	quality	NN
work_27d5tr326feydgcx32t4nhl424	158	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	158	13	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	158	14	also	also	RB
work_27d5tr326feydgcx32t4nhl424	158	15	conduct	conduct	VBP
work_27d5tr326feydgcx32t4nhl424	158	16	manual	manual	JJ
work_27d5tr326feydgcx32t4nhl424	158	17	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	158	18	on	on	IN
work_27d5tr326feydgcx32t4nhl424	158	19	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	158	20	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	158	21	with	with	IN
work_27d5tr326feydgcx32t4nhl424	158	22	reference	reference	NN
work_27d5tr326feydgcx32t4nhl424	158	23	to	to	IN
work_27d5tr326feydgcx32t4nhl424	158	24	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	158	25	Haghighi	Haghighi	NNP
work_27d5tr326feydgcx32t4nhl424	158	26	and	and	CC
work_27d5tr326feydgcx32t4nhl424	158	27	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	158	28	,	,	,
work_27d5tr326feydgcx32t4nhl424	158	29	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	158	30	;	;	:
work_27d5tr326feydgcx32t4nhl424	158	31	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	158	32	and	and	CC
work_27d5tr326feydgcx32t4nhl424	158	33	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	158	34	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	158	35	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	158	36	,	,	,
work_27d5tr326feydgcx32t4nhl424	158	37	2011	2011	CD
work_27d5tr326feydgcx32t4nhl424	158	38	;	;	:
work_27d5tr326feydgcx32t4nhl424	158	39	Delort	Delort	NNP
work_27d5tr326feydgcx32t4nhl424	158	40	and	and	CC
work_27d5tr326feydgcx32t4nhl424	158	41	Alfonseca	Alfonseca	NNP
work_27d5tr326feydgcx32t4nhl424	158	42	,	,	,
work_27d5tr326feydgcx32t4nhl424	158	43	2011	2011	CD
work_27d5tr326feydgcx32t4nhl424	158	44	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	158	45	.	.	.
work_27d5tr326feydgcx32t4nhl424	159	1	5.2	5.2	CD
work_27d5tr326feydgcx32t4nhl424	159	2	Comparison	Comparison	NNP
work_27d5tr326feydgcx32t4nhl424	159	3	with	with	IN
work_27d5tr326feydgcx32t4nhl424	159	4	other	other	JJ
work_27d5tr326feydgcx32t4nhl424	159	5	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	159	6	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	159	7	In	in	IN
work_27d5tr326feydgcx32t4nhl424	159	8	this	this	DT
work_27d5tr326feydgcx32t4nhl424	159	9	subsection	subsection	NN
work_27d5tr326feydgcx32t4nhl424	159	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	159	11	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	159	12	compare	compare	VBP
work_27d5tr326feydgcx32t4nhl424	159	13	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	159	14	model	model	NN
work_27d5tr326feydgcx32t4nhl424	159	15	with	with	IN
work_27d5tr326feydgcx32t4nhl424	159	16	the	the	DT
work_27d5tr326feydgcx32t4nhl424	159	17	following	follow	VBG
work_27d5tr326feydgcx32t4nhl424	159	18	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	159	19	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	159	20	:	:	:
work_27d5tr326feydgcx32t4nhl424	159	21	KL	KL	NNP
work_27d5tr326feydgcx32t4nhl424	159	22	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	159	23	sum	sum	NN
work_27d5tr326feydgcx32t4nhl424	159	24	:	:	:
work_27d5tr326feydgcx32t4nhl424	159	25	It	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	159	26	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	159	27	developed	develop	VBN
work_27d5tr326feydgcx32t4nhl424	159	28	by	by	IN
work_27d5tr326feydgcx32t4nhl424	159	29	Haghighi	Haghighi	NNP
work_27d5tr326feydgcx32t4nhl424	159	30	and	and	CC
work_27d5tr326feydgcx32t4nhl424	159	31	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	159	32	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	159	33	Lin	Lin	NNP
work_27d5tr326feydgcx32t4nhl424	159	34	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	159	35	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	159	36	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	159	37	,	,	,
work_27d5tr326feydgcx32t4nhl424	159	38	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	159	39	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	159	40	by	by	IN
work_27d5tr326feydgcx32t4nhl424	159	41	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	159	42	a	a	DT
work_27d5tr326feydgcx32t4nhl424	159	43	KL-	KL-	NNP
work_27d5tr326feydgcx32t4nhl424	159	44	divergence	divergence	NN
work_27d5tr326feydgcx32t4nhl424	159	45	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	159	46	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	159	47	selection	selection	NN
work_27d5tr326feydgcx32t4nhl424	159	48	strategy	strategy	NN
work_27d5tr326feydgcx32t4nhl424	159	49	.	.	.
work_27d5tr326feydgcx32t4nhl424	160	1	KL(Ps||Qd	KL(Ps||Qd	NNP
work_27d5tr326feydgcx32t4nhl424	160	2	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	160	3	=	=	NFP
work_27d5tr326feydgcx32t4nhl424	160	4	∑	∑	.
work_27d5tr326feydgcx32t4nhl424	160	5	w	w	NNP
work_27d5tr326feydgcx32t4nhl424	160	6	P(w)log	P(w)log	NNP
work_27d5tr326feydgcx32t4nhl424	160	7	P(w	P(w	NNP
work_27d5tr326feydgcx32t4nhl424	160	8	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	160	9	Q(w	q(w	CD
work_27d5tr326feydgcx32t4nhl424	160	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	160	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	160	12	19	19	CD
work_27d5tr326feydgcx32t4nhl424	160	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	160	14	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	160	15	Ps	Ps	NNP
work_27d5tr326feydgcx32t4nhl424	160	16	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	160	17	the	the	DT
work_27d5tr326feydgcx32t4nhl424	160	18	unigram	unigram	JJ
work_27d5tr326feydgcx32t4nhl424	160	19	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	160	20	of	of	IN
work_27d5tr326feydgcx32t4nhl424	160	21	candidate	candidate	NN
work_27d5tr326feydgcx32t4nhl424	160	22	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	160	23	and	and	CC
work_27d5tr326feydgcx32t4nhl424	160	24	Qd	Qd	NNP
work_27d5tr326feydgcx32t4nhl424	160	25	denotes	denote	VBZ
work_27d5tr326feydgcx32t4nhl424	160	26	the	the	DT
work_27d5tr326feydgcx32t4nhl424	160	27	unigram	unigram	JJ
work_27d5tr326feydgcx32t4nhl424	160	28	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	160	29	of	of	IN
work_27d5tr326feydgcx32t4nhl424	160	30	document	document	NN
work_27d5tr326feydgcx32t4nhl424	160	31	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	160	32	.	.	.
work_27d5tr326feydgcx32t4nhl424	161	1	Sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	161	2	with	with	IN
work_27d5tr326feydgcx32t4nhl424	161	3	higher	high	JJR
work_27d5tr326feydgcx32t4nhl424	161	4	ranking	ranking	JJ
work_27d5tr326feydgcx32t4nhl424	161	5	score	score	NN
work_27d5tr326feydgcx32t4nhl424	161	6	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	161	7	selected	select	VBN
work_27d5tr326feydgcx32t4nhl424	161	8	into	into	IN
work_27d5tr326feydgcx32t4nhl424	161	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	161	10	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	161	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	162	1	HierSum	HierSum	NNP
work_27d5tr326feydgcx32t4nhl424	162	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	162	3	A	a	DT
work_27d5tr326feydgcx32t4nhl424	162	4	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	162	5	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	162	6	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	162	7	proposed	propose	VBN
work_27d5tr326feydgcx32t4nhl424	162	8	by	by	IN
work_27d5tr326feydgcx32t4nhl424	162	9	Haghighi	Haghighi	NNP
work_27d5tr326feydgcx32t4nhl424	162	10	and	and	CC
work_27d5tr326feydgcx32t4nhl424	162	11	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	162	12	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	162	13	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	162	14	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	162	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	162	16	where	where	WRB
work_27d5tr326feydgcx32t4nhl424	162	17	unigram	unigram	JJ
work_27d5tr326feydgcx32t4nhl424	162	18	distribution	distribution	NN
work_27d5tr326feydgcx32t4nhl424	162	19	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	162	20	calculated	calculate	VBN
work_27d5tr326feydgcx32t4nhl424	162	21	from	from	IN
work_27d5tr326feydgcx32t4nhl424	162	22	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	162	23	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	162	24	model	model	NN
work_27d5tr326feydgcx32t4nhl424	162	25	in	in	IN
work_27d5tr326feydgcx32t4nhl424	162	26	Equ.(14	Equ.(14	NNP
work_27d5tr326feydgcx32t4nhl424	162	27	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	162	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	163	1	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	163	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	163	3	A	a	DT
work_27d5tr326feydgcx32t4nhl424	163	4	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	163	5	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	163	6	developed	develop	VBN
work_27d5tr326feydgcx32t4nhl424	163	7	by	by	IN
work_27d5tr326feydgcx32t4nhl424	163	8	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	163	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	163	10	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	163	11	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	163	12	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	163	13	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	163	14	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	163	15	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	163	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	164	1	For	for	IN
work_27d5tr326feydgcx32t4nhl424	164	2	fair	fair	JJ
work_27d5tr326feydgcx32t4nhl424	164	3	comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	164	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	164	5	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	164	6	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	164	7	the	the	DT
work_27d5tr326feydgcx32t4nhl424	164	8	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	164	9	pro-	pro-	NN
work_27d5tr326feydgcx32t4nhl424	164	10	precessing	precess	VBG
work_27d5tr326feydgcx32t4nhl424	164	11	methods	method	NNS
work_27d5tr326feydgcx32t4nhl424	164	12	with	with	IN
work_27d5tr326feydgcx32t4nhl424	164	13	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	164	14	model	model	NN
work_27d5tr326feydgcx32t4nhl424	164	15	and	and	CC
work_27d5tr326feydgcx32t4nhl424	164	16	all	all	DT
work_27d5tr326feydgcx32t4nhl424	164	17	sum-	sum-	JJ
work_27d5tr326feydgcx32t4nhl424	164	18	5Here	5here	RB
work_27d5tr326feydgcx32t4nhl424	164	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	164	20	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	164	21	only	only	RB
work_27d5tr326feydgcx32t4nhl424	164	22	use	use	VBP
work_27d5tr326feydgcx32t4nhl424	164	23	the	the	DT
work_27d5tr326feydgcx32t4nhl424	164	24	docset	docset	JJ
work_27d5tr326feydgcx32t4nhl424	164	25	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	164	26	A	a	DT
work_27d5tr326feydgcx32t4nhl424	164	27	data	data	NN
work_27d5tr326feydgcx32t4nhl424	164	28	in	in	IN
work_27d5tr326feydgcx32t4nhl424	164	29	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	164	30	,	,	,
work_27d5tr326feydgcx32t4nhl424	164	31	since	since	IN
work_27d5tr326feydgcx32t4nhl424	164	32	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	164	33	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	164	34	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	164	35	composed	compose	VBN
work_27d5tr326feydgcx32t4nhl424	164	36	of	of	IN
work_27d5tr326feydgcx32t4nhl424	164	37	docset	docset	VBN
work_27d5tr326feydgcx32t4nhl424	164	38	-	-	.
work_27d5tr326feydgcx32t4nhl424	164	39	A	a	NN
work_27d5tr326feydgcx32t4nhl424	164	40	and	and	CC
work_27d5tr326feydgcx32t4nhl424	164	41	docset	docset	NN
work_27d5tr326feydgcx32t4nhl424	164	42	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	164	43	B	b	NN
work_27d5tr326feydgcx32t4nhl424	164	44	data	data	NN
work_27d5tr326feydgcx32t4nhl424	164	45	,	,	,
work_27d5tr326feydgcx32t4nhl424	164	46	and	and	CC
work_27d5tr326feydgcx32t4nhl424	164	47	the	the	DT
work_27d5tr326feydgcx32t4nhl424	164	48	docset-	docset-	JJ
work_27d5tr326feydgcx32t4nhl424	164	49	B	b	NN
work_27d5tr326feydgcx32t4nhl424	164	50	data	data	NN
work_27d5tr326feydgcx32t4nhl424	164	51	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	164	52	mainly	mainly	RB
work_27d5tr326feydgcx32t4nhl424	164	53	for	for	IN
work_27d5tr326feydgcx32t4nhl424	164	54	the	the	DT
work_27d5tr326feydgcx32t4nhl424	164	55	update	update	NN
work_27d5tr326feydgcx32t4nhl424	164	56	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	164	57	task	task	NN
work_27d5tr326feydgcx32t4nhl424	164	58	.	.	.
work_27d5tr326feydgcx32t4nhl424	165	1	6http://tartarus.org/	6http://tartarus.org/	NNP
work_27d5tr326feydgcx32t4nhl424	165	2	martin	martin	NNP
work_27d5tr326feydgcx32t4nhl424	165	3	/	/	SYM
work_27d5tr326feydgcx32t4nhl424	165	4	PorterStemmer/.	PorterStemmer/.	.
work_27d5tr326feydgcx32t4nhl424	166	1	7Jackknife	7jackknife	NN
work_27d5tr326feydgcx32t4nhl424	166	2	scoring	score	VBG
work_27d5tr326feydgcx32t4nhl424	166	3	for	for	IN
work_27d5tr326feydgcx32t4nhl424	166	4	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	166	5	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	166	6	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	166	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	166	8	order	order	NN
work_27d5tr326feydgcx32t4nhl424	166	9	to	to	TO
work_27d5tr326feydgcx32t4nhl424	166	10	compare	compare	VB
work_27d5tr326feydgcx32t4nhl424	166	11	with	with	IN
work_27d5tr326feydgcx32t4nhl424	166	12	the	the	DT
work_27d5tr326feydgcx32t4nhl424	166	13	human	human	JJ
work_27d5tr326feydgcx32t4nhl424	166	14	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	166	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	167	1	94	94	CD
work_27d5tr326feydgcx32t4nhl424	167	2	maries	mary	NNS
work_27d5tr326feydgcx32t4nhl424	167	3	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	167	4	truncated	truncate	VBN
work_27d5tr326feydgcx32t4nhl424	167	5	to	to	IN
work_27d5tr326feydgcx32t4nhl424	167	6	the	the	DT
work_27d5tr326feydgcx32t4nhl424	167	7	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	167	8	length	length	NN
work_27d5tr326feydgcx32t4nhl424	167	9	of	of	IN
work_27d5tr326feydgcx32t4nhl424	167	10	100	100	CD
work_27d5tr326feydgcx32t4nhl424	167	11	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	167	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	168	1	From	from	IN
work_27d5tr326feydgcx32t4nhl424	168	2	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	168	3	1	1	CD
work_27d5tr326feydgcx32t4nhl424	168	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	168	5	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	168	6	2	2	CD
work_27d5tr326feydgcx32t4nhl424	168	7	,	,	,
work_27d5tr326feydgcx32t4nhl424	168	8	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	168	9	can	can	MD
work_27d5tr326feydgcx32t4nhl424	168	10	Methods	Methods	NNP
work_27d5tr326feydgcx32t4nhl424	168	11	ROUGE-1	rouge-1	NN
work_27d5tr326feydgcx32t4nhl424	168	12	ROUGE-2	rouge-2	NN
work_27d5tr326feydgcx32t4nhl424	168	13	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	168	14	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	15	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	168	16	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	168	17	0.3724	0.3724	CD
work_27d5tr326feydgcx32t4nhl424	168	18	0.1030	0.1030	CD
work_27d5tr326feydgcx32t4nhl424	168	19	0.1342	0.1342	CD
work_27d5tr326feydgcx32t4nhl424	168	20	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	168	21	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	22	0.3660	0.3660	CD
work_27d5tr326feydgcx32t4nhl424	168	23	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	24	0.3788	0.3788	CD
work_27d5tr326feydgcx32t4nhl424	168	25	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	26	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	27	0.0999	0.0999	CD
work_27d5tr326feydgcx32t4nhl424	168	28	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	29	0.1061	0.1061	CD
work_27d5tr326feydgcx32t4nhl424	168	30	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	31	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	32	0.1290	0.1290	CD
work_27d5tr326feydgcx32t4nhl424	168	33	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	34	0.1394	0.1394	CD
work_27d5tr326feydgcx32t4nhl424	168	35	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	36	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	168	37	0.3703	0.3703	CD
work_27d5tr326feydgcx32t4nhl424	168	38	0.1007	0.1007	CD
work_27d5tr326feydgcx32t4nhl424	168	39	0.1314	0.1314	CD
work_27d5tr326feydgcx32t4nhl424	168	40	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	41	0.3600	0.3600	CD
work_27d5tr326feydgcx32t4nhl424	168	42	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	43	0.3806	0.3806	CD
work_27d5tr326feydgcx32t4nhl424	168	44	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	45	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	46	0.0952	0.0952	CD
work_27d5tr326feydgcx32t4nhl424	168	47	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	48	0.1059	0.1059	CD
work_27d5tr326feydgcx32t4nhl424	168	49	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	50	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	51	0.1241	0.1241	CD
work_27d5tr326feydgcx32t4nhl424	168	52	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	53	0.1387	0.1387	NN
work_27d5tr326feydgcx32t4nhl424	168	54	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	55	HierSum	HierSum	NNP
work_27d5tr326feydgcx32t4nhl424	168	56	0.3613	0.3613	CD
work_27d5tr326feydgcx32t4nhl424	168	57	0.0948	0.0948	CD
work_27d5tr326feydgcx32t4nhl424	168	58	0.1278	0.1278	CD
work_27d5tr326feydgcx32t4nhl424	168	59	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	60	0.3374	0.3374	CD
work_27d5tr326feydgcx32t4nhl424	168	61	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	62	0.3752	0.3752	CD
work_27d5tr326feydgcx32t4nhl424	168	63	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	64	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	65	0.0899	0.0899	CD
work_27d5tr326feydgcx32t4nhl424	168	66	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	67	0.0998	0.0998	CD
work_27d5tr326feydgcx32t4nhl424	168	68	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	69	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	70	0.1197	0.1197	CD
work_27d5tr326feydgcx32t4nhl424	168	71	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	72	0.1359	0.1359	CD
work_27d5tr326feydgcx32t4nhl424	168	73	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	74	KLsum	KLsum	VBD
work_27d5tr326feydgcx32t4nhl424	168	75	0.3504	0.3504	CD
work_27d5tr326feydgcx32t4nhl424	168	76	0.0917	0.0917	CD
work_27d5tr326feydgcx32t4nhl424	168	77	0.1234	0.1234	CD
work_27d5tr326feydgcx32t4nhl424	168	78	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	79	0.3411	0.3411	CD
work_27d5tr326feydgcx32t4nhl424	168	80	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	81	0.3597	0.3597	NNP
work_27d5tr326feydgcx32t4nhl424	168	82	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	83	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	84	0.0842	0.0842	NNP
work_27d5tr326feydgcx32t4nhl424	168	85	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	86	0.0992	0.0992	NNP
work_27d5tr326feydgcx32t4nhl424	168	87	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	88	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	89	0.1155	0.1155	CD
work_27d5tr326feydgcx32t4nhl424	168	90	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	91	0.1315	0.1315	CD
work_27d5tr326feydgcx32t4nhl424	168	92	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	93	StandLDA	StandLDA	.
work_27d5tr326feydgcx32t4nhl424	168	94	0.3368	0.3368	CD
work_27d5tr326feydgcx32t4nhl424	168	95	0.0797	0.0797	CD
work_27d5tr326feydgcx32t4nhl424	168	96	0.1156	0.1156	CD
work_27d5tr326feydgcx32t4nhl424	168	97	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	98	0.3252	0.3252	CD
work_27d5tr326feydgcx32t4nhl424	168	99	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	100	0.3386	0.3386	NNS
work_27d5tr326feydgcx32t4nhl424	168	101	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	102	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	103	0.0758	0.0758	CD
work_27d5tr326feydgcx32t4nhl424	168	104	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	105	0.0836	0.0836	CD
work_27d5tr326feydgcx32t4nhl424	168	106	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	107	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	108	0.1072	0.1072	IN
work_27d5tr326feydgcx32t4nhl424	168	109	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	110	0.1240	0.1240	CD
work_27d5tr326feydgcx32t4nhl424	168	111	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	112	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	168	113	1	1	CD
work_27d5tr326feydgcx32t4nhl424	168	114	:	:	:
work_27d5tr326feydgcx32t4nhl424	168	115	Comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	168	116	of	of	IN
work_27d5tr326feydgcx32t4nhl424	168	117	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	168	118	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	168	119	on	on	IN
work_27d5tr326feydgcx32t4nhl424	168	120	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	168	121	Methods	Methods	NNP
work_27d5tr326feydgcx32t4nhl424	168	122	ROUGE-1	rouge-1	NN
work_27d5tr326feydgcx32t4nhl424	168	123	ROUGE-2	rouge-2	NN
work_27d5tr326feydgcx32t4nhl424	168	124	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	168	125	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	126	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	168	127	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	168	128	0.3903	0.3903	CD
work_27d5tr326feydgcx32t4nhl424	168	129	0.1223	0.1223	CD
work_27d5tr326feydgcx32t4nhl424	168	130	0.1488	0.1488	CD
work_27d5tr326feydgcx32t4nhl424	168	131	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	168	132	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	133	0.3819	0.3819	CD
work_27d5tr326feydgcx32t4nhl424	168	134	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	135	0.3987	0.3987	CD
work_27d5tr326feydgcx32t4nhl424	168	136	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	137	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	138	0.1167	0.1167	CD
work_27d5tr326feydgcx32t4nhl424	168	139	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	140	0.1279	0.1279	CD
work_27d5tr326feydgcx32t4nhl424	168	141	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	142	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	143	0.1446	0.1446	IN
work_27d5tr326feydgcx32t4nhl424	168	144	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	145	0.1530	0.1530	CD
work_27d5tr326feydgcx32t4nhl424	168	146	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	147	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	168	148	0.3824	0.3824	CD
work_27d5tr326feydgcx32t4nhl424	168	149	0.1173	0.1173	CD
work_27d5tr326feydgcx32t4nhl424	168	150	0.1436	0.1436	CD
work_27d5tr326feydgcx32t4nhl424	168	151	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	152	0.3686	0.3686	CD
work_27d5tr326feydgcx32t4nhl424	168	153	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	154	0.3952	0.3952	CD
work_27d5tr326feydgcx32t4nhl424	168	155	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	156	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	157	0.1132	0.1132	CD
work_27d5tr326feydgcx32t4nhl424	168	158	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	159	0.1214	0.1214	NNS
work_27d5tr326feydgcx32t4nhl424	168	160	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	161	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	162	0.1358	0.1358	CD
work_27d5tr326feydgcx32t4nhl424	168	163	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	164	0.1514	0.1514	CD
work_27d5tr326feydgcx32t4nhl424	168	165	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	166	HierSum	HierSum	NNP
work_27d5tr326feydgcx32t4nhl424	168	167	0.3706	0.3706	CD
work_27d5tr326feydgcx32t4nhl424	168	168	0.1088	0.1088	CD
work_27d5tr326feydgcx32t4nhl424	168	169	0.1386	0.1386	CD
work_27d5tr326feydgcx32t4nhl424	168	170	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	171	0.3624	0.3624	CD
work_27d5tr326feydgcx32t4nhl424	168	172	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	173	0.3788	0.3788	CD
work_27d5tr326feydgcx32t4nhl424	168	174	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	175	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	176	0.0950	0.0950	CD
work_27d5tr326feydgcx32t4nhl424	168	177	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	178	0.1144	0.1144	CD
work_27d5tr326feydgcx32t4nhl424	168	179	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	180	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	181	0.1312	0.1312	CD
work_27d5tr326feydgcx32t4nhl424	168	182	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	183	0.1464	0.1464	CD
work_27d5tr326feydgcx32t4nhl424	168	184	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	185	KLsum	KLsum	NNP
work_27d5tr326feydgcx32t4nhl424	168	186	0.3619	0.3619	CD
work_27d5tr326feydgcx32t4nhl424	168	187	0.0972	0.0972	CD
work_27d5tr326feydgcx32t4nhl424	168	188	0.1299	0.1299	CD
work_27d5tr326feydgcx32t4nhl424	168	189	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	190	0.3510	0.3510	CD
work_27d5tr326feydgcx32t4nhl424	168	191	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	192	0.3728	0.3728	CD
work_27d5tr326feydgcx32t4nhl424	168	193	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	194	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	195	0.0917	0.0917	IN
work_27d5tr326feydgcx32t4nhl424	168	196	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	197	0.1047	0.1047	CD
work_27d5tr326feydgcx32t4nhl424	168	198	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	199	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	200	0.1213	0.1213	CD
work_27d5tr326feydgcx32t4nhl424	168	201	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	202	0.1385	0.1385	NN
work_27d5tr326feydgcx32t4nhl424	168	203	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	204	StandLDA	StandLDA	.
work_27d5tr326feydgcx32t4nhl424	168	205	0.3552	0.3552	CD
work_27d5tr326feydgcx32t4nhl424	168	206	0.0847	0.0847	CD
work_27d5tr326feydgcx32t4nhl424	168	207	0.1214	0.1214	CD
work_27d5tr326feydgcx32t4nhl424	168	208	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	209	0.3447	0.3447	CD
work_27d5tr326feydgcx32t4nhl424	168	210	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	211	0.3657	0.3657	CD
work_27d5tr326feydgcx32t4nhl424	168	212	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	213	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	214	0.0813	0.0813	CD
work_27d5tr326feydgcx32t4nhl424	168	215	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	168	216	0.0881	0.0881	CD
work_27d5tr326feydgcx32t4nhl424	168	217	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	218	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	168	219	0.1141	0.1141	CD
work_27d5tr326feydgcx32t4nhl424	168	220	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	168	221	0.1286	0.1286	CD
work_27d5tr326feydgcx32t4nhl424	168	222	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	168	223	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	168	224	2	2	CD
work_27d5tr326feydgcx32t4nhl424	168	225	:	:	:
work_27d5tr326feydgcx32t4nhl424	168	226	Comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	168	227	of	of	IN
work_27d5tr326feydgcx32t4nhl424	168	228	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	168	229	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	168	230	on	on	IN
work_27d5tr326feydgcx32t4nhl424	168	231	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	168	232	see	see	VBP
work_27d5tr326feydgcx32t4nhl424	168	233	that	that	IN
work_27d5tr326feydgcx32t4nhl424	168	234	among	among	IN
work_27d5tr326feydgcx32t4nhl424	168	235	all	all	PDT
work_27d5tr326feydgcx32t4nhl424	168	236	the	the	DT
work_27d5tr326feydgcx32t4nhl424	168	237	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	168	238	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	168	239	,	,	,
work_27d5tr326feydgcx32t4nhl424	168	240	Hybh-	Hybh-	NNP
work_27d5tr326feydgcx32t4nhl424	168	241	sum	sum	NN
work_27d5tr326feydgcx32t4nhl424	168	242	achieves	achieve	VBZ
work_27d5tr326feydgcx32t4nhl424	168	243	the	the	DT
work_27d5tr326feydgcx32t4nhl424	168	244	best	good	JJS
work_27d5tr326feydgcx32t4nhl424	168	245	result	result	NN
work_27d5tr326feydgcx32t4nhl424	168	246	.	.	.
work_27d5tr326feydgcx32t4nhl424	169	1	This	this	DT
work_27d5tr326feydgcx32t4nhl424	169	2	further	further	JJ
work_27d5tr326feydgcx32t4nhl424	169	3	illus-	illus-	NN
work_27d5tr326feydgcx32t4nhl424	169	4	trates	trate	VBZ
work_27d5tr326feydgcx32t4nhl424	169	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	169	6	advantages	advantage	NNS
work_27d5tr326feydgcx32t4nhl424	169	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	169	8	combining	combine	VBG
work_27d5tr326feydgcx32t4nhl424	169	9	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	169	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	169	11	with	with	IN
work_27d5tr326feydgcx32t4nhl424	169	12	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	169	13	method	method	NN
work_27d5tr326feydgcx32t4nhl424	169	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	170	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	170	2	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	170	3	1	1	CD
work_27d5tr326feydgcx32t4nhl424	170	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	170	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	170	6	can	can	MD
work_27d5tr326feydgcx32t4nhl424	170	7	see	see	VB
work_27d5tr326feydgcx32t4nhl424	170	8	that	that	IN
work_27d5tr326feydgcx32t4nhl424	170	9	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	170	10	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	170	11	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	170	12	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	170	13	model	model	NN
work_27d5tr326feydgcx32t4nhl424	170	14	performs	perform	VBZ
work_27d5tr326feydgcx32t4nhl424	170	15	better	well	JJR
work_27d5tr326feydgcx32t4nhl424	170	16	than	than	IN
work_27d5tr326feydgcx32t4nhl424	170	17	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	170	18	and	and	CC
work_27d5tr326feydgcx32t4nhl424	170	19	the	the	DT
work_27d5tr326feydgcx32t4nhl424	170	20	improvements	improvement	NNS
work_27d5tr326feydgcx32t4nhl424	170	21	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	170	22	3.4	3.4	CD
work_27d5tr326feydgcx32t4nhl424	170	23	%	%	NN
work_27d5tr326feydgcx32t4nhl424	170	24	and	and	CC
work_27d5tr326feydgcx32t4nhl424	170	25	3.7	3.7	CD
work_27d5tr326feydgcx32t4nhl424	170	26	%	%	NN
work_27d5tr326feydgcx32t4nhl424	170	27	with	with	IN
work_27d5tr326feydgcx32t4nhl424	170	28	re-	re-	JJ
work_27d5tr326feydgcx32t4nhl424	170	29	spect	spect	NN
work_27d5tr326feydgcx32t4nhl424	170	30	to	to	IN
work_27d5tr326feydgcx32t4nhl424	170	31	ROUGE-2	ROUGE-2	NNP
work_27d5tr326feydgcx32t4nhl424	170	32	and	and	CC
work_27d5tr326feydgcx32t4nhl424	170	33	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	170	34	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	170	35	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	170	36	on	on	IN
work_27d5tr326feydgcx32t4nhl424	170	37	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	170	38	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	170	39	.	.	.
work_27d5tr326feydgcx32t4nhl424	171	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	171	2	comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	171	3	can	can	MD
work_27d5tr326feydgcx32t4nhl424	171	4	be	be	VB
work_27d5tr326feydgcx32t4nhl424	171	5	extended	extend	VBN
work_27d5tr326feydgcx32t4nhl424	171	6	to	to	IN
work_27d5tr326feydgcx32t4nhl424	171	7	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	171	8	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	171	9	as	as	IN
work_27d5tr326feydgcx32t4nhl424	171	10	shown	show	VBN
work_27d5tr326feydgcx32t4nhl424	171	11	in	in	IN
work_27d5tr326feydgcx32t4nhl424	171	12	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	171	13	2	2	CD
work_27d5tr326feydgcx32t4nhl424	171	14	:	:	:
work_27d5tr326feydgcx32t4nhl424	171	15	the	the	DT
work_27d5tr326feydgcx32t4nhl424	171	16	performance	performance	NN
work_27d5tr326feydgcx32t4nhl424	171	17	of	of	IN
work_27d5tr326feydgcx32t4nhl424	171	18	S-	S-	NNP
work_27d5tr326feydgcx32t4nhl424	171	19	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	171	20	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	171	21	above	above	IN
work_27d5tr326feydgcx32t4nhl424	171	22	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	171	23	by	by	IN
work_27d5tr326feydgcx32t4nhl424	171	24	4.3	4.3	CD
work_27d5tr326feydgcx32t4nhl424	171	25	%	%	NN
work_27d5tr326feydgcx32t4nhl424	171	26	in	in	IN
work_27d5tr326feydgcx32t4nhl424	171	27	ROUGE-2	ROUGE-2	NNP
work_27d5tr326feydgcx32t4nhl424	171	28	and	and	CC
work_27d5tr326feydgcx32t4nhl424	171	29	5.1	5.1	CD
work_27d5tr326feydgcx32t4nhl424	171	30	%	%	NN
work_27d5tr326feydgcx32t4nhl424	171	31	in	in	IN
work_27d5tr326feydgcx32t4nhl424	171	32	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	171	33	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	171	34	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	171	35	.	.	.
work_27d5tr326feydgcx32t4nhl424	172	1	It	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	172	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	172	3	worth	worth	JJ
work_27d5tr326feydgcx32t4nhl424	172	4	explaining	explain	VBG
work_27d5tr326feydgcx32t4nhl424	172	5	that	that	IN
work_27d5tr326feydgcx32t4nhl424	172	6	these	these	DT
work_27d5tr326feydgcx32t4nhl424	172	7	achievements	achievement	NNS
work_27d5tr326feydgcx32t4nhl424	172	8	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	172	9	significant	significant	JJ
work_27d5tr326feydgcx32t4nhl424	172	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	172	11	because	because	IN
work_27d5tr326feydgcx32t4nhl424	172	12	in	in	IN
work_27d5tr326feydgcx32t4nhl424	172	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	172	14	TAC2008	tac2008	NN
work_27d5tr326feydgcx32t4nhl424	172	15	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	172	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	172	17	the	the	DT
work_27d5tr326feydgcx32t4nhl424	172	18	performance	performance	NN
work_27d5tr326feydgcx32t4nhl424	172	19	of	of	IN
work_27d5tr326feydgcx32t4nhl424	172	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	172	21	top	top	JJ
work_27d5tr326feydgcx32t4nhl424	172	22	ranking	rank	VBG
work_27d5tr326feydgcx32t4nhl424	172	23	systems	system	NNS
work_27d5tr326feydgcx32t4nhl424	172	24	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	172	25	very	very	RB
work_27d5tr326feydgcx32t4nhl424	172	26	close	close	JJ
work_27d5tr326feydgcx32t4nhl424	172	27	,	,	,
work_27d5tr326feydgcx32t4nhl424	172	28	i.e.	i.e.	FW
work_27d5tr326feydgcx32t4nhl424	173	1	the	the	DT
work_27d5tr326feydgcx32t4nhl424	173	2	best	good	JJS
work_27d5tr326feydgcx32t4nhl424	173	3	system	system	NN
work_27d5tr326feydgcx32t4nhl424	173	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	173	5	only	only	RB
work_27d5tr326feydgcx32t4nhl424	173	6	4.2	4.2	CD
work_27d5tr326feydgcx32t4nhl424	173	7	%	%	NN
work_27d5tr326feydgcx32t4nhl424	173	8	above	above	IN
work_27d5tr326feydgcx32t4nhl424	173	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	173	10	4th	4th	JJ
work_27d5tr326feydgcx32t4nhl424	173	11	best	good	JJS
work_27d5tr326feydgcx32t4nhl424	173	12	system	system	NN
work_27d5tr326feydgcx32t4nhl424	173	13	on	on	IN
work_27d5tr326feydgcx32t4nhl424	173	14	ROUGE-	ROUGE-	NNP
work_27d5tr326feydgcx32t4nhl424	173	15	2	2	CD
work_27d5tr326feydgcx32t4nhl424	173	16	and	and	CC
work_27d5tr326feydgcx32t4nhl424	173	17	1.2	1.2	CD
work_27d5tr326feydgcx32t4nhl424	173	18	%	%	NN
work_27d5tr326feydgcx32t4nhl424	173	19	on	on	IN
work_27d5tr326feydgcx32t4nhl424	173	20	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	173	21	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	173	22	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	173	23	.	.	.
work_27d5tr326feydgcx32t4nhl424	174	1	5.3	5.3	CD
work_27d5tr326feydgcx32t4nhl424	174	2	Comparison	Comparison	NNP
work_27d5tr326feydgcx32t4nhl424	174	3	with	with	IN
work_27d5tr326feydgcx32t4nhl424	174	4	other	other	JJ
work_27d5tr326feydgcx32t4nhl424	174	5	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	174	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	175	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	175	2	this	this	DT
work_27d5tr326feydgcx32t4nhl424	175	3	subsection	subsection	NN
work_27d5tr326feydgcx32t4nhl424	175	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	175	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	175	6	compare	compare	VBP
work_27d5tr326feydgcx32t4nhl424	175	7	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	175	8	model	model	NN
work_27d5tr326feydgcx32t4nhl424	175	9	with	with	IN
work_27d5tr326feydgcx32t4nhl424	175	10	some	some	DT
work_27d5tr326feydgcx32t4nhl424	175	11	widely	widely	RB
work_27d5tr326feydgcx32t4nhl424	175	12	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	175	13	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	175	14	in	in	IN
work_27d5tr326feydgcx32t4nhl424	175	15	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	175	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	176	1	Manifold	manifold	NN
work_27d5tr326feydgcx32t4nhl424	176	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	176	3	It	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	176	4	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	176	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	176	6	one	one	CD
work_27d5tr326feydgcx32t4nhl424	176	7	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	176	8	layer	layer	NN
work_27d5tr326feydgcx32t4nhl424	176	9	graph	graph	NN
work_27d5tr326feydgcx32t4nhl424	176	10	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	176	11	semi-	semi-	NNP
work_27d5tr326feydgcx32t4nhl424	176	12	supervised	supervise	VBD
work_27d5tr326feydgcx32t4nhl424	176	13	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	176	14	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	176	15	developed	develop	VBN
work_27d5tr326feydgcx32t4nhl424	176	16	by	by	IN
work_27d5tr326feydgcx32t4nhl424	176	17	Wan	Wan	NNP
work_27d5tr326feydgcx32t4nhl424	176	18	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	176	19	al.(2008	al.(2008	NNP
work_27d5tr326feydgcx32t4nhl424	176	20	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	176	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	177	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	177	2	graph	graph	NN
work_27d5tr326feydgcx32t4nhl424	177	3	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	177	4	constructed	construct	VBN
work_27d5tr326feydgcx32t4nhl424	177	5	only	only	RB
work_27d5tr326feydgcx32t4nhl424	177	6	con-	con-	NN
work_27d5tr326feydgcx32t4nhl424	177	7	sidering	sidere	VBG
work_27d5tr326feydgcx32t4nhl424	177	8	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	177	9	relations	relation	NNS
work_27d5tr326feydgcx32t4nhl424	177	10	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	177	11	tf	tf	NNP
work_27d5tr326feydgcx32t4nhl424	177	12	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	177	13	idf	idf	NNP
work_27d5tr326feydgcx32t4nhl424	177	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	177	15	neglects	neglect	VBZ
work_27d5tr326feydgcx32t4nhl424	177	16	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	177	17	information	information	NN
work_27d5tr326feydgcx32t4nhl424	177	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	178	1	LexRank	LexRank	NNP
work_27d5tr326feydgcx32t4nhl424	178	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	178	3	Graph	Graph	NNP
work_27d5tr326feydgcx32t4nhl424	178	4	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	178	5	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	178	6	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	178	7	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	178	8	Erkan	Erkan	NNP
work_27d5tr326feydgcx32t4nhl424	178	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	178	10	Radev	Radev	NNP
work_27d5tr326feydgcx32t4nhl424	178	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	178	12	2004	2004	CD
work_27d5tr326feydgcx32t4nhl424	178	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	178	14	,	,	,
work_27d5tr326feydgcx32t4nhl424	178	15	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	178	16	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	178	17	a	a	DT
work_27d5tr326feydgcx32t4nhl424	178	18	revised	revise	VBN
work_27d5tr326feydgcx32t4nhl424	178	19	version	version	NN
work_27d5tr326feydgcx32t4nhl424	178	20	of	of	IN
work_27d5tr326feydgcx32t4nhl424	178	21	famous	famous	JJ
work_27d5tr326feydgcx32t4nhl424	178	22	web	web	NN
work_27d5tr326feydgcx32t4nhl424	178	23	ranking	rank	VBG
work_27d5tr326feydgcx32t4nhl424	178	24	algorithm	algorithm	JJ
work_27d5tr326feydgcx32t4nhl424	178	25	PageRank	PageRank	NNP
work_27d5tr326feydgcx32t4nhl424	178	26	.	.	.
work_27d5tr326feydgcx32t4nhl424	179	1	It	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	179	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	179	3	an	an	DT
work_27d5tr326feydgcx32t4nhl424	179	4	unsupervised	unsupervised	JJ
work_27d5tr326feydgcx32t4nhl424	179	5	ranking	rank	VBG
work_27d5tr326feydgcx32t4nhl424	179	6	algorithms	algorithm	NNS
work_27d5tr326feydgcx32t4nhl424	179	7	compared	compare	VBN
work_27d5tr326feydgcx32t4nhl424	179	8	with	with	IN
work_27d5tr326feydgcx32t4nhl424	179	9	Manifold	Manifold	NNP
work_27d5tr326feydgcx32t4nhl424	179	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	180	1	SVM	SVM	NNP
work_27d5tr326feydgcx32t4nhl424	180	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	180	3	A	a	DT
work_27d5tr326feydgcx32t4nhl424	180	4	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	180	5	method	method	NN
work_27d5tr326feydgcx32t4nhl424	180	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	180	7	Support	Support	NNP
work_27d5tr326feydgcx32t4nhl424	180	8	Vector	Vector	NNP
work_27d5tr326feydgcx32t4nhl424	180	9	Ma-	Ma-	NNP
work_27d5tr326feydgcx32t4nhl424	180	10	chine	chine	NN
work_27d5tr326feydgcx32t4nhl424	180	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	180	12	SVM	SVM	NNP
work_27d5tr326feydgcx32t4nhl424	180	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	180	14	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	180	15	Vapnik	Vapnik	NNP
work_27d5tr326feydgcx32t4nhl424	180	16	1995	1995	CD
work_27d5tr326feydgcx32t4nhl424	180	17	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	180	18	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	180	19	uses	use	VBZ
work_27d5tr326feydgcx32t4nhl424	180	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	180	21	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	180	22	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	180	23	as	as	IN
work_27d5tr326feydgcx32t4nhl424	180	24	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	180	25	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	180	26	.	.	.
work_27d5tr326feydgcx32t4nhl424	181	1	MEAD	MEAD	NNP
work_27d5tr326feydgcx32t4nhl424	181	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	181	3	A	a	DT
work_27d5tr326feydgcx32t4nhl424	181	4	centroid	centroid	NN
work_27d5tr326feydgcx32t4nhl424	181	5	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	181	6	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	181	7	algorithm	algorithm	NN
work_27d5tr326feydgcx32t4nhl424	181	8	by	by	IN
work_27d5tr326feydgcx32t4nhl424	181	9	Radev	Radev	NNP
work_27d5tr326feydgcx32t4nhl424	181	10	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	181	11	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	181	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	182	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	182	2	2004	2004	CD
work_27d5tr326feydgcx32t4nhl424	182	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	182	4	.	.	.
work_27d5tr326feydgcx32t4nhl424	183	1	Cluster	cluster	NN
work_27d5tr326feydgcx32t4nhl424	183	2	centroids	centroid	NNS
work_27d5tr326feydgcx32t4nhl424	183	3	in	in	IN
work_27d5tr326feydgcx32t4nhl424	183	4	MEAD	MEAD	NNP
work_27d5tr326feydgcx32t4nhl424	183	5	consists	consist	VBZ
work_27d5tr326feydgcx32t4nhl424	183	6	of	of	IN
work_27d5tr326feydgcx32t4nhl424	183	7	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	183	8	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	183	9	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	183	10	central	central	JJ
work_27d5tr326feydgcx32t4nhl424	183	11	not	not	RB
work_27d5tr326feydgcx32t4nhl424	183	12	only	only	RB
work_27d5tr326feydgcx32t4nhl424	183	13	to	to	IN
work_27d5tr326feydgcx32t4nhl424	183	14	one	one	CD
work_27d5tr326feydgcx32t4nhl424	183	15	article	article	NN
work_27d5tr326feydgcx32t4nhl424	183	16	in	in	IN
work_27d5tr326feydgcx32t4nhl424	183	17	a	a	DT
work_27d5tr326feydgcx32t4nhl424	183	18	cluster	cluster	NN
work_27d5tr326feydgcx32t4nhl424	183	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	183	20	but	but	CC
work_27d5tr326feydgcx32t4nhl424	183	21	to	to	IN
work_27d5tr326feydgcx32t4nhl424	183	22	all	all	PDT
work_27d5tr326feydgcx32t4nhl424	183	23	the	the	DT
work_27d5tr326feydgcx32t4nhl424	183	24	articles	article	NNS
work_27d5tr326feydgcx32t4nhl424	183	25	.	.	.
work_27d5tr326feydgcx32t4nhl424	184	1	Similarity	similarity	NN
work_27d5tr326feydgcx32t4nhl424	184	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	184	3	measure	measure	NN
work_27d5tr326feydgcx32t4nhl424	184	4	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	184	5	tf	tf	NNP
work_27d5tr326feydgcx32t4nhl424	184	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	184	7	idf	idf	NNP
work_27d5tr326feydgcx32t4nhl424	184	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	185	1	At	at	IN
work_27d5tr326feydgcx32t4nhl424	185	2	the	the	DT
work_27d5tr326feydgcx32t4nhl424	185	3	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	185	4	time	time	NN
work_27d5tr326feydgcx32t4nhl424	185	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	6	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	185	7	also	also	RB
work_27d5tr326feydgcx32t4nhl424	185	8	present	present	VBP
work_27d5tr326feydgcx32t4nhl424	185	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	185	10	top	top	JJ
work_27d5tr326feydgcx32t4nhl424	185	11	three	three	CD
work_27d5tr326feydgcx32t4nhl424	185	12	participating	participate	VBG
work_27d5tr326feydgcx32t4nhl424	185	13	systems	system	NNS
work_27d5tr326feydgcx32t4nhl424	185	14	with	with	IN
work_27d5tr326feydgcx32t4nhl424	185	15	regard	regard	NN
work_27d5tr326feydgcx32t4nhl424	185	16	to	to	IN
work_27d5tr326feydgcx32t4nhl424	185	17	ROUGE-2	ROUGE-2	NNP
work_27d5tr326feydgcx32t4nhl424	185	18	on	on	IN
work_27d5tr326feydgcx32t4nhl424	185	19	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	185	20	and	and	CC
work_27d5tr326feydgcx32t4nhl424	185	21	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	185	22	for	for	IN
work_27d5tr326feydgcx32t4nhl424	185	23	comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	185	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	25	denoted	denote	VBN
work_27d5tr326feydgcx32t4nhl424	185	26	as	as	IN
work_27d5tr326feydgcx32t4nhl424	185	27	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	185	28	denoted	denote	VBN
work_27d5tr326feydgcx32t4nhl424	185	29	as	as	IN
work_27d5tr326feydgcx32t4nhl424	185	30	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	185	31	1st	1st	NN
work_27d5tr326feydgcx32t4nhl424	185	32	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	33	2nd	2nd	JJ
work_27d5tr326feydgcx32t4nhl424	185	34	and	and	CC
work_27d5tr326feydgcx32t4nhl424	185	35	3rd)(Gillick	3rd)(Gillick	NNP
work_27d5tr326feydgcx32t4nhl424	185	36	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	185	37	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	185	38	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	185	39	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	40	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	185	41	;	;	:
work_27d5tr326feydgcx32t4nhl424	185	42	Zhang	Zhang	NNP
work_27d5tr326feydgcx32t4nhl424	185	43	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	185	44	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	185	45	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	185	46	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	47	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	185	48	;	;	:
work_27d5tr326feydgcx32t4nhl424	185	49	Gillick	Gillick	NNP
work_27d5tr326feydgcx32t4nhl424	185	50	et	et	NNP
work_27d5tr326feydgcx32t4nhl424	185	51	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	185	52	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	185	53	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	54	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	185	55	;	;	:
work_27d5tr326feydgcx32t4nhl424	185	56	Varma	Varma	NNP
work_27d5tr326feydgcx32t4nhl424	185	57	et	et	FW
work_27d5tr326feydgcx32t4nhl424	185	58	al	al	NNP
work_27d5tr326feydgcx32t4nhl424	185	59	.	.	NNP
work_27d5tr326feydgcx32t4nhl424	185	60	,	,	,
work_27d5tr326feydgcx32t4nhl424	185	61	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	185	62	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	185	63	.	.	.
work_27d5tr326feydgcx32t4nhl424	186	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	186	2	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	186	3	scores	score	NNS
work_27d5tr326feydgcx32t4nhl424	186	4	of	of	IN
work_27d5tr326feydgcx32t4nhl424	186	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	186	6	top	top	JJ
work_27d5tr326feydgcx32t4nhl424	186	7	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	186	8	system	system	NN
work_27d5tr326feydgcx32t4nhl424	186	9	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	186	10	directly	directly	RB
work_27d5tr326feydgcx32t4nhl424	186	11	provided	provide	VBN
work_27d5tr326feydgcx32t4nhl424	186	12	by	by	IN
work_27d5tr326feydgcx32t4nhl424	186	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	186	14	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	186	15	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	186	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	187	1	From	from	IN
work_27d5tr326feydgcx32t4nhl424	187	2	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	187	3	3	3	CD
work_27d5tr326feydgcx32t4nhl424	187	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	187	5	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	187	6	4	4	CD
work_27d5tr326feydgcx32t4nhl424	187	7	,	,	,
work_27d5tr326feydgcx32t4nhl424	187	8	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	187	9	can	can	MD
work_27d5tr326feydgcx32t4nhl424	187	10	see	see	VB
work_27d5tr326feydgcx32t4nhl424	187	11	that	that	IN
work_27d5tr326feydgcx32t4nhl424	187	12	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	187	13	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	187	14	outperforms	outperform	VBZ
work_27d5tr326feydgcx32t4nhl424	187	15	the	the	DT
work_27d5tr326feydgcx32t4nhl424	187	16	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	187	17	in	in	IN
work_27d5tr326feydgcx32t4nhl424	187	18	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	187	19	of	of	IN
work_27d5tr326feydgcx32t4nhl424	187	20	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	187	21	metrics	metric	NNS
work_27d5tr326feydgcx32t4nhl424	187	22	consistently	consistently	RB
work_27d5tr326feydgcx32t4nhl424	187	23	.	.	.
work_27d5tr326feydgcx32t4nhl424	188	1	When	when	WRB
work_27d5tr326feydgcx32t4nhl424	188	2	compared	compare	VBN
work_27d5tr326feydgcx32t4nhl424	188	3	with	with	IN
work_27d5tr326feydgcx32t4nhl424	188	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	188	5	standard	standard	JJ
work_27d5tr326feydgcx32t4nhl424	188	6	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	188	7	method	method	NN
work_27d5tr326feydgcx32t4nhl424	188	8	SVM	SVM	NNP
work_27d5tr326feydgcx32t4nhl424	188	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	188	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	188	11	relative	relative	JJ
work_27d5tr326feydgcx32t4nhl424	188	12	improvements	improvement	NNS
work_27d5tr326feydgcx32t4nhl424	188	13	over	over	IN
work_27d5tr326feydgcx32t4nhl424	188	14	the	the	DT
work_27d5tr326feydgcx32t4nhl424	188	15	ROUGE-1	ROUGE-1	NNP
work_27d5tr326feydgcx32t4nhl424	188	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	188	17	ROUGE-2	ROUGE-2	NNP
work_27d5tr326feydgcx32t4nhl424	188	18	and	and	CC
work_27d5tr326feydgcx32t4nhl424	188	19	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	188	20	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	188	21	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	188	22	scores	score	NNS
work_27d5tr326feydgcx32t4nhl424	188	23	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	188	24	4.3	4.3	CD
work_27d5tr326feydgcx32t4nhl424	188	25	%	%	NN
work_27d5tr326feydgcx32t4nhl424	188	26	,	,	,
work_27d5tr326feydgcx32t4nhl424	188	27	13.1	13.1	CD
work_27d5tr326feydgcx32t4nhl424	188	28	%	%	NN
work_27d5tr326feydgcx32t4nhl424	188	29	,	,	,
work_27d5tr326feydgcx32t4nhl424	188	30	8.3	8.3	CD
work_27d5tr326feydgcx32t4nhl424	188	31	%	%	NN
work_27d5tr326feydgcx32t4nhl424	188	32	respec-	respec-	NN
work_27d5tr326feydgcx32t4nhl424	188	33	tively	tively	RB
work_27d5tr326feydgcx32t4nhl424	188	34	on	on	IN
work_27d5tr326feydgcx32t4nhl424	188	35	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	188	36	and	and	CC
work_27d5tr326feydgcx32t4nhl424	188	37	7.2	7.2	CD
work_27d5tr326feydgcx32t4nhl424	188	38	%	%	NN
work_27d5tr326feydgcx32t4nhl424	188	39	,	,	,
work_27d5tr326feydgcx32t4nhl424	188	40	14.9	14.9	CD
work_27d5tr326feydgcx32t4nhl424	188	41	%	%	NN
work_27d5tr326feydgcx32t4nhl424	188	42	,	,	,
work_27d5tr326feydgcx32t4nhl424	188	43	14.3	14.3	CD
work_27d5tr326feydgcx32t4nhl424	188	44	%	%	NN
work_27d5tr326feydgcx32t4nhl424	188	45	on	on	IN
work_27d5tr326feydgcx32t4nhl424	188	46	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	188	47	.	.	.
work_27d5tr326feydgcx32t4nhl424	189	1	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	189	2	model	model	NN
work_27d5tr326feydgcx32t4nhl424	189	3	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	189	4	not	not	RB
work_27d5tr326feydgcx32t4nhl424	189	5	as	as	RB
work_27d5tr326feydgcx32t4nhl424	189	6	good	good	JJ
work_27d5tr326feydgcx32t4nhl424	189	7	as	as	IN
work_27d5tr326feydgcx32t4nhl424	189	8	top	top	JJ
work_27d5tr326feydgcx32t4nhl424	189	9	par-	par-	NN
work_27d5tr326feydgcx32t4nhl424	189	10	ticipating	ticipating	NN
work_27d5tr326feydgcx32t4nhl424	189	11	systems	system	NNS
work_27d5tr326feydgcx32t4nhl424	189	12	on	on	IN
work_27d5tr326feydgcx32t4nhl424	189	13	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	189	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	189	15	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	189	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	190	1	But	but	CC
work_27d5tr326feydgcx32t4nhl424	190	2	considering	consider	VBG
work_27d5tr326feydgcx32t4nhl424	190	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	190	4	fact	fact	NN
work_27d5tr326feydgcx32t4nhl424	190	5	that	that	IN
work_27d5tr326feydgcx32t4nhl424	190	6	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	190	7	model	model	NN
work_27d5tr326feydgcx32t4nhl424	190	8	neither	neither	CC
work_27d5tr326feydgcx32t4nhl424	190	9	uses	use	VBZ
work_27d5tr326feydgcx32t4nhl424	190	10	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	190	11	tence	tence	NN
work_27d5tr326feydgcx32t4nhl424	190	12	compression	compression	NN
work_27d5tr326feydgcx32t4nhl424	190	13	algorithm	algorithm	NNP
work_27d5tr326feydgcx32t4nhl424	190	14	nor	nor	CC
work_27d5tr326feydgcx32t4nhl424	190	15	leverage	leverage	NN
work_27d5tr326feydgcx32t4nhl424	190	16	domain	domain	NN
work_27d5tr326feydgcx32t4nhl424	190	17	knowledge	knowledge	NN
work_27d5tr326feydgcx32t4nhl424	190	18	bases	basis	NNS
work_27d5tr326feydgcx32t4nhl424	190	19	like	like	IN
work_27d5tr326feydgcx32t4nhl424	190	20	Wikipedia	Wikipedia	NNP
work_27d5tr326feydgcx32t4nhl424	190	21	or	or	CC
work_27d5tr326feydgcx32t4nhl424	190	22	training	training	NN
work_27d5tr326feydgcx32t4nhl424	190	23	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	190	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	190	25	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	190	26	small	small	JJ
work_27d5tr326feydgcx32t4nhl424	190	27	difference	difference	NN
work_27d5tr326feydgcx32t4nhl424	190	28	in	in	IN
work_27d5tr326feydgcx32t4nhl424	190	29	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	190	30	scores	score	NNS
work_27d5tr326feydgcx32t4nhl424	190	31	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	190	32	reason-	reason-	NN
work_27d5tr326feydgcx32t4nhl424	190	33	able	able	JJ
work_27d5tr326feydgcx32t4nhl424	190	34	.	.	.
work_27d5tr326feydgcx32t4nhl424	191	1	5.4	5.4	CD
work_27d5tr326feydgcx32t4nhl424	191	2	Manual	Manual	NNP
work_27d5tr326feydgcx32t4nhl424	191	3	Evaluations	Evaluations	NNPS
work_27d5tr326feydgcx32t4nhl424	191	4	In	in	IN
work_27d5tr326feydgcx32t4nhl424	191	5	order	order	NN
work_27d5tr326feydgcx32t4nhl424	191	6	to	to	TO
work_27d5tr326feydgcx32t4nhl424	191	7	obtain	obtain	VB
work_27d5tr326feydgcx32t4nhl424	191	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	191	9	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	191	10	accurate	accurate	JJ
work_27d5tr326feydgcx32t4nhl424	191	11	measure	measure	NN
work_27d5tr326feydgcx32t4nhl424	191	12	of	of	IN
work_27d5tr326feydgcx32t4nhl424	191	13	sum-	sum-	JJ
work_27d5tr326feydgcx32t4nhl424	191	14	mary	mary	NNP
work_27d5tr326feydgcx32t4nhl424	191	15	quality	quality	NN
work_27d5tr326feydgcx32t4nhl424	191	16	for	for	IN
work_27d5tr326feydgcx32t4nhl424	191	17	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	191	18	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	191	19	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	191	20	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	191	21	model	model	NN
work_27d5tr326feydgcx32t4nhl424	191	22	and	and	CC
work_27d5tr326feydgcx32t4nhl424	191	23	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	191	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	191	25	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	191	26	performed	perform	VBD
work_27d5tr326feydgcx32t4nhl424	191	27	a	a	DT
work_27d5tr326feydgcx32t4nhl424	191	28	simple	simple	JJ
work_27d5tr326feydgcx32t4nhl424	191	29	user	user	NN
work_27d5tr326feydgcx32t4nhl424	191	30	study	study	NN
work_27d5tr326feydgcx32t4nhl424	191	31	concerning	concern	VBG
work_27d5tr326feydgcx32t4nhl424	191	32	the	the	DT
work_27d5tr326feydgcx32t4nhl424	191	33	following	following	JJ
work_27d5tr326feydgcx32t4nhl424	191	34	aspects	aspect	NNS
work_27d5tr326feydgcx32t4nhl424	191	35	:	:	:
work_27d5tr326feydgcx32t4nhl424	191	36	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	191	37	1	1	LS
work_27d5tr326feydgcx32t4nhl424	191	38	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	191	39	Overall	overall	JJ
work_27d5tr326feydgcx32t4nhl424	191	40	quality	quality	NN
work_27d5tr326feydgcx32t4nhl424	191	41	:	:	:
work_27d5tr326feydgcx32t4nhl424	191	42	Which	which	WDT
work_27d5tr326feydgcx32t4nhl424	191	43	sum-	sum-	VBD
work_27d5tr326feydgcx32t4nhl424	191	44	mary	mary	NNP
work_27d5tr326feydgcx32t4nhl424	191	45	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	191	46	better	well	JJR
work_27d5tr326feydgcx32t4nhl424	191	47	overall	overall	JJ
work_27d5tr326feydgcx32t4nhl424	191	48	?	?	.
work_27d5tr326feydgcx32t4nhl424	192	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	192	2	2	2	LS
work_27d5tr326feydgcx32t4nhl424	192	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	192	4	Focus	focus	NN
work_27d5tr326feydgcx32t4nhl424	192	5	:	:	:
work_27d5tr326feydgcx32t4nhl424	192	6	Which	which	WDT
work_27d5tr326feydgcx32t4nhl424	192	7	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	192	8	contains	contain	VBZ
work_27d5tr326feydgcx32t4nhl424	192	9	less	less	RBR
work_27d5tr326feydgcx32t4nhl424	192	10	irrelevant	irrelevant	JJ
work_27d5tr326feydgcx32t4nhl424	192	11	content	content	NN
work_27d5tr326feydgcx32t4nhl424	192	12	?	?	.
work_27d5tr326feydgcx32t4nhl424	193	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	193	2	3)Responsiveness	3)responsiveness	NN
work_27d5tr326feydgcx32t4nhl424	193	3	:	:	:
work_27d5tr326feydgcx32t4nhl424	193	4	Which	which	WDT
work_27d5tr326feydgcx32t4nhl424	193	5	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	193	6	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	193	7	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	193	8	responsive	responsive	JJ
work_27d5tr326feydgcx32t4nhl424	193	9	to	to	IN
work_27d5tr326feydgcx32t4nhl424	193	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	193	11	query	query	NN
work_27d5tr326feydgcx32t4nhl424	193	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	194	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	194	2	4	4	LS
work_27d5tr326feydgcx32t4nhl424	194	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	194	4	Non	non	JJ
work_27d5tr326feydgcx32t4nhl424	194	5	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	194	6	Redundancy	redundancy	NN
work_27d5tr326feydgcx32t4nhl424	194	7	:	:	:
work_27d5tr326feydgcx32t4nhl424	194	8	Which	which	WDT
work_27d5tr326feydgcx32t4nhl424	194	9	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	194	10	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	194	11	less	less	RBR
work_27d5tr326feydgcx32t4nhl424	194	12	re-	re-	RB
work_27d5tr326feydgcx32t4nhl424	194	13	dundant	dundant	JJ
work_27d5tr326feydgcx32t4nhl424	194	14	?	?	.
work_27d5tr326feydgcx32t4nhl424	195	1	8	8	CD
work_27d5tr326feydgcx32t4nhl424	195	2	judges	judge	NNS
work_27d5tr326feydgcx32t4nhl424	195	3	who	who	WP
work_27d5tr326feydgcx32t4nhl424	195	4	specialize	specialize	VBP
work_27d5tr326feydgcx32t4nhl424	195	5	in	in	IN
work_27d5tr326feydgcx32t4nhl424	195	6	NLP	NLP	NNP
work_27d5tr326feydgcx32t4nhl424	195	7	partic-	partic-	NN
work_27d5tr326feydgcx32t4nhl424	195	8	ipated	ipate	VBN
work_27d5tr326feydgcx32t4nhl424	195	9	in	in	IN
work_27d5tr326feydgcx32t4nhl424	195	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	195	11	blind	blind	JJ
work_27d5tr326feydgcx32t4nhl424	195	12	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	195	13	task	task	NN
work_27d5tr326feydgcx32t4nhl424	195	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	196	1	Evaluators	evaluator	NNS
work_27d5tr326feydgcx32t4nhl424	196	2	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	196	3	presented	present	VBN
work_27d5tr326feydgcx32t4nhl424	196	4	with	with	IN
work_27d5tr326feydgcx32t4nhl424	196	5	two	two	CD
work_27d5tr326feydgcx32t4nhl424	196	6	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	196	7	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	196	8	by	by	IN
work_27d5tr326feydgcx32t4nhl424	196	9	S	s	NN
work_27d5tr326feydgcx32t4nhl424	196	10	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	11	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	196	12	95	95	CD
work_27d5tr326feydgcx32t4nhl424	196	13	Methods	Methods	NNP
work_27d5tr326feydgcx32t4nhl424	196	14	ROUGE-1	rouge-1	NN
work_27d5tr326feydgcx32t4nhl424	196	15	ROUGE-2	rouge-2	NN
work_27d5tr326feydgcx32t4nhl424	196	16	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	196	17	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	18	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	196	19	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	196	20	0.3724	0.3724	CD
work_27d5tr326feydgcx32t4nhl424	196	21	0.1030	0.1030	CD
work_27d5tr326feydgcx32t4nhl424	196	22	0.1342	0.1342	CD
work_27d5tr326feydgcx32t4nhl424	196	23	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	196	24	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	25	0.3660	0.3660	CD
work_27d5tr326feydgcx32t4nhl424	196	26	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	196	27	0.3788	0.3788	CD
work_27d5tr326feydgcx32t4nhl424	196	28	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	29	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	30	0.0999	0.0999	CD
work_27d5tr326feydgcx32t4nhl424	196	31	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	32	0.1061	0.1061	CD
work_27d5tr326feydgcx32t4nhl424	196	33	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	34	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	35	0.1290	0.1290	CD
work_27d5tr326feydgcx32t4nhl424	196	36	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	37	0.1394	0.1394	CD
work_27d5tr326feydgcx32t4nhl424	196	38	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	39	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	40	1st	1st	NN
work_27d5tr326feydgcx32t4nhl424	196	41	0.3742	0.3742	CD
work_27d5tr326feydgcx32t4nhl424	196	42	0.1039	0.1039	CD
work_27d5tr326feydgcx32t4nhl424	196	43	0.1364	0.1364	CD
work_27d5tr326feydgcx32t4nhl424	196	44	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	45	0.3639	0.3639	CD
work_27d5tr326feydgcx32t4nhl424	196	46	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	47	0.3845	0.3845	CD
work_27d5tr326feydgcx32t4nhl424	196	48	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	49	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	50	0.0974	0.0974	CD
work_27d5tr326feydgcx32t4nhl424	196	51	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	52	0.1104	0.1104	CD
work_27d5tr326feydgcx32t4nhl424	196	53	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	54	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	55	0.1285	0.1285	CD
work_27d5tr326feydgcx32t4nhl424	196	56	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	57	0.1443	0.1443	CD
work_27d5tr326feydgcx32t4nhl424	196	58	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	59	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	60	2nd	2nd	JJ
work_27d5tr326feydgcx32t4nhl424	196	61	0.3717	0.3717	CD
work_27d5tr326feydgcx32t4nhl424	196	62	0.0990	0.0990	CD
work_27d5tr326feydgcx32t4nhl424	196	63	0.1326	0.1326	CD
work_27d5tr326feydgcx32t4nhl424	196	64	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	65	0.3610	0.3610	NNP
work_27d5tr326feydgcx32t4nhl424	196	66	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	67	0.3824	0.3824	CD
work_27d5tr326feydgcx32t4nhl424	196	68	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	69	0.0944	0.0944	CD
work_27d5tr326feydgcx32t4nhl424	196	70	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	71	0.1038	0.1038	CD
work_27d5tr326feydgcx32t4nhl424	196	72	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	73	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	74	0.1269	0.1269	IN
work_27d5tr326feydgcx32t4nhl424	196	75	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	76	0.1385	0.1385	NNS
work_27d5tr326feydgcx32t4nhl424	196	77	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	78	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	79	3rd	3rd	JJ
work_27d5tr326feydgcx32t4nhl424	196	80	0.3710	0.3710	CD
work_27d5tr326feydgcx32t4nhl424	196	81	0.0977	0.0977	CD
work_27d5tr326feydgcx32t4nhl424	196	82	0.1329	0.1329	CD
work_27d5tr326feydgcx32t4nhl424	196	83	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	84	0.3550	0.3550	CD
work_27d5tr326feydgcx32t4nhl424	196	85	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	86	0.3849	0.3849	CD
work_27d5tr326feydgcx32t4nhl424	196	87	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	88	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	89	0.0920	0.0920	CD
work_27d5tr326feydgcx32t4nhl424	196	90	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	196	91	0.1034	0.1034	CD
work_27d5tr326feydgcx32t4nhl424	196	92	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	93	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	94	0.1267	0.1267	CD
work_27d5tr326feydgcx32t4nhl424	196	95	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	96	0.1391	0.1391	CD
work_27d5tr326feydgcx32t4nhl424	196	97	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	98	PageRank	PageRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	99	0.3597	0.3597	CD
work_27d5tr326feydgcx32t4nhl424	196	100	0.0879	0.0879	CD
work_27d5tr326feydgcx32t4nhl424	196	101	0.1221	0.1221	CD
work_27d5tr326feydgcx32t4nhl424	196	102	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	103	0.3499	0.3499	CD
work_27d5tr326feydgcx32t4nhl424	196	104	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	105	0.3695	0.3695	NNP
work_27d5tr326feydgcx32t4nhl424	196	106	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	107	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	108	0.0809	0.0809	NNP
work_27d5tr326feydgcx32t4nhl424	196	109	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	110	0.0950	0.0950	NNP
work_27d5tr326feydgcx32t4nhl424	196	111	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	112	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	113	0.1173	0.1173	CD
work_27d5tr326feydgcx32t4nhl424	196	114	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	115	0.1269	0.1269	NNP
work_27d5tr326feydgcx32t4nhl424	196	116	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	117	Manifold	manifold	NN
work_27d5tr326feydgcx32t4nhl424	196	118	0.3621	0.3621	CD
work_27d5tr326feydgcx32t4nhl424	196	119	0.0931	0.0931	CD
work_27d5tr326feydgcx32t4nhl424	196	120	0.1243	0.1243	CD
work_27d5tr326feydgcx32t4nhl424	196	121	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	122	0.3506	0.3506	CD
work_27d5tr326feydgcx32t4nhl424	196	123	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	196	124	0.3736	0.3736	CD
work_27d5tr326feydgcx32t4nhl424	196	125	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	126	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	127	0.0868	0.0868	CD
work_27d5tr326feydgcx32t4nhl424	196	128	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	129	0.0994	0.0994	CD
work_27d5tr326feydgcx32t4nhl424	196	130	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	131	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	132	0.1206	0.1206	CD
work_27d5tr326feydgcx32t4nhl424	196	133	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	134	0.1280	0.1280	CD
work_27d5tr326feydgcx32t4nhl424	196	135	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	136	SVM	SVM	NNP
work_27d5tr326feydgcx32t4nhl424	196	137	0.3588	0.3588	CD
work_27d5tr326feydgcx32t4nhl424	196	138	0.0921	0.0921	CD
work_27d5tr326feydgcx32t4nhl424	196	139	0.1258	0.1258	CD
work_27d5tr326feydgcx32t4nhl424	196	140	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	141	0.3489	0.3489	CD
work_27d5tr326feydgcx32t4nhl424	196	142	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	143	0.3687	0.3687	CD
work_27d5tr326feydgcx32t4nhl424	196	144	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	145	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	146	0.0882	0.0882	CD
work_27d5tr326feydgcx32t4nhl424	196	147	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	148	0.0960	0.0960	CD
work_27d5tr326feydgcx32t4nhl424	196	149	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	150	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	151	0.1204	0.1204	CD
work_27d5tr326feydgcx32t4nhl424	196	152	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	153	0.1302	0.1302	CD
work_27d5tr326feydgcx32t4nhl424	196	154	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	155	MEAD	MEAD	NNS
work_27d5tr326feydgcx32t4nhl424	196	156	0.3558	0.3558	CD
work_27d5tr326feydgcx32t4nhl424	196	157	0.0917	0.0917	CD
work_27d5tr326feydgcx32t4nhl424	196	158	0.1226	0.1226	CD
work_27d5tr326feydgcx32t4nhl424	196	159	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	160	0.3489	0.3489	CD
work_27d5tr326feydgcx32t4nhl424	196	161	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	162	0.3627	0.3627	CD
work_27d5tr326feydgcx32t4nhl424	196	163	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	164	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	165	0.0882	0.0882	CD
work_27d5tr326feydgcx32t4nhl424	196	166	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	167	0.0952	0.0952	CD
work_27d5tr326feydgcx32t4nhl424	196	168	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	169	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	170	0.1174	0.1174	CD
work_27d5tr326feydgcx32t4nhl424	196	171	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	172	0.1278	0.1278	CD
work_27d5tr326feydgcx32t4nhl424	196	173	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	174	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	196	175	3	3	CD
work_27d5tr326feydgcx32t4nhl424	196	176	:	:	:
work_27d5tr326feydgcx32t4nhl424	196	177	Comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	196	178	with	with	IN
work_27d5tr326feydgcx32t4nhl424	196	179	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	196	180	on	on	IN
work_27d5tr326feydgcx32t4nhl424	196	181	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	196	182	Methods	Methods	NNP
work_27d5tr326feydgcx32t4nhl424	196	183	ROUGE-1	rouge-1	NN
work_27d5tr326feydgcx32t4nhl424	196	184	ROUGE-2	rouge-2	NN
work_27d5tr326feydgcx32t4nhl424	196	185	ROUGE	ROUGE	NNP
work_27d5tr326feydgcx32t4nhl424	196	186	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	187	SU4	SU4	NNP
work_27d5tr326feydgcx32t4nhl424	196	188	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	196	189	0.3903	0.3903	CD
work_27d5tr326feydgcx32t4nhl424	196	190	0.1223	0.1223	CD
work_27d5tr326feydgcx32t4nhl424	196	191	0.1488	0.1488	CD
work_27d5tr326feydgcx32t4nhl424	196	192	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	196	193	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	194	0.3819	0.3819	CD
work_27d5tr326feydgcx32t4nhl424	196	195	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	196	196	0.3987	0.3987	CD
work_27d5tr326feydgcx32t4nhl424	196	197	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	198	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	199	0.1167	0.1167	CD
work_27d5tr326feydgcx32t4nhl424	196	200	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	196	201	0.1279	0.1279	CD
work_27d5tr326feydgcx32t4nhl424	196	202	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	203	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	204	0.1446	0.1446	IN
work_27d5tr326feydgcx32t4nhl424	196	205	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	206	0.1530	0.1530	CD
work_27d5tr326feydgcx32t4nhl424	196	207	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	208	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	209	1st	1st	NN
work_27d5tr326feydgcx32t4nhl424	196	210	0.3917	0.3917	CD
work_27d5tr326feydgcx32t4nhl424	196	211	0.1218	0.1218	CD
work_27d5tr326feydgcx32t4nhl424	196	212	0.1505	0.1505	CD
work_27d5tr326feydgcx32t4nhl424	196	213	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	214	0.3778	0.3778	CD
work_27d5tr326feydgcx32t4nhl424	196	215	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	216	0.4057	0.4057	CD
work_27d5tr326feydgcx32t4nhl424	196	217	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	218	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	219	0.1122	0.1122	NNP
work_27d5tr326feydgcx32t4nhl424	196	220	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	221	0.1314	0.1314	CD
work_27d5tr326feydgcx32t4nhl424	196	222	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	223	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	224	0.1414	0.1414	CD
work_27d5tr326feydgcx32t4nhl424	196	225	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	226	0.1596	0.1596	CD
work_27d5tr326feydgcx32t4nhl424	196	227	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	228	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	229	2nd	2nd	JJ
work_27d5tr326feydgcx32t4nhl424	196	230	0.3914	0.3914	CD
work_27d5tr326feydgcx32t4nhl424	196	231	0.1212	0.1212	CD
work_27d5tr326feydgcx32t4nhl424	196	232	0.1513	0.1513	CD
work_27d5tr326feydgcx32t4nhl424	196	233	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	234	0.3808	0.3808	CD
work_27d5tr326feydgcx32t4nhl424	196	235	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	236	0.4020	0.4020	CD
work_27d5tr326feydgcx32t4nhl424	196	237	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	238	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	239	0.1147	0.1147	CD
work_27d5tr326feydgcx32t4nhl424	196	240	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	241	0.1277	0.1277	CD
work_27d5tr326feydgcx32t4nhl424	196	242	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	243	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	244	0.1455	0.1455	CD
work_27d5tr326feydgcx32t4nhl424	196	245	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	246	0.1571	0.1571	CD
work_27d5tr326feydgcx32t4nhl424	196	247	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	248	SysRank	SysRank	NNP
work_27d5tr326feydgcx32t4nhl424	196	249	3rd	3rd	JJ
work_27d5tr326feydgcx32t4nhl424	196	250	0.3851	0.3851	CD
work_27d5tr326feydgcx32t4nhl424	196	251	0.1084	0.1084	CD
work_27d5tr326feydgcx32t4nhl424	196	252	0.1447	0.1447	CD
work_27d5tr326feydgcx32t4nhl424	196	253	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	254	0.3762	0.3762	CD
work_27d5tr326feydgcx32t4nhl424	196	255	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	256	0.3932	0.3932	CD
work_27d5tr326feydgcx32t4nhl424	196	257	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	258	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	259	0.1025	0.1025	CD
work_27d5tr326feydgcx32t4nhl424	196	260	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	196	261	0.1144	0.1144	CD
work_27d5tr326feydgcx32t4nhl424	196	262	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	263	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	264	0.1398	0.1398	CD
work_27d5tr326feydgcx32t4nhl424	196	265	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	266	0.1496	0.1496	CD
work_27d5tr326feydgcx32t4nhl424	196	267	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	268	PageRank	pagerank	WRB
work_27d5tr326feydgcx32t4nhl424	196	269	0.3616	0.3616	CD
work_27d5tr326feydgcx32t4nhl424	196	270	0.0849	0.0849	CD
work_27d5tr326feydgcx32t4nhl424	196	271	0.1249	0.1249	CD
work_27d5tr326feydgcx32t4nhl424	196	272	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	273	0.3532	0.3532	CD
work_27d5tr326feydgcx32t4nhl424	196	274	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	275	0.3700	0.3700	CD
work_27d5tr326feydgcx32t4nhl424	196	276	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	277	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	278	0.0802	0.0802	CD
work_27d5tr326feydgcx32t4nhl424	196	279	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	280	0.0896	0.0896	CD
work_27d5tr326feydgcx32t4nhl424	196	281	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	282	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	283	0.1221	0.1221	CD
work_27d5tr326feydgcx32t4nhl424	196	284	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	285	0.1277	0.1277	CD
work_27d5tr326feydgcx32t4nhl424	196	286	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	287	Manifold	Manifold	NNP
work_27d5tr326feydgcx32t4nhl424	196	288	0.3713	0.3713	CD
work_27d5tr326feydgcx32t4nhl424	196	289	0.1014	0.1014	CD
work_27d5tr326feydgcx32t4nhl424	196	290	0.1342	0.1342	CD
work_27d5tr326feydgcx32t4nhl424	196	291	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	292	0.3586	0.3586	CD
work_27d5tr326feydgcx32t4nhl424	196	293	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	294	0.3841	0.3841	CD
work_27d5tr326feydgcx32t4nhl424	196	295	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	296	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	297	0.0950	0.0950	CD
work_27d5tr326feydgcx32t4nhl424	196	298	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	299	0.1178	0.1178	CD
work_27d5tr326feydgcx32t4nhl424	196	300	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	301	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	302	0.1299	0.1299	CD
work_27d5tr326feydgcx32t4nhl424	196	303	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	304	0.1385	0.1385	NNS
work_27d5tr326feydgcx32t4nhl424	196	305	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	306	SVM	SVM	NNP
work_27d5tr326feydgcx32t4nhl424	196	307	0.3649	0.3649	CD
work_27d5tr326feydgcx32t4nhl424	196	308	0.1028	0.1028	CD
work_27d5tr326feydgcx32t4nhl424	196	309	0.1319	0.1319	CD
work_27d5tr326feydgcx32t4nhl424	196	310	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	311	0.3536	0.3536	CD
work_27d5tr326feydgcx32t4nhl424	196	312	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	313	0.3762	0.3762	CD
work_27d5tr326feydgcx32t4nhl424	196	314	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	315	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	316	0.0957	0.0957	CD
work_27d5tr326feydgcx32t4nhl424	196	317	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	318	0.1099	0.1099	CD
work_27d5tr326feydgcx32t4nhl424	196	319	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	320	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	321	0.1258	0.1258	CD
work_27d5tr326feydgcx32t4nhl424	196	322	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	323	0.1380	0.1380	CD
work_27d5tr326feydgcx32t4nhl424	196	324	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	325	MEAD	MEAD	NNS
work_27d5tr326feydgcx32t4nhl424	196	326	0.3601	0.3601	CD
work_27d5tr326feydgcx32t4nhl424	196	327	0.1001	0.1001	CD
work_27d5tr326feydgcx32t4nhl424	196	328	0.1287	0.1287	CD
work_27d5tr326feydgcx32t4nhl424	196	329	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	330	0.3536	0.3536	CD
work_27d5tr326feydgcx32t4nhl424	196	331	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	332	0.3666	0.3666	CD
work_27d5tr326feydgcx32t4nhl424	196	333	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	334	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	335	0.0953	0.0953	CD
work_27d5tr326feydgcx32t4nhl424	196	336	-	-	:
work_27d5tr326feydgcx32t4nhl424	196	337	0.1049	0.1049	CD
work_27d5tr326feydgcx32t4nhl424	196	338	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	339	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	196	340	0.1228	0.1228	CD
work_27d5tr326feydgcx32t4nhl424	196	341	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	196	342	0.1346	0.1346	CD
work_27d5tr326feydgcx32t4nhl424	196	343	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	196	344	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	196	345	4	4	CD
work_27d5tr326feydgcx32t4nhl424	196	346	:	:	:
work_27d5tr326feydgcx32t4nhl424	196	347	Comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	196	348	with	with	IN
work_27d5tr326feydgcx32t4nhl424	196	349	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	196	350	on	on	IN
work_27d5tr326feydgcx32t4nhl424	196	351	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	196	352	and	and	CC
work_27d5tr326feydgcx32t4nhl424	196	353	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	196	354	,	,	,
work_27d5tr326feydgcx32t4nhl424	196	355	as	as	RB
work_27d5tr326feydgcx32t4nhl424	196	356	well	well	RB
work_27d5tr326feydgcx32t4nhl424	196	357	as	as	IN
work_27d5tr326feydgcx32t4nhl424	196	358	the	the	DT
work_27d5tr326feydgcx32t4nhl424	196	359	four	four	CD
work_27d5tr326feydgcx32t4nhl424	196	360	questions	question	NNS
work_27d5tr326feydgcx32t4nhl424	196	361	above	above	RB
work_27d5tr326feydgcx32t4nhl424	196	362	.	.	.
work_27d5tr326feydgcx32t4nhl424	197	1	Then	then	RB
work_27d5tr326feydgcx32t4nhl424	197	2	they	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	197	3	need	need	VBP
work_27d5tr326feydgcx32t4nhl424	197	4	to	to	TO
work_27d5tr326feydgcx32t4nhl424	197	5	answer	answer	VB
work_27d5tr326feydgcx32t4nhl424	197	6	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	197	7	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	197	8	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	197	9	better	well	JJR
work_27d5tr326feydgcx32t4nhl424	197	10	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	197	11	tie	tie	NN
work_27d5tr326feydgcx32t4nhl424	197	12	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	197	13	.	.	.
work_27d5tr326feydgcx32t4nhl424	198	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	198	2	randomly	randomly	RB
work_27d5tr326feydgcx32t4nhl424	198	3	select	select	VBP
work_27d5tr326feydgcx32t4nhl424	198	4	20	20	CD
work_27d5tr326feydgcx32t4nhl424	198	5	document	document	NN
work_27d5tr326feydgcx32t4nhl424	198	6	collections	collection	NNS
work_27d5tr326feydgcx32t4nhl424	198	7	from	from	IN
work_27d5tr326feydgcx32t4nhl424	198	8	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	198	9	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	198	10	data	datum	NNS
work_27d5tr326feydgcx32t4nhl424	198	11	and	and	CC
work_27d5tr326feydgcx32t4nhl424	198	12	randomly	randomly	RB
work_27d5tr326feydgcx32t4nhl424	198	13	assign	assign	VB
work_27d5tr326feydgcx32t4nhl424	198	14	two	two	CD
work_27d5tr326feydgcx32t4nhl424	198	15	sum-	sum-	JJ
work_27d5tr326feydgcx32t4nhl424	198	16	maries	mary	NNS
work_27d5tr326feydgcx32t4nhl424	198	17	for	for	IN
work_27d5tr326feydgcx32t4nhl424	198	18	each	each	DT
work_27d5tr326feydgcx32t4nhl424	198	19	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	198	20	to	to	IN
work_27d5tr326feydgcx32t4nhl424	198	21	three	three	CD
work_27d5tr326feydgcx32t4nhl424	198	22	different	different	JJ
work_27d5tr326feydgcx32t4nhl424	198	23	evalua-	evalua-	NN
work_27d5tr326feydgcx32t4nhl424	198	24	tors	tor	NNS
work_27d5tr326feydgcx32t4nhl424	198	25	to	to	TO
work_27d5tr326feydgcx32t4nhl424	198	26	judge	judge	VB
work_27d5tr326feydgcx32t4nhl424	198	27	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	198	28	model	model	NN
work_27d5tr326feydgcx32t4nhl424	198	29	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	198	30	better	well	JJR
work_27d5tr326feydgcx32t4nhl424	198	31	in	in	IN
work_27d5tr326feydgcx32t4nhl424	198	32	each	each	DT
work_27d5tr326feydgcx32t4nhl424	198	33	aspect	aspect	NN
work_27d5tr326feydgcx32t4nhl424	198	34	.	.	.
work_27d5tr326feydgcx32t4nhl424	199	1	As	as	IN
work_27d5tr326feydgcx32t4nhl424	199	2	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	199	3	can	can	MD
work_27d5tr326feydgcx32t4nhl424	199	4	see	see	VB
work_27d5tr326feydgcx32t4nhl424	199	5	from	from	IN
work_27d5tr326feydgcx32t4nhl424	199	6	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	199	7	5	5	CD
work_27d5tr326feydgcx32t4nhl424	199	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	199	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	199	10	two	two	CD
work_27d5tr326feydgcx32t4nhl424	199	11	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	199	12	al-	al-	VBP
work_27d5tr326feydgcx32t4nhl424	199	13	most	most	JJS
work_27d5tr326feydgcx32t4nhl424	199	14	tie	tie	VBP
work_27d5tr326feydgcx32t4nhl424	199	15	with	with	IN
work_27d5tr326feydgcx32t4nhl424	199	16	respect	respect	NN
work_27d5tr326feydgcx32t4nhl424	199	17	to	to	IN
work_27d5tr326feydgcx32t4nhl424	199	18	Non	non	JJ
work_27d5tr326feydgcx32t4nhl424	199	19	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	199	20	redundancy	redundancy	NN
work_27d5tr326feydgcx32t4nhl424	199	21	,	,	,
work_27d5tr326feydgcx32t4nhl424	199	22	mainly	mainly	RB
work_27d5tr326feydgcx32t4nhl424	199	23	because	because	IN
work_27d5tr326feydgcx32t4nhl424	199	24	both	both	DT
work_27d5tr326feydgcx32t4nhl424	199	25	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	199	26	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	199	27	used	use	VBN
work_27d5tr326feydgcx32t4nhl424	199	28	appropriate	appropriate	JJ
work_27d5tr326feydgcx32t4nhl424	199	29	MMR	MMR	NNP
work_27d5tr326feydgcx32t4nhl424	199	30	strategies	strategy	NNS
work_27d5tr326feydgcx32t4nhl424	199	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	200	1	But	but	CC
work_27d5tr326feydgcx32t4nhl424	200	2	as	as	IN
work_27d5tr326feydgcx32t4nhl424	200	3	for	for	IN
work_27d5tr326feydgcx32t4nhl424	200	4	Overall	overall	JJ
work_27d5tr326feydgcx32t4nhl424	200	5	quality	quality	NN
work_27d5tr326feydgcx32t4nhl424	200	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	200	7	Focus	Focus	NNP
work_27d5tr326feydgcx32t4nhl424	200	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	200	9	Our(win	our(win	ADD
work_27d5tr326feydgcx32t4nhl424	200	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	200	11	Hybhsum(win	Hybhsum(win	NNP
work_27d5tr326feydgcx32t4nhl424	200	12	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	200	13	Tie	Tie	NNP
work_27d5tr326feydgcx32t4nhl424	200	14	Overall	Overall	NNP
work_27d5tr326feydgcx32t4nhl424	200	15	37	37	CD
work_27d5tr326feydgcx32t4nhl424	200	16	14	14	CD
work_27d5tr326feydgcx32t4nhl424	200	17	9	9	CD
work_27d5tr326feydgcx32t4nhl424	200	18	Focus	Focus	NNP
work_27d5tr326feydgcx32t4nhl424	200	19	32	32	CD
work_27d5tr326feydgcx32t4nhl424	200	20	18	18	CD
work_27d5tr326feydgcx32t4nhl424	200	21	10	10	CD
work_27d5tr326feydgcx32t4nhl424	200	22	Responsiveness	responsiveness	NN
work_27d5tr326feydgcx32t4nhl424	200	23	33	33	CD
work_27d5tr326feydgcx32t4nhl424	200	24	13	13	CD
work_27d5tr326feydgcx32t4nhl424	200	25	14	14	CD
work_27d5tr326feydgcx32t4nhl424	200	26	Non	non	JJ
work_27d5tr326feydgcx32t4nhl424	200	27	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	200	28	redundancy	redundancy	JJ
work_27d5tr326feydgcx32t4nhl424	200	29	13	13	CD
work_27d5tr326feydgcx32t4nhl424	200	30	11	11	CD
work_27d5tr326feydgcx32t4nhl424	200	31	36	36	CD
work_27d5tr326feydgcx32t4nhl424	200	32	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	200	33	5	5	CD
work_27d5tr326feydgcx32t4nhl424	200	34	:	:	:
work_27d5tr326feydgcx32t4nhl424	200	35	Comparison	comparison	NN
work_27d5tr326feydgcx32t4nhl424	200	36	with	with	IN
work_27d5tr326feydgcx32t4nhl424	200	37	baselines	baseline	NNS
work_27d5tr326feydgcx32t4nhl424	200	38	on	on	IN
work_27d5tr326feydgcx32t4nhl424	200	39	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	200	40	Responsiveness	Responsiveness	NNP
work_27d5tr326feydgcx32t4nhl424	200	41	,	,	,
work_27d5tr326feydgcx32t4nhl424	200	42	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	200	43	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	200	44	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	200	45	model	model	NN
work_27d5tr326feydgcx32t4nhl424	200	46	outputs	output	NNS
work_27d5tr326feydgcx32t4nhl424	200	47	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	200	48	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	200	49	on	on	IN
work_27d5tr326feydgcx32t4nhl424	200	50	t	t	NN
work_27d5tr326feydgcx32t4nhl424	200	51	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	200	52	test	test	NN
work_27d5tr326feydgcx32t4nhl424	200	53	on	on	IN
work_27d5tr326feydgcx32t4nhl424	200	54	95	95	CD
work_27d5tr326feydgcx32t4nhl424	200	55	%	%	NN
work_27d5tr326feydgcx32t4nhl424	200	56	confidence	confidence	NN
work_27d5tr326feydgcx32t4nhl424	200	57	level	level	NN
work_27d5tr326feydgcx32t4nhl424	200	58	.	.	.
work_27d5tr326feydgcx32t4nhl424	201	1	Ta-	Ta-	NNP
work_27d5tr326feydgcx32t4nhl424	201	2	ble	ble	VB
work_27d5tr326feydgcx32t4nhl424	201	3	6	6	CD
work_27d5tr326feydgcx32t4nhl424	201	4	shows	show	VBZ
work_27d5tr326feydgcx32t4nhl424	201	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	201	6	example	example	NN
work_27d5tr326feydgcx32t4nhl424	201	7	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	201	8	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	201	9	re-	re-	RB
work_27d5tr326feydgcx32t4nhl424	201	10	spectively	spectively	RB
work_27d5tr326feydgcx32t4nhl424	201	11	by	by	IN
work_27d5tr326feydgcx32t4nhl424	201	12	two	two	CD
work_27d5tr326feydgcx32t4nhl424	201	13	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	201	14	for	for	IN
work_27d5tr326feydgcx32t4nhl424	201	15	document	document	NN
work_27d5tr326feydgcx32t4nhl424	201	16	collection	collection	NN
work_27d5tr326feydgcx32t4nhl424	201	17	D0803A	d0803a	NN
work_27d5tr326feydgcx32t4nhl424	201	18	-	-	:
work_27d5tr326feydgcx32t4nhl424	201	19	A	a	NN
work_27d5tr326feydgcx32t4nhl424	201	20	in	in	IN
work_27d5tr326feydgcx32t4nhl424	201	21	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	201	22	,	,	,
work_27d5tr326feydgcx32t4nhl424	201	23	whose	whose	WP$
work_27d5tr326feydgcx32t4nhl424	201	24	query	query	NN
work_27d5tr326feydgcx32t4nhl424	201	25	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	201	26	“	"	``
work_27d5tr326feydgcx32t4nhl424	201	27	Describe	describe	VB
work_27d5tr326feydgcx32t4nhl424	201	28	the	the	DT
work_27d5tr326feydgcx32t4nhl424	201	29	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	201	30	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	201	31	accidents	accident	NNS
work_27d5tr326feydgcx32t4nhl424	201	32	in	in	IN
work_27d5tr326feydgcx32t4nhl424	201	33	China	China	NNP
work_27d5tr326feydgcx32t4nhl424	201	34	and	and	CC
work_27d5tr326feydgcx32t4nhl424	201	35	actions	action	NNS
work_27d5tr326feydgcx32t4nhl424	201	36	taken	take	VBN
work_27d5tr326feydgcx32t4nhl424	201	37	“	"	``
work_27d5tr326feydgcx32t4nhl424	201	38	.	.	.
work_27d5tr326feydgcx32t4nhl424	202	1	From	from	IN
work_27d5tr326feydgcx32t4nhl424	202	2	table	table	NN
work_27d5tr326feydgcx32t4nhl424	202	3	6	6	CD
work_27d5tr326feydgcx32t4nhl424	202	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	202	5	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	202	6	can	can	MD
work_27d5tr326feydgcx32t4nhl424	202	7	see	see	VB
work_27d5tr326feydgcx32t4nhl424	202	8	that	that	IN
work_27d5tr326feydgcx32t4nhl424	202	9	each	each	DT
work_27d5tr326feydgcx32t4nhl424	202	10	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	202	11	in	in	IN
work_27d5tr326feydgcx32t4nhl424	202	12	these	these	DT
work_27d5tr326feydgcx32t4nhl424	202	13	two	two	CD
work_27d5tr326feydgcx32t4nhl424	202	14	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	202	15	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	202	16	somewhat	somewhat	RB
work_27d5tr326feydgcx32t4nhl424	202	17	related	related	JJ
work_27d5tr326feydgcx32t4nhl424	202	18	to	to	IN
work_27d5tr326feydgcx32t4nhl424	202	19	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	202	20	of	of	IN
work_27d5tr326feydgcx32t4nhl424	202	21	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	202	22	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	202	23	in	in	IN
work_27d5tr326feydgcx32t4nhl424	202	24	China	China	NNP
work_27d5tr326feydgcx32t4nhl424	202	25	.	.	.
work_27d5tr326feydgcx32t4nhl424	203	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	203	2	also	also	RB
work_27d5tr326feydgcx32t4nhl424	203	3	observe	observe	VBP
work_27d5tr326feydgcx32t4nhl424	203	4	that	that	IN
work_27d5tr326feydgcx32t4nhl424	203	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	203	6	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	203	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	203	8	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	203	9	6(a	6(a	CD
work_27d5tr326feydgcx32t4nhl424	203	10	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	203	11	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	203	12	better	well	JJR
work_27d5tr326feydgcx32t4nhl424	203	13	than	than	IN
work_27d5tr326feydgcx32t4nhl424	203	14	that	that	DT
work_27d5tr326feydgcx32t4nhl424	203	15	in	in	IN
work_27d5tr326feydgcx32t4nhl424	203	16	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	203	17	6(b	6(b	CD
work_27d5tr326feydgcx32t4nhl424	203	18	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	203	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	203	20	tend-	tend-	NNP
work_27d5tr326feydgcx32t4nhl424	203	21	ing	ing	NNP
work_27d5tr326feydgcx32t4nhl424	203	22	to	to	TO
work_27d5tr326feydgcx32t4nhl424	203	23	select	select	VB
work_27d5tr326feydgcx32t4nhl424	203	24	shorter	short	JJR
work_27d5tr326feydgcx32t4nhl424	203	25	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	203	26	and	and	CC
work_27d5tr326feydgcx32t4nhl424	203	27	provide	provide	VB
work_27d5tr326feydgcx32t4nhl424	203	28	more	more	RBR
work_27d5tr326feydgcx32t4nhl424	203	29	in-	in-	JJ
work_27d5tr326feydgcx32t4nhl424	203	30	formation	formation	NN
work_27d5tr326feydgcx32t4nhl424	203	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	204	1	This	this	DT
work_27d5tr326feydgcx32t4nhl424	204	2	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	204	3	because	because	IN
work_27d5tr326feydgcx32t4nhl424	204	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	204	5	in	in	IN
work_27d5tr326feydgcx32t4nhl424	204	6	S	S	NNP
work_27d5tr326feydgcx32t4nhl424	204	7	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	204	8	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	204	9	model	model	NN
work_27d5tr326feydgcx32t4nhl424	204	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	204	11	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	204	12	modeling	modeling	NN
work_27d5tr326feydgcx32t4nhl424	204	13	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	204	14	determined	determine	VBN
work_27d5tr326feydgcx32t4nhl424	204	15	simultaneously	simultaneously	RB
work_27d5tr326feydgcx32t4nhl424	204	16	by	by	IN
work_27d5tr326feydgcx32t4nhl424	204	17	various	various	JJ
work_27d5tr326feydgcx32t4nhl424	204	18	features	feature	NNS
work_27d5tr326feydgcx32t4nhl424	204	19	including	include	VBG
work_27d5tr326feydgcx32t4nhl424	204	20	terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	204	21	and	and	CC
work_27d5tr326feydgcx32t4nhl424	204	22	other	other	JJ
work_27d5tr326feydgcx32t4nhl424	204	23	ones	one	NNS
work_27d5tr326feydgcx32t4nhl424	204	24	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	204	25	as	as	IN
work_27d5tr326feydgcx32t4nhl424	204	26	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	204	27	tence	tence	NN
work_27d5tr326feydgcx32t4nhl424	204	28	length	length	NN
work_27d5tr326feydgcx32t4nhl424	204	29	,	,	,
work_27d5tr326feydgcx32t4nhl424	204	30	sentence	sentence	NN
work_27d5tr326feydgcx32t4nhl424	204	31	position	position	NN
work_27d5tr326feydgcx32t4nhl424	204	32	and	and	CC
work_27d5tr326feydgcx32t4nhl424	204	33	so	so	RB
work_27d5tr326feydgcx32t4nhl424	204	34	on	on	RB
work_27d5tr326feydgcx32t4nhl424	204	35	,	,	,
work_27d5tr326feydgcx32t4nhl424	204	36	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	204	37	can	can	MD
work_27d5tr326feydgcx32t4nhl424	204	38	contribute	contribute	VB
work_27d5tr326feydgcx32t4nhl424	204	39	to	to	IN
work_27d5tr326feydgcx32t4nhl424	204	40	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	204	41	quality	quality	NN
work_27d5tr326feydgcx32t4nhl424	204	42	.	.	.
work_27d5tr326feydgcx32t4nhl424	205	1	As	as	IN
work_27d5tr326feydgcx32t4nhl424	205	2	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	205	3	can	can	MD
work_27d5tr326feydgcx32t4nhl424	205	4	see	see	VB
work_27d5tr326feydgcx32t4nhl424	205	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	205	6	in	in	IN
work_27d5tr326feydgcx32t4nhl424	205	7	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	205	8	6(b	6(b	CD
work_27d5tr326feydgcx32t4nhl424	205	9	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	205	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	205	11	sentences	sentence	NNS
work_27d5tr326feydgcx32t4nhl424	205	12	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	205	13	3	3	CD
work_27d5tr326feydgcx32t4nhl424	205	14	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	205	15	and	and	CC
work_27d5tr326feydgcx32t4nhl424	205	16	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	205	17	5	5	LS
work_27d5tr326feydgcx32t4nhl424	205	18	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	205	19	provide	provide	VB
work_27d5tr326feydgcx32t4nhl424	205	20	some	some	DT
work_27d5tr326feydgcx32t4nhl424	205	21	unimportant	unimportant	JJ
work_27d5tr326feydgcx32t4nhl424	205	22	information	information	NN
work_27d5tr326feydgcx32t4nhl424	205	23	such	such	JJ
work_27d5tr326feydgcx32t4nhl424	205	24	as	as	IN
work_27d5tr326feydgcx32t4nhl424	205	25	“	"	``
work_27d5tr326feydgcx32t4nhl424	205	26	somebody	somebody	NN
work_27d5tr326feydgcx32t4nhl424	205	27	said	say	VBD
work_27d5tr326feydgcx32t4nhl424	205	28	“	"	``
work_27d5tr326feydgcx32t4nhl424	205	29	,	,	,
work_27d5tr326feydgcx32t4nhl424	205	30	though	though	IN
work_27d5tr326feydgcx32t4nhl424	205	31	they	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	205	32	contain	contain	VBP
work_27d5tr326feydgcx32t4nhl424	205	33	some	some	DT
work_27d5tr326feydgcx32t4nhl424	205	34	words	word	NNS
work_27d5tr326feydgcx32t4nhl424	205	35	which	which	WDT
work_27d5tr326feydgcx32t4nhl424	205	36	are	be	VBP
work_27d5tr326feydgcx32t4nhl424	205	37	related	relate	VBN
work_27d5tr326feydgcx32t4nhl424	205	38	to	to	IN
work_27d5tr326feydgcx32t4nhl424	205	39	topics	topic	NNS
work_27d5tr326feydgcx32t4nhl424	205	40	about	about	IN
work_27d5tr326feydgcx32t4nhl424	205	41	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	205	42	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	205	43	.	.	.
work_27d5tr326feydgcx32t4nhl424	206	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	206	2	1)China	1)china	CD
work_27d5tr326feydgcx32t4nhl424	206	3	to	to	TO
work_27d5tr326feydgcx32t4nhl424	206	4	close	close	VB
work_27d5tr326feydgcx32t4nhl424	206	5	at	at	RB
work_27d5tr326feydgcx32t4nhl424	206	6	least	least	RBS
work_27d5tr326feydgcx32t4nhl424	206	7	4,000	4,000	CD
work_27d5tr326feydgcx32t4nhl424	206	8	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	206	9	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	206	10	this	this	DT
work_27d5tr326feydgcx32t4nhl424	206	11	year	year	NN
work_27d5tr326feydgcx32t4nhl424	206	12	:	:	:
work_27d5tr326feydgcx32t4nhl424	206	13	official	official	NN
work_27d5tr326feydgcx32t4nhl424	206	14	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	206	15	2)By	2)by	CD
work_27d5tr326feydgcx32t4nhl424	206	16	Oct.	October	NNP
work_27d5tr326feydgcx32t4nhl424	206	17	10	10	CD
work_27d5tr326feydgcx32t4nhl424	206	18	this	this	DT
work_27d5tr326feydgcx32t4nhl424	206	19	year	year	NN
work_27d5tr326feydgcx32t4nhl424	206	20	there	there	EX
work_27d5tr326feydgcx32t4nhl424	206	21	had	have	VBD
work_27d5tr326feydgcx32t4nhl424	206	22	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	206	23	43	43	CD
work_27d5tr326feydgcx32t4nhl424	206	24	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	206	25	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	206	26	accidents	accident	NNS
work_27d5tr326feydgcx32t4nhl424	206	27	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	206	28	killed	kill	VBD
work_27d5tr326feydgcx32t4nhl424	206	29	10	10	CD
work_27d5tr326feydgcx32t4nhl424	206	30	or	or	CC
work_27d5tr326feydgcx32t4nhl424	206	31	more	more	JJR
work_27d5tr326feydgcx32t4nhl424	206	32	people	people	NNS
work_27d5tr326feydgcx32t4nhl424	206	33	,	,	,
work_27d5tr326feydgcx32t4nhl424	206	34	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	206	35	3)Offi-	3)offi-	CD
work_27d5tr326feydgcx32t4nhl424	206	36	cials	cial	NNS
work_27d5tr326feydgcx32t4nhl424	206	37	had	have	VBD
work_27d5tr326feydgcx32t4nhl424	206	38	stakes	stake	NNS
work_27d5tr326feydgcx32t4nhl424	206	39	in	in	IN
work_27d5tr326feydgcx32t4nhl424	206	40	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	206	41	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	206	42	.	.	.
work_27d5tr326feydgcx32t4nhl424	207	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	207	2	4)All	4)all	CD
work_27d5tr326feydgcx32t4nhl424	207	3	the	the	DT
work_27d5tr326feydgcx32t4nhl424	207	4	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	207	5	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	207	6	will	will	MD
work_27d5tr326feydgcx32t4nhl424	207	7	be	be	VB
work_27d5tr326feydgcx32t4nhl424	207	8	closed	close	VBN
work_27d5tr326feydgcx32t4nhl424	207	9	down	down	RP
work_27d5tr326feydgcx32t4nhl424	207	10	this	this	DT
work_27d5tr326feydgcx32t4nhl424	207	11	year	year	NN
work_27d5tr326feydgcx32t4nhl424	207	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	208	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	208	2	5	5	CD
work_27d5tr326feydgcx32t4nhl424	208	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	208	4	In	in	IN
work_27d5tr326feydgcx32t4nhl424	208	5	the	the	DT
work_27d5tr326feydgcx32t4nhl424	208	6	first	first	JJ
work_27d5tr326feydgcx32t4nhl424	208	7	eight	eight	CD
work_27d5tr326feydgcx32t4nhl424	208	8	months	month	NNS
work_27d5tr326feydgcx32t4nhl424	208	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	208	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	208	11	death	death	NN
work_27d5tr326feydgcx32t4nhl424	208	12	toll	toll	NN
work_27d5tr326feydgcx32t4nhl424	208	13	of	of	IN
work_27d5tr326feydgcx32t4nhl424	208	14	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	208	15	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	208	16	accidents	accident	NNS
work_27d5tr326feydgcx32t4nhl424	208	17	rose	rise	VBD
work_27d5tr326feydgcx32t4nhl424	208	18	8.5	8.5	CD
work_27d5tr326feydgcx32t4nhl424	208	19	percent	percent	NN
work_27d5tr326feydgcx32t4nhl424	208	20	last	last	JJ
work_27d5tr326feydgcx32t4nhl424	208	21	year	year	NN
work_27d5tr326feydgcx32t4nhl424	208	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	209	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	209	2	6	6	LS
work_27d5tr326feydgcx32t4nhl424	209	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	209	4	The	the	DT
work_27d5tr326feydgcx32t4nhl424	209	5	government	government	NN
work_27d5tr326feydgcx32t4nhl424	209	6	has	have	VBZ
work_27d5tr326feydgcx32t4nhl424	209	7	issued	issue	VBN
work_27d5tr326feydgcx32t4nhl424	209	8	a	a	DT
work_27d5tr326feydgcx32t4nhl424	209	9	series	series	NN
work_27d5tr326feydgcx32t4nhl424	209	10	of	of	IN
work_27d5tr326feydgcx32t4nhl424	209	11	regulations	regulation	NNS
work_27d5tr326feydgcx32t4nhl424	209	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	209	13	measures	measure	NNS
work_27d5tr326feydgcx32t4nhl424	209	14	to	to	TO
work_27d5tr326feydgcx32t4nhl424	209	15	improve	improve	VB
work_27d5tr326feydgcx32t4nhl424	209	16	the	the	DT
work_27d5tr326feydgcx32t4nhl424	209	17	coun.try	coun.try	NNP
work_27d5tr326feydgcx32t4nhl424	209	18	’s	’s	POS
work_27d5tr326feydgcx32t4nhl424	209	19	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	209	20	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	209	21	safety	safety	NN
work_27d5tr326feydgcx32t4nhl424	209	22	situation	situation	NN
work_27d5tr326feydgcx32t4nhl424	209	23	.	.	.
work_27d5tr326feydgcx32t4nhl424	210	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	210	2	7)The	7)the	CD
work_27d5tr326feydgcx32t4nhl424	210	3	mining	mining	NN
work_27d5tr326feydgcx32t4nhl424	210	4	safety	safety	NN
work_27d5tr326feydgcx32t4nhl424	210	5	technology	technology	NN
work_27d5tr326feydgcx32t4nhl424	210	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	210	7	equipments	equipment	NNS
work_27d5tr326feydgcx32t4nhl424	210	8	have	have	VBP
work_27d5tr326feydgcx32t4nhl424	210	9	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	210	10	sold	sell	VBN
work_27d5tr326feydgcx32t4nhl424	210	11	to	to	IN
work_27d5tr326feydgcx32t4nhl424	210	12	countries	country	NNS
work_27d5tr326feydgcx32t4nhl424	210	13	.	.	.
work_27d5tr326feydgcx32t4nhl424	211	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	211	2	8)More	8)more	CD
work_27d5tr326feydgcx32t4nhl424	211	3	than	than	IN
work_27d5tr326feydgcx32t4nhl424	211	4	6,000	6,000	CD
work_27d5tr326feydgcx32t4nhl424	211	5	miners	miner	NNS
work_27d5tr326feydgcx32t4nhl424	211	6	died	die	VBD
work_27d5tr326feydgcx32t4nhl424	211	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	211	8	accidents	accident	NNS
work_27d5tr326feydgcx32t4nhl424	211	9	in	in	IN
work_27d5tr326feydgcx32t4nhl424	211	10	China	China	NNP
work_27d5tr326feydgcx32t4nhl424	211	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	211	12	1	1	CD
work_27d5tr326feydgcx32t4nhl424	211	13	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	211	14	In	in	IN
work_27d5tr326feydgcx32t4nhl424	211	15	the	the	DT
work_27d5tr326feydgcx32t4nhl424	211	16	first	first	JJ
work_27d5tr326feydgcx32t4nhl424	211	17	eight	eight	CD
work_27d5tr326feydgcx32t4nhl424	211	18	months	month	NNS
work_27d5tr326feydgcx32t4nhl424	211	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	211	20	the	the	DT
work_27d5tr326feydgcx32t4nhl424	211	21	death	death	NN
work_27d5tr326feydgcx32t4nhl424	211	22	toll	toll	NN
work_27d5tr326feydgcx32t4nhl424	211	23	of	of	IN
work_27d5tr326feydgcx32t4nhl424	211	24	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	211	25	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	211	26	accidents	accident	NNS
work_27d5tr326feydgcx32t4nhl424	211	27	across	across	IN
work_27d5tr326feydgcx32t4nhl424	211	28	China	China	NNP
work_27d5tr326feydgcx32t4nhl424	211	29	rose	rise	VBD
work_27d5tr326feydgcx32t4nhl424	211	30	8.5	8.5	CD
work_27d5tr326feydgcx32t4nhl424	211	31	percent	percent	NN
work_27d5tr326feydgcx32t4nhl424	211	32	from	from	IN
work_27d5tr326feydgcx32t4nhl424	211	33	the	the	DT
work_27d5tr326feydgcx32t4nhl424	211	34	same	same	JJ
work_27d5tr326feydgcx32t4nhl424	211	35	period	period	NN
work_27d5tr326feydgcx32t4nhl424	211	36	last	last	JJ
work_27d5tr326feydgcx32t4nhl424	211	37	year	year	NN
work_27d5tr326feydgcx32t4nhl424	211	38	.	.	.
work_27d5tr326feydgcx32t4nhl424	212	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	212	2	2)China	2)China	NNP
work_27d5tr326feydgcx32t4nhl424	212	3	will	will	MD
work_27d5tr326feydgcx32t4nhl424	212	4	close	close	VB
work_27d5tr326feydgcx32t4nhl424	212	5	down	down	RP
work_27d5tr326feydgcx32t4nhl424	212	6	a	a	DT
work_27d5tr326feydgcx32t4nhl424	212	7	number	number	NN
work_27d5tr326feydgcx32t4nhl424	212	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	212	9	ill	ill	RB
work_27d5tr326feydgcx32t4nhl424	212	10	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	212	11	operated	operate	VBN
work_27d5tr326feydgcx32t4nhl424	212	12	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	212	13	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	212	14	at	at	IN
work_27d5tr326feydgcx32t4nhl424	212	15	the	the	DT
work_27d5tr326feydgcx32t4nhl424	212	16	end	end	NN
work_27d5tr326feydgcx32t4nhl424	212	17	of	of	IN
work_27d5tr326feydgcx32t4nhl424	212	18	this	this	DT
work_27d5tr326feydgcx32t4nhl424	212	19	month	month	NN
work_27d5tr326feydgcx32t4nhl424	212	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	212	21	said	say	VBD
work_27d5tr326feydgcx32t4nhl424	212	22	a	a	DT
work_27d5tr326feydgcx32t4nhl424	212	23	work	work	NN
work_27d5tr326feydgcx32t4nhl424	212	24	safety	safety	NN
work_27d5tr326feydgcx32t4nhl424	212	25	official	official	NN
work_27d5tr326feydgcx32t4nhl424	212	26	here	here	RB
work_27d5tr326feydgcx32t4nhl424	212	27	Monday	Monday	NNP
work_27d5tr326feydgcx32t4nhl424	212	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	213	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	213	2	3	3	LS
work_27d5tr326feydgcx32t4nhl424	213	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	213	4	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	213	5	Yizhong	Yizhong	NNP
work_27d5tr326feydgcx32t4nhl424	213	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	213	7	director	director	NN
work_27d5tr326feydgcx32t4nhl424	213	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	213	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	213	10	National	National	NNP
work_27d5tr326feydgcx32t4nhl424	213	11	Bureau	Bureau	NNP
work_27d5tr326feydgcx32t4nhl424	213	12	of	of	IN
work_27d5tr326feydgcx32t4nhl424	213	13	Production	Production	NNP
work_27d5tr326feydgcx32t4nhl424	213	14	Safety	Safety	NNP
work_27d5tr326feydgcx32t4nhl424	213	15	Supervision	Supervision	NNP
work_27d5tr326feydgcx32t4nhl424	213	16	and	and	CC
work_27d5tr326feydgcx32t4nhl424	213	17	Administration	Administration	NNP
work_27d5tr326feydgcx32t4nhl424	213	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	213	19	has	have	VBZ
work_27d5tr326feydgcx32t4nhl424	213	20	said	say	VBN
work_27d5tr326feydgcx32t4nhl424	213	21	the	the	DT
work_27d5tr326feydgcx32t4nhl424	213	22	collusion	collusion	NN
work_27d5tr326feydgcx32t4nhl424	213	23	between	between	IN
work_27d5tr326feydgcx32t4nhl424	213	24	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	213	25	owners	owner	NNS
work_27d5tr326feydgcx32t4nhl424	213	26	and	and	CC
work_27d5tr326feydgcx32t4nhl424	213	27	officials	official	NNS
work_27d5tr326feydgcx32t4nhl424	213	28	is	be	VBZ
work_27d5tr326feydgcx32t4nhl424	213	29	to	to	TO
work_27d5tr326feydgcx32t4nhl424	213	30	be	be	VB
work_27d5tr326feydgcx32t4nhl424	213	31	condemned	condemn	VBN
work_27d5tr326feydgcx32t4nhl424	213	32	.	.	.
work_27d5tr326feydgcx32t4nhl424	214	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	214	2	4)from	4)from	CD
work_27d5tr326feydgcx32t4nhl424	214	3	January	January	NNP
work_27d5tr326feydgcx32t4nhl424	214	4	to	to	IN
work_27d5tr326feydgcx32t4nhl424	214	5	September	September	NNP
work_27d5tr326feydgcx32t4nhl424	214	6	this	this	DT
work_27d5tr326feydgcx32t4nhl424	214	7	year	year	NN
work_27d5tr326feydgcx32t4nhl424	214	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	214	9	4,228	4,228	CD
work_27d5tr326feydgcx32t4nhl424	214	10	people	people	NNS
work_27d5tr326feydgcx32t4nhl424	214	11	were	be	VBD
work_27d5tr326feydgcx32t4nhl424	214	12	killed	kill	VBN
work_27d5tr326feydgcx32t4nhl424	214	13	in	in	IN
work_27d5tr326feydgcx32t4nhl424	214	14	2,337	2,337	CD
work_27d5tr326feydgcx32t4nhl424	214	15	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	214	16	mine	mine	NN
work_27d5tr326feydgcx32t4nhl424	214	17	accidents	accident	NNS
work_27d5tr326feydgcx32t4nhl424	214	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	215	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	215	2	5	5	LS
work_27d5tr326feydgcx32t4nhl424	215	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	215	4	Chen	Chen	NNP
work_27d5tr326feydgcx32t4nhl424	215	5	said	say	VBD
work_27d5tr326feydgcx32t4nhl424	215	6	officials	official	NNS
work_27d5tr326feydgcx32t4nhl424	215	7	who	who	WP
work_27d5tr326feydgcx32t4nhl424	215	8	refused	refuse	VBD
work_27d5tr326feydgcx32t4nhl424	215	9	to	to	TO
work_27d5tr326feydgcx32t4nhl424	215	10	register	register	VB
work_27d5tr326feydgcx32t4nhl424	215	11	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	215	12	stakes	stake	NNS
work_27d5tr326feydgcx32t4nhl424	215	13	in	in	IN
work_27d5tr326feydgcx32t4nhl424	215	14	coal	coal	NN
work_27d5tr326feydgcx32t4nhl424	215	15	mines	mine	NNS
work_27d5tr326feydgcx32t4nhl424	215	16	within	within	IN
work_27d5tr326feydgcx32t4nhl424	215	17	the	the	DT
work_27d5tr326feydgcx32t4nhl424	215	18	required	require	VBN
work_27d5tr326feydgcx32t4nhl424	215	19	time	time	NN
work_27d5tr326feydgcx32t4nhl424	215	20	Table	table	NN
work_27d5tr326feydgcx32t4nhl424	215	21	6	6	CD
work_27d5tr326feydgcx32t4nhl424	215	22	:	:	:
work_27d5tr326feydgcx32t4nhl424	215	23	Example	example	NN
work_27d5tr326feydgcx32t4nhl424	215	24	summary	summary	NN
work_27d5tr326feydgcx32t4nhl424	215	25	text	text	NN
work_27d5tr326feydgcx32t4nhl424	215	26	generated	generate	VBN
work_27d5tr326feydgcx32t4nhl424	215	27	by	by	IN
work_27d5tr326feydgcx32t4nhl424	215	28	systems	system	NNS
work_27d5tr326feydgcx32t4nhl424	215	29	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	215	30	a)S	a)S	NNP
work_27d5tr326feydgcx32t4nhl424	215	31	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	215	32	sLDA	slda	NN
work_27d5tr326feydgcx32t4nhl424	215	33	and	and	CC
work_27d5tr326feydgcx32t4nhl424	215	34	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	215	35	b	b	NN
work_27d5tr326feydgcx32t4nhl424	215	36	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	215	37	Hybhsum	Hybhsum	NNP
work_27d5tr326feydgcx32t4nhl424	215	38	.	.	.
work_27d5tr326feydgcx32t4nhl424	216	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	216	2	D0803A	D0803A	NNP
work_27d5tr326feydgcx32t4nhl424	216	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	216	4	A	a	NN
work_27d5tr326feydgcx32t4nhl424	216	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	216	6	TAC2008	TAC2008	NNP
work_27d5tr326feydgcx32t4nhl424	216	7	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	216	8	96	96	CD
work_27d5tr326feydgcx32t4nhl424	216	9	6	6	CD
work_27d5tr326feydgcx32t4nhl424	216	10	Conclusion	Conclusion	NNP
work_27d5tr326feydgcx32t4nhl424	216	11	In	in	IN
work_27d5tr326feydgcx32t4nhl424	216	12	this	this	DT
work_27d5tr326feydgcx32t4nhl424	216	13	paper	paper	NN
work_27d5tr326feydgcx32t4nhl424	216	14	,	,	,
work_27d5tr326feydgcx32t4nhl424	216	15	we	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	216	16	propose	propose	VBP
work_27d5tr326feydgcx32t4nhl424	216	17	a	a	DT
work_27d5tr326feydgcx32t4nhl424	216	18	novel	novel	NN
work_27d5tr326feydgcx32t4nhl424	216	19	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	216	20	ap-	ap-	CD
work_27d5tr326feydgcx32t4nhl424	216	21	proach	proach	NN
work_27d5tr326feydgcx32t4nhl424	216	22	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	216	23	on	on	IN
work_27d5tr326feydgcx32t4nhl424	216	24	revised	revised	JJ
work_27d5tr326feydgcx32t4nhl424	216	25	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	216	26	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	216	27	model	model	NN
work_27d5tr326feydgcx32t4nhl424	216	28	for	for	IN
work_27d5tr326feydgcx32t4nhl424	216	29	query	query	NN
work_27d5tr326feydgcx32t4nhl424	216	30	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	216	31	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	216	32	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	216	33	document	document	NN
work_27d5tr326feydgcx32t4nhl424	216	34	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	216	35	.	.	.
work_27d5tr326feydgcx32t4nhl424	217	1	Our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	217	2	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	217	3	naturally	naturally	RB
work_27d5tr326feydgcx32t4nhl424	217	4	combines	combine	VBZ
work_27d5tr326feydgcx32t4nhl424	217	5	Bayesian	bayesian	JJ
work_27d5tr326feydgcx32t4nhl424	217	6	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	217	7	model	model	NN
work_27d5tr326feydgcx32t4nhl424	217	8	with	with	IN
work_27d5tr326feydgcx32t4nhl424	217	9	supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	217	10	method	method	NN
work_27d5tr326feydgcx32t4nhl424	217	11	and	and	CC
work_27d5tr326feydgcx32t4nhl424	217	12	enjoy	enjoy	VB
work_27d5tr326feydgcx32t4nhl424	217	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	217	14	advantages	advantage	NNS
work_27d5tr326feydgcx32t4nhl424	217	15	of	of	IN
work_27d5tr326feydgcx32t4nhl424	217	16	both	both	DT
work_27d5tr326feydgcx32t4nhl424	217	17	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	217	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	218	1	Experiments	experiment	NNS
work_27d5tr326feydgcx32t4nhl424	218	2	on	on	IN
work_27d5tr326feydgcx32t4nhl424	218	3	benchmark	benchmark	JJ
work_27d5tr326feydgcx32t4nhl424	218	4	demon-	demon-	NN
work_27d5tr326feydgcx32t4nhl424	218	5	strate	strate	NNP
work_27d5tr326feydgcx32t4nhl424	218	6	good	good	JJ
work_27d5tr326feydgcx32t4nhl424	218	7	performance	performance	NN
work_27d5tr326feydgcx32t4nhl424	218	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	218	9	our	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	218	10	model	model	NN
work_27d5tr326feydgcx32t4nhl424	218	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	219	1	Acknowledgments	acknowledgment	NNS
work_27d5tr326feydgcx32t4nhl424	219	2	This	this	DT
work_27d5tr326feydgcx32t4nhl424	219	3	research	research	NN
work_27d5tr326feydgcx32t4nhl424	219	4	work	work	NN
work_27d5tr326feydgcx32t4nhl424	219	5	has	have	VBZ
work_27d5tr326feydgcx32t4nhl424	219	6	been	be	VBN
work_27d5tr326feydgcx32t4nhl424	219	7	supported	support	VBN
work_27d5tr326feydgcx32t4nhl424	219	8	by	by	IN
work_27d5tr326feydgcx32t4nhl424	219	9	NSFC	NSFC	NNP
work_27d5tr326feydgcx32t4nhl424	219	10	grants	grant	NNS
work_27d5tr326feydgcx32t4nhl424	219	11	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	219	12	No.90920011	No.90920011	NNP
work_27d5tr326feydgcx32t4nhl424	219	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	219	14	No.61273278	No.61273278	NNP
work_27d5tr326feydgcx32t4nhl424	219	15	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	219	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	219	17	National	National	NNP
work_27d5tr326feydgcx32t4nhl424	219	18	Key	Key	NNP
work_27d5tr326feydgcx32t4nhl424	219	19	Technology	Technology	NNP
work_27d5tr326feydgcx32t4nhl424	219	20	R&D	R&D	NNP
work_27d5tr326feydgcx32t4nhl424	219	21	Program	Program	NNP
work_27d5tr326feydgcx32t4nhl424	219	22	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	219	23	No:2011BAH1B0403	No:2011BAH1B0403	NNPS
work_27d5tr326feydgcx32t4nhl424	219	24	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	219	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	219	26	and	and	CC
work_27d5tr326feydgcx32t4nhl424	219	27	National	National	NNP
work_27d5tr326feydgcx32t4nhl424	219	28	High	High	NNP
work_27d5tr326feydgcx32t4nhl424	219	29	Tech-	Tech-	NNP
work_27d5tr326feydgcx32t4nhl424	219	30	nology	nology	DT
work_27d5tr326feydgcx32t4nhl424	219	31	R&D	R&D	NNP
work_27d5tr326feydgcx32t4nhl424	219	32	Program	Program	NNP
work_27d5tr326feydgcx32t4nhl424	219	33	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	219	34	No.2012AA011101	No.2012AA011101	NNP
work_27d5tr326feydgcx32t4nhl424	219	35	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	219	36	.	.	.
work_27d5tr326feydgcx32t4nhl424	220	1	We	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	220	2	also	also	RB
work_27d5tr326feydgcx32t4nhl424	220	3	thank	thank	VBP
work_27d5tr326feydgcx32t4nhl424	220	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	220	5	three	three	CD
work_27d5tr326feydgcx32t4nhl424	220	6	anonymous	anonymous	JJ
work_27d5tr326feydgcx32t4nhl424	220	7	reviewers	reviewer	NNS
work_27d5tr326feydgcx32t4nhl424	220	8	for	for	IN
work_27d5tr326feydgcx32t4nhl424	220	9	their	-PRON-	PRP$
work_27d5tr326feydgcx32t4nhl424	220	10	helpful	helpful	JJ
work_27d5tr326feydgcx32t4nhl424	220	11	comments	comment	NNS
work_27d5tr326feydgcx32t4nhl424	220	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	221	1	Corresponding	correspond	VBG
work_27d5tr326feydgcx32t4nhl424	221	2	author	author	NN
work_27d5tr326feydgcx32t4nhl424	221	3	:	:	:
work_27d5tr326feydgcx32t4nhl424	221	4	Sujian	sujian	JJ
work_27d5tr326feydgcx32t4nhl424	221	5	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	221	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	222	1	References	reference	NNS
work_27d5tr326feydgcx32t4nhl424	222	2	David	David	NNP
work_27d5tr326feydgcx32t4nhl424	222	3	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	222	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	222	5	Jon	Jon	NNP
work_27d5tr326feydgcx32t4nhl424	222	6	McAuliffe	McAuliffe	NNP
work_27d5tr326feydgcx32t4nhl424	222	7	.	.	.
work_27d5tr326feydgcx32t4nhl424	223	1	Supervised	supervise	VBN
work_27d5tr326feydgcx32t4nhl424	223	2	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	223	3	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	223	4	.	.	.
work_27d5tr326feydgcx32t4nhl424	224	1	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	224	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	225	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	225	2	Neural	Neural	NNP
work_27d5tr326feydgcx32t4nhl424	225	3	Information	Information	NNP
work_27d5tr326feydgcx32t4nhl424	225	4	Processing	Processing	NNP
work_27d5tr326feydgcx32t4nhl424	225	5	Systems	Systems	NNPS
work_27d5tr326feydgcx32t4nhl424	225	6	David	David	NNP
work_27d5tr326feydgcx32t4nhl424	225	7	Blei	Blei	NNP
work_27d5tr326feydgcx32t4nhl424	225	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	225	9	Andrew	Andrew	NNP
work_27d5tr326feydgcx32t4nhl424	225	10	Ng	Ng	NNP
work_27d5tr326feydgcx32t4nhl424	225	11	and	and	CC
work_27d5tr326feydgcx32t4nhl424	225	12	Micheal	Micheal	NNP
work_27d5tr326feydgcx32t4nhl424	225	13	Jordan	Jordan	NNP
work_27d5tr326feydgcx32t4nhl424	225	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	226	1	Latent	latent	JJ
work_27d5tr326feydgcx32t4nhl424	226	2	dirichlet	dirichlet	NN
work_27d5tr326feydgcx32t4nhl424	226	3	allocation	allocation	NN
work_27d5tr326feydgcx32t4nhl424	226	4	.	.	.
work_27d5tr326feydgcx32t4nhl424	227	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	227	2	The	the	DT
work_27d5tr326feydgcx32t4nhl424	227	3	Journal	Journal	NNP
work_27d5tr326feydgcx32t4nhl424	227	4	of	of	IN
work_27d5tr326feydgcx32t4nhl424	227	5	Machine	Machine	NNP
work_27d5tr326feydgcx32t4nhl424	227	6	Learn-	Learn-	NNP
work_27d5tr326feydgcx32t4nhl424	227	7	ing	ing	NNP
work_27d5tr326feydgcx32t4nhl424	227	8	Research	Research	NNP
work_27d5tr326feydgcx32t4nhl424	227	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	227	10	page	page	NN
work_27d5tr326feydgcx32t4nhl424	227	11	:	:	:
work_27d5tr326feydgcx32t4nhl424	227	12	993	993	CD
work_27d5tr326feydgcx32t4nhl424	227	13	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	227	14	1022	1022	CD
work_27d5tr326feydgcx32t4nhl424	227	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	228	1	Charles	Charles	NNP
work_27d5tr326feydgcx32t4nhl424	228	2	Broyden	Broyden	NNP
work_27d5tr326feydgcx32t4nhl424	228	3	.	.	.
work_27d5tr326feydgcx32t4nhl424	229	1	1965	1965	CD
work_27d5tr326feydgcx32t4nhl424	229	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	230	1	A	a	DT
work_27d5tr326feydgcx32t4nhl424	230	2	class	class	NN
work_27d5tr326feydgcx32t4nhl424	230	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	230	4	methods	method	NNS
work_27d5tr326feydgcx32t4nhl424	230	5	for	for	IN
work_27d5tr326feydgcx32t4nhl424	230	6	solv-	solv-	NNP
work_27d5tr326feydgcx32t4nhl424	230	7	ing	e	VBG
work_27d5tr326feydgcx32t4nhl424	230	8	nonlinear	nonlinear	JJ
work_27d5tr326feydgcx32t4nhl424	230	9	simultaneous	simultaneous	JJ
work_27d5tr326feydgcx32t4nhl424	230	10	equations	equation	NNS
work_27d5tr326feydgcx32t4nhl424	230	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	231	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	231	2	Math	Math	NNP
work_27d5tr326feydgcx32t4nhl424	231	3	.	.	.
work_27d5tr326feydgcx32t4nhl424	232	1	Comp	Comp	NNP
work_27d5tr326feydgcx32t4nhl424	232	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	233	1	volume	volume	NN
work_27d5tr326feydgcx32t4nhl424	233	2	19	19	CD
work_27d5tr326feydgcx32t4nhl424	233	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	233	4	page	page	NN
work_27d5tr326feydgcx32t4nhl424	233	5	577	577	CD
work_27d5tr326feydgcx32t4nhl424	233	6	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	233	7	593	593	CD
work_27d5tr326feydgcx32t4nhl424	233	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	234	1	Jaime	Jaime	NNP
work_27d5tr326feydgcx32t4nhl424	234	2	Carbonell	Carbonell	NNP
work_27d5tr326feydgcx32t4nhl424	234	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	234	4	Jade	Jade	NNP
work_27d5tr326feydgcx32t4nhl424	234	5	Goldstein	Goldstein	NNP
work_27d5tr326feydgcx32t4nhl424	234	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	235	1	1998	1998	CD
work_27d5tr326feydgcx32t4nhl424	235	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	236	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	236	2	use	use	NN
work_27d5tr326feydgcx32t4nhl424	236	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	236	4	MMR	MMR	NNP
work_27d5tr326feydgcx32t4nhl424	236	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	236	6	diversity	diversity	NN
work_27d5tr326feydgcx32t4nhl424	236	7	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	236	8	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	236	9	reranking	reranke	VBG
work_27d5tr326feydgcx32t4nhl424	236	10	for	for	IN
work_27d5tr326feydgcx32t4nhl424	236	11	reordering	reorder	VBG
work_27d5tr326feydgcx32t4nhl424	236	12	doc-	doc-	NN
work_27d5tr326feydgcx32t4nhl424	236	13	uments	ument	NNS
work_27d5tr326feydgcx32t4nhl424	236	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	236	15	producing	produce	VBG
work_27d5tr326feydgcx32t4nhl424	236	16	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	236	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	237	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	237	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	237	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	237	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	237	5	21st	21st	JJ
work_27d5tr326feydgcx32t4nhl424	237	6	annual	annual	JJ
work_27d5tr326feydgcx32t4nhl424	237	7	international	international	JJ
work_27d5tr326feydgcx32t4nhl424	237	8	ACM	ACM	NNP
work_27d5tr326feydgcx32t4nhl424	237	9	SIGIR	SIGIR	NNP
work_27d5tr326feydgcx32t4nhl424	237	10	conference	conference	NN
work_27d5tr326feydgcx32t4nhl424	237	11	on	on	IN
work_27d5tr326feydgcx32t4nhl424	237	12	Research	research	NN
work_27d5tr326feydgcx32t4nhl424	237	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	237	14	development	development	NN
work_27d5tr326feydgcx32t4nhl424	237	15	in	in	IN
work_27d5tr326feydgcx32t4nhl424	237	16	information	information	NN
work_27d5tr326feydgcx32t4nhl424	237	17	retrieval	retrieval	NN
work_27d5tr326feydgcx32t4nhl424	237	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	238	1	Asli	Asli	NNP
work_27d5tr326feydgcx32t4nhl424	238	2	Celikyilmaz	Celikyilmaz	NNP
work_27d5tr326feydgcx32t4nhl424	238	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	238	4	Dilek	Dilek	NNP
work_27d5tr326feydgcx32t4nhl424	238	5	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	238	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	238	7	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	238	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	239	1	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	239	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	240	1	A	a	DT
work_27d5tr326feydgcx32t4nhl424	240	2	Hy-	Hy-	NNP
work_27d5tr326feydgcx32t4nhl424	240	3	brid	brid	JJ
work_27d5tr326feydgcx32t4nhl424	240	4	hierarchical	hierarchical	JJ
work_27d5tr326feydgcx32t4nhl424	240	5	model	model	NN
work_27d5tr326feydgcx32t4nhl424	240	6	for	for	IN
work_27d5tr326feydgcx32t4nhl424	240	7	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	240	8	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	240	9	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	240	10	summa-	summa-	JJ
work_27d5tr326feydgcx32t4nhl424	240	11	rization	rization	NN
work_27d5tr326feydgcx32t4nhl424	240	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	241	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	241	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	241	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	241	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	241	5	48th	48th	JJ
work_27d5tr326feydgcx32t4nhl424	241	6	Annual	Annual	NNP
work_27d5tr326feydgcx32t4nhl424	241	7	Meeting	Meeting	NNP
work_27d5tr326feydgcx32t4nhl424	241	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	241	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	241	10	Association	Association	NNP
work_27d5tr326feydgcx32t4nhl424	241	11	for	for	IN
work_27d5tr326feydgcx32t4nhl424	241	12	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	241	13	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	241	14	.	.	.
work_27d5tr326feydgcx32t4nhl424	242	1	page	page	NN
work_27d5tr326feydgcx32t4nhl424	242	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	242	3	815	815	CD
work_27d5tr326feydgcx32t4nhl424	242	4	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	242	5	825	825	CD
work_27d5tr326feydgcx32t4nhl424	242	6	Jade	Jade	NNP
work_27d5tr326feydgcx32t4nhl424	242	7	Goldstein	Goldstein	NNP
work_27d5tr326feydgcx32t4nhl424	242	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	242	9	Mark	Mark	NNP
work_27d5tr326feydgcx32t4nhl424	242	10	Kantrowitz	Kantrowitz	NNP
work_27d5tr326feydgcx32t4nhl424	242	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	242	12	Vibhu	Vibhu	NNP
work_27d5tr326feydgcx32t4nhl424	242	13	Mittal	Mittal	NNP
work_27d5tr326feydgcx32t4nhl424	242	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	242	15	Jaime	Jaime	NNP
work_27d5tr326feydgcx32t4nhl424	242	16	Carbonell	Carbonell	NNP
work_27d5tr326feydgcx32t4nhl424	242	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	243	1	1999	1999	CD
work_27d5tr326feydgcx32t4nhl424	243	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	244	1	Summarizing	Summarizing	NNP
work_27d5tr326feydgcx32t4nhl424	244	2	Text	Text	NNP
work_27d5tr326feydgcx32t4nhl424	244	3	Docu-	docu-	NN
work_27d5tr326feydgcx32t4nhl424	244	4	ments	ment	NNS
work_27d5tr326feydgcx32t4nhl424	244	5	:	:	:
work_27d5tr326feydgcx32t4nhl424	244	6	Sentence	Sentence	NNP
work_27d5tr326feydgcx32t4nhl424	244	7	Selection	Selection	NNP
work_27d5tr326feydgcx32t4nhl424	244	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	244	9	Evaluation	Evaluation	NNP
work_27d5tr326feydgcx32t4nhl424	244	10	Metrics	Metrics	NNP
work_27d5tr326feydgcx32t4nhl424	244	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	245	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	245	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	245	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	245	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	245	5	22nd	22nd	JJ
work_27d5tr326feydgcx32t4nhl424	245	6	annual	annual	JJ
work_27d5tr326feydgcx32t4nhl424	245	7	international	international	JJ
work_27d5tr326feydgcx32t4nhl424	245	8	ACM	ACM	NNP
work_27d5tr326feydgcx32t4nhl424	245	9	SI-	SI-	NNP
work_27d5tr326feydgcx32t4nhl424	245	10	GIR	GIR	NNP
work_27d5tr326feydgcx32t4nhl424	245	11	conference	conference	NN
work_27d5tr326feydgcx32t4nhl424	245	12	on	on	IN
work_27d5tr326feydgcx32t4nhl424	245	13	Research	research	NN
work_27d5tr326feydgcx32t4nhl424	245	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	245	15	development	development	NN
work_27d5tr326feydgcx32t4nhl424	245	16	in	in	IN
work_27d5tr326feydgcx32t4nhl424	245	17	infor-	infor-	NNP
work_27d5tr326feydgcx32t4nhl424	245	18	mation	mation	NNP
work_27d5tr326feydgcx32t4nhl424	245	19	retrieval	retrieval	NNP
work_27d5tr326feydgcx32t4nhl424	245	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	245	21	page	page	NN
work_27d5tr326feydgcx32t4nhl424	245	22	:	:	:
work_27d5tr326feydgcx32t4nhl424	245	23	121	121	CD
work_27d5tr326feydgcx32t4nhl424	245	24	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	245	25	128	128	CD
work_27d5tr326feydgcx32t4nhl424	245	26	.	.	.
work_27d5tr326feydgcx32t4nhl424	246	1	Amit	Amit	NNP
work_27d5tr326feydgcx32t4nhl424	246	2	Grubber	Grubber	NNP
work_27d5tr326feydgcx32t4nhl424	246	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	246	4	Micheal	Micheal	NNP
work_27d5tr326feydgcx32t4nhl424	246	5	Rosen	Rosen	NNP
work_27d5tr326feydgcx32t4nhl424	246	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	246	7	zvi	zvi	NNP
work_27d5tr326feydgcx32t4nhl424	246	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	246	9	Yair	Yair	NNP
work_27d5tr326feydgcx32t4nhl424	246	10	Weiss	Weiss	NNP
work_27d5tr326feydgcx32t4nhl424	246	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	247	1	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	247	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	248	1	Hidden	hide	VBN
work_27d5tr326feydgcx32t4nhl424	248	2	Topic	Topic	NNP
work_27d5tr326feydgcx32t4nhl424	248	3	Markov	Markov	NNP
work_27d5tr326feydgcx32t4nhl424	248	4	Model	Model	NNP
work_27d5tr326feydgcx32t4nhl424	248	5	.	.	.
work_27d5tr326feydgcx32t4nhl424	249	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	249	2	Artificial	Artificial	NNP
work_27d5tr326feydgcx32t4nhl424	249	3	Intelligence	Intelligence	NNP
work_27d5tr326feydgcx32t4nhl424	249	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	249	5	Statistics	Statistics	NNPS
work_27d5tr326feydgcx32t4nhl424	249	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	250	1	Hal	Hal	NNP
work_27d5tr326feydgcx32t4nhl424	250	2	Daume	Daume	NNP
work_27d5tr326feydgcx32t4nhl424	250	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	250	4	Daniel	Daniel	NNP
work_27d5tr326feydgcx32t4nhl424	250	5	Marcu	Marcu	NNP
work_27d5tr326feydgcx32t4nhl424	250	6	H.	H.	NNP
work_27d5tr326feydgcx32t4nhl424	250	7	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	250	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	251	1	Bayesian	Bayesian	NNP
work_27d5tr326feydgcx32t4nhl424	251	2	Query-	Query-	NNP
work_27d5tr326feydgcx32t4nhl424	251	3	Focused	focus	VBD
work_27d5tr326feydgcx32t4nhl424	251	4	Summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	251	5	.	.	.
work_27d5tr326feydgcx32t4nhl424	252	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	252	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	252	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	252	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	252	5	21st	21st	JJ
work_27d5tr326feydgcx32t4nhl424	252	6	International	International	NNP
work_27d5tr326feydgcx32t4nhl424	252	7	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	252	8	on	on	IN
work_27d5tr326feydgcx32t4nhl424	252	9	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	252	10	Linguis-	Linguis-	NNP
work_27d5tr326feydgcx32t4nhl424	252	11	tics	tic	NNS
work_27d5tr326feydgcx32t4nhl424	252	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	252	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	252	14	44th	44th	JJ
work_27d5tr326feydgcx32t4nhl424	252	15	annual	annual	JJ
work_27d5tr326feydgcx32t4nhl424	252	16	meeting	meeting	NN
work_27d5tr326feydgcx32t4nhl424	252	17	of	of	IN
work_27d5tr326feydgcx32t4nhl424	252	18	the	the	DT
work_27d5tr326feydgcx32t4nhl424	252	19	Association	Association	NNP
work_27d5tr326feydgcx32t4nhl424	252	20	for	for	IN
work_27d5tr326feydgcx32t4nhl424	252	21	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	252	22	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	252	23	,	,	,
work_27d5tr326feydgcx32t4nhl424	252	24	page	page	NN
work_27d5tr326feydgcx32t4nhl424	252	25	305	305	CD
work_27d5tr326feydgcx32t4nhl424	252	26	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	252	27	312	312	CD
work_27d5tr326feydgcx32t4nhl424	252	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	253	1	Gune	Gune	NNP
work_27d5tr326feydgcx32t4nhl424	253	2	Erkan	Erkan	NNP
work_27d5tr326feydgcx32t4nhl424	253	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	253	4	Dragomir	Dragomir	NNP
work_27d5tr326feydgcx32t4nhl424	253	5	Radev	Radev	NNP
work_27d5tr326feydgcx32t4nhl424	253	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	254	1	2004	2004	CD
work_27d5tr326feydgcx32t4nhl424	254	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	255	1	Lexrank	Lexrank	NNP
work_27d5tr326feydgcx32t4nhl424	255	2	:	:	:
work_27d5tr326feydgcx32t4nhl424	255	3	graph-	graph-	NNP
work_27d5tr326feydgcx32t4nhl424	255	4	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	255	5	lexical	lexical	JJ
work_27d5tr326feydgcx32t4nhl424	255	6	centrality	centrality	NN
work_27d5tr326feydgcx32t4nhl424	255	7	as	as	IN
work_27d5tr326feydgcx32t4nhl424	255	8	salience	salience	NN
work_27d5tr326feydgcx32t4nhl424	255	9	in	in	IN
work_27d5tr326feydgcx32t4nhl424	255	10	text	text	NN
work_27d5tr326feydgcx32t4nhl424	255	11	summariza-	summariza-	JJ
work_27d5tr326feydgcx32t4nhl424	255	12	tion	tion	NN
work_27d5tr326feydgcx32t4nhl424	255	13	.	.	.
work_27d5tr326feydgcx32t4nhl424	256	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	256	2	J.	J.	NNP
work_27d5tr326feydgcx32t4nhl424	256	3	Artif	Artif	NNP
work_27d5tr326feydgcx32t4nhl424	256	4	.	.	.
work_27d5tr326feydgcx32t4nhl424	257	1	Intell	intell	UH
work_27d5tr326feydgcx32t4nhl424	257	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	258	1	Res	Res	NNP
work_27d5tr326feydgcx32t4nhl424	258	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	259	1	(	(	-LRB-
work_27d5tr326feydgcx32t4nhl424	259	2	JAIR	JAIR	NNP
work_27d5tr326feydgcx32t4nhl424	259	3	)	)	-RRB-
work_27d5tr326feydgcx32t4nhl424	259	4	,	,	,
work_27d5tr326feydgcx32t4nhl424	259	5	page	page	NN
work_27d5tr326feydgcx32t4nhl424	259	6	457	457	CD
work_27d5tr326feydgcx32t4nhl424	259	7	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	259	8	479	479	CD
work_27d5tr326feydgcx32t4nhl424	259	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	260	1	Dan	Dan	NNP
work_27d5tr326feydgcx32t4nhl424	260	2	Gillick	Gillick	NNP
work_27d5tr326feydgcx32t4nhl424	260	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	260	4	Benoit	Benoit	NNP
work_27d5tr326feydgcx32t4nhl424	260	5	Favre	Favre	NNP
work_27d5tr326feydgcx32t4nhl424	260	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	260	7	Dilek	Dilek	NNP
work_27d5tr326feydgcx32t4nhl424	260	8	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	260	9	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	260	10	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	260	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	260	12	The	the	DT
work_27d5tr326feydgcx32t4nhl424	260	13	ICSI	ICSI	NNP
work_27d5tr326feydgcx32t4nhl424	260	14	Summarization	Summarization	NNP
work_27d5tr326feydgcx32t4nhl424	260	15	System	System	NNP
work_27d5tr326feydgcx32t4nhl424	260	16	at	at	IN
work_27d5tr326feydgcx32t4nhl424	260	17	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	260	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	260	19	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	260	20	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	260	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	261	1	Dan	Dan	NNP
work_27d5tr326feydgcx32t4nhl424	261	2	Gillick	Gillick	NNP
work_27d5tr326feydgcx32t4nhl424	261	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	261	4	Benoit	Benoit	NNP
work_27d5tr326feydgcx32t4nhl424	261	5	Favre	Favre	NNP
work_27d5tr326feydgcx32t4nhl424	261	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	261	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	261	8	Dilek	Dilek	NNP
work_27d5tr326feydgcx32t4nhl424	261	9	Hakkani	Hakkani	NNP
work_27d5tr326feydgcx32t4nhl424	261	10	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	261	11	Tur	Tur	NNP
work_27d5tr326feydgcx32t4nhl424	261	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	261	13	Berndt	Berndt	NNP
work_27d5tr326feydgcx32t4nhl424	261	14	Bohnet	Bohnet	NNP
work_27d5tr326feydgcx32t4nhl424	261	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	261	16	Yang	Yang	NNP
work_27d5tr326feydgcx32t4nhl424	261	17	Liu	Liu	NNP
work_27d5tr326feydgcx32t4nhl424	261	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	261	19	Shasha	Shasha	NNP
work_27d5tr326feydgcx32t4nhl424	261	20	Xie	Xie	NNP
work_27d5tr326feydgcx32t4nhl424	261	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	262	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	262	2	ICSI	ICSI	NNP
work_27d5tr326feydgcx32t4nhl424	262	3	/	/	SYM
work_27d5tr326feydgcx32t4nhl424	262	4	UTD	UTD	NNP
work_27d5tr326feydgcx32t4nhl424	262	5	Summarization	Summarization	NNP
work_27d5tr326feydgcx32t4nhl424	262	6	System	System	NNP
work_27d5tr326feydgcx32t4nhl424	262	7	at	at	IN
work_27d5tr326feydgcx32t4nhl424	262	8	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	262	9	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	262	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	263	1	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	263	2	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	263	3	Aria	Aria	NNP
work_27d5tr326feydgcx32t4nhl424	263	4	Haghighi	Haghighi	NNP
work_27d5tr326feydgcx32t4nhl424	263	5	and	and	CC
work_27d5tr326feydgcx32t4nhl424	263	6	Lucy	Lucy	NNP
work_27d5tr326feydgcx32t4nhl424	263	7	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	263	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	264	1	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	264	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	265	1	Exploring	explore	VBG
work_27d5tr326feydgcx32t4nhl424	265	2	content	content	NN
work_27d5tr326feydgcx32t4nhl424	265	3	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	265	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	265	5	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	265	6	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	265	7	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	265	8	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	265	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	266	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	266	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	266	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	266	4	Human	Human	NNP
work_27d5tr326feydgcx32t4nhl424	266	5	Language	Language	NNP
work_27d5tr326feydgcx32t4nhl424	266	6	Technologies	Technologies	NNPS
work_27d5tr326feydgcx32t4nhl424	266	7	:	:	:
work_27d5tr326feydgcx32t4nhl424	266	8	The	the	DT
work_27d5tr326feydgcx32t4nhl424	266	9	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	266	10	Annual	Annual	NNP
work_27d5tr326feydgcx32t4nhl424	266	11	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	266	12	of	of	IN
work_27d5tr326feydgcx32t4nhl424	266	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	266	14	North	North	NNP
work_27d5tr326feydgcx32t4nhl424	266	15	American	American	NNP
work_27d5tr326feydgcx32t4nhl424	266	16	Chap-	Chap-	NNP
work_27d5tr326feydgcx32t4nhl424	266	17	ter	ter	NN
work_27d5tr326feydgcx32t4nhl424	266	18	of	of	IN
work_27d5tr326feydgcx32t4nhl424	266	19	the	the	DT
work_27d5tr326feydgcx32t4nhl424	266	20	Association	Association	NNP
work_27d5tr326feydgcx32t4nhl424	266	21	for	for	IN
work_27d5tr326feydgcx32t4nhl424	266	22	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	266	23	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	266	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	266	25	pages	page	VBZ
work_27d5tr326feydgcx32t4nhl424	266	26	362370	362370	CD
work_27d5tr326feydgcx32t4nhl424	266	27	.	.	.
work_27d5tr326feydgcx32t4nhl424	267	1	Feng	Feng	NNP
work_27d5tr326feydgcx32t4nhl424	267	2	Jin	Jin	NNP
work_27d5tr326feydgcx32t4nhl424	267	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	267	4	Minlie	Minlie	NNP
work_27d5tr326feydgcx32t4nhl424	267	5	Huang	Huang	NNP
work_27d5tr326feydgcx32t4nhl424	267	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	267	7	and	and	CC
work_27d5tr326feydgcx32t4nhl424	267	8	Xiaoyan	Xiaoyan	NNP
work_27d5tr326feydgcx32t4nhl424	267	9	Zhu	Zhu	NNP
work_27d5tr326feydgcx32t4nhl424	267	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	268	1	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	268	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	269	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	269	2	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	269	3	systems	system	NNS
work_27d5tr326feydgcx32t4nhl424	269	4	at	at	IN
work_27d5tr326feydgcx32t4nhl424	269	5	tac	tac	NN
work_27d5tr326feydgcx32t4nhl424	269	6	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	269	7	.	.	.
work_27d5tr326feydgcx32t4nhl424	270	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	270	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	270	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	270	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	270	5	third	third	JJ
work_27d5tr326feydgcx32t4nhl424	270	6	Text	Text	NNP
work_27d5tr326feydgcx32t4nhl424	270	7	Analysis	Analysis	NNP
work_27d5tr326feydgcx32t4nhl424	270	8	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	270	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	270	10	TAC-2010	TAC-2010	NNP
work_27d5tr326feydgcx32t4nhl424	270	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	271	1	Liangda	Liangda	NNP
work_27d5tr326feydgcx32t4nhl424	271	2	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	271	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	271	4	Ke	Ke	NNP
work_27d5tr326feydgcx32t4nhl424	271	5	Zhou	Zhou	NNP
work_27d5tr326feydgcx32t4nhl424	271	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	271	7	Gui	Gui	NNP
work_27d5tr326feydgcx32t4nhl424	271	8	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	271	9	Rong	Rong	NNP
work_27d5tr326feydgcx32t4nhl424	271	10	Xue	Xue	NNP
work_27d5tr326feydgcx32t4nhl424	271	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	271	12	Hongyuan	Hongyuan	NNP
work_27d5tr326feydgcx32t4nhl424	271	13	Zha	Zha	NNP
work_27d5tr326feydgcx32t4nhl424	271	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	271	15	Yong	Yong	NNP
work_27d5tr326feydgcx32t4nhl424	271	16	Yu	Yu	NNP
work_27d5tr326feydgcx32t4nhl424	271	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	272	1	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	272	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	273	1	Enhancing	enhance	VBG
work_27d5tr326feydgcx32t4nhl424	273	2	diversity	diversity	NN
work_27d5tr326feydgcx32t4nhl424	273	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	273	4	coverage	coverage	NN
work_27d5tr326feydgcx32t4nhl424	273	5	and	and	CC
work_27d5tr326feydgcx32t4nhl424	273	6	bal-	bal-	NN
work_27d5tr326feydgcx32t4nhl424	273	7	ance	ance	NN
work_27d5tr326feydgcx32t4nhl424	273	8	for	for	IN
work_27d5tr326feydgcx32t4nhl424	273	9	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	273	10	through	through	IN
work_27d5tr326feydgcx32t4nhl424	273	11	structure	structure	NN
work_27d5tr326feydgcx32t4nhl424	273	12	learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	273	13	.	.	.
work_27d5tr326feydgcx32t4nhl424	274	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	274	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	274	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	274	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	274	5	18th	18th	JJ
work_27d5tr326feydgcx32t4nhl424	274	6	international	international	JJ
work_27d5tr326feydgcx32t4nhl424	274	7	conference	conference	NN
work_27d5tr326feydgcx32t4nhl424	274	8	on	on	IN
work_27d5tr326feydgcx32t4nhl424	274	9	World	World	NNP
work_27d5tr326feydgcx32t4nhl424	274	10	wide	wide	JJ
work_27d5tr326feydgcx32t4nhl424	274	11	web	web	NN
work_27d5tr326feydgcx32t4nhl424	274	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	274	13	page	page	NN
work_27d5tr326feydgcx32t4nhl424	274	14	71	71	CD
work_27d5tr326feydgcx32t4nhl424	274	15	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	274	16	80	80	CD
work_27d5tr326feydgcx32t4nhl424	274	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	275	1	Chin	Chin	NNP
work_27d5tr326feydgcx32t4nhl424	275	2	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	275	3	Yew	Yew	NNP
work_27d5tr326feydgcx32t4nhl424	275	4	Lin	Lin	NNP
work_27d5tr326feydgcx32t4nhl424	275	5	,	,	,
work_27d5tr326feydgcx32t4nhl424	275	6	Guihong	Guihong	NNP
work_27d5tr326feydgcx32t4nhl424	275	7	Gao	Gao	NNP
work_27d5tr326feydgcx32t4nhl424	275	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	275	9	Jianfeng	Jianfeng	NNP
work_27d5tr326feydgcx32t4nhl424	275	10	Gao	Gao	NNP
work_27d5tr326feydgcx32t4nhl424	275	11	and	and	CC
work_27d5tr326feydgcx32t4nhl424	275	12	Jian	Jian	NNP
work_27d5tr326feydgcx32t4nhl424	275	13	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	275	14	Yun	Yun	NNP
work_27d5tr326feydgcx32t4nhl424	275	15	Nie	Nie	NNP
work_27d5tr326feydgcx32t4nhl424	275	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	276	1	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	276	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	277	1	An	an	DT
work_27d5tr326feydgcx32t4nhl424	277	2	information	information	NN
work_27d5tr326feydgcx32t4nhl424	277	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	277	4	theoretic	theoretic	JJ
work_27d5tr326feydgcx32t4nhl424	277	5	approach	approach	NN
work_27d5tr326feydgcx32t4nhl424	277	6	to	to	IN
work_27d5tr326feydgcx32t4nhl424	277	7	au-	au-	CC
work_27d5tr326feydgcx32t4nhl424	277	8	tomatic	tomatic	JJ
work_27d5tr326feydgcx32t4nhl424	277	9	evaluation	evaluation	NN
work_27d5tr326feydgcx32t4nhl424	277	10	of	of	IN
work_27d5tr326feydgcx32t4nhl424	277	11	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	277	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	278	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	278	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	278	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	278	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	278	5	main	main	JJ
work_27d5tr326feydgcx32t4nhl424	278	6	conference	conference	NN
work_27d5tr326feydgcx32t4nhl424	278	7	on	on	IN
work_27d5tr326feydgcx32t4nhl424	278	8	Human	Human	NNP
work_27d5tr326feydgcx32t4nhl424	278	9	Language	Language	NNP
work_27d5tr326feydgcx32t4nhl424	278	10	Technology	Technology	NNP
work_27d5tr326feydgcx32t4nhl424	278	11	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	278	12	of	of	IN
work_27d5tr326feydgcx32t4nhl424	278	13	the	the	DT
work_27d5tr326feydgcx32t4nhl424	278	14	North	north	JJ
work_27d5tr326feydgcx32t4nhl424	278	15	American	american	JJ
work_27d5tr326feydgcx32t4nhl424	278	16	Chapter	Chapter	NNP
work_27d5tr326feydgcx32t4nhl424	278	17	of	of	IN
work_27d5tr326feydgcx32t4nhl424	278	18	the	the	DT
work_27d5tr326feydgcx32t4nhl424	278	19	As-	As-	NNP
work_27d5tr326feydgcx32t4nhl424	278	20	sociation	sociation	NN
work_27d5tr326feydgcx32t4nhl424	278	21	of	of	IN
work_27d5tr326feydgcx32t4nhl424	278	22	Computational	Computational	NNP
work_27d5tr326feydgcx32t4nhl424	278	23	Linguistics	Linguistics	NNP
work_27d5tr326feydgcx32t4nhl424	278	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	278	25	page:462	page:462	NNP
work_27d5tr326feydgcx32t4nhl424	278	26	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	278	27	470	470	CD
work_27d5tr326feydgcx32t4nhl424	278	28	.	.	.
work_27d5tr326feydgcx32t4nhl424	279	1	Annie	Annie	NNP
work_27d5tr326feydgcx32t4nhl424	279	2	Louis	Louis	NNP
work_27d5tr326feydgcx32t4nhl424	279	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	279	4	Aravind	Aravind	NNP
work_27d5tr326feydgcx32t4nhl424	279	5	Joshi	Joshi	NNP
work_27d5tr326feydgcx32t4nhl424	279	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	279	7	Ani	Ani	NNP
work_27d5tr326feydgcx32t4nhl424	279	8	Nenkova	Nenkova	NNP
work_27d5tr326feydgcx32t4nhl424	279	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	280	1	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	280	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	281	1	Dis-	Dis-	NNP
work_27d5tr326feydgcx32t4nhl424	281	2	course	course	NN
work_27d5tr326feydgcx32t4nhl424	281	3	indicators	indicator	NNS
work_27d5tr326feydgcx32t4nhl424	281	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	281	5	content	content	NN
work_27d5tr326feydgcx32t4nhl424	281	6	selection	selection	NN
work_27d5tr326feydgcx32t4nhl424	281	7	in	in	IN
work_27d5tr326feydgcx32t4nhl424	281	8	summariza-	summariza-	JJ
work_27d5tr326feydgcx32t4nhl424	281	9	tion	tion	NN
work_27d5tr326feydgcx32t4nhl424	281	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	282	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	282	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	282	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	282	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	282	5	11th	11th	JJ
work_27d5tr326feydgcx32t4nhl424	282	6	Annual	Annual	NNP
work_27d5tr326feydgcx32t4nhl424	282	7	Meeting	Meeting	NNP
work_27d5tr326feydgcx32t4nhl424	282	8	of	of	IN
work_27d5tr326feydgcx32t4nhl424	282	9	the	the	DT
work_27d5tr326feydgcx32t4nhl424	282	10	Special	Special	NNP
work_27d5tr326feydgcx32t4nhl424	282	11	Interest	Interest	NNP
work_27d5tr326feydgcx32t4nhl424	282	12	Group	Group	NNP
work_27d5tr326feydgcx32t4nhl424	282	13	on	on	IN
work_27d5tr326feydgcx32t4nhl424	282	14	Discourse	Discourse	NNP
work_27d5tr326feydgcx32t4nhl424	282	15	and	and	CC
work_27d5tr326feydgcx32t4nhl424	282	16	Dialogue	Dialogue	NNP
work_27d5tr326feydgcx32t4nhl424	282	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	282	18	page:147	page:147	NNP
work_27d5tr326feydgcx32t4nhl424	282	19	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	282	20	156	156	CD
work_27d5tr326feydgcx32t4nhl424	282	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	283	1	Tengfei	Tengfei	NNP
work_27d5tr326feydgcx32t4nhl424	283	2	Ma	Ma	NNP
work_27d5tr326feydgcx32t4nhl424	283	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	283	4	Xiaojun	Xiaojun	NNP
work_27d5tr326feydgcx32t4nhl424	283	5	Wan	Wan	NNP
work_27d5tr326feydgcx32t4nhl424	283	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	284	1	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	284	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	285	1	Multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	285	2	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	285	3	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	285	4	sum-	sum-	JJ
work_27d5tr326feydgcx32t4nhl424	285	5	marization	marization	NN
work_27d5tr326feydgcx32t4nhl424	285	6	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	285	7	minimum	minimum	JJ
work_27d5tr326feydgcx32t4nhl424	285	8	distortion	distortion	NN
work_27d5tr326feydgcx32t4nhl424	285	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	285	10	in	in	IN
work_27d5tr326feydgcx32t4nhl424	285	11	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	285	12	of	of	IN
work_27d5tr326feydgcx32t4nhl424	285	13	International	International	NNP
work_27d5tr326feydgcx32t4nhl424	285	14	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	285	15	of	of	IN
work_27d5tr326feydgcx32t4nhl424	285	16	Data	Data	NNP
work_27d5tr326feydgcx32t4nhl424	285	17	Mining	Mining	NNP
work_27d5tr326feydgcx32t4nhl424	285	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	286	1	page	page	NN
work_27d5tr326feydgcx32t4nhl424	286	2	354363	354363	CD
work_27d5tr326feydgcx32t4nhl424	286	3	.	.	.
work_27d5tr326feydgcx32t4nhl424	287	1	Rebecca	Rebecca	NNP
work_27d5tr326feydgcx32t4nhl424	287	2	Mason	Mason	NNP
work_27d5tr326feydgcx32t4nhl424	287	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	287	4	Eugene	Eugene	NNP
work_27d5tr326feydgcx32t4nhl424	287	5	Charniak	Charniak	NNP
work_27d5tr326feydgcx32t4nhl424	287	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	288	1	2011	2011	CD
work_27d5tr326feydgcx32t4nhl424	288	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	289	1	Extractive	extractive	JJ
work_27d5tr326feydgcx32t4nhl424	289	2	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	289	3	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	289	4	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	289	5	summaries	summary	NNS
work_27d5tr326feydgcx32t4nhl424	289	6	should	should	MD
work_27d5tr326feydgcx32t4nhl424	289	7	explicitly	explicitly	RB
work_27d5tr326feydgcx32t4nhl424	289	8	not	not	RB
work_27d5tr326feydgcx32t4nhl424	289	9	con-	con-	NN
work_27d5tr326feydgcx32t4nhl424	289	10	tain	tain	NN
work_27d5tr326feydgcx32t4nhl424	289	11	document	document	NN
work_27d5tr326feydgcx32t4nhl424	289	12	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	289	13	specific	specific	JJ
work_27d5tr326feydgcx32t4nhl424	289	14	content	content	NN
work_27d5tr326feydgcx32t4nhl424	289	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	290	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	290	2	proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	290	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	290	4	ACL	ACL	NNP
work_27d5tr326feydgcx32t4nhl424	290	5	HLT	HLT	NNP
work_27d5tr326feydgcx32t4nhl424	290	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	290	7	page:49	page:49	NNP
work_27d5tr326feydgcx32t4nhl424	290	8	-	-	:
work_27d5tr326feydgcx32t4nhl424	290	9	54	54	CD
work_27d5tr326feydgcx32t4nhl424	290	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	291	1	Ani	Ani	NNP
work_27d5tr326feydgcx32t4nhl424	291	2	Nenkova	Nenkova	NNP
work_27d5tr326feydgcx32t4nhl424	291	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	291	4	Lucy	Lucy	NNP
work_27d5tr326feydgcx32t4nhl424	291	5	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	291	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	292	1	The	the	DT
work_27d5tr326feydgcx32t4nhl424	292	2	impact	impact	NN
work_27d5tr326feydgcx32t4nhl424	292	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	292	4	fre-	fre-	NN
work_27d5tr326feydgcx32t4nhl424	292	5	quency	quency	NN
work_27d5tr326feydgcx32t4nhl424	292	6	on	on	IN
work_27d5tr326feydgcx32t4nhl424	292	7	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	292	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	293	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	293	2	Tech	Tech	NNP
work_27d5tr326feydgcx32t4nhl424	293	3	.	.	.
work_27d5tr326feydgcx32t4nhl424	294	1	Report	Report	NNP
work_27d5tr326feydgcx32t4nhl424	294	2	MSR	MSR	NNP
work_27d5tr326feydgcx32t4nhl424	294	3	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	294	4	TR-	TR-	NNP
work_27d5tr326feydgcx32t4nhl424	294	5	2005	2005	CD
work_27d5tr326feydgcx32t4nhl424	294	6	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	294	7	101	101	CD
work_27d5tr326feydgcx32t4nhl424	294	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	294	9	Microsoft	Microsoft	NNP
work_27d5tr326feydgcx32t4nhl424	294	10	Research	Research	NNP
work_27d5tr326feydgcx32t4nhl424	294	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	294	12	Redwood	Redwood	NNP
work_27d5tr326feydgcx32t4nhl424	294	13	,	,	,
work_27d5tr326feydgcx32t4nhl424	294	14	Washing-	Washing-	NNP
work_27d5tr326feydgcx32t4nhl424	294	15	ton	ton	NN
work_27d5tr326feydgcx32t4nhl424	294	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	294	17	2005	2005	CD
work_27d5tr326feydgcx32t4nhl424	294	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	295	1	Ani	Ani	NNP
work_27d5tr326feydgcx32t4nhl424	295	2	Nenkova	Nenkova	NNP
work_27d5tr326feydgcx32t4nhl424	295	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	295	4	Lucy	Lucy	NNP
work_27d5tr326feydgcx32t4nhl424	295	5	Vanderwende	Vanderwende	NNP
work_27d5tr326feydgcx32t4nhl424	295	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	295	7	Kathleen	Kathleen	NNP
work_27d5tr326feydgcx32t4nhl424	295	8	McKe-	mcke-	RB
work_27d5tr326feydgcx32t4nhl424	295	9	own	own	JJ
work_27d5tr326feydgcx32t4nhl424	295	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	296	1	2006	2006	CD
work_27d5tr326feydgcx32t4nhl424	296	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	297	1	A	a	DT
work_27d5tr326feydgcx32t4nhl424	297	2	compositional	compositional	JJ
work_27d5tr326feydgcx32t4nhl424	297	3	context	context	NN
work_27d5tr326feydgcx32t4nhl424	297	4	sensitive	sensitive	JJ
work_27d5tr326feydgcx32t4nhl424	297	5	multi-	multi-	FW
work_27d5tr326feydgcx32t4nhl424	297	6	document	document	NN
work_27d5tr326feydgcx32t4nhl424	297	7	summarizer	summarizer	NNP
work_27d5tr326feydgcx32t4nhl424	297	8	:	:	:
work_27d5tr326feydgcx32t4nhl424	297	9	exploring	explore	VBG
work_27d5tr326feydgcx32t4nhl424	297	10	the	the	DT
work_27d5tr326feydgcx32t4nhl424	297	11	factors	factor	NNS
work_27d5tr326feydgcx32t4nhl424	297	12	that	that	WDT
work_27d5tr326feydgcx32t4nhl424	297	13	inu-	inu-	JJ
work_27d5tr326feydgcx32t4nhl424	297	14	ence	ence	NN
work_27d5tr326feydgcx32t4nhl424	297	15	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	297	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	298	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	298	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	298	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	298	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	298	5	29th	29th	NN
work_27d5tr326feydgcx32t4nhl424	298	6	an-	an-	CC
work_27d5tr326feydgcx32t4nhl424	298	7	nual	nual	JJ
work_27d5tr326feydgcx32t4nhl424	298	8	International	International	NNP
work_27d5tr326feydgcx32t4nhl424	298	9	ACM	ACM	NNP
work_27d5tr326feydgcx32t4nhl424	298	10	SIGIR	SIGIR	NNP
work_27d5tr326feydgcx32t4nhl424	298	11	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	298	12	on	on	IN
work_27d5tr326feydgcx32t4nhl424	298	13	Re-	Re-	NNP
work_27d5tr326feydgcx32t4nhl424	298	14	97	97	CD
work_27d5tr326feydgcx32t4nhl424	298	15	search	search	NN
work_27d5tr326feydgcx32t4nhl424	298	16	and	and	CC
work_27d5tr326feydgcx32t4nhl424	298	17	Development	development	NN
work_27d5tr326feydgcx32t4nhl424	298	18	in	in	IN
work_27d5tr326feydgcx32t4nhl424	298	19	Information	Information	NNP
work_27d5tr326feydgcx32t4nhl424	298	20	Retrieval	Retrieval	NNP
work_27d5tr326feydgcx32t4nhl424	298	21	,	,	,
work_27d5tr326feydgcx32t4nhl424	298	22	page	page	NN
work_27d5tr326feydgcx32t4nhl424	298	23	573	573	CD
work_27d5tr326feydgcx32t4nhl424	298	24	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	298	25	580	580	CD
work_27d5tr326feydgcx32t4nhl424	298	26	.	.	.
work_27d5tr326feydgcx32t4nhl424	299	1	Miles	Miles	NNP
work_27d5tr326feydgcx32t4nhl424	299	2	Osborne	Osborne	NNP
work_27d5tr326feydgcx32t4nhl424	299	3	.	.	.
work_27d5tr326feydgcx32t4nhl424	300	1	2002	2002	CD
work_27d5tr326feydgcx32t4nhl424	300	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	301	1	Using	use	VBG
work_27d5tr326feydgcx32t4nhl424	301	2	maximum	maximum	NN
work_27d5tr326feydgcx32t4nhl424	301	3	entropy	entropy	NN
work_27d5tr326feydgcx32t4nhl424	301	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	301	5	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	301	6	tence	tence	NN
work_27d5tr326feydgcx32t4nhl424	301	7	extraction	extraction	NN
work_27d5tr326feydgcx32t4nhl424	301	8	.	.	.
work_27d5tr326feydgcx32t4nhl424	302	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	302	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	302	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	302	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	302	5	ACL-02	acl-02	NN
work_27d5tr326feydgcx32t4nhl424	302	6	Work-	Work-	NNP
work_27d5tr326feydgcx32t4nhl424	302	7	shop	shop	NN
work_27d5tr326feydgcx32t4nhl424	302	8	on	on	IN
work_27d5tr326feydgcx32t4nhl424	302	9	Automatic	Automatic	NNP
work_27d5tr326feydgcx32t4nhl424	302	10	Summarization	Summarization	NNP
work_27d5tr326feydgcx32t4nhl424	302	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	302	12	Volume	volume	NN
work_27d5tr326feydgcx32t4nhl424	302	13	4	4	CD
work_27d5tr326feydgcx32t4nhl424	302	14	page:1-	page:1-	CD
work_27d5tr326feydgcx32t4nhl424	302	15	8	8	CD
work_27d5tr326feydgcx32t4nhl424	302	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	303	1	Jahna	Jahna	NNP
work_27d5tr326feydgcx32t4nhl424	303	2	Otterbacher	Otterbacher	NNP
work_27d5tr326feydgcx32t4nhl424	303	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	303	4	Gunes	Gunes	NNPS
work_27d5tr326feydgcx32t4nhl424	303	5	Erkan	Erkan	NNP
work_27d5tr326feydgcx32t4nhl424	303	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	303	7	Dragomir	Dragomir	NNP
work_27d5tr326feydgcx32t4nhl424	303	8	Radev	Radev	NNP
work_27d5tr326feydgcx32t4nhl424	303	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	304	1	2005	2005	CD
work_27d5tr326feydgcx32t4nhl424	304	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	305	1	Using	use	VBG
work_27d5tr326feydgcx32t4nhl424	305	2	random	random	JJ
work_27d5tr326feydgcx32t4nhl424	305	3	walks	walk	NNS
work_27d5tr326feydgcx32t4nhl424	305	4	for	for	IN
work_27d5tr326feydgcx32t4nhl424	305	5	question	question	NN
work_27d5tr326feydgcx32t4nhl424	305	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	305	7	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	305	8	sen-	sen-	NN
work_27d5tr326feydgcx32t4nhl424	305	9	tence	tence	NN
work_27d5tr326feydgcx32t4nhl424	305	10	retrieval	retrieval	NN
work_27d5tr326feydgcx32t4nhl424	305	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	306	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	306	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	306	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	306	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	306	5	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	306	6	on	on	IN
work_27d5tr326feydgcx32t4nhl424	306	7	Human	Human	NNP
work_27d5tr326feydgcx32t4nhl424	306	8	Language	Language	NNP
work_27d5tr326feydgcx32t4nhl424	306	9	Technology	Technology	NNP
work_27d5tr326feydgcx32t4nhl424	306	10	and	and	CC
work_27d5tr326feydgcx32t4nhl424	306	11	Empirical	Empirical	NNP
work_27d5tr326feydgcx32t4nhl424	306	12	Methods	Methods	NNPS
work_27d5tr326feydgcx32t4nhl424	306	13	in	in	IN
work_27d5tr326feydgcx32t4nhl424	306	14	Natural	Natural	NNP
work_27d5tr326feydgcx32t4nhl424	306	15	Language	Language	NNP
work_27d5tr326feydgcx32t4nhl424	306	16	Processing	Processing	NNP
work_27d5tr326feydgcx32t4nhl424	306	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	306	18	page	page	NN
work_27d5tr326feydgcx32t4nhl424	306	19	915	915	CD
work_27d5tr326feydgcx32t4nhl424	306	20	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	306	21	922	922	CD
work_27d5tr326feydgcx32t4nhl424	306	22	You	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	306	23	Ouyang	Ouyang	NNP
work_27d5tr326feydgcx32t4nhl424	306	24	,	,	,
work_27d5tr326feydgcx32t4nhl424	306	25	Wenjie	Wenjie	NNP
work_27d5tr326feydgcx32t4nhl424	306	26	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	306	27	,	,	,
work_27d5tr326feydgcx32t4nhl424	306	28	Sujian	Sujian	NNP
work_27d5tr326feydgcx32t4nhl424	306	29	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	306	30	and	and	CC
work_27d5tr326feydgcx32t4nhl424	306	31	Qin	Qin	NNP
work_27d5tr326feydgcx32t4nhl424	306	32	Lua	Lua	NNP
work_27d5tr326feydgcx32t4nhl424	306	33	.	.	.
work_27d5tr326feydgcx32t4nhl424	307	1	2011	2011	CD
work_27d5tr326feydgcx32t4nhl424	307	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	308	1	Applying	apply	VBG
work_27d5tr326feydgcx32t4nhl424	308	2	regression	regression	NN
work_27d5tr326feydgcx32t4nhl424	308	3	models	model	NNS
work_27d5tr326feydgcx32t4nhl424	308	4	to	to	IN
work_27d5tr326feydgcx32t4nhl424	308	5	query	query	NN
work_27d5tr326feydgcx32t4nhl424	308	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	308	7	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	308	8	multi-	multi-	JJ
work_27d5tr326feydgcx32t4nhl424	308	9	document	document	NN
work_27d5tr326feydgcx32t4nhl424	308	10	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	308	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	309	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	309	2	Information	Information	NNP
work_27d5tr326feydgcx32t4nhl424	309	3	Processing	Processing	NNP
work_27d5tr326feydgcx32t4nhl424	309	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	309	5	Management	Management	NNP
work_27d5tr326feydgcx32t4nhl424	309	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	309	7	page	page	NN
work_27d5tr326feydgcx32t4nhl424	309	8	227	227	CD
work_27d5tr326feydgcx32t4nhl424	309	9	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	309	10	237	237	CD
work_27d5tr326feydgcx32t4nhl424	309	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	310	1	You	-PRON-	PRP
work_27d5tr326feydgcx32t4nhl424	310	2	Ouyang	Ouyang	NNP
work_27d5tr326feydgcx32t4nhl424	310	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	310	4	Sujian	Sujian	NNP
work_27d5tr326feydgcx32t4nhl424	310	5	.	.	.
work_27d5tr326feydgcx32t4nhl424	311	1	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	311	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	311	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	311	4	Wenjie	Wenjie	NNP
work_27d5tr326feydgcx32t4nhl424	311	5	.	.	.
work_27d5tr326feydgcx32t4nhl424	312	1	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	312	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	313	1	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	313	2	,	,	,
work_27d5tr326feydgcx32t4nhl424	313	3	Develop-	Develop-	NNP
work_27d5tr326feydgcx32t4nhl424	313	4	ing	ing	NNP
work_27d5tr326feydgcx32t4nhl424	313	5	learning	learning	NN
work_27d5tr326feydgcx32t4nhl424	313	6	strategies	strategy	NNS
work_27d5tr326feydgcx32t4nhl424	313	7	for	for	IN
work_27d5tr326feydgcx32t4nhl424	313	8	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	313	9	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	313	10	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	313	11	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	313	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	314	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	314	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	314	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	314	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	314	5	sixteenth	sixteenth	JJ
work_27d5tr326feydgcx32t4nhl424	314	6	ACM	ACM	NNP
work_27d5tr326feydgcx32t4nhl424	314	7	conference	conference	NN
work_27d5tr326feydgcx32t4nhl424	314	8	on	on	IN
work_27d5tr326feydgcx32t4nhl424	314	9	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	314	10	on	on	IN
work_27d5tr326feydgcx32t4nhl424	314	11	information	information	NN
work_27d5tr326feydgcx32t4nhl424	314	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	314	13	knowledge	knowledge	NN
work_27d5tr326feydgcx32t4nhl424	314	14	manage-	manage-	NNP
work_27d5tr326feydgcx32t4nhl424	314	15	ment	ment	JJ
work_27d5tr326feydgcx32t4nhl424	314	16	,	,	,
work_27d5tr326feydgcx32t4nhl424	314	17	page	page	NN
work_27d5tr326feydgcx32t4nhl424	314	18	:	:	:
work_27d5tr326feydgcx32t4nhl424	314	19	7986	7986	CD
work_27d5tr326feydgcx32t4nhl424	314	20	.	.	.
work_27d5tr326feydgcx32t4nhl424	315	1	Daniel	Daniel	NNP
work_27d5tr326feydgcx32t4nhl424	315	2	Ramage	Ramage	NNP
work_27d5tr326feydgcx32t4nhl424	315	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	315	4	David	David	NNP
work_27d5tr326feydgcx32t4nhl424	315	5	Hall	Hall	NNP
work_27d5tr326feydgcx32t4nhl424	315	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	315	7	Ramesh	Ramesh	NNP
work_27d5tr326feydgcx32t4nhl424	315	8	Nallapati	Nallapati	NNP
work_27d5tr326feydgcx32t4nhl424	315	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	315	10	Christopher	Christopher	NNP
work_27d5tr326feydgcx32t4nhl424	315	11	Manning	Manning	NNP
work_27d5tr326feydgcx32t4nhl424	315	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	316	1	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	316	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	317	1	Labeled	Labeled	NNP
work_27d5tr326feydgcx32t4nhl424	317	2	LDA	LDA	NNP
work_27d5tr326feydgcx32t4nhl424	317	3	:	:	:
work_27d5tr326feydgcx32t4nhl424	317	4	A	a	DT
work_27d5tr326feydgcx32t4nhl424	317	5	super-	super-	NN
work_27d5tr326feydgcx32t4nhl424	317	6	vised	vise	VBN
work_27d5tr326feydgcx32t4nhl424	317	7	topic	topic	JJ
work_27d5tr326feydgcx32t4nhl424	317	8	model	model	NN
work_27d5tr326feydgcx32t4nhl424	317	9	for	for	IN
work_27d5tr326feydgcx32t4nhl424	317	10	credit	credit	NN
work_27d5tr326feydgcx32t4nhl424	317	11	attribution	attribution	NN
work_27d5tr326feydgcx32t4nhl424	317	12	in	in	IN
work_27d5tr326feydgcx32t4nhl424	317	13	multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	317	14	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	317	15	labeled	labeled	JJ
work_27d5tr326feydgcx32t4nhl424	317	16	corpora	corpora	NN
work_27d5tr326feydgcx32t4nhl424	317	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	318	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	318	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	318	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	318	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	318	5	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	318	6	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	318	7	on	on	IN
work_27d5tr326feydgcx32t4nhl424	318	8	Empirical	Empirical	NNP
work_27d5tr326feydgcx32t4nhl424	318	9	Methods	Methods	NNPS
work_27d5tr326feydgcx32t4nhl424	318	10	in	in	IN
work_27d5tr326feydgcx32t4nhl424	318	11	Natural	Natural	NNP
work_27d5tr326feydgcx32t4nhl424	318	12	Language	Language	NNP
work_27d5tr326feydgcx32t4nhl424	318	13	Processing	Processing	NNP
work_27d5tr326feydgcx32t4nhl424	318	14	,	,	,
work_27d5tr326feydgcx32t4nhl424	318	15	Vol	Vol	NNP
work_27d5tr326feydgcx32t4nhl424	318	16	1	1	CD
work_27d5tr326feydgcx32t4nhl424	318	17	,	,	,
work_27d5tr326feydgcx32t4nhl424	318	18	page	page	NN
work_27d5tr326feydgcx32t4nhl424	318	19	248	248	CD
work_27d5tr326feydgcx32t4nhl424	318	20	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	318	21	256	256	CD
work_27d5tr326feydgcx32t4nhl424	318	22	.	.	.
work_27d5tr326feydgcx32t4nhl424	319	1	Dou	Dou	NNP
work_27d5tr326feydgcx32t4nhl424	319	2	She	She	NNP
work_27d5tr326feydgcx32t4nhl424	319	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	319	4	Jian	Jian	NNP
work_27d5tr326feydgcx32t4nhl424	319	5	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	319	6	Tao	Tao	NNP
work_27d5tr326feydgcx32t4nhl424	319	7	Sun	Sun	NNP
work_27d5tr326feydgcx32t4nhl424	319	8	,	,	,
work_27d5tr326feydgcx32t4nhl424	319	9	Hua	Hua	NNP
work_27d5tr326feydgcx32t4nhl424	319	10	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	319	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	319	12	Qiang	Qiang	NNP
work_27d5tr326feydgcx32t4nhl424	319	13	Yang	Yang	NNP
work_27d5tr326feydgcx32t4nhl424	319	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	319	15	Zheng	Zheng	NNP
work_27d5tr326feydgcx32t4nhl424	319	16	Chen	Chen	NNP
work_27d5tr326feydgcx32t4nhl424	319	17	.	.	.
work_27d5tr326feydgcx32t4nhl424	320	1	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	320	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	321	1	Document	document	NN
work_27d5tr326feydgcx32t4nhl424	321	2	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	321	3	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	321	4	conditional	conditional	JJ
work_27d5tr326feydgcx32t4nhl424	321	5	random	random	JJ
work_27d5tr326feydgcx32t4nhl424	321	6	elds	eld	NNS
work_27d5tr326feydgcx32t4nhl424	321	7	.	.	.
work_27d5tr326feydgcx32t4nhl424	322	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	322	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	322	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	322	4	Inter-	Inter-	NNP
work_27d5tr326feydgcx32t4nhl424	322	5	national	national	NNP
work_27d5tr326feydgcx32t4nhl424	322	6	Joint	Joint	NNP
work_27d5tr326feydgcx32t4nhl424	322	7	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	322	8	on	on	IN
work_27d5tr326feydgcx32t4nhl424	322	9	Artificial	Artificial	NNP
work_27d5tr326feydgcx32t4nhl424	322	10	Intelligence	Intelligence	NNP
work_27d5tr326feydgcx32t4nhl424	322	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	322	12	page	page	NN
work_27d5tr326feydgcx32t4nhl424	322	13	:	:	:
work_27d5tr326feydgcx32t4nhl424	322	14	28622867	28622867	CD
work_27d5tr326feydgcx32t4nhl424	322	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	323	1	V.	V.	NNP
work_27d5tr326feydgcx32t4nhl424	323	2	Varma	Varma	NNP
work_27d5tr326feydgcx32t4nhl424	323	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	4	V.	V.	NNP
work_27d5tr326feydgcx32t4nhl424	323	5	Bharat	Bharat	NNP
work_27d5tr326feydgcx32t4nhl424	323	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	7	S.	S.	NNP
work_27d5tr326feydgcx32t4nhl424	323	8	Kovelamudi	Kovelamudi	NNP
work_27d5tr326feydgcx32t4nhl424	323	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	10	P.	P.	NNP
work_27d5tr326feydgcx32t4nhl424	323	11	Bysani	Bysani	NNP
work_27d5tr326feydgcx32t4nhl424	323	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	13	S.	S.	NNP
work_27d5tr326feydgcx32t4nhl424	323	14	GSK	GSK	NNP
work_27d5tr326feydgcx32t4nhl424	323	15	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	16	K.	K.	NNP
work_27d5tr326feydgcx32t4nhl424	323	17	Kumar	Kumar	NNP
work_27d5tr326feydgcx32t4nhl424	323	18	N	N	NNP
work_27d5tr326feydgcx32t4nhl424	323	19	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	20	K.	K.	NNP
work_27d5tr326feydgcx32t4nhl424	323	21	Reddy	Reddy	NNP
work_27d5tr326feydgcx32t4nhl424	323	22	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	23	N.	N.	NNP
work_27d5tr326feydgcx32t4nhl424	323	24	Maganti	Maganti	NNP
work_27d5tr326feydgcx32t4nhl424	323	25	,	,	,
work_27d5tr326feydgcx32t4nhl424	323	26	IIIT	IIIT	NNP
work_27d5tr326feydgcx32t4nhl424	323	27	Hyderabad	Hyderabad	NNP
work_27d5tr326feydgcx32t4nhl424	323	28	at	at	IN
work_27d5tr326feydgcx32t4nhl424	323	29	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	323	30	2009	2009	CD
work_27d5tr326feydgcx32t4nhl424	323	31	.	.	.
work_27d5tr326feydgcx32t4nhl424	324	1	TAC2009	TAC2009	NNP
work_27d5tr326feydgcx32t4nhl424	324	2	Xiaojun	Xiaojun	NNP
work_27d5tr326feydgcx32t4nhl424	324	3	Wan	Wan	NNP
work_27d5tr326feydgcx32t4nhl424	324	4	and	and	CC
work_27d5tr326feydgcx32t4nhl424	324	5	Jianwu	Jianwu	NNP
work_27d5tr326feydgcx32t4nhl424	324	6	Yang	Yang	NNP
work_27d5tr326feydgcx32t4nhl424	324	7	.	.	.
work_27d5tr326feydgcx32t4nhl424	325	1	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	325	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	326	1	Multi	multi	JJ
work_27d5tr326feydgcx32t4nhl424	326	2	-	-	JJ
work_27d5tr326feydgcx32t4nhl424	326	3	document	document	JJ
work_27d5tr326feydgcx32t4nhl424	326	4	Summarization	Summarization	NNP
work_27d5tr326feydgcx32t4nhl424	326	5	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	326	6	cluster	cluster	NN
work_27d5tr326feydgcx32t4nhl424	326	7	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	326	8	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	326	9	link	link	NN
work_27d5tr326feydgcx32t4nhl424	326	10	analysis	analysis	NN
work_27d5tr326feydgcx32t4nhl424	326	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	327	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	327	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	327	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	327	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	327	5	31st	31st	JJ
work_27d5tr326feydgcx32t4nhl424	327	6	annual	annual	JJ
work_27d5tr326feydgcx32t4nhl424	327	7	international	international	JJ
work_27d5tr326feydgcx32t4nhl424	327	8	ACM	ACM	NNP
work_27d5tr326feydgcx32t4nhl424	327	9	SI-	SI-	NNP
work_27d5tr326feydgcx32t4nhl424	327	10	GIR	GIR	NNP
work_27d5tr326feydgcx32t4nhl424	327	11	conference	conference	NN
work_27d5tr326feydgcx32t4nhl424	327	12	on	on	IN
work_27d5tr326feydgcx32t4nhl424	327	13	Research	research	NN
work_27d5tr326feydgcx32t4nhl424	327	14	and	and	CC
work_27d5tr326feydgcx32t4nhl424	327	15	development	development	NN
work_27d5tr326feydgcx32t4nhl424	327	16	in	in	IN
work_27d5tr326feydgcx32t4nhl424	327	17	in-	in-	NNP
work_27d5tr326feydgcx32t4nhl424	327	18	formation	formation	NN
work_27d5tr326feydgcx32t4nhl424	327	19	retrieval	retrieval	NN
work_27d5tr326feydgcx32t4nhl424	327	20	,	,	,
work_27d5tr326feydgcx32t4nhl424	327	21	page	page	NN
work_27d5tr326feydgcx32t4nhl424	327	22	:	:	:
work_27d5tr326feydgcx32t4nhl424	327	23	299	299	CD
work_27d5tr326feydgcx32t4nhl424	327	24	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	327	25	306	306	CD
work_27d5tr326feydgcx32t4nhl424	327	26	.	.	.
work_27d5tr326feydgcx32t4nhl424	328	1	Xiaojun	Xiaojun	NNP
work_27d5tr326feydgcx32t4nhl424	328	2	Wan	Wan	NNP
work_27d5tr326feydgcx32t4nhl424	328	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	328	4	Jianwu	Jianwu	NNP
work_27d5tr326feydgcx32t4nhl424	328	5	Yang	Yang	NNP
work_27d5tr326feydgcx32t4nhl424	328	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	328	7	Jianguo	Jianguo	NNP
work_27d5tr326feydgcx32t4nhl424	328	8	Xiao	Xiao	NNP
work_27d5tr326feydgcx32t4nhl424	328	9	.	.	.
work_27d5tr326feydgcx32t4nhl424	329	1	2007	2007	CD
work_27d5tr326feydgcx32t4nhl424	329	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	330	1	Manifold	manifold	NN
work_27d5tr326feydgcx32t4nhl424	330	2	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	330	3	ranking	rank	VBG
work_27d5tr326feydgcx32t4nhl424	330	4	based	base	VBN
work_27d5tr326feydgcx32t4nhl424	330	5	topic	topic	NN
work_27d5tr326feydgcx32t4nhl424	330	6	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	330	7	focused	focus	VBN
work_27d5tr326feydgcx32t4nhl424	330	8	multi-	multi-	JJ
work_27d5tr326feydgcx32t4nhl424	330	9	document	document	NN
work_27d5tr326feydgcx32t4nhl424	330	10	summarization	summarization	NN
work_27d5tr326feydgcx32t4nhl424	330	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	331	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	331	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	331	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	331	4	In-	In-	NNP
work_27d5tr326feydgcx32t4nhl424	331	5	ternational	ternational	JJ
work_27d5tr326feydgcx32t4nhl424	331	6	Joint	Joint	NNP
work_27d5tr326feydgcx32t4nhl424	331	7	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	331	8	on	on	IN
work_27d5tr326feydgcx32t4nhl424	331	9	Artificial	Artificial	NNP
work_27d5tr326feydgcx32t4nhl424	331	10	Intelligence	Intelligence	NNP
work_27d5tr326feydgcx32t4nhl424	331	11	,	,	,
work_27d5tr326feydgcx32t4nhl424	331	12	page	page	NN
work_27d5tr326feydgcx32t4nhl424	331	13	2903	2903	CD
work_27d5tr326feydgcx32t4nhl424	331	14	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	331	15	2908	2908	CD
work_27d5tr326feydgcx32t4nhl424	331	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	332	1	Furu	Furu	NNP
work_27d5tr326feydgcx32t4nhl424	332	2	Wei	Wei	NNP
work_27d5tr326feydgcx32t4nhl424	332	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	332	4	Wenjie	Wenjie	NNP
work_27d5tr326feydgcx32t4nhl424	332	5	Li	Li	NNP
work_27d5tr326feydgcx32t4nhl424	332	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	332	7	Qin	Qin	NNP
work_27d5tr326feydgcx32t4nhl424	332	8	Lu	Lu	NNP
work_27d5tr326feydgcx32t4nhl424	332	9	and	and	CC
work_27d5tr326feydgcx32t4nhl424	332	10	Yanxiang	Yanxiang	NNP
work_27d5tr326feydgcx32t4nhl424	332	11	He	He	NNP
work_27d5tr326feydgcx32t4nhl424	332	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	333	1	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	333	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	334	1	Ex-	ex-	XX
work_27d5tr326feydgcx32t4nhl424	334	2	ploiting	ploite	VBG
work_27d5tr326feydgcx32t4nhl424	334	3	Query	query	JJ
work_27d5tr326feydgcx32t4nhl424	334	4	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	334	5	Sensitive	sensitive	JJ
work_27d5tr326feydgcx32t4nhl424	334	6	Similarity	Similarity	NNP
work_27d5tr326feydgcx32t4nhl424	334	7	for	for	IN
work_27d5tr326feydgcx32t4nhl424	334	8	Graph	Graph	NNP
work_27d5tr326feydgcx32t4nhl424	334	9	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	334	10	Based	base	VBN
work_27d5tr326feydgcx32t4nhl424	334	11	Query	Query	NNP
work_27d5tr326feydgcx32t4nhl424	334	12	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	334	13	Oriented	Oriented	NNP
work_27d5tr326feydgcx32t4nhl424	334	14	Summarization	Summarization	NNP
work_27d5tr326feydgcx32t4nhl424	334	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	335	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	335	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	335	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	335	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	335	5	31st	31st	JJ
work_27d5tr326feydgcx32t4nhl424	335	6	annual	annual	JJ
work_27d5tr326feydgcx32t4nhl424	335	7	International	International	NNP
work_27d5tr326feydgcx32t4nhl424	335	8	ACM	ACM	NNP
work_27d5tr326feydgcx32t4nhl424	335	9	SIGIR	SIGIR	NNP
work_27d5tr326feydgcx32t4nhl424	335	10	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	335	11	on	on	IN
work_27d5tr326feydgcx32t4nhl424	335	12	Research	Research	NNP
work_27d5tr326feydgcx32t4nhl424	335	13	and	and	CC
work_27d5tr326feydgcx32t4nhl424	335	14	Development	Development	NNP
work_27d5tr326feydgcx32t4nhl424	335	15	in	in	IN
work_27d5tr326feydgcx32t4nhl424	335	16	Information	Information	NNP
work_27d5tr326feydgcx32t4nhl424	335	17	Retrieval	Retrieval	NNP
work_27d5tr326feydgcx32t4nhl424	335	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	335	19	page	page	NN
work_27d5tr326feydgcx32t4nhl424	335	20	283	283	CD
work_27d5tr326feydgcx32t4nhl424	335	21	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	335	22	290	290	CD
work_27d5tr326feydgcx32t4nhl424	335	23	.	.	.
work_27d5tr326feydgcx32t4nhl424	336	1	Jin	Jin	NNP
work_27d5tr326feydgcx32t4nhl424	336	2	Zhang	Zhang	NNP
work_27d5tr326feydgcx32t4nhl424	336	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	336	4	Xueqi	Xueqi	NNP
work_27d5tr326feydgcx32t4nhl424	336	5	Cheng	Cheng	NNP
work_27d5tr326feydgcx32t4nhl424	336	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	336	7	Hongbo	Hongbo	NNP
work_27d5tr326feydgcx32t4nhl424	336	8	Xu	Xu	NNP
work_27d5tr326feydgcx32t4nhl424	336	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	336	10	Xiaolei	Xiaolei	NNP
work_27d5tr326feydgcx32t4nhl424	336	11	Wang	Wang	NNP
work_27d5tr326feydgcx32t4nhl424	336	12	,	,	,
work_27d5tr326feydgcx32t4nhl424	336	13	Yil-	Yil-	NNP
work_27d5tr326feydgcx32t4nhl424	336	14	ing	ing	NNP
work_27d5tr326feydgcx32t4nhl424	336	15	Zeng	Zeng	NNP
work_27d5tr326feydgcx32t4nhl424	336	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	337	1	ICTCAS	ICTCAS	NNP
work_27d5tr326feydgcx32t4nhl424	337	2	’s	’s	POS
work_27d5tr326feydgcx32t4nhl424	337	3	ICTGrasper	ictgrasper	NN
work_27d5tr326feydgcx32t4nhl424	337	4	at	at	IN
work_27d5tr326feydgcx32t4nhl424	337	5	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	337	6	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	337	7	:	:	:
work_27d5tr326feydgcx32t4nhl424	337	8	Sum-	Sum-	NNP
work_27d5tr326feydgcx32t4nhl424	337	9	marizing	marize	VBG
work_27d5tr326feydgcx32t4nhl424	337	10	Dynamic	Dynamic	NNP
work_27d5tr326feydgcx32t4nhl424	337	11	Information	Information	NNP
work_27d5tr326feydgcx32t4nhl424	337	12	with	with	IN
work_27d5tr326feydgcx32t4nhl424	337	13	Signature	Signature	NNP
work_27d5tr326feydgcx32t4nhl424	337	14	Terms	term	NNS
work_27d5tr326feydgcx32t4nhl424	337	15	Based	base	VBD
work_27d5tr326feydgcx32t4nhl424	337	16	Content	Content	NNP
work_27d5tr326feydgcx32t4nhl424	337	17	Filtering	Filtering	NNP
work_27d5tr326feydgcx32t4nhl424	337	18	,	,	,
work_27d5tr326feydgcx32t4nhl424	337	19	TAC	TAC	NNP
work_27d5tr326feydgcx32t4nhl424	337	20	2008	2008	CD
work_27d5tr326feydgcx32t4nhl424	337	21	.	.	.
work_27d5tr326feydgcx32t4nhl424	338	1	Dengzhong	dengzhong	JJ
work_27d5tr326feydgcx32t4nhl424	338	2	Zhou	Zhou	NNP
work_27d5tr326feydgcx32t4nhl424	338	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	338	4	Jason	Jason	NNP
work_27d5tr326feydgcx32t4nhl424	338	5	Weston	Weston	NNP
work_27d5tr326feydgcx32t4nhl424	338	6	,	,	,
work_27d5tr326feydgcx32t4nhl424	338	7	Arthur	Arthur	NNP
work_27d5tr326feydgcx32t4nhl424	338	8	Gretton	Gretton	NNP
work_27d5tr326feydgcx32t4nhl424	338	9	,	,	,
work_27d5tr326feydgcx32t4nhl424	338	10	Olivier	Olivier	NNP
work_27d5tr326feydgcx32t4nhl424	338	11	Bousquet	Bousquet	NNP
work_27d5tr326feydgcx32t4nhl424	338	12	and	and	CC
work_27d5tr326feydgcx32t4nhl424	338	13	Bernhard	Bernhard	NNP
work_27d5tr326feydgcx32t4nhl424	338	14	Schlkopf	Schlkopf	NNP
work_27d5tr326feydgcx32t4nhl424	338	15	.	.	.
work_27d5tr326feydgcx32t4nhl424	339	1	2003	2003	CD
work_27d5tr326feydgcx32t4nhl424	339	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	340	1	Ranking	rank	VBG
work_27d5tr326feydgcx32t4nhl424	340	2	on	on	IN
work_27d5tr326feydgcx32t4nhl424	340	3	Data	Data	NNP
work_27d5tr326feydgcx32t4nhl424	340	4	Manifolds	Manifolds	NNP
work_27d5tr326feydgcx32t4nhl424	340	5	.	.	.
work_27d5tr326feydgcx32t4nhl424	341	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	341	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	341	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	341	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	341	5	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	341	6	on	on	IN
work_27d5tr326feydgcx32t4nhl424	341	7	Advances	Advances	NNPS
work_27d5tr326feydgcx32t4nhl424	341	8	in	in	IN
work_27d5tr326feydgcx32t4nhl424	341	9	Neural	Neural	NNP
work_27d5tr326feydgcx32t4nhl424	341	10	Information	Information	NNP
work_27d5tr326feydgcx32t4nhl424	341	11	Processing	Processing	NNP
work_27d5tr326feydgcx32t4nhl424	341	12	Systems	Systems	NNPS
work_27d5tr326feydgcx32t4nhl424	341	13	,	,	,
work_27d5tr326feydgcx32t4nhl424	341	14	page	page	NN
work_27d5tr326feydgcx32t4nhl424	341	15	169	169	CD
work_27d5tr326feydgcx32t4nhl424	341	16	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	341	17	176	176	CD
work_27d5tr326feydgcx32t4nhl424	341	18	.	.	.
work_27d5tr326feydgcx32t4nhl424	342	1	Jun	Jun	NNP
work_27d5tr326feydgcx32t4nhl424	342	2	Zhu	Zhu	NNP
work_27d5tr326feydgcx32t4nhl424	342	3	and	and	CC
work_27d5tr326feydgcx32t4nhl424	342	4	Eric	Eric	NNP
work_27d5tr326feydgcx32t4nhl424	342	5	Xing	Xing	NNP
work_27d5tr326feydgcx32t4nhl424	342	6	.	.	.
work_27d5tr326feydgcx32t4nhl424	343	1	2010	2010	CD
work_27d5tr326feydgcx32t4nhl424	343	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	344	1	Conditional	Conditional	NNP
work_27d5tr326feydgcx32t4nhl424	344	2	Topic	Topic	NNP
work_27d5tr326feydgcx32t4nhl424	344	3	Random	Random	NNP
work_27d5tr326feydgcx32t4nhl424	344	4	Fields	Fields	NNPS
work_27d5tr326feydgcx32t4nhl424	344	5	.	.	.
work_27d5tr326feydgcx32t4nhl424	345	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	345	2	Proceedings	proceeding	NNS
work_27d5tr326feydgcx32t4nhl424	345	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	345	4	the	the	DT
work_27d5tr326feydgcx32t4nhl424	345	5	27th	27th	JJ
work_27d5tr326feydgcx32t4nhl424	345	6	International	International	NNP
work_27d5tr326feydgcx32t4nhl424	345	7	Con-	Con-	NNP
work_27d5tr326feydgcx32t4nhl424	345	8	ference	ference	NN
work_27d5tr326feydgcx32t4nhl424	345	9	on	on	IN
work_27d5tr326feydgcx32t4nhl424	345	10	Machine	Machine	NNP
work_27d5tr326feydgcx32t4nhl424	345	11	Learning	Learning	NNP
work_27d5tr326feydgcx32t4nhl424	345	12	.	.	.
work_27d5tr326feydgcx32t4nhl424	346	1	Xiaojin	Xiaojin	NNP
work_27d5tr326feydgcx32t4nhl424	346	2	Zhu	Zhu	NNP
work_27d5tr326feydgcx32t4nhl424	346	3	,	,	,
work_27d5tr326feydgcx32t4nhl424	346	4	Zoubin	Zoubin	NNP
work_27d5tr326feydgcx32t4nhl424	346	5	Ghahramani	Ghahramani	NNP
work_27d5tr326feydgcx32t4nhl424	346	6	and	and	CC
work_27d5tr326feydgcx32t4nhl424	346	7	John	John	NNP
work_27d5tr326feydgcx32t4nhl424	346	8	Laf-	Laf-	NNP
work_27d5tr326feydgcx32t4nhl424	346	9	ferty	ferty	NN
work_27d5tr326feydgcx32t4nhl424	346	10	.	.	.
work_27d5tr326feydgcx32t4nhl424	347	1	2003	2003	CD
work_27d5tr326feydgcx32t4nhl424	347	2	.	.	.
work_27d5tr326feydgcx32t4nhl424	348	1	Semi	semi	RB
work_27d5tr326feydgcx32t4nhl424	348	2	-	-	HYPH
work_27d5tr326feydgcx32t4nhl424	348	3	supervised	supervised	JJ
work_27d5tr326feydgcx32t4nhl424	348	4	Learning	Learning	NNP
work_27d5tr326feydgcx32t4nhl424	348	5	using	use	VBG
work_27d5tr326feydgcx32t4nhl424	348	6	Gaussian	Gaussian	NNP
work_27d5tr326feydgcx32t4nhl424	348	7	Fields	Fields	NNPS
work_27d5tr326feydgcx32t4nhl424	348	8	and	and	CC
work_27d5tr326feydgcx32t4nhl424	348	9	Harmonic	Harmonic	NNP
work_27d5tr326feydgcx32t4nhl424	348	10	Functions	Functions	NNP
work_27d5tr326feydgcx32t4nhl424	348	11	.	.	.
work_27d5tr326feydgcx32t4nhl424	349	1	In	in	IN
work_27d5tr326feydgcx32t4nhl424	349	2	Proceedings	Proceedings	NNP
work_27d5tr326feydgcx32t4nhl424	349	3	of	of	IN
work_27d5tr326feydgcx32t4nhl424	349	4	In-	In-	NNP
work_27d5tr326feydgcx32t4nhl424	349	5	ternational	ternational	JJ
work_27d5tr326feydgcx32t4nhl424	349	6	Conference	Conference	NNP
work_27d5tr326feydgcx32t4nhl424	349	7	of	of	IN
work_27d5tr326feydgcx32t4nhl424	349	8	Machine	Machine	NNP
work_27d5tr326feydgcx32t4nhl424	349	9	Learning	Learning	NNP
work_27d5tr326feydgcx32t4nhl424	349	10	,	,	,
work_27d5tr326feydgcx32t4nhl424	349	11	page	page	NN
work_27d5tr326feydgcx32t4nhl424	349	12	:	:	:
work_27d5tr326feydgcx32t4nhl424	349	13	912	912	CD
work_27d5tr326feydgcx32t4nhl424	349	14	-	-	SYM
work_27d5tr326feydgcx32t4nhl424	349	15	919	919	CD
work_27d5tr326feydgcx32t4nhl424	349	16	.	.	.
work_27d5tr326feydgcx32t4nhl424	350	1	98	98	CD