id	sid	tid	token	lemma	pos
ajst-20212	1	1	academic	academic	ADJ
ajst-20212	1	2	journal	journal	NOUN
ajst-20212	1	3	of	of	ADP
ajst-20212	1	4	science	science	NOUN
ajst-20212	1	5	and	and	CCONJ
ajst-20212	1	6	technology	technology	NOUN
ajst-20212	1	7	issn	issn	NOUN
ajst-20212	1	8	:	:	PUNCT
ajst-20212	1	9	2771	2771	NUM
ajst-20212	1	10	-	-	SYM
ajst-20212	1	11	3032	3032	NUM
ajst-20212	1	12	|	|	NOUN
ajst-20212	1	13	vol	vol	NOUN
ajst-20212	1	14	.	.	PROPN
ajst-20212	2	1	10	10	NUM
ajst-20212	2	2	,	,	PUNCT
ajst-20212	2	3	no	no	INTJ
ajst-20212	2	4	.	.	NOUN
ajst-20212	2	5	2	2	NUM
ajst-20212	2	6	,	,	PUNCT
ajst-20212	2	7	2024	2024	NUM
ajst-20212	2	8	167	167	NUM
ajst-20212	2	9	attribute‐level	attribute‐level	NOUN
ajst-20212	2	10	sentiment	sentiment	NOUN
ajst-20212	2	11	analysis	analysis	NOUN
ajst-20212	2	12	incorporating	incorporate	VERB
ajst-20212	2	13	auxiliary	auxiliary	ADJ
ajst-20212	2	14	information	information	NOUN
ajst-20212	2	15	huiyang	huiyang	PROPN
ajst-20212	2	16	yin1	yin1	PROPN
ajst-20212	2	17	,	,	PUNCT
ajst-20212	2	18	jin	jin	PROPN
ajst-20212	2	19	hou1	hou1	PROPN
ajst-20212	2	20	,	,	PUNCT
ajst-20212	2	21	2	2	NUM
ajst-20212	2	22	,	,	PUNCT
ajst-20212	2	23	*	*	PUNCT
ajst-20212	2	24	1	1	NUM
ajst-20212	2	25	school	school	NOUN
ajst-20212	2	26	of	of	ADP
ajst-20212	2	27	automation	automation	NOUN
ajst-20212	2	28	and	and	CCONJ
ajst-20212	2	29	information	information	NOUN
ajst-20212	2	30	engineering	engineering	NOUN
ajst-20212	2	31	,	,	PUNCT
ajst-20212	2	32	sichuan	sichuan	PROPN
ajst-20212	2	33	university	university	PROPN
ajst-20212	2	34	of	of	ADP
ajst-20212	2	35	science	science	PROPN
ajst-20212	2	36	&	&	CCONJ
ajst-20212	2	37	engineering	engineering	PROPN
ajst-20212	2	38	,	,	PUNCT
ajst-20212	2	39	zigong	zigong	PROPN
ajst-20212	2	40	,	,	PUNCT
ajst-20212	2	41	china	china	PROPN
ajst-20212	2	42	2	2	NUM
ajst-20212	2	43	artificial	artificial	ADJ
ajst-20212	2	44	intelligence	intelligence	NOUN
ajst-20212	2	45	key	key	NOUN
ajst-20212	2	46	laboratory	laboratory	NOUN
ajst-20212	2	47	of	of	ADP
ajst-20212	2	48	sichuan	sichuan	PROPN
ajst-20212	2	49	province	province	PROPN
ajst-20212	2	50	,	,	PUNCT
ajst-20212	2	51	yibin	yibin	PROPN
ajst-20212	2	52	,	,	PUNCT
ajst-20212	2	53	china	china	PROPN
ajst-20212	2	54	*	*	PUNCT
ajst-20212	2	55	corresponding	correspond	VERB
ajst-20212	2	56	author	author	NOUN
ajst-20212	2	57	:	:	PUNCT
ajst-20212	2	58	jin	jin	PROPN
ajst-20212	2	59	hou	hou	PROPN
ajst-20212	2	60	(	(	PUNCT
ajst-20212	2	61	email	email	NOUN
ajst-20212	2	62	:	:	PUNCT
ajst-20212	2	63	houjin828@sina.com	houjin828@sina.com	X
ajst-20212	2	64	)	)	PUNCT
ajst-20212	2	65	abstract	abstract	NOUN
ajst-20212	2	66	:	:	PUNCT
ajst-20212	2	67	in	in	ADP
ajst-20212	2	68	order	order	NOUN
ajst-20212	2	69	to	to	PART
ajst-20212	2	70	obtain	obtain	VERB
ajst-20212	2	71	the	the	DET
ajst-20212	2	72	sentiment	sentiment	NOUN
ajst-20212	2	73	tendency	tendency	NOUN
ajst-20212	2	74	of	of	ADP
ajst-20212	2	75	multiple	multiple	ADJ
ajst-20212	2	76	attributes	attribute	NOUN
ajst-20212	2	77	contained	contain	VERB
ajst-20212	2	78	in	in	ADP
ajst-20212	2	79	the	the	DET
ajst-20212	2	80	data	datum	NOUN
ajst-20212	2	81	reviews	review	NOUN
ajst-20212	2	82	,	,	PUNCT
ajst-20212	2	83	this	this	DET
ajst-20212	2	84	paper	paper	NOUN
ajst-20212	2	85	uses	use	VERB
ajst-20212	2	86	an	an	DET
ajst-20212	2	87	attribute	attribute	NOUN
ajst-20212	2	88	-	-	PUNCT
ajst-20212	2	89	level	level	NOUN
ajst-20212	2	90	sentiment	sentiment	NOUN
ajst-20212	2	91	analysis	analysis	NOUN
ajst-20212	2	92	model	model	NOUN
ajst-20212	2	93	rbt	rbt	PROPN
ajst-20212	2	94	-	-	PUNCT
ajst-20212	2	95	biatt	biatt	NOUN
ajst-20212	2	96	-	-	PUNCT
ajst-20212	2	97	gcn	gcn	NOUN
ajst-20212	2	98	,	,	PUNCT
ajst-20212	2	99	which	which	PRON
ajst-20212	2	100	incorporates	incorporate	VERB
ajst-20212	2	101	auxiliary	auxiliary	ADJ
ajst-20212	2	102	information	information	NOUN
ajst-20212	2	103	and	and	CCONJ
ajst-20212	2	104	attention	attention	NOUN
ajst-20212	2	105	mechanism	mechanism	NOUN
ajst-20212	2	106	,	,	PUNCT
ajst-20212	2	107	to	to	PART
ajst-20212	2	108	analyze	analyze	VERB
ajst-20212	2	109	the	the	DET
ajst-20212	2	110	attribute	attribute	NOUN
ajst-20212	2	111	-	-	PUNCT
ajst-20212	2	112	level	level	NOUN
ajst-20212	2	113	sentiment	sentiment	NOUN
ajst-20212	2	114	analysis	analysis	NOUN
ajst-20212	2	115	of	of	ADP
ajst-20212	2	116	e	e	NOUN
ajst-20212	2	117	-	-	NOUN
ajst-20212	2	118	commerce	commerce	NOUN
ajst-20212	2	119	reviews	review	NOUN
ajst-20212	2	120	of	of	ADP
ajst-20212	2	121	citrus	citrus	NOUN
ajst-20212	2	122	.	.	PUNCT
ajst-20212	3	1	the	the	DET
ajst-20212	3	2	model	model	NOUN
ajst-20212	3	3	incorporates	incorporate	VERB
ajst-20212	3	4	auxiliary	auxiliary	ADJ
ajst-20212	3	5	information	information	NOUN
ajst-20212	3	6	,	,	PUNCT
ajst-20212	3	7	attention	attention	NOUN
ajst-20212	3	8	mechanism	mechanism	NOUN
ajst-20212	3	9	,	,	PUNCT
ajst-20212	3	10	and	and	CCONJ
ajst-20212	3	11	gcn	gcn	NOUN
ajst-20212	3	12	network	network	NOUN
ajst-20212	3	13	on	on	ADP
ajst-20212	3	14	the	the	DET
ajst-20212	3	15	basis	basis	NOUN
ajst-20212	3	16	of	of	ADP
ajst-20212	3	17	the	the	DET
ajst-20212	3	18	previous	previous	ADJ
ajst-20212	3	19	one	one	NUM
ajst-20212	3	20	,	,	PUNCT
ajst-20212	3	21	and	and	CCONJ
ajst-20212	3	22	the	the	DET
ajst-20212	3	23	final	final	ADJ
ajst-20212	3	24	macro	macro	ADJ
ajst-20212	3	25	f1	f1	NOUN
ajst-20212	3	26	value	value	NOUN
ajst-20212	3	27	obtained	obtain	VERB
ajst-20212	3	28	can	can	AUX
ajst-20212	3	29	reach	reach	VERB
ajst-20212	3	30	78.21	78.21	NUM
ajst-20212	3	31	%	%	NOUN
ajst-20212	3	32	,	,	PUNCT
ajst-20212	3	33	which	which	PRON
ajst-20212	3	34	achieves	achieve	VERB
ajst-20212	3	35	good	good	ADJ
ajst-20212	3	36	results	result	NOUN
ajst-20212	3	37	in	in	ADP
ajst-20212	3	38	comparison	comparison	NOUN
ajst-20212	3	39	with	with	ADP
ajst-20212	3	40	other	other	ADJ
ajst-20212	3	41	models	model	NOUN
ajst-20212	3	42	.	.	PUNCT
ajst-20212	4	1	keywords	keyword	NOUN
ajst-20212	4	2	:	:	PUNCT
ajst-20212	4	3	emotional	emotional	ADJ
ajst-20212	4	4	classification	classification	NOUN
ajst-20212	4	5	,	,	PUNCT
ajst-20212	4	6	aspect	aspect	NOUN
ajst-20212	4	7	-	-	PUNCT
ajst-20212	4	8	level	level	NOUN
ajst-20212	4	9	emotional	emotional	ADJ
ajst-20212	4	10	analysis	analysis	NOUN
ajst-20212	4	11	,	,	PUNCT
ajst-20212	4	12	attention	attention	NOUN
ajst-20212	4	13	mechanism	mechanism	NOUN
ajst-20212	4	14	.	.	PUNCT
ajst-20212	5	1	1	1	X
ajst-20212	5	2	.	.	X
ajst-20212	5	3	introduction	introduction	NOUN
ajst-20212	5	4	attribute	attribute	NOUN
ajst-20212	5	5	-	-	PUNCT
ajst-20212	5	6	level	level	NOUN
ajst-20212	5	7	sentiment	sentiment	NOUN
ajst-20212	5	8	analysis	analysis	NOUN
ajst-20212	5	9	is	be	AUX
ajst-20212	5	10	a	a	DET
ajst-20212	5	11	fine	fine	ADV
ajst-20212	5	12	-	-	PUNCT
ajst-20212	5	13	grained	grain	VERB
ajst-20212	5	14	sentiment	sentiment	NOUN
ajst-20212	5	15	analysis	analysis	NOUN
ajst-20212	5	16	method	method	NOUN
ajst-20212	5	17	whose	whose	DET
ajst-20212	5	18	goal	goal	NOUN
ajst-20212	5	19	is	be	AUX
ajst-20212	5	20	to	to	PART
ajst-20212	5	21	identify	identify	VERB
ajst-20212	5	22	the	the	DET
ajst-20212	5	23	sentiment	sentiment	NOUN
ajst-20212	5	24	polarity	polarity	NOUN
ajst-20212	5	25	of	of	ADP
ajst-20212	5	26	each	each	DET
ajst-20212	5	27	attribute	attribute	NOUN
ajst-20212	5	28	in	in	ADP
ajst-20212	5	29	a	a	DET
ajst-20212	5	30	statement	statement	NOUN
ajst-20212	5	31	.	.	PUNCT
ajst-20212	6	1	for	for	ADP
ajst-20212	6	2	example	example	NOUN
ajst-20212	6	3	,	,	PUNCT
ajst-20212	6	4	for	for	ADP
ajst-20212	6	5	the	the	DET
ajst-20212	6	6	sentence	sentence	NOUN
ajst-20212	6	7	"	"	PUNCT
ajst-20212	6	8	the	the	DET
ajst-20212	6	9	property	property	NOUN
ajst-20212	6	10	is	be	AUX
ajst-20212	6	11	far	far	ADV
ajst-20212	6	12	away	away	ADV
ajst-20212	6	13	from	from	ADP
ajst-20212	6	14	the	the	DET
ajst-20212	6	15	city	city	NOUN
ajst-20212	6	16	,	,	PUNCT
ajst-20212	6	17	but	but	CCONJ
ajst-20212	6	18	the	the	DET
ajst-20212	6	19	transportation	transportation	NOUN
ajst-20212	6	20	is	be	AUX
ajst-20212	6	21	quite	quite	ADV
ajst-20212	6	22	convenient	convenient	ADJ
ajst-20212	6	23	"	"	PUNCT
ajst-20212	6	24	,	,	PUNCT
ajst-20212	6	25	when	when	SCONJ
ajst-20212	6	26	the	the	DET
ajst-20212	6	27	attribute	attribute	NOUN
ajst-20212	6	28	word	word	NOUN
ajst-20212	6	29	is	be	AUX
ajst-20212	6	30	"	"	PUNCT
ajst-20212	6	31	distance	distance	NOUN
ajst-20212	6	32	"	"	PUNCT
ajst-20212	6	33	,	,	PUNCT
ajst-20212	6	34	the	the	DET
ajst-20212	6	35	sentiment	sentiment	NOUN
ajst-20212	6	36	polarity	polarity	NOUN
ajst-20212	6	37	is	be	AUX
ajst-20212	6	38	negative	negative	ADJ
ajst-20212	6	39	;	;	PUNCT
ajst-20212	6	40	while	while	SCONJ
ajst-20212	6	41	when	when	SCONJ
ajst-20212	6	42	the	the	DET
ajst-20212	6	43	attribute	attribute	NOUN
ajst-20212	6	44	word	word	NOUN
ajst-20212	6	45	is	be	AUX
ajst-20212	6	46	"	"	PUNCT
ajst-20212	6	47	transportation	transportation	NOUN
ajst-20212	6	48	"	"	PUNCT
ajst-20212	6	49	,	,	PUNCT
ajst-20212	6	50	the	the	DET
ajst-20212	6	51	sentiment	sentiment	NOUN
ajst-20212	6	52	polarity	polarity	NOUN
ajst-20212	6	53	is	be	AUX
ajst-20212	6	54	positive.when	positive.when	SCONJ
ajst-20212	6	55	the	the	DET
ajst-20212	6	56	attribute	attribute	NOUN
ajst-20212	6	57	word	word	NOUN
ajst-20212	6	58	is	be	AUX
ajst-20212	6	59	"	"	PUNCT
ajst-20212	6	60	transportation	transportation	NOUN
ajst-20212	6	61	"	"	PUNCT
ajst-20212	6	62	,	,	PUNCT
ajst-20212	6	63	the	the	DET
ajst-20212	6	64	sentiment	sentiment	NOUN
ajst-20212	6	65	polarity	polarity	NOUN
ajst-20212	6	66	is	be	AUX
ajst-20212	6	67	negative	negative	ADJ
ajst-20212	6	68	;	;	PUNCT
ajst-20212	6	69	while	while	SCONJ
ajst-20212	6	70	when	when	SCONJ
ajst-20212	6	71	the	the	DET
ajst-20212	6	72	attribute	attribute	NOUN
ajst-20212	6	73	word	word	NOUN
ajst-20212	6	74	is	be	AUX
ajst-20212	6	75	"	"	PUNCT
ajst-20212	6	76	transportation	transportation	NOUN
ajst-20212	6	77	"	"	PUNCT
ajst-20212	6	78	,	,	PUNCT
ajst-20212	6	79	the	the	DET
ajst-20212	6	80	sentiment	sentiment	NOUN
ajst-20212	6	81	polarity	polarity	NOUN
ajst-20212	6	82	is	be	AUX
ajst-20212	6	83	positive	positive	ADJ
ajst-20212	6	84	.	.	PUNCT
ajst-20212	7	1	fu[1]et	fu[1]et	PROPN
ajst-20212	8	1	al	al	PROPN
ajst-20212	8	2	.	.	PROPN
ajst-20212	8	3	used	use	VERB
ajst-20212	8	4	kl	kl	PROPN
ajst-20212	8	5	scatter	scatter	NOUN
ajst-20212	8	6	to	to	PART
ajst-20212	8	7	explore	explore	VERB
ajst-20212	8	8	the	the	DET
ajst-20212	8	9	association	association	NOUN
ajst-20212	8	10	between	between	ADP
ajst-20212	8	11	a	a	DET
ajst-20212	8	12	given	give	VERB
ajst-20212	8	13	passage	passage	NOUN
ajst-20212	8	14	and	and	CCONJ
ajst-20212	8	15	a	a	DET
ajst-20212	8	16	topic	topic	NOUN
ajst-20212	8	17	model.nadeem	model.nadeem	X
ajst-20212	9	1	akthar	akthar	NOUN
ajst-20212	10	1	[	[	X
ajst-20212	10	2	2	2	NUM
ajst-20212	10	3	]	]	PUNCT
ajst-20212	10	4	et	et	PROPN
ajst-20212	10	5	al	al	PROPN
ajst-20212	10	6	.	.	PROPN
ajst-20212	10	7	used	use	VERB
ajst-20212	10	8	a	a	DET
ajst-20212	10	9	topic	topic	NOUN
ajst-20212	10	10	model	model	NOUN
ajst-20212	10	11	called	call	VERB
ajst-20212	10	12	mallet.huaishao	mallet.huaishao	PROPN
ajst-20212	10	13	luo	luo	PROPN
ajst-20212	11	1	[	[	X
ajst-20212	11	2	3	3	X
ajst-20212	11	3	]	]	PUNCT
ajst-20212	11	4	et	et	PROPN
ajst-20212	11	5	al	al	PROPN
ajst-20212	11	6	.	.	PROPN
ajst-20212	11	7	proposed	propose	VERB
ajst-20212	11	8	a	a	DET
ajst-20212	11	9	bidirectional	bidirectional	ADJ
ajst-20212	11	10	dependency	dependency	NOUN
ajst-20212	11	11	tree	tree	NOUN
ajst-20212	11	12	network	network	NOUN
ajst-20212	11	13	that	that	PRON
ajst-20212	11	14	acquires	acquire	VERB
ajst-20212	11	15	top	top	ADJ
ajst-20212	11	16	-	-	PUNCT
ajst-20212	11	17	down	down	NOUN
ajst-20212	11	18	and	and	CCONJ
ajst-20212	11	19	bottom	bottom	NOUN
ajst-20212	11	20	-	-	PUNCT
ajst-20212	11	21	up	up	ADP
ajst-20212	11	22	representations	representation	NOUN
ajst-20212	11	23	and	and	CCONJ
ajst-20212	11	24	merges	merge	VERB
ajst-20212	11	25	them	they	PRON
ajst-20212	11	26	,	,	PUNCT
ajst-20212	11	27	and	and	CCONJ
ajst-20212	11	28	then	then	ADV
ajst-20212	11	29	learns	learn	VERB
ajst-20212	11	30	feature	feature	NOUN
ajst-20212	11	31	representations	representation	NOUN
ajst-20212	11	32	through	through	ADP
ajst-20212	11	33	neural	neural	ADJ
ajst-20212	11	34	networks	network	NOUN
ajst-20212	11	35	for	for	ADP
ajst-20212	11	36	attribute	attribute	NOUN
ajst-20212	11	37	extraction	extraction	NOUN
ajst-20212	11	38	.	.	PUNCT
ajst-20212	12	1	feature	feature	NOUN
ajst-20212	12	2	representation	representation	NOUN
ajst-20212	12	3	for	for	ADP
ajst-20212	12	4	attribute	attribute	NOUN
ajst-20212	12	5	extraction.yuliya	extraction.yuliya	X
ajst-20212	13	1	[	[	X
ajst-20212	13	2	4	4	NUM
ajst-20212	13	3	]	]	PUNCT
ajst-20212	13	4	used	use	VERB
ajst-20212	13	5	crf	crf	PROPN
ajst-20212	13	6	model	model	NOUN
ajst-20212	13	7	for	for	ADP
ajst-20212	13	8	attribute	attribute	NOUN
ajst-20212	13	9	extraction	extraction	NOUN
ajst-20212	13	10	and	and	CCONJ
ajst-20212	13	11	achieved	achieve	VERB
ajst-20212	13	12	good	good	ADJ
ajst-20212	13	13	results	result	NOUN
ajst-20212	13	14	on	on	ADP
ajst-20212	13	15	restaurant	restaurant	NOUN
ajst-20212	13	16	and	and	CCONJ
ajst-20212	13	17	automobile	automobile	NOUN
ajst-20212	13	18	domain	domain	NOUN
ajst-20212	13	19	datasets.hu	datasets.hu	PROPN
ajst-20212	13	20	xu	xu	PROPN
ajst-20212	14	1	[	[	X
ajst-20212	14	2	5	5	NUM
ajst-20212	14	3	]	]	PUNCT
ajst-20212	14	4	used	use	VERB
ajst-20212	14	5	a	a	DET
ajst-20212	14	6	bidirectional	bidirectional	ADJ
ajst-20212	14	7	convolutional	convolutional	ADJ
ajst-20212	14	8	neural	neural	ADJ
ajst-20212	14	9	network	network	NOUN
ajst-20212	14	10	for	for	ADP
ajst-20212	14	11	attribute	attribute	NOUN
ajst-20212	14	12	word	word	NOUN
ajst-20212	14	13	extraction	extraction	NOUN
ajst-20212	14	14	and	and	CCONJ
ajst-20212	14	15	achieved	achieve	VERB
ajst-20212	14	16	significant	significant	ADJ
ajst-20212	14	17	performance.peng	performance.peng	PROPN
ajst-20212	14	18	chen	chen	PROPN
ajst-20212	15	1	[	[	X
ajst-20212	15	2	6	6	NUM
ajst-20212	15	3	]	]	PUNCT
ajst-20212	15	4	proposed	propose	VERB
ajst-20212	15	5	recurrent	recurrent	ADJ
ajst-20212	15	6	neural	neural	ADJ
ajst-20212	15	7	network	network	NOUN
ajst-20212	15	8	based	base	VERB
ajst-20212	15	9	on	on	ADP
ajst-20212	15	10	multiple	multiple	ADJ
ajst-20212	15	11	attention	attention	NOUN
ajst-20212	15	12	mechanism	mechanism	NOUN
ajst-20212	15	13	and	and	CCONJ
ajst-20212	15	14	applied	apply	VERB
ajst-20212	15	15	it	it	PRON
ajst-20212	15	16	to	to	ADP
ajst-20212	15	17	attribute	attribute	NOUN
ajst-20212	15	18	-	-	PUNCT
ajst-20212	15	19	level	level	NOUN
ajst-20212	15	20	sentiment	sentiment	NOUN
ajst-20212	15	21	analysis	analysis	NOUN
ajst-20212	15	22	.	.	PUNCT
ajst-20212	16	1	in	in	ADP
ajst-20212	16	2	recent	recent	ADJ
ajst-20212	16	3	years	year	NOUN
ajst-20212	16	4	,	,	PUNCT
ajst-20212	16	5	with	with	ADP
ajst-20212	16	6	the	the	DET
ajst-20212	16	7	development	development	NOUN
ajst-20212	16	8	of	of	ADP
ajst-20212	16	9	graph	graph	NOUN
ajst-20212	16	10	neural	neural	ADJ
ajst-20212	16	11	networks	network	NOUN
ajst-20212	16	12	(	(	PUNCT
ajst-20212	16	13	gnn	gnn	PROPN
ajst-20212	16	14	)	)	PUNCT
ajst-20212	16	15	,	,	PUNCT
ajst-20212	16	16	more	more	ADJ
ajst-20212	16	17	attempts	attempt	NOUN
ajst-20212	16	18	have	have	AUX
ajst-20212	16	19	been	be	AUX
ajst-20212	16	20	made	make	VERB
ajst-20212	16	21	to	to	PART
ajst-20212	16	22	utilize	utilize	VERB
ajst-20212	16	23	graph	graph	NOUN
ajst-20212	16	24	neural	neural	ADJ
ajst-20212	16	25	networks	network	NOUN
ajst-20212	16	26	to	to	PART
ajst-20212	16	27	solve	solve	VERB
ajst-20212	16	28	problems	problem	NOUN
ajst-20212	16	29	in	in	ADP
ajst-20212	16	30	attribute	attribute	NOUN
ajst-20212	16	31	-	-	PUNCT
ajst-20212	16	32	level	level	NOUN
ajst-20212	16	33	sentiment	sentiment	NOUN
ajst-20212	16	34	analysis	analysis	NOUN
ajst-20212	16	35	[	[	X
ajst-20212	16	36	7	7	NUM
ajst-20212	16	37	]	]	PUNCT
ajst-20212	16	38	.	.	PUNCT
ajst-20212	17	1	ruifan	ruifan	PROPN
ajst-20212	17	2	li	li	PROPN
ajst-20212	18	1	[	[	X
ajst-20212	18	2	8	8	NUM
ajst-20212	18	3	]	]	PUNCT
ajst-20212	18	4	and	and	CCONJ
ajst-20212	18	5	others	other	NOUN
ajst-20212	18	6	proposed	propose	VERB
ajst-20212	18	7	a	a	DET
ajst-20212	18	8	dual	dual	ADJ
ajst-20212	18	9	graph	graph	NOUN
ajst-20212	18	10	convolutional	convolutional	ADJ
ajst-20212	18	11	network	network	NOUN
ajst-20212	18	12	model	model	NOUN
ajst-20212	18	13	for	for	ADP
ajst-20212	18	14	attribute	attribute	NOUN
ajst-20212	18	15	sentiment	sentiment	NOUN
ajst-20212	18	16	classification	classification	NOUN
ajst-20212	18	17	task	task	NOUN
ajst-20212	18	18	,	,	PUNCT
ajst-20212	18	19	capturing	capture	VERB
ajst-20212	18	20	syntactic	syntactic	ADJ
ajst-20212	18	21	knowledge	knowledge	NOUN
ajst-20212	18	22	and	and	CCONJ
ajst-20212	18	23	semantic	semantic	ADJ
ajst-20212	18	24	relevance	relevance	NOUN
ajst-20212	18	25	respectively	respectively	ADV
ajst-20212	18	26	,	,	PUNCT
ajst-20212	18	27	and	and	CCONJ
ajst-20212	18	28	proposed	propose	VERB
ajst-20212	18	29	two	two	NUM
ajst-20212	18	30	regularization	regularization	NOUN
ajst-20212	18	31	methods	method	NOUN
ajst-20212	18	32	to	to	PART
ajst-20212	18	33	constrain	constrain	VERB
ajst-20212	18	34	the	the	DET
ajst-20212	18	35	model	model	NOUN
ajst-20212	18	36	,	,	PUNCT
ajst-20212	18	37	which	which	PRON
ajst-20212	18	38	led	lead	VERB
ajst-20212	18	39	to	to	ADP
ajst-20212	18	40	a	a	DET
ajst-20212	18	41	better	well	ADJ
ajst-20212	18	42	generalization	generalization	NOUN
ajst-20212	18	43	ability.zeguan	ability.zeguan	PROPN
ajst-20212	18	44	xiao	xiao	PROPN
ajst-20212	19	1	[	[	X
ajst-20212	19	2	9	9	NUM
ajst-20212	19	3	]	]	PUNCT
ajst-20212	19	4	proposed	propose	VERB
ajst-20212	19	5	a	a	DET
ajst-20212	19	6	bert	bert	NOUN
ajst-20212	19	7	-	-	PUNCT
ajst-20212	19	8	based	base	VERB
ajst-20212	19	9	graph	graph	NOUN
ajst-20212	19	10	convolutional	convolutional	ADJ
ajst-20212	19	11	neural	neural	ADJ
ajst-20212	19	12	network	network	NOUN
ajst-20212	19	13	for	for	ADP
ajst-20212	19	14	attribute	attribute	NOUN
ajst-20212	19	15	-	-	PUNCT
ajst-20212	19	16	level	level	NOUN
ajst-20212	19	17	sentiment	sentiment	NOUN
ajst-20212	19	18	analysis	analysis	NOUN
ajst-20212	19	19	.	.	PUNCT
ajst-20212	20	1	ruyan	ruyan	PROPN
ajst-20212	20	2	wang	wang	PROPN
ajst-20212	20	3	et	et	PROPN
ajst-20212	20	4	al	al	PROPN
ajst-20212	20	5	.	.	PROPN
ajst-20212	20	6	proposed	propose	VERB
ajst-20212	20	7	a	a	DET
ajst-20212	20	8	multi	multi	ADJ
ajst-20212	20	9	-	-	ADJ
ajst-20212	20	10	interaction	interaction	NOUN
ajst-20212	20	11	graph	graph	NOUN
ajst-20212	20	12	convolutional	convolutional	ADJ
ajst-20212	20	13	network	network	NOUN
ajst-20212	20	14	model	model	NOUN
ajst-20212	20	15	,	,	PUNCT
ajst-20212	20	16	which	which	PRON
ajst-20212	20	17	takes	take	VERB
ajst-20212	20	18	the	the	DET
ajst-20212	20	19	positional	positional	ADJ
ajst-20212	20	20	features	feature	NOUN
ajst-20212	20	21	of	of	ADP
ajst-20212	20	22	each	each	DET
ajst-20212	20	23	word	word	NOUN
ajst-20212	20	24	in	in	ADP
ajst-20212	20	25	the	the	DET
ajst-20212	20	26	text	text	NOUN
ajst-20212	20	27	as	as	ADP
ajst-20212	20	28	the	the	DET
ajst-20212	20	29	input	input	NOUN
ajst-20212	20	30	to	to	ADP
ajst-20212	20	31	each	each	DET
ajst-20212	20	32	layer	layer	NOUN
ajst-20212	20	33	of	of	ADP
ajst-20212	20	34	the	the	DET
ajst-20212	20	35	gcn	gcn	NOUN
ajst-20212	20	36	,	,	PUNCT
ajst-20212	20	37	and	and	CCONJ
ajst-20212	20	38	at	at	ADP
ajst-20212	20	39	the	the	DET
ajst-20212	20	40	same	same	ADJ
ajst-20212	20	41	time	time	NOUN
ajst-20212	20	42	weights	weight	VERB
ajst-20212	20	43	the	the	DET
ajst-20212	20	44	adjacency	adjacency	NOUN
ajst-20212	20	45	matrices	matrix	NOUN
ajst-20212	20	46	of	of	ADP
ajst-20212	20	47	the	the	DET
ajst-20212	20	48	gcn	gcn	NOUN
ajst-20212	20	49	using	use	VERB
ajst-20212	20	50	the	the	DET
ajst-20212	20	51	dependent	dependent	ADJ
ajst-20212	20	52	syntactic	syntactic	ADJ
ajst-20212	20	53	tree	tree	NOUN
ajst-20212	20	54	,	,	PUNCT
ajst-20212	20	55	and	and	CCONJ
ajst-20212	20	56	the	the	DET
ajst-20212	20	57	syntactic	syntactic	ADJ
ajst-20212	20	58	distance	distance	NOUN
ajst-20212	20	59	features	feature	NOUN
ajst-20212	20	60	of	of	ADP
ajst-20212	20	61	the	the	DET
ajst-20212	20	62	text	text	NOUN
ajst-20212	20	63	words	word	NOUN
ajst-20212	20	64	,	,	PUNCT
ajst-20212	20	65	and	and	CCONJ
ajst-20212	20	66	handles	handle	VERB
ajst-20212	20	67	the	the	DET
ajst-20212	20	68	semantic	semantic	ADJ
ajst-20212	20	69	and	and	CCONJ
ajst-20212	20	70	grammatical	grammatical	ADJ
ajst-20212	20	71	information	information	NOUN
ajst-20212	20	72	between	between	ADP
ajst-20212	20	73	words	word	NOUN
ajst-20212	20	74	through	through	ADP
ajst-20212	20	75	the	the	DET
ajst-20212	20	76	interactions	interaction	NOUN
ajst-20212	20	77	of	of	ADP
ajst-20212	20	78	semantic	semantic	ADJ
ajst-20212	20	79	and	and	CCONJ
ajst-20212	20	80	syntactic	syntactic	ADJ
ajst-20212	20	81	rules	rule	NOUN
ajst-20212	20	82	.	.	PUNCT
ajst-20212	21	1	2	2	X
ajst-20212	21	2	.	.	X
ajst-20212	21	3	related	relate	VERB
ajst-20212	21	4	theory	theory	NOUN
ajst-20212	21	5	and	and	CCONJ
ajst-20212	21	6	technology	technology	NOUN
ajst-20212	21	7	2.1	2.1	NUM
ajst-20212	21	8	.	.	PUNCT
ajst-20212	22	1	bert	bert	PROPN
ajst-20212	22	2	pre	pre	ADJ
ajst-20212	22	3	-	-	ADJ
ajst-20212	22	4	training	training	ADJ
ajst-20212	22	5	models	model	NOUN
ajst-20212	22	6	is	be	AUX
ajst-20212	22	7	a	a	DET
ajst-20212	22	8	current	current	ADJ
ajst-20212	22	9	hot	hot	ADJ
ajst-20212	22	10	topic	topic	NOUN
ajst-20212	22	11	in	in	ADP
ajst-20212	22	12	nlp	nlp	NOUN
ajst-20212	22	13	,	,	PUNCT
ajst-20212	22	14	where	where	SCONJ
ajst-20212	22	15	unsupervised	unsupervised	ADJ
ajst-20212	22	16	learning	learning	NOUN
ajst-20212	22	17	equips	equip	VERB
ajst-20212	22	18	models	model	NOUN
ajst-20212	22	19	with	with	ADP
ajst-20212	22	20	certain	certain	ADJ
ajst-20212	22	21	prior	prior	ADJ
ajst-20212	22	22	knowledge	knowledge	NOUN
ajst-20212	22	23	to	to	PART
ajst-20212	22	24	assist	assist	VERB
ajst-20212	22	25	downstream	downstream	ADJ
ajst-20212	22	26	models	model	NOUN
ajst-20212	22	27	in	in	ADP
ajst-20212	22	28	semantic	semantic	ADJ
ajst-20212	22	29	representation	representation	NOUN
ajst-20212	22	30	of	of	ADP
ajst-20212	22	31	the	the	DET
ajst-20212	22	32	target	target	NOUN
ajst-20212	22	33	task	task	NOUN
ajst-20212	22	34	,	,	PUNCT
ajst-20212	22	35	thus	thus	ADV
ajst-20212	22	36	avoiding	avoid	VERB
ajst-20212	22	37	overfitting	overfitte	VERB
ajst-20212	22	38	.	.	PUNCT
ajst-20212	23	1	bert	bert	PROPN
ajst-20212	23	2	learns	learn	VERB
ajst-20212	23	3	highly	highly	ADV
ajst-20212	23	4	enriched	enrich	VERB
ajst-20212	23	5	language	language	NOUN
ajst-20212	23	6	representations	representation	NOUN
ajst-20212	23	7	by	by	ADP
ajst-20212	23	8	performing	perform	VERB
ajst-20212	23	9	pre	pre	ADJ
ajst-20212	23	10	-	-	ADJ
ajst-20212	23	11	training	training	ADJ
ajst-20212	23	12	tasks	task	NOUN
ajst-20212	23	13	of	of	ADP
ajst-20212	23	14	masked	mask	VERB
ajst-20212	23	15	language	language	NOUN
ajst-20212	23	16	model	model	NOUN
ajst-20212	23	17	(	(	PUNCT
ajst-20212	23	18	mlm	mlm	PROPN
ajst-20212	23	19	)	)	PUNCT
ajst-20212	23	20	and	and	CCONJ
ajst-20212	23	21	next	next	ADJ
ajst-20212	23	22	sentence	sentence	NOUN
ajst-20212	23	23	prediction	prediction	NOUN
ajst-20212	23	24	(	(	PUNCT
ajst-20212	23	25	nsp	nsp	PROPN
ajst-20212	23	26	)	)	PUNCT
ajst-20212	23	27	on	on	ADP
ajst-20212	23	28	a	a	DET
ajst-20212	23	29	large	large	ADJ
ajst-20212	23	30	-	-	PUNCT
ajst-20212	23	31	scale	scale	NOUN
ajst-20212	23	32	text	text	NOUN
ajst-20212	23	33	corpus	corpus	NOUN
ajst-20212	23	34	.	.	PUNCT
ajst-20212	24	1	these	these	DET
ajst-20212	24	2	pre	pre	ADJ
ajst-20212	24	3	-	-	ADJ
ajst-20212	24	4	trained	train	VERB
ajst-20212	24	5	language	language	NOUN
ajst-20212	24	6	representations	representation	NOUN
ajst-20212	24	7	can	can	AUX
ajst-20212	24	8	be	be	AUX
ajst-20212	24	9	fine	fine	ADV
ajst-20212	24	10	-	-	PUNCT
ajst-20212	24	11	tuned	tune	VERB
ajst-20212	24	12	and	and	CCONJ
ajst-20212	24	13	applied	apply	VERB
ajst-20212	24	14	to	to	ADP
ajst-20212	24	15	various	various	ADJ
ajst-20212	24	16	types	type	NOUN
ajst-20212	24	17	of	of	ADP
ajst-20212	24	18	natural	natural	ADJ
ajst-20212	24	19	language	language	NOUN
ajst-20212	24	20	processing	processing	NOUN
ajst-20212	24	21	tasks	task	NOUN
ajst-20212	24	22	,	,	PUNCT
ajst-20212	24	23	such	such	ADJ
ajst-20212	24	24	as	as	ADP
ajst-20212	24	25	sentiment	sentiment	NOUN
ajst-20212	24	26	analysis	analysis	NOUN
ajst-20212	24	27	,	,	PUNCT
ajst-20212	24	28	named	name	VERB
ajst-20212	24	29	entity	entity	NOUN
ajst-20212	24	30	recognition	recognition	NOUN
ajst-20212	24	31	,	,	PUNCT
ajst-20212	24	32	etc	etc	X
ajst-20212	24	33	.	.	X
ajst-20212	24	34	,	,	PUNCT
ajst-20212	24	35	where	where	SCONJ
ajst-20212	24	36	they	they	PRON
ajst-20212	24	37	show	show	VERB
ajst-20212	24	38	excellent	excellent	ADJ
ajst-20212	24	39	performance	performance	NOUN
ajst-20212	24	40	.	.	PUNCT
ajst-20212	25	1	bert	bert	PROPN
ajst-20212	25	2	is	be	AUX
ajst-20212	25	3	known	know	VERB
ajst-20212	25	4	for	for	ADP
ajst-20212	25	5	its	its	PRON
ajst-20212	25	6	strong	strong	ADJ
ajst-20212	25	7	representational	representational	ADJ
ajst-20212	25	8	capabilities	capability	NOUN
ajst-20212	25	9	and	and	CCONJ
ajst-20212	25	10	versatility	versatility	NOUN
ajst-20212	25	11	,	,	PUNCT
ajst-20212	25	12	and	and	CCONJ
ajst-20212	25	13	can	can	AUX
ajst-20212	25	14	be	be	AUX
ajst-20212	25	15	adapted	adapt	VERB
ajst-20212	25	16	to	to	ADP
ajst-20212	25	17	a	a	DET
ajst-20212	25	18	wide	wide	ADJ
ajst-20212	25	19	range	range	NOUN
ajst-20212	25	20	of	of	ADP
ajst-20212	25	21	natural	natural	ADJ
ajst-20212	25	22	language	language	NOUN
ajst-20212	25	23	processing	processing	NOUN
ajst-20212	25	24	tasks	task	NOUN
ajst-20212	25	25	and	and	CCONJ
ajst-20212	25	26	achieve	achieve	VERB
ajst-20212	25	27	excellent	excellent	ADJ
ajst-20212	25	28	performance	performance	NOUN
ajst-20212	25	29	in	in	ADP
ajst-20212	25	30	many	many	ADJ
ajst-20212	25	31	challenging	challenging	ADJ
ajst-20212	25	32	tasks	task	NOUN
ajst-20212	25	33	.	.	PUNCT
ajst-20212	26	1	in	in	ADP
ajst-20212	26	2	addition	addition	NOUN
ajst-20212	26	3	,	,	PUNCT
ajst-20212	26	4	with	with	ADP
ajst-20212	26	5	its	its	PRON
ajst-20212	26	6	open	open	ADJ
ajst-20212	26	7	source	source	NOUN
ajst-20212	26	8	and	and	CCONJ
ajst-20212	26	9	easy	easy	ADJ
ajst-20212	26	10	-	-	PUNCT
ajst-20212	26	11	to	to	PART
ajst-20212	26	12	-	-	PUNCT
ajst-20212	26	13	use	use	NOUN
ajst-20212	26	14	features	feature	NOUN
ajst-20212	26	15	,	,	PUNCT
ajst-20212	26	16	bert	bert	PROPN
ajst-20212	26	17	has	have	AUX
ajst-20212	26	18	become	become	VERB
ajst-20212	26	19	one	one	NUM
ajst-20212	26	20	of	of	ADP
ajst-20212	26	21	the	the	DET
ajst-20212	26	22	important	important	ADJ
ajst-20212	26	23	milestones	milestone	NOUN
ajst-20212	26	24	in	in	ADP
ajst-20212	26	25	the	the	DET
ajst-20212	26	26	field	field	NOUN
ajst-20212	26	27	of	of	ADP
ajst-20212	26	28	natural	natural	ADJ
ajst-20212	26	29	language	language	NOUN
ajst-20212	26	30	processing	processing	NOUN
ajst-20212	26	31	,	,	PUNCT
ajst-20212	26	32	and	and	CCONJ
ajst-20212	26	33	the	the	DET
ajst-20212	26	34	basic	basic	ADJ
ajst-20212	26	35	structure	structure	NOUN
ajst-20212	26	36	of	of	ADP
ajst-20212	26	37	bert	bert	PROPN
ajst-20212	26	38	is	be	AUX
ajst-20212	26	39	shown	show	VERB
ajst-20212	26	40	in	in	ADP
ajst-20212	26	41	figure	figure	NOUN
ajst-20212	26	42	1	1	NUM
ajst-20212	26	43	.	.	PUNCT
ajst-20212	26	44	figure	figure	NOUN
ajst-20212	26	45	1	1	NUM
ajst-20212	26	46	.	.	PUNCT
ajst-20212	27	1	structure	structure	NOUN
ajst-20212	27	2	of	of	ADP
ajst-20212	27	3	bert	bert	PROPN
ajst-20212	27	4	168	168	NUM
ajst-20212	27	5	2.2	2.2	NUM
ajst-20212	27	6	.	.	PUNCT
ajst-20212	28	1	robrta	robrta	PROPN
ajst-20212	28	2	roberta	roberta	PROPN
ajst-20212	28	3	is	be	AUX
ajst-20212	28	4	an	an	DET
ajst-20212	28	5	improved	improved	ADJ
ajst-20212	28	6	pre	pre	ADJ
ajst-20212	28	7	-	-	ADJ
ajst-20212	28	8	trained	train	VERB
ajst-20212	28	9	language	language	NOUN
ajst-20212	28	10	model	model	NOUN
ajst-20212	28	11	introduced	introduce	VERB
ajst-20212	28	12	by	by	ADP
ajst-20212	28	13	facebook	facebook	PROPN
ajst-20212	28	14	ai	ai	VERB
ajst-20212	28	15	in	in	ADP
ajst-20212	28	16	2019	2019	NUM
ajst-20212	28	17	to	to	PART
ajst-20212	28	18	optimize	optimize	VERB
ajst-20212	28	19	and	and	CCONJ
ajst-20212	28	20	enhance	enhance	VERB
ajst-20212	28	21	the	the	DET
ajst-20212	28	22	original	original	ADJ
ajst-20212	28	23	bert	bert	NOUN
ajst-20212	28	24	model.roberta	model.roberta	NOUN
ajst-20212	28	25	is	be	AUX
ajst-20212	28	26	based	base	VERB
ajst-20212	28	27	on	on	ADP
ajst-20212	28	28	the	the	DET
ajst-20212	28	29	transformer	transformer	NOUN
ajst-20212	28	30	architecture	architecture	NOUN
ajst-20212	28	31	and	and	CCONJ
ajst-20212	28	32	uses	use	VERB
ajst-20212	28	33	large	large	ADJ
ajst-20212	28	34	-	-	PUNCT
ajst-20212	28	35	scale	scale	NOUN
ajst-20212	28	36	unsupervised	unsupervised	ADJ
ajst-20212	28	37	training	training	NOUN
ajst-20212	28	38	to	to	PART
ajst-20212	28	39	learn	learn	VERB
ajst-20212	28	40	language	language	NOUN
ajst-20212	28	41	representations	representation	NOUN
ajst-20212	28	42	.	.	PUNCT
ajst-20212	29	1	unlike	unlike	ADP
ajst-20212	29	2	bert	bert	PROPN
ajst-20212	29	3	,	,	PUNCT
ajst-20212	29	4	roberta	roberta	PROPN
ajst-20212	29	5	uses	use	VERB
ajst-20212	29	6	larger	large	ADJ
ajst-20212	29	7	batch	batch	NOUN
ajst-20212	29	8	training	training	NOUN
ajst-20212	29	9	,	,	PUNCT
ajst-20212	29	10	longer	long	ADV
ajst-20212	29	11	training	training	NOUN
ajst-20212	29	12	time	time	NOUN
ajst-20212	29	13	,	,	PUNCT
ajst-20212	29	14	and	and	CCONJ
ajst-20212	29	15	more	more	ADJ
ajst-20212	29	16	data	datum	NOUN
ajst-20212	29	17	for	for	ADP
ajst-20212	29	18	pre	pre	NOUN
ajst-20212	29	19	-	-	NOUN
ajst-20212	29	20	training	training	NOUN
ajst-20212	29	21	to	to	PART
ajst-20212	29	22	achieve	achieve	VERB
ajst-20212	29	23	better	well	ADJ
ajst-20212	29	24	results	result	NOUN
ajst-20212	29	25	in	in	ADP
ajst-20212	29	26	various	various	ADJ
ajst-20212	29	27	natural	natural	ADJ
ajst-20212	29	28	language	language	NOUN
ajst-20212	29	29	processing	processing	NOUN
ajst-20212	29	30	tasks	task	NOUN
ajst-20212	29	31	.	.	PUNCT
ajst-20212	30	1	in	in	ADP
ajst-20212	30	2	addition	addition	NOUN
ajst-20212	30	3	,	,	PUNCT
ajst-20212	30	4	roberta	roberta	PROPN
ajst-20212	30	5	employs	employ	VERB
ajst-20212	30	6	dynamic	dynamic	ADJ
ajst-20212	30	7	masking	masking	NOUN
ajst-20212	30	8	technique	technique	NOUN
ajst-20212	30	9	to	to	PART
ajst-20212	30	10	re	re	VERB
ajst-20212	30	11	-	-	VERB
ajst-20212	30	12	select	select	VERB
ajst-20212	30	13	the	the	DET
ajst-20212	30	14	mask	mask	NOUN
ajst-20212	30	15	position	position	NOUN
ajst-20212	30	16	randomly	randomly	ADV
ajst-20212	30	17	in	in	ADP
ajst-20212	30	18	each	each	DET
ajst-20212	30	19	iteration	iteration	NOUN
ajst-20212	30	20	instead	instead	ADV
ajst-20212	30	21	of	of	ADP
ajst-20212	30	22	fixing	fix	VERB
ajst-20212	30	23	the	the	DET
ajst-20212	30	24	mask	mask	NOUN
ajst-20212	30	25	position	position	NOUN
ajst-20212	30	26	like	like	ADP
ajst-20212	30	27	bert	bert	NOUN
ajst-20212	30	28	,	,	PUNCT
ajst-20212	30	29	and	and	CCONJ
ajst-20212	30	30	experimentally	experimentally	ADV
ajst-20212	30	31	demonstrates	demonstrate	VERB
ajst-20212	30	32	that	that	SCONJ
ajst-20212	30	33	the	the	DET
ajst-20212	30	34	nsp	nsp	PROPN
ajst-20212	30	35	task	task	NOUN
ajst-20212	30	36	has	have	VERB
ajst-20212	30	37	no	no	DET
ajst-20212	30	38	effect	effect	NOUN
ajst-20212	30	39	as	as	ADP
ajst-20212	30	40	selection	selection	NOUN
ajst-20212	30	41	removal	removal	NOUN
ajst-20212	30	42	in	in	ADP
ajst-20212	30	43	natural	natural	ADJ
ajst-20212	30	44	language	language	NOUN
ajst-20212	30	45	processing	processing	NOUN
ajst-20212	30	46	.	.	PUNCT
ajst-20212	31	1	2.3	2.3	NUM
ajst-20212	31	2	.	.	PUNCT
ajst-20212	31	3	bigru	bigru	NOUN
ajst-20212	31	4	in	in	ADP
ajst-20212	31	5	2014	2014	NUM
ajst-20212	31	6	k	k	PROPN
ajst-20212	32	1	cho	cho	PROPN
ajst-20212	32	2	et	et	PROPN
ajst-20212	32	3	al	al	PROPN
ajst-20212	32	4	.	.	PROPN
ajst-20212	32	5	proposed	propose	VERB
ajst-20212	32	6	gated	gate	VERB
ajst-20212	32	7	recurrent	recurrent	ADJ
ajst-20212	32	8	unit	unit	NOUN
ajst-20212	32	9	(	(	PUNCT
ajst-20212	32	10	gru	gru	PROPN
ajst-20212	32	11	)	)	PUNCT
ajst-20212	32	12	based	base	VERB
ajst-20212	32	13	on	on	ADP
ajst-20212	32	14	lstm	lstm	PROPN
ajst-20212	32	15	.	.	PUNCT
ajst-20212	33	1	the	the	DET
ajst-20212	33	2	gated	gate	VERB
ajst-20212	33	3	recurrent	recurrent	ADJ
ajst-20212	33	4	unit	unit	NOUN
ajst-20212	33	5	is	be	AUX
ajst-20212	33	6	a	a	DET
ajst-20212	33	7	recurrent	recurrent	ADJ
ajst-20212	33	8	neural	neural	ADJ
ajst-20212	33	9	network	network	NOUN
ajst-20212	33	10	structure	structure	NOUN
ajst-20212	33	11	similar	similar	ADJ
ajst-20212	33	12	to	to	ADP
ajst-20212	33	13	a	a	DET
ajst-20212	33	14	long	long	ADJ
ajst-20212	33	15	short	short	ADJ
ajst-20212	33	16	-	-	PUNCT
ajst-20212	33	17	term	term	NOUN
ajst-20212	33	18	memory	memory	NOUN
ajst-20212	33	19	network	network	NOUN
ajst-20212	33	20	,	,	PUNCT
ajst-20212	33	21	which	which	PRON
ajst-20212	33	22	is	be	AUX
ajst-20212	33	23	designed	design	VERB
ajst-20212	33	24	to	to	PART
ajst-20212	33	25	solve	solve	VERB
ajst-20212	33	26	the	the	DET
ajst-20212	33	27	gradient	gradient	NOUN
ajst-20212	33	28	vanishing	vanishing	NOUN
ajst-20212	33	29	and	and	CCONJ
ajst-20212	33	30	information	information	NOUN
ajst-20212	33	31	transfer	transfer	NOUN
ajst-20212	33	32	problems	problem	NOUN
ajst-20212	33	33	in	in	ADP
ajst-20212	33	34	modeling	model	VERB
ajst-20212	33	35	long	long	ADJ
ajst-20212	33	36	sequence	sequence	NOUN
ajst-20212	33	37	data	datum	NOUN
ajst-20212	33	38	.	.	PUNCT
ajst-20212	34	1	the	the	DET
ajst-20212	34	2	gru	gru	NOUN
ajst-20212	34	3	consists	consist	VERB
ajst-20212	34	4	of	of	ADP
ajst-20212	34	5	two	two	NUM
ajst-20212	34	6	important	important	ADJ
ajst-20212	34	7	gating	gate	VERB
ajst-20212	34	8	units	unit	NOUN
ajst-20212	34	9	:	:	PUNCT
ajst-20212	34	10	the	the	DET
ajst-20212	34	11	reset	reset	NOUN
ajst-20212	34	12	gate	gate	NOUN
ajst-20212	34	13	and	and	CCONJ
ajst-20212	34	14	the	the	DET
ajst-20212	34	15	update	update	NOUN
ajst-20212	34	16	gate	gate	NOUN
ajst-20212	34	17	.	.	PUNCT
ajst-20212	35	1	the	the	DET
ajst-20212	35	2	reset	reset	NOUN
ajst-20212	35	3	gate	gate	NOUN
ajst-20212	35	4	controls	control	VERB
ajst-20212	35	5	how	how	SCONJ
ajst-20212	35	6	to	to	PART
ajst-20212	35	7	integrate	integrate	VERB
ajst-20212	35	8	past	past	ADJ
ajst-20212	35	9	information	information	NOUN
ajst-20212	35	10	and	and	CCONJ
ajst-20212	35	11	the	the	DET
ajst-20212	35	12	update	update	NOUN
ajst-20212	35	13	gate	gate	NOUN
ajst-20212	35	14	controls	control	VERB
ajst-20212	35	15	how	how	SCONJ
ajst-20212	35	16	to	to	PART
ajst-20212	35	17	integrate	integrate	VERB
ajst-20212	35	18	new	new	ADJ
ajst-20212	35	19	input	input	NOUN
ajst-20212	35	20	information	information	NOUN
ajst-20212	35	21	.	.	PUNCT
ajst-20212	36	1	gru	gru	PROPN
ajst-20212	36	2	also	also	ADV
ajst-20212	36	3	uses	use	VERB
ajst-20212	36	4	"	"	PUNCT
ajst-20212	36	5	candidate	candidate	NOUN
ajst-20212	36	6	values	value	NOUN
ajst-20212	36	7	"	"	PUNCT
ajst-20212	36	8	to	to	PART
ajst-20212	36	9	update	update	VERB
ajst-20212	36	10	the	the	DET
ajst-20212	36	11	current	current	ADJ
ajst-20212	36	12	internal	internal	ADJ
ajst-20212	36	13	state	state	NOUN
ajst-20212	36	14	while	while	SCONJ
ajst-20212	36	15	preserving	preserve	VERB
ajst-20212	36	16	the	the	DET
ajst-20212	36	17	old	old	ADJ
ajst-20212	36	18	state	state	NOUN
ajst-20212	36	19	information	information	NOUN
ajst-20212	36	20	.	.	PUNCT
ajst-20212	37	1	compared	compare	VERB
ajst-20212	37	2	with	with	ADP
ajst-20212	37	3	lstm	lstm	PROPN
ajst-20212	37	4	,	,	PUNCT
ajst-20212	37	5	gru	gru	PROPN
ajst-20212	37	6	has	have	VERB
ajst-20212	37	7	a	a	DET
ajst-20212	37	8	simpler	simple	ADJ
ajst-20212	37	9	structure	structure	NOUN
ajst-20212	37	10	and	and	CCONJ
ajst-20212	37	11	higher	high	ADJ
ajst-20212	37	12	computational	computational	ADJ
ajst-20212	37	13	efficiency	efficiency	NOUN
ajst-20212	37	14	,	,	PUNCT
ajst-20212	37	15	and	and	CCONJ
ajst-20212	37	16	therefore	therefore	ADV
ajst-20212	37	17	has	have	VERB
ajst-20212	37	18	better	well	ADJ
ajst-20212	37	19	performance	performance	NOUN
ajst-20212	37	20	in	in	ADP
ajst-20212	37	21	some	some	DET
ajst-20212	37	22	scenarios.gru	scenarios.gru	NOUN
ajst-20212	37	23	is	be	AUX
ajst-20212	37	24	widely	widely	ADV
ajst-20212	37	25	used	use	VERB
ajst-20212	37	26	in	in	ADP
ajst-20212	37	27	natural	natural	ADJ
ajst-20212	37	28	language	language	NOUN
ajst-20212	37	29	processing	processing	NOUN
ajst-20212	37	30	,	,	PUNCT
ajst-20212	37	31	speech	speech	NOUN
ajst-20212	37	32	recognition	recognition	NOUN
ajst-20212	37	33	,	,	PUNCT
ajst-20212	37	34	machine	machine	NOUN
ajst-20212	37	35	translation	translation	NOUN
ajst-20212	37	36	and	and	CCONJ
ajst-20212	37	37	other	other	ADJ
ajst-20212	37	38	tasks	task	NOUN
ajst-20212	37	39	,	,	PUNCT
ajst-20212	37	40	and	and	CCONJ
ajst-20212	37	41	has	have	AUX
ajst-20212	37	42	become	become	VERB
ajst-20212	37	43	one	one	NUM
ajst-20212	37	44	of	of	ADP
ajst-20212	37	45	the	the	DET
ajst-20212	37	46	most	most	ADV
ajst-20212	37	47	important	important	ADJ
ajst-20212	37	48	models	model	NOUN
ajst-20212	37	49	for	for	ADP
ajst-20212	37	50	processing	process	VERB
ajst-20212	37	51	sequential	sequential	ADJ
ajst-20212	37	52	data	datum	NOUN
ajst-20212	37	53	.	.	PUNCT
ajst-20212	38	1	its	its	PRON
ajst-20212	38	2	design	design	NOUN
ajst-20212	38	3	concept	concept	NOUN
ajst-20212	38	4	is	be	AUX
ajst-20212	38	5	to	to	PART
ajst-20212	38	6	optimize	optimize	VERB
ajst-20212	38	7	the	the	DET
ajst-20212	38	8	information	information	NOUN
ajst-20212	38	9	flow	flow	VERB
ajst-20212	38	10	through	through	ADP
ajst-20212	38	11	the	the	DET
ajst-20212	38	12	gating	gate	VERB
ajst-20212	38	13	mechanism	mechanism	NOUN
ajst-20212	38	14	and	and	CCONJ
ajst-20212	38	15	improve	improve	VERB
ajst-20212	38	16	the	the	DET
ajst-20212	38	17	modeling	modeling	NOUN
ajst-20212	38	18	ability	ability	NOUN
ajst-20212	38	19	for	for	ADP
ajst-20212	38	20	long	long	ADJ
ajst-20212	38	21	sequence	sequence	NOUN
ajst-20212	38	22	data	datum	NOUN
ajst-20212	38	23	.	.	PUNCT
ajst-20212	39	1	the	the	DET
ajst-20212	39	2	internal	internal	ADJ
ajst-20212	39	3	structure	structure	NOUN
ajst-20212	39	4	of	of	ADP
ajst-20212	39	5	the	the	DET
ajst-20212	39	6	gru	gru	NOUN
ajst-20212	39	7	is	be	AUX
ajst-20212	39	8	shown	show	VERB
ajst-20212	39	9	in	in	ADP
ajst-20212	39	10	figure	figure	NOUN
ajst-20212	39	11	2	2	NUM
ajst-20212	39	12	.	.	PUNCT
ajst-20212	39	13	figure	figure	NOUN
ajst-20212	39	14	2	2	NUM
ajst-20212	39	15	.	.	PUNCT
ajst-20212	39	16	internal	internal	ADJ
ajst-20212	39	17	framework	framework	NOUN
ajst-20212	39	18	of	of	ADP
ajst-20212	39	19	gru	gru	NOUN
ajst-20212	39	20	here	here	ADV
ajst-20212	39	21	,	,	PUNCT
ajst-20212	39	22	r	r	NOUN
ajst-20212	39	23	and	and	CCONJ
ajst-20212	39	24	z	z	NOUN
ajst-20212	39	25	control	control	NOUN
ajst-20212	39	26	the	the	DET
ajst-20212	39	27	reset	reset	NOUN
ajst-20212	39	28	gate	gate	NOUN
ajst-20212	39	29	and	and	CCONJ
ajst-20212	39	30	update	update	NOUN
ajst-20212	39	31	gate	gate	NOUN
ajst-20212	39	32	,	,	PUNCT
ajst-20212	39	33	respectively	respectively	ADV
ajst-20212	39	34	;	;	PUNCT
ajst-20212	39	35	their	their	PRON
ajst-20212	39	36	states	state	NOUN
ajst-20212	39	37	are	be	AUX
ajst-20212	39	38	jointly	jointly	ADV
ajst-20212	39	39	determined	determine	VERB
ajst-20212	39	40	by	by	ADP
ajst-20212	39	41	the	the	DET
ajst-20212	39	42	hidden	hidden	ADJ
ajst-20212	39	43	state	state	NOUN
ajst-20212	39	44	of	of	ADP
ajst-20212	39	45	the	the	DET
ajst-20212	39	46	previous	previous	ADJ
ajst-20212	39	47	node	node	NOUN
ajst-20212	39	48	and	and	CCONJ
ajst-20212	39	49	the	the	DET
ajst-20212	39	50	inputs	input	NOUN
ajst-20212	39	51	of	of	ADP
ajst-20212	39	52	the	the	DET
ajst-20212	39	53	current	current	ADJ
ajst-20212	39	54	node	node	NOUN
ajst-20212	39	55	.	.	PUNCT
ajst-20212	40	1	σ	σ	NOUN
ajst-20212	40	2	represents	represent	VERB
ajst-20212	40	3	the	the	DET
ajst-20212	40	4	sigmoid	sigmoid	NOUN
ajst-20212	40	5	function	function	NOUN
ajst-20212	40	6	,	,	PUNCT
ajst-20212	40	7	which	which	PRON
ajst-20212	40	8	is	be	AUX
ajst-20212	40	9	used	use	VERB
ajst-20212	40	10	to	to	PART
ajst-20212	40	11	convert	convert	VERB
ajst-20212	40	12	the	the	DET
ajst-20212	40	13	data	datum	NOUN
ajst-20212	40	14	to	to	ADP
ajst-20212	40	15	a	a	DET
ajst-20212	40	16	value	value	NOUN
ajst-20212	40	17	between	between	ADP
ajst-20212	40	18	0	0	NUM
ajst-20212	40	19	and	and	CCONJ
ajst-20212	40	20	1	1	NUM
ajst-20212	40	21	,	,	PUNCT
ajst-20212	40	22	and	and	CCONJ
ajst-20212	40	23	serves	serve	VERB
ajst-20212	40	24	as	as	ADP
ajst-20212	40	25	a	a	DET
ajst-20212	40	26	gating	gate	VERB
ajst-20212	40	27	signal	signal	NOUN
ajst-20212	40	28	.	.	PUNCT
ajst-20212	41	1	r	r	NOUN
ajst-20212	41	2	and	and	CCONJ
ajst-20212	41	3	z	z	NOUN
ajst-20212	41	4	are	be	AUX
ajst-20212	41	5	expressed	express	VERB
ajst-20212	41	6	as	as	ADP
ajst-20212	41	7	in	in	ADP
ajst-20212	41	8	eqs.1	eqs.1	NOUN
ajst-20212	41	9	and	and	CCONJ
ajst-20212	41	10	2	2	NUM
ajst-20212	41	11	:	:	SYM
ajst-20212	41	12	1	1	NUM
ajst-20212	41	13	(	(	PUNCT
ajst-20212	41	14	[	[	PUNCT
ajst-20212	41	15	,	,	PUNCT
ajst-20212	41	16	]	]	X
ajst-20212	41	17	)	)	PUNCT
ajst-20212	41	18	r	r	NOUN
ajst-20212	41	19	t	t	PROPN
ajst-20212	41	20	tr	tr	NOUN
ajst-20212	41	21	w	w	NOUN
ajst-20212	41	22	x	x	PROPN
ajst-20212	41	23	h	h	NOUN
ajst-20212	41	24			ADJ
ajst-20212	41	25	(	(	PUNCT
ajst-20212	41	26	1	1	NUM
ajst-20212	41	27	)	)	PUNCT
ajst-20212	41	28	1	1	NUM
ajst-20212	41	29	(	(	PUNCT
ajst-20212	41	30	[	[	PUNCT
ajst-20212	41	31	,	,	PUNCT
ajst-20212	41	32	]	]	X
ajst-20212	41	33	)	)	PUNCT
ajst-20212	41	34	z	z	NOUN
ajst-20212	41	35	t	t	PROPN
ajst-20212	41	36	tz	tz	PROPN
ajst-20212	41	37	w	w	PROPN
ajst-20212	41	38	x	x	PROPN
ajst-20212	41	39	h	h	NOUN
ajst-20212	41	40			ADJ
ajst-20212	41	41	(	(	PUNCT
ajst-20212	41	42	2	2	X
ajst-20212	41	43	)	)	PUNCT
ajst-20212	41	44	the	the	DET
ajst-20212	41	45	reset	reset	NOUN
ajst-20212	41	46	gate	gate	NOUN
ajst-20212	41	47	determines	determine	VERB
ajst-20212	41	48	how	how	SCONJ
ajst-20212	41	49	much	much	ADJ
ajst-20212	41	50	of	of	ADP
ajst-20212	41	51	the	the	DET
ajst-20212	41	52	historical	historical	ADJ
ajst-20212	41	53	information	information	NOUN
ajst-20212	41	54	will	will	AUX
ajst-20212	41	55	be	be	AUX
ajst-20212	41	56	forgotten	forget	VERB
ajst-20212	41	57	by	by	ADP
ajst-20212	41	58	the	the	DET
ajst-20212	41	59	time	time	NOUN
ajst-20212	41	60	it	it	PRON
ajst-20212	41	61	passes	pass	VERB
ajst-20212	41	62	to	to	ADP
ajst-20212	41	63	the	the	DET
ajst-20212	41	64	next	next	ADJ
ajst-20212	41	65	time	time	NOUN
ajst-20212	41	66	step	step	NOUN
ajst-20212	41	67	and	and	CCONJ
ajst-20212	41	68	thus	thus	ADV
ajst-20212	41	69	is	be	AUX
ajst-20212	41	70	used	use	VERB
ajst-20212	41	71	to	to	PART
ajst-20212	41	72	reset	reset	VERB
ajst-20212	41	73	the	the	DET
ajst-20212	41	74	stored	store	VERB
ajst-20212	41	75	information.the	information.the	DET
ajst-20212	41	76	gru	gru	NOUN
ajst-20212	41	77	directly	directly	ADV
ajst-20212	41	78	utilizes	utilize	VERB
ajst-20212	41	79	the	the	DET
ajst-20212	41	80	hidden	hidden	ADJ
ajst-20212	41	81	unit	unit	NOUN
ajst-20212	41	82	to	to	PART
ajst-20212	41	83	record	record	VERB
ajst-20212	41	84	the	the	DET
ajst-20212	41	85	historical	historical	ADJ
ajst-20212	41	86	state	state	NOUN
ajst-20212	41	87	and	and	CCONJ
ajst-20212	41	88	adjusts	adjust	VERB
ajst-20212	41	89	the	the	DET
ajst-20212	41	90	trade	trade	NOUN
ajst-20212	41	91	-	-	PUNCT
ajst-20212	41	92	off	off	ADP
ajst-20212	41	93	ratio	ratio	NOUN
ajst-20212	41	94	between	between	ADP
ajst-20212	41	95	the	the	DET
ajst-20212	41	96	current	current	ADJ
ajst-20212	41	97	information	information	NOUN
ajst-20212	41	98	and	and	CCONJ
ajst-20212	41	99	the	the	DET
ajst-20212	41	100	memorized	memorized	ADJ
ajst-20212	41	101	information	information	NOUN
ajst-20212	41	102	by	by	ADP
ajst-20212	41	103	using	use	VERB
ajst-20212	41	104	the	the	DET
ajst-20212	41	105	reset	reset	NOUN
ajst-20212	41	106	gate	gate	NOUN
ajst-20212	41	107	in	in	ADP
ajst-20212	41	108	order	order	NOUN
ajst-20212	41	109	to	to	PART
ajst-20212	41	110	generate	generate	VERB
ajst-20212	41	111	the	the	DET
ajst-20212	41	112	new	new	ADJ
ajst-20212	41	113	stored	store	VERB
ajst-20212	41	114	information	information	NOUN
ajst-20212	41	115	and	and	CCONJ
ajst-20212	41	116	continue	continue	VERB
ajst-20212	41	117	to	to	PART
ajst-20212	41	118	pass	pass	VERB
ajst-20212	41	119	forward	forward	ADV
ajst-20212	41	120	.	.	PUNCT
ajst-20212	42	1	the	the	DET
ajst-20212	42	2	output	output	NOUN
ajst-20212	42	3	of	of	ADP
ajst-20212	42	4	the	the	DET
ajst-20212	42	5	reset	reset	ADJ
ajst-20212	42	6	gate	gate	NOUN
ajst-20212	42	7	ranges	range	VERB
ajst-20212	42	8	between	between	ADP
ajst-20212	42	9	[	[	X
ajst-20212	42	10	0	0	NUM
ajst-20212	42	11	,	,	PUNCT
ajst-20212	42	12	1	1	NUM
ajst-20212	42	13	]	]	PUNCT
ajst-20212	42	14	,	,	PUNCT
ajst-20212	42	15	when	when	SCONJ
ajst-20212	42	16	the	the	DET
ajst-20212	42	17	output	output	NOUN
ajst-20212	42	18	is	be	AUX
ajst-20212	42	19	0	0	NUM
ajst-20212	42	20	it	it	PRON
ajst-20212	42	21	means	mean	VERB
ajst-20212	42	22	that	that	SCONJ
ajst-20212	42	23	all	all	DET
ajst-20212	42	24	the	the	DET
ajst-20212	42	25	memorized	memorized	ADJ
ajst-20212	42	26	information	information	NOUN
ajst-20212	42	27	is	be	AUX
ajst-20212	42	28	reset	reset	ADJ
ajst-20212	42	29	,	,	PUNCT
ajst-20212	42	30	and	and	CCONJ
ajst-20212	42	31	when	when	SCONJ
ajst-20212	42	32	the	the	DET
ajst-20212	42	33	output	output	NOUN
ajst-20212	42	34	is	be	AUX
ajst-20212	42	35	1	1	NUM
ajst-20212	42	36	it	it	PRON
ajst-20212	42	37	means	mean	VERB
ajst-20212	42	38	that	that	SCONJ
ajst-20212	42	39	all	all	DET
ajst-20212	42	40	the	the	DET
ajst-20212	42	41	memorized	memorized	ADJ
ajst-20212	42	42	information	information	NOUN
ajst-20212	42	43	is	be	AUX
ajst-20212	42	44	passed	pass	VERB
ajst-20212	42	45	all	all	DET
ajst-20212	42	46	the	the	DET
ajst-20212	42	47	way	way	NOUN
ajst-20212	42	48	through	through	ADV
ajst-20212	42	49	.	.	PUNCT
ajst-20212	43	1	after	after	ADP
ajst-20212	43	2	obtaining	obtain	VERB
ajst-20212	43	3	the	the	DET
ajst-20212	43	4	gating	gate	VERB
ajst-20212	43	5	information	information	NOUN
ajst-20212	43	6	,	,	PUNCT
ajst-20212	43	7	r	r	NOUN
ajst-20212	43	8	is	be	AUX
ajst-20212	43	9	first	first	ADV
ajst-20212	43	10	used	use	VERB
ajst-20212	43	11	to	to	PART
ajst-20212	43	12	control	control	VERB
ajst-20212	43	13	the	the	DET
ajst-20212	43	14	generated	generate	VERB
ajst-20212	43	15	"	"	PUNCT
ajst-20212	43	16	reset	reset	VERB
ajst-20212	43	17	"	"	PUNCT
ajst-20212	43	18	data	datum	NOUN
ajst-20212	43	19	,	,	PUNCT
ajst-20212	43	20	which	which	PRON
ajst-20212	43	21	is	be	AUX
ajst-20212	43	22	processed	process	VERB
ajst-20212	43	23	in	in	ADP
ajst-20212	43	24	the	the	DET
ajst-20212	43	25	following	follow	VERB
ajst-20212	43	26	equations	equation	NOUN
ajst-20212	43	27	3	3	NUM
ajst-20212	43	28	and	and	CCONJ
ajst-20212	43	29	4	4	NUM
ajst-20212	43	30	:	:	PUNCT
ajst-20212	43	31	'	'	PUNCT
ajst-20212	43	32	1	1	NUM
ajst-20212	43	33	1	1	NUM
ajst-20212	43	34	t	t	NOUN
ajst-20212	43	35	th	th	NOUN
ajst-20212	43	36	h	h	NOUN
ajst-20212	43	37	r	r	NOUN
ajst-20212	43	38			ADJ
ajst-20212	43	39			PROPN
ajst-20212	43	40	(	(	PUNCT
ajst-20212	43	41	3	3	NUM
ajst-20212	43	42	)	)	PUNCT
ajst-20212	43	43	'	'	PUNCT
ajst-20212	44	1	1tanh	1tanh	NUM
ajst-20212	44	2	(	(	PUNCT
ajst-20212	44	3	[	[	PUNCT
ajst-20212	44	4	,	,	PUNCT
ajst-20212	44	5	]	]	X
ajst-20212	44	6	)	)	PUNCT
ajst-20212	44	7	t	t	PROPN
ajst-20212	44	8	th	th	X
ajst-20212	44	9	w	w	NOUN
ajst-20212	44	10	x	x	PUNCT
ajst-20212	44	11	h	h	NOUN
ajst-20212	44	12			ADJ
ajst-20212	44	13	(	(	PUNCT
ajst-20212	44	14	4	4	X
ajst-20212	44	15	)	)	PUNCT
ajst-20212	44	16	the	the	DET
ajst-20212	44	17	update	update	NOUN
ajst-20212	44	18	gate	gate	NOUN
ajst-20212	44	19	determines	determine	VERB
ajst-20212	44	20	how	how	SCONJ
ajst-20212	44	21	much	much	ADJ
ajst-20212	44	22	past	past	ADJ
ajst-20212	44	23	information	information	NOUN
ajst-20212	44	24	will	will	AUX
ajst-20212	44	25	be	be	AUX
ajst-20212	44	26	passed	pass	VERB
ajst-20212	44	27	on	on	ADP
ajst-20212	44	28	to	to	ADP
ajst-20212	44	29	the	the	DET
ajst-20212	44	30	future	future	NOUN
ajst-20212	44	31	,	,	PUNCT
ajst-20212	44	32	similar	similar	ADJ
ajst-20212	44	33	to	to	ADP
ajst-20212	44	34	the	the	DET
ajst-20212	44	35	input	input	NOUN
ajst-20212	44	36	and	and	CCONJ
ajst-20212	44	37	forget	forget	VERB
ajst-20212	44	38	gates	gate	NOUN
ajst-20212	44	39	in	in	ADP
ajst-20212	44	40	lstm	lstm	NOUN
ajst-20212	44	41	,	,	PUNCT
ajst-20212	44	42	and	and	CCONJ
ajst-20212	44	43	it	it	PRON
ajst-20212	44	44	is	be	AUX
ajst-20212	44	45	used	use	VERB
ajst-20212	44	46	to	to	PART
ajst-20212	44	47	compute	compute	VERB
ajst-20212	44	48	the	the	DET
ajst-20212	44	49	output	output	NOUN
ajst-20212	44	50	of	of	ADP
ajst-20212	44	51	the	the	DET
ajst-20212	44	52	hidden	hidden	ADJ
ajst-20212	44	53	state	state	NOUN
ajst-20212	44	54	at	at	ADP
ajst-20212	44	55	the	the	DET
ajst-20212	44	56	current	current	ADJ
ajst-20212	44	57	moment	moment	NOUN
ajst-20212	44	58	.	.	PUNCT
ajst-20212	45	1	the	the	DET
ajst-20212	45	2	data	datum	NOUN
ajst-20212	45	3	processing	processing	NOUN
ajst-20212	45	4	of	of	ADP
ajst-20212	45	5	the	the	DET
ajst-20212	45	6	update	update	NOUN
ajst-20212	45	7	gate	gate	NOUN
ajst-20212	45	8	is	be	AUX
ajst-20212	45	9	similar	similar	ADJ
ajst-20212	45	10	to	to	ADP
ajst-20212	45	11	that	that	PRON
ajst-20212	45	12	of	of	ADP
ajst-20212	45	13	the	the	DET
ajst-20212	45	14	reset	reset	NOUN
ajst-20212	45	15	gate	gate	NOUN
ajst-20212	45	16	,	,	PUNCT
ajst-20212	45	17	but	but	CCONJ
ajst-20212	45	18	the	the	DET
ajst-20212	45	19	values	value	NOUN
ajst-20212	45	20	and	and	CCONJ
ajst-20212	45	21	roles	role	NOUN
ajst-20212	45	22	of	of	ADP
ajst-20212	45	23	the	the	DET
ajst-20212	45	24	weight	weight	NOUN
ajst-20212	45	25	matrix	matrix	NOUN
ajst-20212	45	26	are	be	AUX
ajst-20212	45	27	different	different	ADJ
ajst-20212	45	28	.	.	PUNCT
ajst-20212	46	1	by	by	ADP
ajst-20212	46	2	utilizing	utilize	VERB
ajst-20212	46	3	z	z	NOUN
ajst-20212	46	4	for	for	ADP
ajst-20212	46	5	updating	updating	NOUN
ajst-20212	46	6	,	,	PUNCT
ajst-20212	46	7	the	the	DET
ajst-20212	46	8	update	update	NOUN
ajst-20212	46	9	expression	expression	NOUN
ajst-20212	46	10	is	be	AUX
ajst-20212	46	11	shown	show	VERB
ajst-20212	46	12	in	in	ADP
ajst-20212	46	13	eq.5	eq.5	PROPN
ajst-20212	46	14	:	:	PUNCT
ajst-20212	46	15	'	'	PUNCT
ajst-20212	46	16	1(1	1(1	NUM
ajst-20212	46	17	)	)	PUNCT
ajst-20212	46	18	t	t	NOUN
ajst-20212	46	19	th	th	X
ajst-20212	46	20	z	z	NOUN
ajst-20212	46	21	h	h	NOUN
ajst-20212	47	1	z	z	PROPN
ajst-20212	47	2	h	h	PROPN
ajst-20212	47	3			PROPN
ajst-20212	47	4			PRON
ajst-20212	47	5			PROPN
ajst-20212	47	6	(	(	PUNCT
ajst-20212	47	7	5	5	NUM
ajst-20212	47	8	)	)	PUNCT
ajst-20212	47	9	unidirectional	unidirectional	ADJ
ajst-20212	47	10	gru	gru	PROPN
ajst-20212	47	11	can	can	AUX
ajst-20212	47	12	only	only	ADV
ajst-20212	47	13	capture	capture	VERB
ajst-20212	47	14	forward	forward	ADV
ajst-20212	47	15	sequence	sequence	NOUN
ajst-20212	47	16	features	feature	NOUN
ajst-20212	47	17	and	and	CCONJ
ajst-20212	47	18	can	can	AUX
ajst-20212	47	19	not	not	PART
ajst-20212	47	20	acquire	acquire	VERB
ajst-20212	47	21	backward	backward	ADJ
ajst-20212	47	22	information	information	NOUN
ajst-20212	47	23	of	of	ADP
ajst-20212	47	24	text	text	NOUN
ajst-20212	47	25	.	.	PUNCT
ajst-20212	48	1	therefore	therefore	ADV
ajst-20212	48	2	,	,	PUNCT
ajst-20212	48	3	some	some	DET
ajst-20212	48	4	scholars	scholar	NOUN
ajst-20212	48	5	have	have	AUX
ajst-20212	48	6	proposed	propose	VERB
ajst-20212	48	7	bigru	bigru	NOUN
ajst-20212	48	8	to	to	PART
ajst-20212	48	9	capture	capture	VERB
ajst-20212	48	10	long	long	ADJ
ajst-20212	48	11	text	text	NOUN
ajst-20212	48	12	sequence	sequence	NOUN
ajst-20212	48	13	features.bigru	features.bigru	NOUN
ajst-20212	48	14	is	be	AUX
ajst-20212	48	15	a	a	DET
ajst-20212	48	16	neural	neural	ADJ
ajst-20212	48	17	network	network	NOUN
ajst-20212	48	18	structure	structure	NOUN
ajst-20212	48	19	that	that	PRON
ajst-20212	48	20	synthesizes	synthesize	VERB
ajst-20212	48	21	both	both	CCONJ
ajst-20212	48	22	forward	forward	ADJ
ajst-20212	48	23	and	and	CCONJ
ajst-20212	48	24	backward	backward	ADJ
ajst-20212	48	25	information	information	NOUN
ajst-20212	48	26	,	,	PUNCT
ajst-20212	48	27	combining	combine	VERB
ajst-20212	48	28	the	the	DET
ajst-20212	48	29	advantages	advantage	NOUN
ajst-20212	48	30	of	of	ADP
ajst-20212	48	31	bi	bi	NOUN
ajst-20212	48	32	-	-	NOUN
ajst-20212	48	33	directionality	directionality	NOUN
ajst-20212	48	34	and	and	CCONJ
ajst-20212	48	35	gating	gate	VERB
ajst-20212	48	36	mechanism	mechanism	NOUN
ajst-20212	48	37	,	,	PUNCT
ajst-20212	48	38	which	which	PRON
ajst-20212	48	39	helps	help	VERB
ajst-20212	48	40	to	to	PART
ajst-20212	48	41	process	process	VERB
ajst-20212	48	42	sequence	sequence	NOUN
ajst-20212	48	43	data	datum	NOUN
ajst-20212	48	44	efficiently	efficiently	ADV
ajst-20212	48	45	and	and	CCONJ
ajst-20212	48	46	capture	capture	VERB
ajst-20212	48	47	the	the	DET
ajst-20212	48	48	long	long	ADJ
ajst-20212	48	49	term	term	NOUN
ajst-20212	48	50	dependencies	dependency	NOUN
ajst-20212	48	51	in	in	ADP
ajst-20212	48	52	it	it	PRON
ajst-20212	48	53	.	.	PUNCT
ajst-20212	49	1	the	the	DET
ajst-20212	49	2	structure	structure	NOUN
ajst-20212	49	3	of	of	ADP
ajst-20212	49	4	bigru	bigru	NOUN
ajst-20212	49	5	is	be	AUX
ajst-20212	49	6	shown	show	VERB
ajst-20212	49	7	in	in	ADP
ajst-20212	49	8	figure	figure	NOUN
ajst-20212	49	9	3	3	NUM
ajst-20212	49	10	.	.	PUNCT
ajst-20212	49	11	figure	figure	VERB
ajst-20212	49	12	3	3	NUM
ajst-20212	49	13	.	.	PUNCT
ajst-20212	49	14	framework	framework	NOUN
ajst-20212	49	15	of	of	ADP
ajst-20212	49	16	bigru	bigru	NOUN
ajst-20212	49	17	2.4	2.4	NUM
ajst-20212	49	18	.	.	PUNCT
ajst-20212	50	1	textcnn	textcnn	VERB
ajst-20212	50	2	the	the	DET
ajst-20212	50	3	input	input	NOUN
ajst-20212	50	4	to	to	ADP
ajst-20212	50	5	textcnn	textcnn	PROPN
ajst-20212	50	6	is	be	AUX
ajst-20212	50	7	usually	usually	ADV
ajst-20212	50	8	converting	convert	VERB
ajst-20212	50	9	the	the	DET
ajst-20212	50	10	text	text	NOUN
ajst-20212	50	11	into	into	ADP
ajst-20212	50	12	a	a	DET
ajst-20212	50	13	sequence	sequence	NOUN
ajst-20212	50	14	of	of	ADP
ajst-20212	50	15	word	word	NOUN
ajst-20212	50	16	embedding	embed	VERB
ajst-20212	50	17	vectors	vector	NOUN
ajst-20212	50	18	and	and	CCONJ
ajst-20212	50	19	then	then	ADV
ajst-20212	50	20	feeding	feed	VERB
ajst-20212	50	21	these	these	DET
ajst-20212	50	22	vectors	vector	NOUN
ajst-20212	50	23	into	into	ADP
ajst-20212	50	24	a	a	DET
ajst-20212	50	25	convolutional	convolutional	ADJ
ajst-20212	50	26	layer	layer	NOUN
ajst-20212	50	27	for	for	ADP
ajst-20212	50	28	feature	feature	NOUN
ajst-20212	50	29	extraction	extraction	NOUN
ajst-20212	50	30	.	.	PUNCT
ajst-20212	51	1	convolutional	convolutional	ADJ
ajst-20212	51	2	operations	operation	NOUN
ajst-20212	51	3	help	help	VERB
ajst-20212	51	4	in	in	ADP
ajst-20212	51	5	capturing	capture	VERB
ajst-20212	51	6	combinations	combination	NOUN
ajst-20212	51	7	of	of	ADP
ajst-20212	51	8	words	word	NOUN
ajst-20212	51	9	in	in	ADP
ajst-20212	51	10	different	different	ADJ
ajst-20212	51	11	length	length	NOUN
ajst-20212	51	12	ranges	range	NOUN
ajst-20212	51	13	,	,	PUNCT
ajst-20212	51	14	thus	thus	ADV
ajst-20212	51	15	learning	learn	VERB
ajst-20212	51	16	local	local	ADJ
ajst-20212	51	17	features	feature	NOUN
ajst-20212	51	18	169	169	NUM
ajst-20212	51	19	in	in	ADP
ajst-20212	51	20	the	the	DET
ajst-20212	51	21	text	text	NOUN
ajst-20212	51	22	.	.	PUNCT
ajst-20212	52	1	next	next	ADJ
ajst-20212	52	2	the	the	DET
ajst-20212	52	3	features	feature	NOUN
ajst-20212	52	4	extracted	extract	VERB
ajst-20212	52	5	from	from	ADP
ajst-20212	52	6	each	each	DET
ajst-20212	52	7	feature	feature	NOUN
ajst-20212	52	8	map	map	NOUN
ajst-20212	52	9	are	be	AUX
ajst-20212	52	10	aggregated	aggregate	VERB
ajst-20212	52	11	by	by	ADP
ajst-20212	52	12	pooling	pool	VERB
ajst-20212	52	13	layer	layer	NOUN
ajst-20212	52	14	and	and	CCONJ
ajst-20212	52	15	finally	finally	ADV
ajst-20212	52	16	the	the	DET
ajst-20212	52	17	obtained	obtain	VERB
ajst-20212	52	18	features	feature	NOUN
ajst-20212	52	19	are	be	AUX
ajst-20212	52	20	fed	feed	VERB
ajst-20212	52	21	to	to	ADP
ajst-20212	52	22	the	the	DET
ajst-20212	52	23	output	output	NOUN
ajst-20212	52	24	layer	layer	NOUN
ajst-20212	52	25	for	for	ADP
ajst-20212	52	26	classification	classification	NOUN
ajst-20212	52	27	prediction	prediction	NOUN
ajst-20212	52	28	.	.	PUNCT
ajst-20212	53	1	textcnn	textcnn	PROPN
ajst-20212	53	2	is	be	AUX
ajst-20212	53	3	simple	simple	ADJ
ajst-20212	53	4	and	and	CCONJ
ajst-20212	53	5	efficient	efficient	ADJ
ajst-20212	53	6	and	and	CCONJ
ajst-20212	53	7	is	be	AUX
ajst-20212	53	8	suitable	suitable	ADJ
ajst-20212	53	9	for	for	ADP
ajst-20212	53	10	various	various	ADJ
ajst-20212	53	11	text	text	NOUN
ajst-20212	53	12	classification	classification	NOUN
ajst-20212	53	13	tasks	task	NOUN
ajst-20212	53	14	such	such	ADJ
ajst-20212	53	15	as	as	ADP
ajst-20212	53	16	sentiment	sentiment	NOUN
ajst-20212	53	17	analysis	analysis	NOUN
ajst-20212	53	18	,	,	PUNCT
ajst-20212	53	19	text	text	NOUN
ajst-20212	53	20	categorization	categorization	NOUN
ajst-20212	53	21	,	,	PUNCT
ajst-20212	53	22	spam	spam	NOUN
ajst-20212	53	23	filtering	filtering	NOUN
ajst-20212	53	24	and	and	CCONJ
ajst-20212	53	25	so	so	ADV
ajst-20212	53	26	on	on	ADV
ajst-20212	53	27	.	.	PUNCT
ajst-20212	54	1	it	it	PRON
ajst-20212	54	2	performs	perform	VERB
ajst-20212	54	3	well	well	ADV
ajst-20212	54	4	in	in	ADP
ajst-20212	54	5	processing	process	VERB
ajst-20212	54	6	short	short	ADJ
ajst-20212	54	7	or	or	CCONJ
ajst-20212	54	8	fixed	fix	VERB
ajst-20212	54	9	-	-	PUNCT
ajst-20212	54	10	length	length	NOUN
ajst-20212	54	11	texts	text	NOUN
ajst-20212	54	12	and	and	CCONJ
ajst-20212	54	13	can	can	AUX
ajst-20212	54	14	be	be	AUX
ajst-20212	54	15	trained	train	VERB
ajst-20212	54	16	and	and	CCONJ
ajst-20212	54	17	deployed	deploy	VERB
ajst-20212	54	18	quickly	quickly	ADV
ajst-20212	54	19	,	,	PUNCT
ajst-20212	54	20	so	so	CCONJ
ajst-20212	54	21	it	it	PRON
ajst-20212	54	22	is	be	AUX
ajst-20212	54	23	widely	widely	ADV
ajst-20212	54	24	used	use	VERB
ajst-20212	54	25	in	in	ADP
ajst-20212	54	26	the	the	DET
ajst-20212	54	27	field	field	NOUN
ajst-20212	54	28	of	of	ADP
ajst-20212	54	29	text	text	NOUN
ajst-20212	54	30	processing.the	processing.the	NOUN
ajst-20212	54	31	structure	structure	NOUN
ajst-20212	54	32	of	of	ADP
ajst-20212	54	33	the	the	DET
ajst-20212	54	34	textcnn	textcnn	PROPN
ajst-20212	54	35	model	model	NOUN
ajst-20212	54	36	is	be	AUX
ajst-20212	54	37	shown	show	VERB
ajst-20212	54	38	in	in	ADP
ajst-20212	54	39	figure	figure	NOUN
ajst-20212	54	40	4	4	NUM
ajst-20212	54	41	below	below	ADV
ajst-20212	54	42	.	.	PUNCT
ajst-20212	55	1	figure	figure	VERB
ajst-20212	55	2	4	4	NUM
ajst-20212	55	3	.	.	PUNCT
ajst-20212	56	1	framework	framework	NOUN
ajst-20212	56	2	of	of	ADP
ajst-20212	56	3	textcnn	textcnn	PROPN
ajst-20212	56	4	2.5	2.5	NUM
ajst-20212	56	5	.	.	PUNCT
ajst-20212	57	1	gcn	gcn	NOUN
ajst-20212	57	2	graph	graph	VERB
ajst-20212	57	3	convolutional	convolutional	ADJ
ajst-20212	57	4	neural	neural	ADJ
ajst-20212	57	5	network	network	NOUN
ajst-20212	57	6	(	(	PUNCT
ajst-20212	57	7	gcn	gcn	NOUN
ajst-20212	57	8	)	)	PUNCT
ajst-20212	57	9	is	be	AUX
ajst-20212	57	10	a	a	DET
ajst-20212	57	11	deep	deep	ADJ
ajst-20212	57	12	learning	learning	NOUN
ajst-20212	57	13	model	model	NOUN
ajst-20212	57	14	specialized	specialize	VERB
ajst-20212	57	15	in	in	ADP
ajst-20212	57	16	processing	processing	NOUN
ajst-20212	57	17	graph	graph	NOUN
ajst-20212	57	18	data	datum	NOUN
ajst-20212	57	19	.	.	PUNCT
ajst-20212	58	1	unlike	unlike	ADP
ajst-20212	58	2	traditional	traditional	ADJ
ajst-20212	58	3	convolutional	convolutional	ADJ
ajst-20212	58	4	neural	neural	ADJ
ajst-20212	58	5	networks	network	NOUN
ajst-20212	58	6	(	(	PUNCT
ajst-20212	58	7	cnn	cnn	PROPN
ajst-20212	58	8	)	)	PUNCT
ajst-20212	58	9	and	and	CCONJ
ajst-20212	58	10	recurrent	recurrent	ADJ
ajst-20212	58	11	neural	neural	ADJ
ajst-20212	58	12	networks	network	NOUN
ajst-20212	58	13	(	(	PUNCT
ajst-20212	58	14	rnn	rnn	PROPN
ajst-20212	58	15	)	)	PUNCT
ajst-20212	58	16	that	that	PRON
ajst-20212	58	17	focus	focus	VERB
ajst-20212	58	18	on	on	ADP
ajst-20212	58	19	processing	process	VERB
ajst-20212	58	20	gridded	gridde	VERB
ajst-20212	58	21	or	or	CCONJ
ajst-20212	58	22	sequential	sequential	ADJ
ajst-20212	58	23	data	datum	NOUN
ajst-20212	58	24	,	,	PUNCT
ajst-20212	58	25	gcn	gcn	NOUN
ajst-20212	58	26	focuses	focus	VERB
ajst-20212	58	27	on	on	ADP
ajst-20212	58	28	graph	graph	NOUN
ajst-20212	58	29	structural	structural	ADJ
ajst-20212	58	30	information	information	NOUN
ajst-20212	58	31	in	in	ADP
ajst-20212	58	32	unstructured	unstructured	ADJ
ajst-20212	58	33	data	datum	NOUN
ajst-20212	58	34	.	.	PUNCT
ajst-20212	59	1	3	3	X
ajst-20212	59	2	.	.	X
ajst-20212	59	3	rbt	rbt	PROPN
ajst-20212	59	4	-	-	PUNCT
ajst-20212	59	5	biatt	biatt	NOUN
ajst-20212	59	6	-	-	PUNCT
ajst-20212	59	7	gcn	gcn	NOUN
ajst-20212	59	8	(	(	PUNCT
ajst-20212	59	9	rbg	rbg	NOUN
ajst-20212	59	10	)	)	PUNCT
ajst-20212	59	11	attributelevel	attributelevel	NOUN
ajst-20212	59	12	sentiment	sentiment	NOUN
ajst-20212	59	13	analysis	analysis	NOUN
ajst-20212	59	14	with	with	ADP
ajst-20212	59	15	the	the	DET
ajst-20212	59	16	rapid	rapid	ADJ
ajst-20212	59	17	development	development	NOUN
ajst-20212	59	18	of	of	ADP
ajst-20212	59	19	social	social	ADJ
ajst-20212	59	20	media	medium	NOUN
ajst-20212	59	21	and	and	CCONJ
ajst-20212	59	22	ecommerce	ecommerce	NOUN
ajst-20212	59	23	platforms	platform	NOUN
ajst-20212	59	24	,	,	PUNCT
ajst-20212	59	25	people	people	NOUN
ajst-20212	59	26	are	be	AUX
ajst-20212	59	27	more	more	ADV
ajst-20212	59	28	inclined	inclined	ADJ
ajst-20212	59	29	to	to	PART
ajst-20212	59	30	evaluate	evaluate	VERB
ajst-20212	59	31	things	thing	NOUN
ajst-20212	59	32	from	from	ADP
ajst-20212	59	33	multiple	multiple	ADJ
ajst-20212	59	34	dimensions	dimension	NOUN
ajst-20212	59	35	and	and	CCONJ
ajst-20212	59	36	attributes	attribute	NOUN
ajst-20212	59	37	.	.	PUNCT
ajst-20212	60	1	therefore	therefore	ADV
ajst-20212	60	2	,	,	PUNCT
ajst-20212	60	3	in	in	ADP
ajst-20212	60	4	today	today	NOUN
ajst-20212	60	5	's	's	PART
ajst-20212	60	6	review	review	NOUN
ajst-20212	60	7	text	text	NOUN
ajst-20212	60	8	,	,	PUNCT
ajst-20212	60	9	it	it	PRON
ajst-20212	60	10	is	be	AUX
ajst-20212	60	11	no	no	ADV
ajst-20212	60	12	longer	long	ADV
ajst-20212	60	13	accurate	accurate	ADJ
ajst-20212	60	14	enough	enough	ADV
ajst-20212	60	15	to	to	PART
ajst-20212	60	16	only	only	ADV
ajst-20212	60	17	provide	provide	VERB
ajst-20212	60	18	the	the	DET
ajst-20212	60	19	sentiment	sentiment	NOUN
ajst-20212	60	20	polarity	polarity	NOUN
ajst-20212	60	21	of	of	ADP
ajst-20212	60	22	a	a	DET
ajst-20212	60	23	paragraph	paragraph	NOUN
ajst-20212	60	24	or	or	CCONJ
ajst-20212	60	25	sentence	sentence	NOUN
ajst-20212	60	26	,	,	PUNCT
ajst-20212	60	27	and	and	CCONJ
ajst-20212	60	28	it	it	PRON
ajst-20212	60	29	is	be	AUX
ajst-20212	60	30	necessary	necessary	ADJ
ajst-20212	60	31	to	to	PART
ajst-20212	60	32	discriminate	discriminate	VERB
ajst-20212	60	33	the	the	DET
ajst-20212	60	34	sentiment	sentiment	NOUN
ajst-20212	60	35	polarity	polarity	NOUN
ajst-20212	60	36	based	base	VERB
ajst-20212	60	37	on	on	ADP
ajst-20212	60	38	specific	specific	ADJ
ajst-20212	60	39	attributes	attribute	NOUN
ajst-20212	60	40	.	.	PUNCT
ajst-20212	61	1	for	for	ADP
ajst-20212	61	2	example	example	NOUN
ajst-20212	61	3	,	,	PUNCT
ajst-20212	61	4	in	in	ADP
ajst-20212	61	5	the	the	DET
ajst-20212	61	6	sentence	sentence	NOUN
ajst-20212	61	7	"	"	PUNCT
ajst-20212	61	8	the	the	DET
ajst-20212	61	9	fruit	fruit	NOUN
ajst-20212	61	10	tasted	taste	VERB
ajst-20212	61	11	very	very	ADV
ajst-20212	61	12	good	good	ADJ
ajst-20212	61	13	,	,	PUNCT
ajst-20212	61	14	but	but	CCONJ
ajst-20212	61	15	the	the	DET
ajst-20212	61	16	express	express	ADJ
ajst-20212	61	17	delivery	delivery	NOUN
ajst-20212	61	18	was	be	AUX
ajst-20212	61	19	too	too	ADV
ajst-20212	61	20	slow	slow	ADJ
ajst-20212	61	21	"	"	PUNCT
ajst-20212	61	22	,	,	PUNCT
ajst-20212	61	23	the	the	DET
ajst-20212	61	24	attribute	attribute	NOUN
ajst-20212	61	25	words	word	VERB
ajst-20212	61	26	"	"	PUNCT
ajst-20212	61	27	fruit	fruit	NOUN
ajst-20212	61	28	"	"	PUNCT
ajst-20212	61	29	and	and	CCONJ
ajst-20212	61	30	"	"	PUNCT
ajst-20212	61	31	express	express	ADJ
ajst-20212	61	32	delivery	delivery	NOUN
ajst-20212	61	33	"	"	PUNCT
ajst-20212	61	34	are	be	AUX
ajst-20212	61	35	made	make	VERB
ajst-20212	61	36	to	to	PART
ajst-20212	61	37	be	be	AUX
ajst-20212	61	38	"	"	PUNCT
ajst-20212	61	39	positive	positive	ADJ
ajst-20212	61	40	"	"	PUNCT
ajst-20212	61	41	and	and	CCONJ
ajst-20212	61	42	"	"	PUNCT
ajst-20212	61	43	negative	negative	ADJ
ajst-20212	61	44	"	"	PUNCT
ajst-20212	61	45	sentiment	sentiment	NOUN
ajst-20212	61	46	polarity	polarity	NOUN
ajst-20212	61	47	.	.	PUNCT
ajst-20212	62	1	in	in	ADP
ajst-20212	62	2	order	order	NOUN
ajst-20212	62	3	to	to	PART
ajst-20212	62	4	perform	perform	VERB
ajst-20212	62	5	attribute	attribute	NOUN
ajst-20212	62	6	-	-	PUNCT
ajst-20212	62	7	level	level	NOUN
ajst-20212	62	8	sentiment	sentiment	NOUN
ajst-20212	62	9	analysis	analysis	NOUN
ajst-20212	62	10	on	on	ADP
ajst-20212	62	11	reviews	review	NOUN
ajst-20212	62	12	,	,	PUNCT
ajst-20212	62	13	this	this	DET
ajst-20212	62	14	section	section	NOUN
ajst-20212	62	15	proposes	propose	VERB
ajst-20212	62	16	a	a	DET
ajst-20212	62	17	model	model	NOUN
ajst-20212	62	18	rbt	rbt	PROPN
ajst-20212	62	19	-	-	PUNCT
ajst-20212	62	20	biatt	biatt	NOUN
ajst-20212	62	21	-	-	PUNCT
ajst-20212	62	22	gcn	gcn	NOUN
ajst-20212	62	23	that	that	PRON
ajst-20212	62	24	incorporates	incorporate	VERB
ajst-20212	62	25	auxiliary	auxiliary	ADJ
ajst-20212	62	26	information	information	NOUN
ajst-20212	62	27	,	,	PUNCT
ajst-20212	62	28	and	and	CCONJ
ajst-20212	62	29	sets	set	VERB
ajst-20212	62	30	up	up	ADP
ajst-20212	62	31	a	a	DET
ajst-20212	62	32	large	large	ADJ
ajst-20212	62	33	number	number	NOUN
ajst-20212	62	34	of	of	ADP
ajst-20212	62	35	baseline	baseline	ADJ
ajst-20212	62	36	comparison	comparison	NOUN
ajst-20212	62	37	experiments	experiment	NOUN
ajst-20212	62	38	as	as	ADV
ajst-20212	62	39	well	well	ADV
ajst-20212	62	40	as	as	ADP
ajst-20212	62	41	ablation	ablation	NOUN
ajst-20212	62	42	experiments	experiment	NOUN
ajst-20212	62	43	to	to	PART
ajst-20212	62	44	demonstrate	demonstrate	VERB
ajst-20212	62	45	the	the	DET
ajst-20212	62	46	effectiveness	effectiveness	NOUN
ajst-20212	62	47	of	of	ADP
ajst-20212	62	48	the	the	DET
ajst-20212	62	49	model	model	NOUN
ajst-20212	62	50	.	.	PUNCT
ajst-20212	63	1	the	the	DET
ajst-20212	63	2	model	model	NOUN
ajst-20212	63	3	includes	include	VERB
ajst-20212	63	4	input	input	NOUN
ajst-20212	63	5	,	,	PUNCT
ajst-20212	63	6	hidden	hidden	ADJ
ajst-20212	63	7	and	and	CCONJ
ajst-20212	63	8	output	output	NOUN
ajst-20212	63	9	layers	layer	NOUN
ajst-20212	63	10	.	.	PUNCT
ajst-20212	64	1	in	in	ADP
ajst-20212	64	2	addition	addition	NOUN
ajst-20212	64	3	to	to	ADP
ajst-20212	64	4	this	this	PRON
ajst-20212	64	5	,	,	PUNCT
ajst-20212	64	6	a	a	DET
ajst-20212	64	7	two	two	NUM
ajst-20212	64	8	-	-	PUNCT
ajst-20212	64	9	way	way	NOUN
ajst-20212	64	10	attention	attention	NOUN
ajst-20212	64	11	mechanism	mechanism	NOUN
ajst-20212	64	12	,	,	PUNCT
ajst-20212	64	13	lexical	lexical	ADJ
ajst-20212	64	14	labeling	labeling	NOUN
ajst-20212	64	15	information	information	NOUN
ajst-20212	64	16	and	and	CCONJ
ajst-20212	64	17	location	location	NOUN
ajst-20212	64	18	coding	code	VERB
ajst-20212	64	19	information	information	NOUN
ajst-20212	64	20	are	be	AUX
ajst-20212	64	21	added	add	VERB
ajst-20212	64	22	to	to	ADP
ajst-20212	64	23	the	the	DET
ajst-20212	64	24	graph	graph	NOUN
ajst-20212	64	25	convolutional	convolutional	ADJ
ajst-20212	64	26	neural	neural	ADJ
ajst-20212	64	27	network	network	NOUN
ajst-20212	64	28	.	.	PUNCT
ajst-20212	65	1	the	the	DET
ajst-20212	65	2	overall	overall	ADJ
ajst-20212	65	3	framework	framework	NOUN
ajst-20212	65	4	is	be	AUX
ajst-20212	65	5	shown	show	VERB
ajst-20212	65	6	in	in	ADP
ajst-20212	65	7	figure	figure	NOUN
ajst-20212	65	8	5	5	NUM
ajst-20212	65	9	.	.	PUNCT
ajst-20212	65	10	figure	figure	NOUN
ajst-20212	65	11	5	5	NUM
ajst-20212	65	12	.	.	PUNCT
ajst-20212	65	13	structural	structural	ADJ
ajst-20212	65	14	diagram	diagram	NOUN
ajst-20212	65	15	of	of	ADP
ajst-20212	65	16	rbt	rbt	PROPN
ajst-20212	65	17	-	-	PUNCT
ajst-20212	65	18	biatt	biatt	NOUN
ajst-20212	65	19	-	-	PUNCT
ajst-20212	65	20	gcn	gcn	NOUN
ajst-20212	65	21	3.1	3.1	NUM
ajst-20212	65	22	.	.	PUNCT
ajst-20212	65	23	model	model	NOUN
ajst-20212	65	24	building	building	NOUN
ajst-20212	65	25	input	input	NOUN
ajst-20212	65	26	layer	layer	NOUN
ajst-20212	65	27	:	:	PUNCT
ajst-20212	65	28	get	get	VERB
ajst-20212	65	29	the	the	DET
ajst-20212	65	30	comment	comment	NOUN
ajst-20212	65	31	text	text	NOUN
ajst-20212	65	32	s	s	X
ajst-20212	65	33	,	,	PUNCT
ajst-20212	65	34	the	the	DET
ajst-20212	65	35	text	text	NOUN
ajst-20212	65	36	consists	consist	VERB
ajst-20212	65	37	of	of	ADP
ajst-20212	65	38	a	a	DET
ajst-20212	65	39	sequence	sequence	NOUN
ajst-20212	65	40	of	of	ADP
ajst-20212	65	41	n	n	DET
ajst-20212	65	42	consecutive	consecutive	ADJ
ajst-20212	65	43	words	word	NOUN
ajst-20212	65	44	as	as	ADP
ajst-20212	65	45	in	in	ADP
ajst-20212	65	46	equation	equation	NOUN
ajst-20212	65	47	6	6	NUM
ajst-20212	65	48	.	.	PUNCT
ajst-20212	66	1	c	c	NOUN
ajst-20212	66	2	1	1	NUM
ajst-20212	66	3	2	2	NUM
ajst-20212	66	4	3	3	NUM
ajst-20212	66	5	n	n	CCONJ
ajst-20212	66	6	{	{	PUNCT
ajst-20212	66	7	,	,	PUNCT
ajst-20212	66	8	,	,	PUNCT
ajst-20212	66	9	,	,	PUNCT
ajst-20212	66	10	,	,	PUNCT
ajst-20212	66	11	s	s	PART
ajst-20212	66	12	}	}	PUNCT
ajst-20212	66	13	c	c	NOUN
ajst-20212	66	14	c	c	NOUN
ajst-20212	66	15	c	c	NOUN
ajst-20212	66	16	cs	cs	PROPN
ajst-20212	66	17	s	s	PROPN
ajst-20212	66	18	s	s	NOUN
ajst-20212	66	19	s	s	NOUN
ajst-20212	66	20			NUM
ajst-20212	66	21	(	(	PUNCT
ajst-20212	66	22	6	6	NUM
ajst-20212	66	23	)	)	PUNCT
ajst-20212	66	24	in	in	ADP
ajst-20212	66	25	attribute	attribute	NOUN
ajst-20212	66	26	-	-	PUNCT
ajst-20212	66	27	level	level	NOUN
ajst-20212	66	28	sentiment	sentiment	NOUN
ajst-20212	66	29	analysis	analysis	NOUN
ajst-20212	66	30	,	,	PUNCT
ajst-20212	66	31	it	it	PRON
ajst-20212	66	32	is	be	AUX
ajst-20212	66	33	necessary	necessary	ADJ
ajst-20212	66	34	to	to	PART
ajst-20212	66	35	discriminate	discriminate	VERB
ajst-20212	66	36	the	the	DET
ajst-20212	66	37	sentiment	sentiment	NOUN
ajst-20212	66	38	polarity	polarity	NOUN
ajst-20212	66	39	of	of	ADP
ajst-20212	66	40	a	a	DET
ajst-20212	66	41	given	give	VERB
ajst-20212	66	42	attribute	attribute	NOUN
ajst-20212	66	43	word	word	NOUN
ajst-20212	66	44	.	.	PUNCT
ajst-20212	67	1	the	the	DET
ajst-20212	67	2	lexical	lexical	ADJ
ajst-20212	67	3	properties	property	NOUN
ajst-20212	67	4	before	before	ADP
ajst-20212	67	5	and	and	CCONJ
ajst-20212	67	6	after	after	ADP
ajst-20212	67	7	the	the	DET
ajst-20212	67	8	attribute	attribute	NOUN
ajst-20212	67	9	word	word	NOUN
ajst-20212	67	10	may	may	AUX
ajst-20212	67	11	affect	affect	VERB
ajst-20212	67	12	the	the	DET
ajst-20212	67	13	sentiment	sentiment	NOUN
ajst-20212	67	14	polarity	polarity	NOUN
ajst-20212	67	15	.	.	PUNCT
ajst-20212	68	1	based	base	VERB
ajst-20212	68	2	on	on	ADP
ajst-20212	68	3	the	the	DET
ajst-20212	68	4	text	text	NOUN
ajst-20212	68	5	sequence	sequence	NOUN
ajst-20212	68	6	,	,	PUNCT
ajst-20212	68	7	lexical	lexical	ADJ
ajst-20212	68	8	annotation	annotation	NOUN
ajst-20212	68	9	is	be	AUX
ajst-20212	68	10	performed	perform	VERB
ajst-20212	68	11	using	use	VERB
ajst-20212	68	12	a	a	DET
ajst-20212	68	13	lexical	lexical	ADJ
ajst-20212	68	14	annotation	annotation	NOUN
ajst-20212	68	15	tool	tool	NOUN
ajst-20212	68	16	.	.	PUNCT
ajst-20212	69	1	as	as	ADP
ajst-20212	69	2	in	in	ADP
ajst-20212	69	3	equations	equation	NOUN
ajst-20212	69	4	7	7	NUM
ajst-20212	69	5	:	:	SYM
ajst-20212	69	6	1	1	NUM
ajst-20212	69	7	2	2	NUM
ajst-20212	69	8	3	3	NUM
ajst-20212	69	9	{	{	PUNCT
ajst-20212	69	10	,	,	PUNCT
ajst-20212	69	11	,	,	PUNCT
ajst-20212	69	12	,	,	PUNCT
ajst-20212	69	13	,	,	PUNCT
ajst-20212	69	14	}	}	PUNCT
ajst-20212	69	15	p	p	NOUN
ajst-20212	69	16	p	p	X
ajst-20212	69	17	p	p	X
ajst-20212	69	18	p	p	X
ajst-20212	69	19	p	p	NOUN
ajst-20212	70	1	ns	ns	PROPN
ajst-20212	70	2	s	s	NOUN
ajst-20212	70	3	s	s	NOUN
ajst-20212	70	4	s	s	NOUN
ajst-20212	70	5	s	s	NOUN
ajst-20212	70	6			NOUN
ajst-20212	70	7	(	(	PUNCT
ajst-20212	70	8	7	7	X
ajst-20212	70	9	)	)	PUNCT
ajst-20212	70	10	assuming	assume	VERB
ajst-20212	70	11	that	that	SCONJ
ajst-20212	70	12	there	there	PRON
ajst-20212	70	13	are	be	VERB
ajst-20212	70	14	j	j	PROPN
ajst-20212	70	15	given	give	VERB
ajst-20212	70	16	attribute	attribute	NOUN
ajst-20212	70	17	words	word	NOUN
ajst-20212	70	18	in	in	ADP
ajst-20212	70	19	the	the	DET
ajst-20212	70	20	comment	comment	NOUN
ajst-20212	70	21	text	text	NOUN
ajst-20212	70	22	s	s	NOUN
ajst-20212	70	23	,	,	PUNCT
ajst-20212	70	24	then	then	ADV
ajst-20212	70	25	the	the	DET
ajst-20212	70	26	set	set	NOUN
ajst-20212	70	27	of	of	ADP
ajst-20212	70	28	attribute	attribute	NOUN
ajst-20212	70	29	words	word	NOUN
ajst-20212	70	30	is	be	AUX
ajst-20212	70	31	as	as	ADP
ajst-20212	70	32	in	in	ADP
ajst-20212	70	33	equation	equation	NOUN
ajst-20212	70	34	8	8	NUM
ajst-20212	70	35	:	:	SYM
ajst-20212	70	36	1	1	NUM
ajst-20212	70	37	2	2	NUM
ajst-20212	70	38	{	{	PUNCT
ajst-20212	70	39	,	,	PUNCT
ajst-20212	70	40	,	,	PUNCT
ajst-20212	70	41	,	,	PUNCT
ajst-20212	70	42	}	}	PUNCT
ajst-20212	70	43	a	a	DET
ajst-20212	70	44	a	a	DET
ajst-20212	70	45	a	a	DET
ajst-20212	70	46	ajs	ajs	PROPN
ajst-20212	70	47	s	s	PART
ajst-20212	70	48	s	s	PROPN
ajst-20212	70	49	s	s	PROPN
ajst-20212	70	50			NOUN
ajst-20212	70	51	(	(	PUNCT
ajst-20212	70	52	8)	8)	NUM
ajst-20212	70	53	each	each	DET
ajst-20212	70	54	attribute	attribute	NOUN
ajst-20212	70	55	word	word	NOUN
ajst-20212	70	56	is	be	AUX
ajst-20212	70	57	a	a	DET
ajst-20212	70	58	continuous	continuous	ADJ
ajst-20212	70	59	subsequence	subsequence	NOUN
ajst-20212	70	60	in	in	ADP
ajst-20212	70	61	a	a	DET
ajst-20212	70	62	sentence	sentence	NOUN
ajst-20212	70	63	,	,	PUNCT
ajst-20212	70	64	then	then	ADV
ajst-20212	70	65	the	the	DET
ajst-20212	70	66	individual	individual	ADJ
ajst-20212	70	67	attribute	attribute	NOUN
ajst-20212	70	68	word	word	NOUN
ajst-20212	70	69	sequence	sequence	NOUN
ajst-20212	70	70	is	be	AUX
ajst-20212	70	71	represented	represent	VERB
ajst-20212	70	72	as	as	ADP
ajst-20212	70	73	in	in	ADP
ajst-20212	70	74	equation	equation	NOUN
ajst-20212	70	75	9	9	NUM
ajst-20212	70	76	:	:	SYM
ajst-20212	70	77	1	1	NUM
ajst-20212	70	78	2	2	NUM
ajst-20212	70	79	{	{	PUNCT
ajst-20212	70	80	,	,	PUNCT
ajst-20212	70	81	,	,	PUNCT
ajst-20212	70	82	,	,	PUNCT
ajst-20212	70	83	}	}	PUNCT
ajst-20212	70	84	ai	ai	VERB
ajst-20212	70	85	ai	ai	INTJ
ajst-20212	70	86	ai	ai	INTJ
ajst-20212	70	87	ai	ai	INTJ
ajst-20212	70	88	mis	mis	PROPN
ajst-20212	70	89	s	s	PROPN
ajst-20212	70	90	s	s	PROPN
ajst-20212	70	91	s	s	PROPN
ajst-20212	70	92			NOUN
ajst-20212	70	93	(	(	PUNCT
ajst-20212	70	94	9	9	NUM
ajst-20212	70	95	)	)	PUNCT
ajst-20212	70	96	hidden	hidden	ADJ
ajst-20212	70	97	layer	layer	NOUN
ajst-20212	70	98	:	:	PUNCT
ajst-20212	70	99	the	the	DET
ajst-20212	70	100	hidden	hide	VERB
ajst-20212	70	101	layer	layer	NOUN
ajst-20212	70	102	uses	use	VERB
ajst-20212	70	103	the	the	DET
ajst-20212	70	104	bigru	bigru	PROPN
ajst-20212	70	105	-	-	PUNCT
ajst-20212	70	106	textcnn	textcnn	ADJ
ajst-20212	70	107	model	model	NOUN
ajst-20212	70	108	to	to	PART
ajst-20212	70	109	obtain	obtain	VERB
ajst-20212	70	110	global	global	ADJ
ajst-20212	70	111	semantic	semantic	ADJ
ajst-20212	70	112	information	information	NOUN
ajst-20212	70	113	and	and	CCONJ
ajst-20212	70	114	local	local	ADJ
ajst-20212	70	115	semantic	semantic	ADJ
ajst-20212	70	116	information	information	NOUN
ajst-20212	70	117	respectively	respectively	ADV
ajst-20212	70	118	.	.	PUNCT
ajst-20212	71	1	biatt	biatt	PROPN
ajst-20212	71	2	layer	layer	NOUN
ajst-20212	71	3	:	:	PUNCT
ajst-20212	71	4	aspect	aspect	VERB
ajst-20212	71	5	attention	attention	NOUN
ajst-20212	71	6	:	:	PUNCT
ajst-20212	71	7	the	the	DET
ajst-20212	71	8	hidden	hide	VERB
ajst-20212	71	9	layer	layer	NOUN
ajst-20212	71	10	output	output	NOUN
ajst-20212	71	11	is	be	AUX
ajst-20212	71	12	obtained	obtain	VERB
ajst-20212	71	13	through	through	ADP
ajst-20212	71	14	bigru	bigru	NOUN
ajst-20212	71	15	-	-	PUNCT
ajst-20212	71	16	textcnn	textcnn	VERB
ajst-20212	71	17	,	,	PUNCT
ajst-20212	71	18	combined	combine	VERB
ajst-20212	71	19	with	with	ADP
ajst-20212	71	20	the	the	DET
ajst-20212	71	21	context	context	NOUN
ajst-20212	71	22	information	information	NOUN
ajst-20212	71	23	,	,	PUNCT
ajst-20212	71	24	and	and	CCONJ
ajst-20212	71	25	the	the	DET
ajst-20212	71	26	new	new	ADJ
ajst-20212	71	27	attribute	attribute	NOUN
ajst-20212	71	28	word	word	NOUN
ajst-20212	71	29	representation	representation	NOUN
ajst-20212	71	30	is	be	AUX
ajst-20212	71	31	obtained	obtain	VERB
ajst-20212	71	32	through	through	ADP
ajst-20212	71	33	the	the	DET
ajst-20212	71	34	attention	attention	NOUN
ajst-20212	71	35	mechanism	mechanism	NOUN
ajst-20212	71	36	.	.	PUNCT
ajst-20212	72	1	170	170	NUM
ajst-20212	72	2	context	context	NOUN
ajst-20212	72	3	-	-	PUNCT
ajst-20212	72	4	position	position	NOUN
ajst-20212	72	5	attention	attention	NOUN
ajst-20212	72	6	:	:	PUNCT
ajst-20212	72	7	on	on	ADP
ajst-20212	72	8	the	the	DET
ajst-20212	72	9	basis	basis	NOUN
ajst-20212	72	10	of	of	ADP
ajst-20212	72	11	the	the	DET
ajst-20212	72	12	new	new	ADJ
ajst-20212	72	13	attribute	attribute	NOUN
ajst-20212	72	14	word	word	NOUN
ajst-20212	72	15	representation	representation	NOUN
ajst-20212	72	16	,	,	PUNCT
ajst-20212	72	17	combined	combine	VERB
ajst-20212	72	18	with	with	ADP
ajst-20212	72	19	the	the	DET
ajst-20212	72	20	position	position	NOUN
ajst-20212	72	21	encoding	encode	VERB
ajst-20212	72	22	information	information	NOUN
ajst-20212	72	23	,	,	PUNCT
ajst-20212	72	24	after	after	ADP
ajst-20212	72	25	modeling	model	VERB
ajst-20212	72	26	through	through	ADP
ajst-20212	72	27	the	the	DET
ajst-20212	72	28	attention	attention	NOUN
ajst-20212	72	29	mechanism	mechanism	NOUN
ajst-20212	72	30	,	,	PUNCT
ajst-20212	72	31	an	an	DET
ajst-20212	72	32	attribute	attribute	NOUN
ajst-20212	72	33	representation	representation	NOUN
ajst-20212	72	34	that	that	PRON
ajst-20212	72	35	integrates	integrate	VERB
ajst-20212	72	36	lexical	lexical	ADJ
ajst-20212	72	37	,	,	PUNCT
ajst-20212	72	38	positional	positional	ADJ
ajst-20212	72	39	,	,	PUNCT
ajst-20212	72	40	and	and	CCONJ
ajst-20212	72	41	contextual	contextual	ADJ
ajst-20212	72	42	information	information	NOUN
ajst-20212	72	43	is	be	AUX
ajst-20212	72	44	finally	finally	ADV
ajst-20212	72	45	obtained	obtain	VERB
ajst-20212	72	46	,	,	PUNCT
ajst-20212	72	47	which	which	PRON
ajst-20212	72	48	is	be	AUX
ajst-20212	72	49	used	use	VERB
ajst-20212	72	50	for	for	ADP
ajst-20212	72	51	subsequent	subsequent	ADJ
ajst-20212	72	52	input	input	NOUN
ajst-20212	72	53	to	to	PART
ajst-20212	72	54	graph	graph	VERB
ajst-20212	72	55	convolutional	convolutional	ADJ
ajst-20212	72	56	neural	neural	ADJ
ajst-20212	72	57	network	network	NOUN
ajst-20212	72	58	for	for	ADP
ajst-20212	72	59	sentiment	sentiment	NOUN
ajst-20212	72	60	classification	classification	NOUN
ajst-20212	72	61	.	.	PUNCT
ajst-20212	73	1	gcn	gcn	ADJ
ajst-20212	73	2	layer	layer	NOUN
ajst-20212	73	3	:	:	PUNCT
ajst-20212	73	4	in	in	ADP
ajst-20212	73	5	order	order	NOUN
ajst-20212	73	6	to	to	PART
ajst-20212	73	7	capture	capture	VERB
ajst-20212	73	8	the	the	DET
ajst-20212	73	9	sentiment	sentiment	NOUN
ajst-20212	73	10	dependencies	dependency	NOUN
ajst-20212	73	11	between	between	ADP
ajst-20212	73	12	attributes	attribute	NOUN
ajst-20212	73	13	in	in	ADP
ajst-20212	73	14	the	the	DET
ajst-20212	73	15	comment	comment	NOUN
ajst-20212	73	16	text	text	NOUN
ajst-20212	73	17	,	,	PUNCT
ajst-20212	73	18	each	each	DET
ajst-20212	73	19	attribute	attribute	NOUN
ajst-20212	73	20	in	in	ADP
ajst-20212	73	21	the	the	DET
ajst-20212	73	22	comment	comment	NOUN
ajst-20212	73	23	text	text	NOUN
ajst-20212	73	24	is	be	AUX
ajst-20212	73	25	considered	consider	VERB
ajst-20212	73	26	as	as	ADP
ajst-20212	73	27	a	a	DET
ajst-20212	73	28	node	node	NOUN
ajst-20212	73	29	,	,	PUNCT
ajst-20212	73	30	and	and	CCONJ
ajst-20212	73	31	then	then	ADV
ajst-20212	73	32	edges	edge	NOUN
ajst-20212	73	33	are	be	AUX
ajst-20212	73	34	constructed	construct	VERB
ajst-20212	73	35	based	base	VERB
ajst-20212	73	36	on	on	ADP
ajst-20212	73	37	the	the	DET
ajst-20212	73	38	adjacencies	adjacency	NOUN
ajst-20212	73	39	between	between	ADP
ajst-20212	73	40	the	the	DET
ajst-20212	73	41	attributes	attribute	NOUN
ajst-20212	73	42	,	,	PUNCT
ajst-20212	73	43	and	and	CCONJ
ajst-20212	73	44	these	these	DET
ajst-20212	73	45	edges	edge	NOUN
ajst-20212	73	46	represent	represent	VERB
ajst-20212	73	47	the	the	DET
ajst-20212	73	48	sentiment	sentiment	NOUN
ajst-20212	73	49	dependencies	dependency	NOUN
ajst-20212	73	50	between	between	ADP
ajst-20212	73	51	the	the	DET
ajst-20212	73	52	attributes	attribute	NOUN
ajst-20212	73	53	.	.	PUNCT
ajst-20212	74	1	the	the	DET
ajst-20212	74	2	generated	generate	VERB
ajst-20212	74	3	graph	graph	NOUN
ajst-20212	74	4	structure	structure	NOUN
ajst-20212	74	5	includes	include	VERB
ajst-20212	74	6	these	these	DET
ajst-20212	74	7	nodes	node	NOUN
ajst-20212	74	8	and	and	CCONJ
ajst-20212	74	9	edges	edge	NOUN
ajst-20212	74	10	as	as	ADP
ajst-20212	74	11	inputs	input	NOUN
ajst-20212	74	12	to	to	ADP
ajst-20212	74	13	the	the	DET
ajst-20212	74	14	graph	graph	NOUN
ajst-20212	74	15	convolutional	convolutional	ADJ
ajst-20212	74	16	neural	neural	ADJ
ajst-20212	74	17	network	network	NOUN
ajst-20212	74	18	(	(	PUNCT
ajst-20212	74	19	gcn	gcn	NOUN
ajst-20212	74	20	)	)	PUNCT
ajst-20212	74	21	.	.	PUNCT
ajst-20212	75	1	3.2	3.2	NUM
ajst-20212	75	2	.	.	PUNCT
ajst-20212	75	3	criteria	criterion	NOUN
ajst-20212	75	4	for	for	ADP
ajst-20212	75	5	evaluation	evaluation	NOUN
ajst-20212	75	6	this	this	DET
ajst-20212	75	7	chapter	chapter	NOUN
ajst-20212	75	8	focuses	focus	VERB
ajst-20212	75	9	on	on	ADP
ajst-20212	75	10	measuring	measure	VERB
ajst-20212	75	11	model	model	NOUN
ajst-20212	75	12	performance	performance	NOUN
ajst-20212	75	13	using	use	VERB
ajst-20212	75	14	f1	f1	NOUN
ajst-20212	75	15	(	(	PUNCT
ajst-20212	75	16	f1	f1	NOUN
ajst-20212	75	17	-	-	PUNCT
ajst-20212	75	18	score	score	NOUN
ajst-20212	75	19	)	)	PUNCT
ajst-20212	75	20	values	value	NOUN
ajst-20212	75	21	as	as	ADV
ajst-20212	75	22	well	well	ADV
ajst-20212	75	23	as	as	ADP
ajst-20212	75	24	macro	macro	ADJ
ajst-20212	75	25	-	-	ADJ
ajst-20212	75	26	averaged	average	VERB
ajst-20212	75	27	f1	f1	NOUN
ajst-20212	75	28	(	(	PUNCT
ajst-20212	75	29	macro_f1	macro_f1	NOUN
ajst-20212	75	30	)	)	PUNCT
ajst-20212	75	31	values	value	NOUN
ajst-20212	75	32	.	.	PUNCT
ajst-20212	76	1	the	the	DET
ajst-20212	76	2	f1	f1	PROPN
ajst-20212	76	3	values	value	NOUN
ajst-20212	76	4	are	be	AUX
ajst-20212	76	5	calculated	calculate	VERB
ajst-20212	76	6	using	use	VERB
ajst-20212	76	7	the	the	DET
ajst-20212	76	8	following	follow	VERB
ajst-20212	76	9	formula	formula	NOUN
ajst-20212	76	10	:	:	PUNCT
ajst-20212	76	11	2	2	NUM
ajst-20212	76	12	*	*	SYM
ajst-20212	76	13	*	*	PUNCT
ajst-20212	76	14	1	1	NUM
ajst-20212	76	15	*	*	SYM
ajst-20212	76	16	100	100	NUM
ajst-20212	76	17	%	%	NOUN
ajst-20212	76	18	p	p	NOUN
ajst-20212	76	19	r	r	NOUN
ajst-20212	76	20	f	f	NOUN
ajst-20212	76	21	p	p	NOUN
ajst-20212	76	22	r	r	NOUN
ajst-20212	76	23			PROPN
ajst-20212	76	24			PUNCT
ajst-20212	76	25	(	(	PUNCT
ajst-20212	76	26	10	10	NUM
ajst-20212	76	27	)	)	PUNCT
ajst-20212	76	28	where	where	SCONJ
ajst-20212	76	29	r	r	NOUN
ajst-20212	76	30	precision	precision	NOUN
ajst-20212	76	31	rate	rate	NOUN
ajst-20212	76	32	and	and	CCONJ
ajst-20212	76	33	r	r	NOUN
ajst-20212	76	34	is	be	AUX
ajst-20212	76	35	the	the	DET
ajst-20212	76	36	recall	recall	NOUN
ajst-20212	76	37	rate	rate	NOUN
ajst-20212	76	38	.	.	PUNCT
ajst-20212	77	1	*	*	PUNCT
ajst-20212	77	2	100	100	NUM
ajst-20212	77	3	%	%	NOUN
ajst-20212	77	4	tp	tp	ADP
ajst-20212	77	5	p	p	NOUN
ajst-20212	77	6	tp	tp	ADP
ajst-20212	77	7	fp	fp	PROPN
ajst-20212	77	8			PROPN
ajst-20212	77	9			PUNCT
ajst-20212	77	10	(	(	PUNCT
ajst-20212	77	11	11	11	NUM
ajst-20212	77	12	)	)	PUNCT
ajst-20212	77	13	*	*	PUNCT
ajst-20212	77	14	100	100	NUM
ajst-20212	77	15	%	%	NOUN
ajst-20212	77	16	tp	tp	ADP
ajst-20212	77	17	r	r	NOUN
ajst-20212	77	18	tp	tp	NOUN
ajst-20212	77	19	fn	fn	PROPN
ajst-20212	77	20			PROPN
ajst-20212	77	21			PUNCT
ajst-20212	77	22	(	(	PUNCT
ajst-20212	77	23	11	11	NUM
ajst-20212	77	24	)	)	PUNCT
ajst-20212	77	25	where	where	SCONJ
ajst-20212	77	26	tp	tp	NOUN
ajst-20212	77	27	means	mean	VERB
ajst-20212	77	28	that	that	SCONJ
ajst-20212	77	29	the	the	DET
ajst-20212	77	30	prediction	prediction	NOUN
ajst-20212	77	31	is	be	AUX
ajst-20212	77	32	positive	positive	ADJ
ajst-20212	77	33	and	and	CCONJ
ajst-20212	77	34	the	the	DET
ajst-20212	77	35	actual	actual	ADJ
ajst-20212	77	36	value	value	NOUN
ajst-20212	77	37	is	be	AUX
ajst-20212	77	38	also	also	ADV
ajst-20212	77	39	positive;fp	positive;fp	NOUN
ajst-20212	77	40	means	mean	NOUN
ajst-20212	77	41	that	that	SCONJ
ajst-20212	77	42	the	the	DET
ajst-20212	77	43	prediction	prediction	NOUN
ajst-20212	77	44	is	be	AUX
ajst-20212	77	45	positive	positive	ADJ
ajst-20212	77	46	but	but	CCONJ
ajst-20212	77	47	the	the	DET
ajst-20212	77	48	actual	actual	ADJ
ajst-20212	77	49	value	value	NOUN
ajst-20212	77	50	is	be	AUX
ajst-20212	77	51	negative	negative	ADJ
ajst-20212	77	52	;	;	PUNCT
ajst-20212	77	53	fn	fn	NOUN
ajst-20212	77	54	means	mean	VERB
ajst-20212	77	55	that	that	SCONJ
ajst-20212	77	56	the	the	DET
ajst-20212	77	57	prediction	prediction	NOUN
ajst-20212	77	58	is	be	AUX
ajst-20212	77	59	negative	negative	ADJ
ajst-20212	77	60	but	but	CCONJ
ajst-20212	77	61	the	the	DET
ajst-20212	77	62	actual	actual	ADJ
ajst-20212	77	63	value	value	NOUN
ajst-20212	77	64	is	be	AUX
ajst-20212	77	65	positive	positive	ADJ
ajst-20212	77	66	.	.	PUNCT
ajst-20212	78	1	the	the	DET
ajst-20212	78	2	macro_f1	macro_f1	ADJ
ajst-20212	78	3	calculation	calculation	NOUN
ajst-20212	78	4	formula	formula	NOUN
ajst-20212	78	5	is	be	AUX
ajst-20212	78	6	shown	show	VERB
ajst-20212	78	7	in	in	ADP
ajst-20212	78	8	12	12	NUM
ajst-20212	78	9	.	.	PUNCT
ajst-20212	79	1	1	1	NUM
ajst-20212	79	2	1	1	NUM
ajst-20212	79	3	_	_	SYM
ajst-20212	79	4	1	1	NUM
ajst-20212	79	5	1	1	NUM
ajst-20212	79	6	x	x	X
ajst-20212	79	7	i	i	PRON
ajst-20212	80	1	i	i	PRON
ajst-20212	80	2	macro	macro	VERB
ajst-20212	81	1	f	f	PROPN
ajst-20212	82	1	f	f	PROPN
ajst-20212	82	2	x	x	PROPN
ajst-20212	82	3			NUM
ajst-20212	82	4			ADJ
ajst-20212	82	5			X
ajst-20212	82	6	(	(	PUNCT
ajst-20212	82	7	12	12	NUM
ajst-20212	82	8	)	)	PUNCT
ajst-20212	82	9	3.3	3.3	NUM
ajst-20212	82	10	.	.	PUNCT
ajst-20212	83	1	data	datum	NOUN
ajst-20212	83	2	set	set	VERB
ajst-20212	83	3	this	this	DET
ajst-20212	83	4	chapter	chapter	NOUN
ajst-20212	83	5	first	first	ADJ
ajst-20212	83	6	experiments	experiment	NOUN
ajst-20212	83	7	with	with	ADP
ajst-20212	83	8	the	the	DET
ajst-20212	83	9	model	model	NOUN
ajst-20212	83	10	using	use	VERB
ajst-20212	83	11	the	the	DET
ajst-20212	83	12	publicly	publicly	ADV
ajst-20212	83	13	available	available	ADJ
ajst-20212	83	14	dataset	dataset	NOUN
ajst-20212	83	15	ai	ai	VERB
ajst-20212	83	16	challenger2018	challenger2018	NOUN
ajst-20212	83	17	to	to	PART
ajst-20212	83	18	validate	validate	VERB
ajst-20212	83	19	the	the	DET
ajst-20212	83	20	effectiveness	effectiveness	NOUN
ajst-20212	83	21	of	of	ADP
ajst-20212	83	22	the	the	DET
ajst-20212	83	23	model	model	NOUN
ajst-20212	83	24	.	.	PUNCT
ajst-20212	84	1	this	this	DET
ajst-20212	84	2	dataset	dataset	NOUN
ajst-20212	84	3	is	be	AUX
ajst-20212	84	4	user	user	NOUN
ajst-20212	84	5	reviews	review	NOUN
ajst-20212	84	6	in	in	ADP
ajst-20212	84	7	the	the	DET
ajst-20212	84	8	field	field	NOUN
ajst-20212	84	9	of	of	ADP
ajst-20212	84	10	food	food	NOUN
ajst-20212	84	11	and	and	CCONJ
ajst-20212	84	12	beverage	beverage	NOUN
ajst-20212	84	13	,	,	PUNCT
ajst-20212	84	14	which	which	PRON
ajst-20212	84	15	is	be	AUX
ajst-20212	84	16	often	often	ADV
ajst-20212	84	17	used	use	VERB
ajst-20212	84	18	in	in	ADP
ajst-20212	84	19	chinese	chinese	ADJ
ajst-20212	84	20	attribute	attribute	NOUN
ajst-20212	84	21	-	-	PUNCT
ajst-20212	84	22	level	level	NOUN
ajst-20212	84	23	sentiment	sentiment	NOUN
ajst-20212	84	24	analysis	analysis	NOUN
ajst-20212	84	25	tasks	task	NOUN
ajst-20212	84	26	,	,	PUNCT
ajst-20212	84	27	and	and	CCONJ
ajst-20212	84	28	can	can	AUX
ajst-20212	84	29	be	be	AUX
ajst-20212	84	30	used	use	VERB
ajst-20212	84	31	very	very	ADV
ajst-20212	84	32	reliably	reliably	ADV
ajst-20212	84	33	in	in	ADP
ajst-20212	84	34	related	related	ADJ
ajst-20212	84	35	research	research	NOUN
ajst-20212	84	36	.	.	PUNCT
ajst-20212	85	1	the	the	DET
ajst-20212	85	2	dataset	dataset	NOUN
ajst-20212	85	3	is	be	AUX
ajst-20212	85	4	divided	divide	VERB
ajst-20212	85	5	and	and	CCONJ
ajst-20212	85	6	has	have	AUX
ajst-20212	85	7	been	be	AUX
ajst-20212	85	8	labeled	label	VERB
ajst-20212	85	9	with	with	ADP
ajst-20212	85	10	4	4	NUM
ajst-20212	85	11	types	type	NOUN
ajst-20212	85	12	of	of	ADP
ajst-20212	85	13	sentiment	sentiment	NOUN
ajst-20212	85	14	tendencies	tendency	NOUN
ajst-20212	85	15	for	for	ADP
ajst-20212	85	16	20	20	NUM
ajst-20212	85	17	finegrained	finegrained	ADJ
ajst-20212	85	18	attributes	attribute	NOUN
ajst-20212	85	19	,	,	PUNCT
ajst-20212	85	20	with	with	ADP
ajst-20212	85	21	the	the	DET
ajst-20212	85	22	sentiment	sentiment	NOUN
ajst-20212	85	23	tendency	tendency	NOUN
ajst-20212	85	24	labels	label	VERB
ajst-20212	85	25	:	:	PUNCT
ajst-20212	85	26	{	{	PUNCT
ajst-20212	85	27	unmentioned	unmentioned	ADJ
ajst-20212	85	28	:	:	PUNCT
ajst-20212	85	29	-2	-2	ADJ
ajst-20212	85	30	,	,	PUNCT
ajst-20212	85	31	negative	negative	ADJ
ajst-20212	85	32	:	:	PUNCT
ajst-20212	85	33	-1	-1	ADJ
ajst-20212	85	34	,	,	PUNCT
ajst-20212	85	35	neutral	neutral	ADJ
ajst-20212	85	36	:	:	PUNCT
ajst-20212	85	37	0	0	NUM
ajst-20212	85	38	,	,	PUNCT
ajst-20212	85	39	positive	positive	ADJ
ajst-20212	85	40	:	:	PUNCT
ajst-20212	85	41	1	1	NUM
ajst-20212	85	42	}	}	PUNCT
ajst-20212	85	43	.	.	PUNCT
ajst-20212	86	1	in	in	ADP
ajst-20212	86	2	the	the	DET
ajst-20212	86	3	experiments	experiment	NOUN
ajst-20212	86	4	in	in	ADP
ajst-20212	86	5	this	this	DET
ajst-20212	86	6	chapter	chapter	NOUN
ajst-20212	86	7	,	,	PUNCT
ajst-20212	86	8	four	four	NUM
ajst-20212	86	9	of	of	ADP
ajst-20212	86	10	these	these	DET
ajst-20212	86	11	fine	fine	ADV
ajst-20212	86	12	-	-	PUNCT
ajst-20212	86	13	grained	grain	VERB
ajst-20212	86	14	attributes	attribute	NOUN
ajst-20212	86	15	have	have	AUX
ajst-20212	86	16	been	be	AUX
ajst-20212	86	17	selected	select	VERB
ajst-20212	86	18	for	for	ADP
ajst-20212	86	19	the	the	DET
ajst-20212	86	20	study	study	NOUN
ajst-20212	86	21	.	.	PUNCT
ajst-20212	87	1	3.4	3.4	NUM
ajst-20212	87	2	.	.	PUNCT
ajst-20212	87	3	baseline	baseline	PROPN
ajst-20212	87	4	modeling	model	VERB
ajst-20212	87	5	experiments	experiment	NOUN
ajst-20212	87	6	atae	atae	NOUN
ajst-20212	87	7	-	-	PUNCT
ajst-20212	87	8	lstm	lstm	NOUN
ajst-20212	87	9	:	:	PUNCT
ajst-20212	87	10	a	a	DET
ajst-20212	87	11	neural	neural	ADJ
ajst-20212	87	12	network	network	NOUN
ajst-20212	87	13	model	model	NOUN
ajst-20212	87	14	for	for	ADP
ajst-20212	87	15	attribute	attribute	NOUN
ajst-20212	87	16	-	-	PUNCT
ajst-20212	87	17	level	level	NOUN
ajst-20212	87	18	sentiment	sentiment	NOUN
ajst-20212	87	19	analysis	analysis	NOUN
ajst-20212	87	20	,	,	PUNCT
ajst-20212	87	21	by	by	ADP
ajst-20212	87	22	introducing	introduce	VERB
ajst-20212	87	23	an	an	DET
ajst-20212	87	24	attentional	attentional	ADJ
ajst-20212	87	25	mechanism	mechanism	NOUN
ajst-20212	87	26	,	,	PUNCT
ajst-20212	87	27	the	the	DET
ajst-20212	87	28	model	model	NOUN
ajst-20212	87	29	is	be	AUX
ajst-20212	87	30	able	able	ADJ
ajst-20212	87	31	to	to	PART
ajst-20212	87	32	dynamically	dynamically	ADV
ajst-20212	87	33	adjust	adjust	VERB
ajst-20212	87	34	the	the	DET
ajst-20212	87	35	level	level	NOUN
ajst-20212	87	36	of	of	ADP
ajst-20212	87	37	attention	attention	NOUN
ajst-20212	87	38	paid	pay	VERB
ajst-20212	87	39	to	to	ADP
ajst-20212	87	40	the	the	DET
ajst-20212	87	41	target	target	NOUN
ajst-20212	87	42	and	and	CCONJ
ajst-20212	87	43	related	relate	VERB
ajst-20212	87	44	words	word	NOUN
ajst-20212	87	45	in	in	ADP
ajst-20212	87	46	order	order	NOUN
ajst-20212	87	47	to	to	PART
ajst-20212	87	48	capture	capture	VERB
ajst-20212	87	49	the	the	DET
ajst-20212	87	50	emotional	emotional	ADJ
ajst-20212	87	51	connection	connection	NOUN
ajst-20212	87	52	between	between	ADP
ajst-20212	87	53	the	the	DET
ajst-20212	87	54	target	target	NOUN
ajst-20212	87	55	and	and	CCONJ
ajst-20212	87	56	the	the	DET
ajst-20212	87	57	expression	expression	NOUN
ajst-20212	87	58	more	more	ADV
ajst-20212	87	59	efficiently.the	efficiently.the	DET
ajst-20212	87	60	lstm	lstm	ADJ
ajst-20212	87	61	layer	layer	NOUN
ajst-20212	87	62	is	be	AUX
ajst-20212	87	63	used	use	VERB
ajst-20212	87	64	to	to	PART
ajst-20212	87	65	process	process	VERB
ajst-20212	87	66	sequential	sequential	ADJ
ajst-20212	87	67	data	datum	NOUN
ajst-20212	87	68	,	,	PUNCT
ajst-20212	87	69	which	which	PRON
ajst-20212	87	70	helps	help	VERB
ajst-20212	87	71	the	the	DET
ajst-20212	87	72	model	model	NOUN
ajst-20212	87	73	to	to	PART
ajst-20212	87	74	understand	understand	VERB
ajst-20212	87	75	long	long	ADJ
ajst-20212	87	76	-	-	PUNCT
ajst-20212	87	77	distance	distance	NOUN
ajst-20212	87	78	dependencies	dependency	NOUN
ajst-20212	87	79	.	.	PUNCT
ajst-20212	88	1	ian	ian	PROPN
ajst-20212	88	2	:	:	PUNCT
ajst-20212	88	3	interactive	interactive	ADJ
ajst-20212	88	4	attention	attention	NOUN
ajst-20212	88	5	network	network	NOUN
ajst-20212	88	6	for	for	ADP
ajst-20212	88	7	atsa	atsa	NOUN
ajst-20212	88	8	tasks	task	NOUN
ajst-20212	88	9	,	,	PUNCT
ajst-20212	88	10	based	base	VERB
ajst-20212	88	11	on	on	ADP
ajst-20212	88	12	lstm	lstm	NOUN
ajst-20212	88	13	and	and	CCONJ
ajst-20212	88	14	attention	attention	NOUN
ajst-20212	88	15	mechanisms	mechanism	NOUN
ajst-20212	88	16	.	.	PUNCT
ajst-20212	89	1	aen	aen	PROPN
ajst-20212	89	2	:	:	PUNCT
ajst-20212	89	3	the	the	DET
ajst-20212	89	4	model	model	NOUN
ajst-20212	89	5	mainly	mainly	ADV
ajst-20212	89	6	consists	consist	VERB
ajst-20212	89	7	of	of	ADP
ajst-20212	89	8	an	an	DET
ajst-20212	89	9	embedding	embed	VERB
ajst-20212	89	10	layer	layer	NOUN
ajst-20212	89	11	,	,	PUNCT
ajst-20212	89	12	an	an	DET
ajst-20212	89	13	attentional	attentional	ADJ
ajst-20212	89	14	coding	code	VERB
ajst-20212	89	15	layer	layer	NOUN
ajst-20212	89	16	,	,	PUNCT
ajst-20212	89	17	an	an	DET
ajst-20212	89	18	attention	attention	NOUN
ajst-20212	89	19	layer	layer	NOUN
ajst-20212	89	20	for	for	ADP
ajst-20212	89	21	specific	specific	ADJ
ajst-20212	89	22	aspect	aspect	NOUN
ajst-20212	89	23	words	word	NOUN
ajst-20212	89	24	and	and	CCONJ
ajst-20212	89	25	an	an	DET
ajst-20212	89	26	output	output	NOUN
ajst-20212	89	27	layer	layer	NOUN
ajst-20212	89	28	.	.	PUNCT
ajst-20212	90	1	bert	bert	NOUN
ajst-20212	90	2	-	-	PUNCT
ajst-20212	90	3	lstm	lstm	PROPN
ajst-20212	90	4	:	:	PUNCT
ajst-20212	90	5	by	by	ADP
ajst-20212	90	6	combining	combine	VERB
ajst-20212	90	7	bert	bert	PROPN
ajst-20212	90	8	and	and	CCONJ
ajst-20212	90	9	lstm	lstm	NOUN
ajst-20212	90	10	to	to	PART
ajst-20212	90	11	increase	increase	VERB
ajst-20212	90	12	the	the	DET
ajst-20212	90	13	utilization	utilization	NOUN
ajst-20212	90	14	of	of	ADP
ajst-20212	90	15	textual	textual	ADJ
ajst-20212	90	16	information	information	NOUN
ajst-20212	90	17	,	,	PUNCT
ajst-20212	90	18	and	and	CCONJ
ajst-20212	90	19	finally	finally	ADV
ajst-20212	90	20	the	the	DET
ajst-20212	90	21	output	output	NOUN
ajst-20212	90	22	results	result	NOUN
ajst-20212	90	23	of	of	ADP
ajst-20212	90	24	the	the	DET
ajst-20212	90	25	fully	fully	ADV
ajst-20212	90	26	connected	connect	VERB
ajst-20212	90	27	layer	layer	NOUN
ajst-20212	90	28	are	be	AUX
ajst-20212	90	29	used	use	VERB
ajst-20212	90	30	to	to	PART
ajst-20212	90	31	judge	judge	VERB
ajst-20212	90	32	the	the	DET
ajst-20212	90	33	sentiment	sentiment	NOUN
ajst-20212	90	34	classification	classification	NOUN
ajst-20212	90	35	.	.	PUNCT
ajst-20212	91	1	table	table	NOUN
ajst-20212	91	2	1	1	NUM
ajst-20212	91	3	shows	show	VERB
ajst-20212	91	4	the	the	DET
ajst-20212	91	5	experimental	experimental	ADJ
ajst-20212	91	6	results	result	NOUN
ajst-20212	91	7	of	of	ADP
ajst-20212	91	8	each	each	DET
ajst-20212	91	9	model	model	NOUN
ajst-20212	91	10	.	.	PUNCT
ajst-20212	92	1	in	in	ADP
ajst-20212	92	2	table	table	NOUN
ajst-20212	92	3	1	1	NUM
ajst-20212	92	4	p	p	NOUN
ajst-20212	92	5	,	,	PUNCT
ajst-20212	92	6	u	u	NOUN
ajst-20212	92	7	,	,	PUNCT
ajst-20212	92	8	and	and	CCONJ
ajst-20212	92	9	n	n	PRON
ajst-20212	92	10	are	be	AUX
ajst-20212	92	11	negative	negative	ADJ
ajst-20212	92	12	,	,	PUNCT
ajst-20212	92	13	neutral	neutral	ADJ
ajst-20212	92	14	,	,	PUNCT
ajst-20212	92	15	and	and	CCONJ
ajst-20212	92	16	positive	positive	ADJ
ajst-20212	92	17	,	,	PUNCT
ajst-20212	92	18	respectively	respectively	ADV
ajst-20212	92	19	.	.	PUNCT
ajst-20212	92	20	table	table	NOUN
ajst-20212	92	21	1	1	NUM
ajst-20212	92	22	.	.	PUNCT
ajst-20212	93	1	experimental	experimental	ADJ
ajst-20212	93	2	results	result	NOUN
ajst-20212	93	3	for	for	ADP
ajst-20212	93	4	each	each	DET
ajst-20212	93	5	model	model	NOUN
ajst-20212	93	6	model	model	NOUN
ajst-20212	93	7	p	p	NOUN
ajst-20212	93	8	-	-	PUNCT
ajst-20212	93	9	f1	f1	NOUN
ajst-20212	93	10	(	(	PUNCT
ajst-20212	93	11	%	%	INTJ
ajst-20212	93	12	)	)	PUNCT
ajst-20212	93	13	u	u	NOUN
ajst-20212	93	14	-	-	NOUN
ajst-20212	93	15	f1	f1	ADJ
ajst-20212	93	16	(	(	PUNCT
ajst-20212	93	17	%	%	NOUN
ajst-20212	93	18	)	)	PUNCT
ajst-20212	93	19	n	n	CCONJ
ajst-20212	93	20	-	-	PUNCT
ajst-20212	93	21	f1	f1	NOUN
ajst-20212	93	22	(	(	PUNCT
ajst-20212	93	23	%	%	NOUN
ajst-20212	93	24	)	)	PUNCT
ajst-20212	93	25	m	m	PROPN
ajst-20212	93	26	-	-	PUNCT
ajst-20212	93	27	f	f	X
ajst-20212	93	28	(	(	PUNCT
ajst-20212	93	29	%	%	INTJ
ajst-20212	93	30	)	)	PUNCT
ajst-20212	93	31	atae	atae	NOUN
ajst-20212	93	32	-	-	PUNCT
ajst-20212	93	33	lstm	lstm	NOUN
ajst-20212	93	34	76.83	76.83	NUM
ajst-20212	93	35	45.51	45.51	NUM
ajst-20212	94	1	71.81	71.81	NUM
ajst-20212	94	2	64.72	64.72	NUM
ajst-20212	94	3	ian	ian	PROPN
ajst-20212	94	4	70.34	70.34	NUM
ajst-20212	94	5	55.14	55.14	NUM
ajst-20212	94	6	61.49	61.49	NUM
ajst-20212	94	7	62.32	62.32	NUM
ajst-20212	94	8	aen	aen	PROPN
ajst-20212	94	9	73.96	73.96	NUM
ajst-20212	94	10	45.32	45.32	NUM
ajst-20212	94	11	70.79	70.79	NUM
ajst-20212	94	12	63.36	63.36	NUM
ajst-20212	94	13	bert	bert	NOUN
ajst-20212	94	14	-	-	PUNCT
ajst-20212	94	15	lstm	lstm	NOUN
ajst-20212	94	16	84.10	84.10	NUM
ajst-20212	94	17	54.11	54.11	NUM
ajst-20212	94	18	81.87	81.87	NUM
ajst-20212	94	19	73.36	73.36	NUM
ajst-20212	94	20	rbt	rbt	PROPN
ajst-20212	94	21	-	-	PUNCT
ajst-20212	94	22	biatt	biatt	NOUN
ajst-20212	94	23	-	-	PUNCT
ajst-20212	94	24	gcn	gcn	NOUN
ajst-20212	94	25	89.26	89.26	NUM
ajst-20212	94	26	59.15	59.15	NUM
ajst-20212	94	27	86.22	86.22	NUM
ajst-20212	94	28	78.21	78.21	NUM
ajst-20212	94	29	as	as	SCONJ
ajst-20212	94	30	can	can	AUX
ajst-20212	94	31	be	be	AUX
ajst-20212	94	32	seen	see	VERB
ajst-20212	94	33	from	from	ADP
ajst-20212	94	34	table	table	NOUN
ajst-20212	94	35	1	1	NUM
ajst-20212	95	1	,	,	PUNCT
ajst-20212	95	2	the	the	DET
ajst-20212	95	3	bert	bert	NOUN
ajst-20212	95	4	-	-	PUNCT
ajst-20212	95	5	lstm	lstm	PROPN
ajst-20212	95	6	and	and	CCONJ
ajst-20212	95	7	rbtbiatt	rbtbiatt	NOUN
ajst-20212	95	8	-	-	PUNCT
ajst-20212	95	9	gcn	gcn	NOUN
ajst-20212	95	10	results	result	NOUN
ajst-20212	95	11	are	be	AUX
ajst-20212	95	12	greatly	greatly	ADV
ajst-20212	95	13	improved	improve	VERB
ajst-20212	95	14	compared	compare	VERB
ajst-20212	95	15	to	to	ADP
ajst-20212	95	16	the	the	DET
ajst-20212	95	17	other	other	ADJ
ajst-20212	95	18	three	three	NUM
ajst-20212	95	19	models	model	NOUN
ajst-20212	95	20	.	.	PUNCT
ajst-20212	96	1	this	this	PRON
ajst-20212	96	2	is	be	AUX
ajst-20212	96	3	obviously	obviously	ADV
ajst-20212	96	4	the	the	DET
ajst-20212	96	5	pre	pre	ADJ
ajst-20212	96	6	-	-	ADJ
ajst-20212	96	7	training	training	ADJ
ajst-20212	96	8	model	model	NOUN
ajst-20212	96	9	plays	play	VERB
ajst-20212	96	10	a	a	DET
ajst-20212	96	11	role	role	NOUN
ajst-20212	96	12	,	,	PUNCT
ajst-20212	96	13	which	which	PRON
ajst-20212	96	14	shows	show	VERB
ajst-20212	96	15	that	that	SCONJ
ajst-20212	96	16	the	the	DET
ajst-20212	96	17	use	use	NOUN
ajst-20212	96	18	of	of	ADP
ajst-20212	96	19	linguistic	linguistic	ADJ
ajst-20212	96	20	pretraining	pretraine	VERB
ajst-20212	96	21	model	model	NOUN
ajst-20212	96	22	can	can	AUX
ajst-20212	96	23	really	really	ADV
ajst-20212	96	24	improve	improve	VERB
ajst-20212	96	25	the	the	DET
ajst-20212	96	26	performance	performance	NOUN
ajst-20212	96	27	of	of	ADP
ajst-20212	96	28	the	the	DET
ajst-20212	96	29	model	model	NOUN
ajst-20212	96	30	in	in	ADP
ajst-20212	96	31	sentiment	sentiment	NOUN
ajst-20212	96	32	analysis	analysis	NOUN
ajst-20212	96	33	.	.	PUNCT
ajst-20212	97	1	the	the	DET
ajst-20212	97	2	rbt	rbt	PROPN
ajst-20212	97	3	-	-	PUNCT
ajst-20212	97	4	biatt	biatt	NOUN
ajst-20212	97	5	-	-	PUNCT
ajst-20212	97	6	gcn	gcn	NOUN
ajst-20212	97	7	,	,	PUNCT
ajst-20212	97	8	on	on	ADP
ajst-20212	97	9	the	the	DET
ajst-20212	97	10	other	other	ADJ
ajst-20212	97	11	hand	hand	NOUN
ajst-20212	97	12	,	,	PUNCT
ajst-20212	97	13	is	be	AUX
ajst-20212	97	14	optimal	optimal	ADJ
ajst-20212	97	15	in	in	ADP
ajst-20212	97	16	all	all	DET
ajst-20212	97	17	metrics	metric	NOUN
ajst-20212	97	18	,	,	PUNCT
ajst-20212	97	19	indicating	indicate	VERB
ajst-20212	97	20	that	that	SCONJ
ajst-20212	97	21	the	the	DET
ajst-20212	97	22	model	model	NOUN
ajst-20212	97	23	used	use	VERB
ajst-20212	97	24	in	in	ADP
ajst-20212	97	25	this	this	DET
ajst-20212	97	26	chapter	chapter	NOUN
ajst-20212	97	27	works	work	VERB
ajst-20212	97	28	better	well	ADV
ajst-20212	97	29	in	in	ADP
ajst-20212	97	30	the	the	DET
ajst-20212	97	31	attribute	attribute	NOUN
ajst-20212	97	32	-	-	PUNCT
ajst-20212	97	33	level	level	NOUN
ajst-20212	97	34	sentiment	sentiment	NOUN
ajst-20212	97	35	analysis	analysis	NOUN
ajst-20212	97	36	experiments	experiment	NOUN
ajst-20212	97	37	.	.	PUNCT
ajst-20212	98	1	3.5	3.5	NUM
ajst-20212	98	2	.	.	PUNCT
ajst-20212	98	3	ablation	ablation	NOUN
ajst-20212	98	4	experiment	experiment	NOUN
ajst-20212	98	5	in	in	ADP
ajst-20212	98	6	order	order	NOUN
ajst-20212	98	7	to	to	PART
ajst-20212	98	8	better	well	ADV
ajst-20212	98	9	study	study	VERB
ajst-20212	98	10	the	the	DET
ajst-20212	98	11	effectiveness	effectiveness	NOUN
ajst-20212	98	12	of	of	ADP
ajst-20212	98	13	the	the	DET
ajst-20212	98	14	rbt	rbt	NOUN
ajst-20212	98	15	-	-	PUNCT
ajst-20212	98	16	biattgcn	biattgcn	NOUN
ajst-20212	98	17	(	(	PUNCT
ajst-20212	98	18	rbg	rbg	NOUN
ajst-20212	98	19	)	)	PUNCT
ajst-20212	98	20	model	model	NOUN
ajst-20212	98	21	,	,	PUNCT
ajst-20212	98	22	this	this	DET
ajst-20212	98	23	paper	paper	NOUN
ajst-20212	98	24	designs	design	VERB
ajst-20212	98	25	relevant	relevant	ADJ
ajst-20212	98	26	ablation	ablation	NOUN
ajst-20212	98	27	experiments	experiment	NOUN
ajst-20212	98	28	to	to	PART
ajst-20212	98	29	remove	remove	VERB
ajst-20212	98	30	the	the	DET
ajst-20212	98	31	lexical	lexical	ADJ
ajst-20212	98	32	information	information	NOUN
ajst-20212	98	33	,	,	PUNCT
ajst-20212	98	34	positional	positional	ADJ
ajst-20212	98	35	coding	code	VERB
ajst-20212	98	36	information	information	NOUN
ajst-20212	98	37	,	,	PUNCT
ajst-20212	98	38	and	and	CCONJ
ajst-20212	98	39	gcn	gcn	NOUN
ajst-20212	98	40	layer	layer	NOUN
ajst-20212	98	41	in	in	ADP
ajst-20212	98	42	the	the	DET
ajst-20212	98	43	model	model	NOUN
ajst-20212	98	44	while	while	SCONJ
ajst-20212	98	45	keeping	keep	VERB
ajst-20212	98	46	the	the	DET
ajst-20212	98	47	rest	rest	NOUN
ajst-20212	98	48	of	of	ADP
ajst-20212	98	49	the	the	DET
ajst-20212	98	50	model	model	NOUN
ajst-20212	98	51	unchanged	unchanged	ADJ
ajst-20212	98	52	,	,	PUNCT
ajst-20212	98	53	respectively	respectively	ADV
ajst-20212	98	54	.	.	PUNCT
ajst-20212	99	1	they	they	PRON
ajst-20212	99	2	are	be	AUX
ajst-20212	99	3	denoted	denote	VERB
ajst-20212	99	4	as	as	ADP
ajst-20212	99	5	rbg	rbg	NOUN
ajst-20212	99	6	/	/	SYM
ajst-20212	99	7	pos	pos	NOUN
ajst-20212	99	8	-	-	PUNCT
ajst-20212	99	9	tag	tag	NOUN
ajst-20212	99	10	,	,	PUNCT
ajst-20212	99	11	rbg	rbg	NOUN
ajst-20212	99	12	/	/	SYM
ajst-20212	99	13	postion	postion	NOUN
ajst-20212	99	14	,	,	PUNCT
ajst-20212	99	15	and	and	CCONJ
ajst-20212	99	16	rbg	rbg	NOUN
ajst-20212	99	17	/	/	SYM
ajst-20212	99	18	gcn	gcn	NOUN
ajst-20212	99	19	,	,	PUNCT
ajst-20212	99	20	respectively	respectively	ADV
ajst-20212	99	21	.	.	PUNCT
ajst-20212	100	1	the	the	DET
ajst-20212	100	2	experimental	experimental	ADJ
ajst-20212	100	3	results	result	NOUN
ajst-20212	100	4	are	be	AUX
ajst-20212	100	5	shown	show	VERB
ajst-20212	100	6	in	in	ADP
ajst-20212	100	7	table	table	NOUN
ajst-20212	100	8	2	2	NUM
ajst-20212	100	9	.	.	PUNCT
ajst-20212	100	10	table	table	NOUN
ajst-20212	100	11	2	2	NUM
ajst-20212	100	12	.	.	PUNCT
ajst-20212	100	13	model	model	NOUN
ajst-20212	100	14	ablation	ablation	NOUN
ajst-20212	100	15	experiments	experiment	NOUN
ajst-20212	100	16	model	model	VERB
ajst-20212	100	17	marco_acc	marco_acc	PROPN
ajst-20212	100	18	(	(	PUNCT
ajst-20212	100	19	%	%	INTJ
ajst-20212	100	20	)	)	PUNCT
ajst-20212	100	21	marco_f1	marco_f1	PROPN
ajst-20212	100	22	(	(	PUNCT
ajst-20212	100	23	%	%	INTJ
ajst-20212	100	24	)	)	PUNCT
ajst-20212	100	25	rbg	rbg	NOUN
ajst-20212	100	26	/	/	SYM
ajst-20212	100	27	pos	pos	NOUN
ajst-20212	100	28	-	-	PUNCT
ajst-20212	100	29	tag	tag	NOUN
ajst-20212	100	30	80.07	80.07	NUM
ajst-20212	100	31	74.10	74.10	NUM
ajst-20212	100	32	rbg	rbg	NOUN
ajst-20212	100	33	/	/	SYM
ajst-20212	100	34	postion	postion	NOUN
ajst-20212	100	35	79.08	79.08	NUM
ajst-20212	100	36	73.24	73.24	NUM
ajst-20212	100	37	rbg	rbg	NOUN
ajst-20212	100	38	/	/	SYM
ajst-20212	100	39	gcn	gcn	NOUN
ajst-20212	100	40	77.62	77.62	NUM
ajst-20212	100	41	72.51	72.51	NUM
ajst-20212	100	42	rbg	rbg	NOUN
ajst-20212	100	43	80.23	80.23	NUM
ajst-20212	100	44	78.21	78.21	NUM
ajst-20212	100	45	as	as	SCONJ
ajst-20212	100	46	can	can	AUX
ajst-20212	100	47	be	be	AUX
ajst-20212	100	48	seen	see	VERB
ajst-20212	100	49	from	from	ADP
ajst-20212	100	50	table	table	NOUN
ajst-20212	100	51	2	2	NUM
ajst-20212	100	52	,	,	PUNCT
ajst-20212	100	53	the	the	DET
ajst-20212	100	54	results	result	NOUN
ajst-20212	100	55	of	of	ADP
ajst-20212	100	56	all	all	DET
ajst-20212	100	57	three	three	NUM
ajst-20212	100	58	sets	set	NOUN
ajst-20212	100	59	of	of	ADP
ajst-20212	100	60	ablation	ablation	NOUN
ajst-20212	100	61	experiments	experiment	NOUN
ajst-20212	100	62	have	have	AUX
ajst-20212	100	63	decreased	decrease	VERB
ajst-20212	100	64	compared	compare	VERB
ajst-20212	100	65	to	to	ADP
ajst-20212	100	66	the	the	DET
ajst-20212	100	67	final	final	ADJ
ajst-20212	100	68	model	model	NOUN
ajst-20212	100	69	.	.	PUNCT
ajst-20212	101	1	after	after	ADP
ajst-20212	101	2	removing	remove	VERB
ajst-20212	101	3	the	the	DET
ajst-20212	101	4	lexical	lexical	ADJ
ajst-20212	101	5	labeling	labeling	NOUN
ajst-20212	101	6	information	information	NOUN
ajst-20212	101	7	,	,	PUNCT
ajst-20212	101	8	the	the	DET
ajst-20212	101	9	metrics	metric	NOUN
ajst-20212	101	10	of	of	ADP
ajst-20212	101	11	the	the	DET
ajst-20212	101	12	model	model	NOUN
ajst-20212	101	13	have	have	AUX
ajst-20212	101	14	decreased	decrease	VERB
ajst-20212	101	15	,	,	PUNCT
ajst-20212	101	16	indicating	indicate	VERB
ajst-20212	101	17	that	that	SCONJ
ajst-20212	101	18	the	the	DET
ajst-20212	101	19	lexical	lexical	ADJ
ajst-20212	101	20	labeling	labeling	NOUN
ajst-20212	101	21	information	information	NOUN
ajst-20212	101	22	can	can	AUX
ajst-20212	101	23	play	play	VERB
ajst-20212	101	24	a	a	DET
ajst-20212	101	25	role	role	NOUN
ajst-20212	101	26	in	in	ADP
ajst-20212	101	27	attribute	attribute	NOUN
ajst-20212	101	28	-	-	PUNCT
ajst-20212	101	29	level	level	NOUN
ajst-20212	101	30	sentiment	sentiment	NOUN
ajst-20212	101	31	analysis	analysis	NOUN
ajst-20212	101	32	,	,	PUNCT
ajst-20212	101	33	and	and	CCONJ
ajst-20212	101	34	the	the	DET
ajst-20212	101	35	metrics	metric	NOUN
ajst-20212	101	36	have	have	AUX
ajst-20212	101	37	decreased	decrease	VERB
ajst-20212	101	38	more	more	ADV
ajst-20212	101	39	after	after	ADP
ajst-20212	101	40	removing	remove	VERB
ajst-20212	101	41	the	the	DET
ajst-20212	101	42	positional	positional	ADJ
ajst-20212	101	43	coding	code	VERB
ajst-20212	101	44	information	information	NOUN
ajst-20212	101	45	,	,	PUNCT
ajst-20212	101	46	indicating	indicate	VERB
ajst-20212	101	47	that	that	SCONJ
ajst-20212	101	48	the	the	DET
ajst-20212	101	49	positional	positional	ADJ
ajst-20212	101	50	coding	code	VERB
ajst-20212	101	51	information	information	NOUN
ajst-20212	101	52	plays	play	VERB
ajst-20212	101	53	a	a	DET
ajst-20212	101	54	greater	great	ADJ
ajst-20212	101	55	role	role	NOUN
ajst-20212	101	56	compared	compare	VERB
ajst-20212	101	57	to	to	ADP
ajst-20212	101	58	the	the	DET
ajst-20212	101	59	lexical	lexical	ADJ
ajst-20212	101	60	labeling	labeling	NOUN
ajst-20212	101	61	information	information	NOUN
ajst-20212	101	62	.	.	PUNCT
ajst-20212	102	1	after	after	ADP
ajst-20212	102	2	removing	remove	VERB
ajst-20212	102	3	the	the	DET
ajst-20212	102	4	gcn	gcn	NOUN
ajst-20212	102	5	,	,	PUNCT
ajst-20212	102	6	the	the	DET
ajst-20212	102	7	model	model	NOUN
ajst-20212	102	8	is	be	AUX
ajst-20212	102	9	the	the	DET
ajst-20212	102	10	least	least	ADV
ajst-20212	102	11	effective	effective	ADJ
ajst-20212	102	12	,	,	PUNCT
ajst-20212	102	13	indicating	indicate	VERB
ajst-20212	102	14	that	that	SCONJ
ajst-20212	102	15	the	the	DET
ajst-20212	102	16	gcn	gcn	NOUN
ajst-20212	102	17	has	have	VERB
ajst-20212	102	18	a	a	DET
ajst-20212	102	19	pivotal	pivotal	ADJ
ajst-20212	102	20	position	position	NOUN
ajst-20212	102	21	among	among	ADP
ajst-20212	102	22	the	the	DET
ajst-20212	102	23	three	three	NUM
ajst-20212	102	24	modules	module	NOUN
ajst-20212	102	25	.	.	PUNCT
ajst-20212	103	1	171	171	NUM
ajst-20212	103	2	3.6	3.6	NUM
ajst-20212	103	3	.	.	PUNCT
ajst-20212	104	1	gcn	gcn	NOUN
ajst-20212	104	2	layer	layer	NOUN
ajst-20212	104	3	comparison	comparison	NOUN
ajst-20212	104	4	experiment	experiment	NOUN
ajst-20212	104	5	based	base	VERB
ajst-20212	104	6	on	on	ADP
ajst-20212	104	7	the	the	DET
ajst-20212	104	8	results	result	NOUN
ajst-20212	104	9	presented	present	VERB
ajst-20212	104	10	in	in	ADP
ajst-20212	104	11	figures	figure	NOUN
ajst-20212	104	12	6	6	NUM
ajst-20212	104	13	,	,	PUNCT
ajst-20212	104	14	experiments	experiment	NOUN
ajst-20212	104	15	on	on	ADP
ajst-20212	104	16	the	the	DET
ajst-20212	104	17	dataset	dataset	NOUN
ajst-20212	104	18	show	show	NOUN
ajst-20212	104	19	that	that	SCONJ
ajst-20212	104	20	the	the	DET
ajst-20212	104	21	model	model	NOUN
ajst-20212	104	22	performs	perform	VERB
ajst-20212	104	23	best	good	ADJ
ajst-20212	104	24	when	when	SCONJ
ajst-20212	104	25	the	the	DET
ajst-20212	104	26	number	number	NOUN
ajst-20212	104	27	of	of	ADP
ajst-20212	104	28	layers	layer	NOUN
ajst-20212	104	29	of	of	ADP
ajst-20212	104	30	the	the	DET
ajst-20212	104	31	gcn	gcn	NOUN
ajst-20212	104	32	is	be	AUX
ajst-20212	104	33	2	2	NUM
ajst-20212	104	34	.	.	PUNCT
ajst-20212	104	35	with	with	ADP
ajst-20212	104	36	a	a	DET
ajst-20212	104	37	gradual	gradual	ADJ
ajst-20212	104	38	increase	increase	NOUN
ajst-20212	104	39	in	in	ADP
ajst-20212	104	40	the	the	DET
ajst-20212	104	41	number	number	NOUN
ajst-20212	104	42	of	of	ADP
ajst-20212	104	43	layers	layer	NOUN
ajst-20212	104	44	,	,	PUNCT
ajst-20212	104	45	the	the	DET
ajst-20212	104	46	number	number	NOUN
ajst-20212	104	47	of	of	ADP
ajst-20212	104	48	model	model	NOUN
ajst-20212	104	49	training	training	NOUN
ajst-20212	104	50	parameters	parameter	NOUN
ajst-20212	104	51	increases	increase	VERB
ajst-20212	104	52	leading	lead	VERB
ajst-20212	104	53	to	to	ADP
ajst-20212	104	54	overfitting	overfitte	VERB
ajst-20212	104	55	and	and	CCONJ
ajst-20212	104	56	thus	thus	ADV
ajst-20212	104	57	a	a	DET
ajst-20212	104	58	decrease	decrease	NOUN
ajst-20212	104	59	in	in	ADP
ajst-20212	104	60	performance	performance	NOUN
ajst-20212	104	61	on	on	ADP
ajst-20212	104	62	the	the	DET
ajst-20212	104	63	dataset	dataset	NOUN
ajst-20212	104	64	.	.	PUNCT
ajst-20212	105	1	figure	figure	NOUN
ajst-20212	105	2	5	5	NUM
ajst-20212	105	3	.	.	PUNCT
ajst-20212	105	4	gcn	gcn	NOUN
ajst-20212	105	5	layer	layer	NOUN
ajst-20212	105	6	comparison	comparison	NOUN
ajst-20212	105	7	4	4	NUM
ajst-20212	105	8	.	.	PUNCT
ajst-20212	106	1	conclusion	conclusion	NOUN
ajst-20212	106	2	in	in	ADP
ajst-20212	106	3	this	this	DET
ajst-20212	106	4	paper	paper	NOUN
ajst-20212	106	5	,	,	PUNCT
ajst-20212	106	6	a	a	DET
ajst-20212	106	7	rbt	rbt	PROPN
ajst-20212	106	8	-	-	PUNCT
ajst-20212	106	9	biatt	biatt	NOUN
ajst-20212	106	10	-	-	PUNCT
ajst-20212	106	11	gcn	gcn	NOUN
ajst-20212	106	12	attribute	attribute	NOUN
ajst-20212	106	13	-	-	PUNCT
ajst-20212	106	14	level	level	NOUN
ajst-20212	106	15	sentiment	sentiment	NOUN
ajst-20212	106	16	analysis	analysis	NOUN
ajst-20212	106	17	research	research	NOUN
ajst-20212	106	18	model	model	NOUN
ajst-20212	106	19	with	with	ADP
ajst-20212	106	20	fusion	fusion	NOUN
ajst-20212	106	21	of	of	ADP
ajst-20212	106	22	auxiliary	auxiliary	ADJ
ajst-20212	106	23	information	information	NOUN
ajst-20212	106	24	is	be	AUX
ajst-20212	106	25	proposed	propose	VERB
ajst-20212	106	26	,	,	PUNCT
ajst-20212	106	27	in	in	ADP
ajst-20212	106	28	order	order	NOUN
ajst-20212	106	29	to	to	PART
ajst-20212	106	30	ensure	ensure	VERB
ajst-20212	106	31	the	the	DET
ajst-20212	106	32	reliability	reliability	NOUN
ajst-20212	106	33	of	of	ADP
ajst-20212	106	34	the	the	DET
ajst-20212	106	35	experiments	experiment	NOUN
ajst-20212	106	36	,	,	PUNCT
ajst-20212	106	37	firstly	firstly	ADV
ajst-20212	106	38	,	,	PUNCT
ajst-20212	106	39	the	the	DET
ajst-20212	106	40	experiments	experiment	NOUN
ajst-20212	106	41	are	be	AUX
ajst-20212	106	42	carried	carry	VERB
ajst-20212	106	43	out	out	ADP
ajst-20212	106	44	by	by	ADP
ajst-20212	106	45	using	use	VERB
ajst-20212	106	46	the	the	DET
ajst-20212	106	47	public	public	ADJ
ajst-20212	106	48	dataset	dataset	NOUN
ajst-20212	106	49	,	,	PUNCT
ajst-20212	106	50	the	the	DET
ajst-20212	106	51	model	model	NOUN
ajst-20212	106	52	is	be	AUX
ajst-20212	106	53	compared	compare	VERB
ajst-20212	106	54	with	with	ADP
ajst-20212	106	55	other	other	ADJ
ajst-20212	106	56	models	model	NOUN
ajst-20212	106	57	,	,	PUNCT
ajst-20212	106	58	and	and	CCONJ
ajst-20212	106	59	then	then	ADV
ajst-20212	106	60	the	the	DET
ajst-20212	106	61	ablation	ablation	NOUN
ajst-20212	106	62	experiments	experiment	NOUN
ajst-20212	106	63	of	of	ADP
ajst-20212	106	64	the	the	DET
ajst-20212	106	65	model	model	NOUN
ajst-20212	106	66	are	be	AUX
ajst-20212	106	67	designed	design	VERB
ajst-20212	106	68	,	,	PUNCT
ajst-20212	106	69	which	which	PRON
ajst-20212	106	70	proves	prove	VERB
ajst-20212	106	71	that	that	SCONJ
ajst-20212	106	72	the	the	DET
ajst-20212	106	73	model	model	NOUN
ajst-20212	106	74	can	can	AUX
ajst-20212	106	75	achieve	achieve	VERB
ajst-20212	106	76	better	well	ADJ
ajst-20212	106	77	results	result	NOUN
ajst-20212	106	78	in	in	ADP
ajst-20212	106	79	attribute	attribute	NOUN
ajst-20212	106	80	-	-	PUNCT
ajst-20212	106	81	level	level	NOUN
ajst-20212	106	82	sentiment	sentiment	NOUN
ajst-20212	106	83	analysis	analysis	NOUN
ajst-20212	106	84	research	research	NOUN
ajst-20212	106	85	,	,	PUNCT
ajst-20212	106	86	and	and	CCONJ
ajst-20212	106	87	proves	prove	VERB
ajst-20212	106	88	that	that	SCONJ
ajst-20212	106	89	the	the	DET
ajst-20212	106	90	gcn	gcn	NOUN
ajst-20212	106	91	and	and	CCONJ
ajst-20212	106	92	the	the	DET
ajst-20212	106	93	auxiliary	auxiliary	ADJ
ajst-20212	106	94	information	information	NOUN
ajst-20212	106	95	have	have	VERB
ajst-20212	106	96	the	the	DET
ajst-20212	106	97	effect	effect	NOUN
ajst-20212	106	98	of	of	ADP
ajst-20212	106	99	improving	improve	VERB
ajst-20212	106	100	the	the	DET
ajst-20212	106	101	attribute	attribute	NOUN
ajst-20212	106	102	-	-	PUNCT
ajst-20212	106	103	level	level	NOUN
ajst-20212	106	104	sentiment	sentiment	NOUN
ajst-20212	106	105	analysis	analysis	NOUN
ajst-20212	106	106	.	.	PUNCT
ajst-20212	107	1	references	reference	NOUN
ajst-20212	107	2	[	[	X
ajst-20212	107	3	1	1	NUM
ajst-20212	107	4	]	]	X
ajst-20212	107	5	fu	fu	NOUN
ajst-20212	107	6	x	x	NOUN
ajst-20212	107	7	,	,	PUNCT
ajst-20212	107	8	liu	liu	PROPN
ajst-20212	107	9	g	g	PROPN
ajst-20212	107	10	,	,	PUNCT
ajst-20212	107	11	guo	guo	PROPN
ajst-20212	107	12	y	y	PROPN
ajst-20212	107	13	,	,	PUNCT
ajst-20212	107	14	et	et	PROPN
ajst-20212	107	15	al	al	PROPN
ajst-20212	107	16	.	.	PUNCT
ajst-20212	108	1	multi	multi	ADJ
ajst-20212	108	2	-	-	ADJ
ajst-20212	108	3	aspect	aspect	ADJ
ajst-20212	108	4	blog	blog	NOUN
ajst-20212	108	5	sentiment	sentiment	NOUN
ajst-20212	108	6	analysis	analysis	NOUN
ajst-20212	108	7	based	base	VERB
ajst-20212	108	8	on	on	ADP
ajst-20212	108	9	lda	lda	PROPN
ajst-20212	108	10	topic	topic	NOUN
ajst-20212	108	11	model	model	NOUN
ajst-20212	108	12	and	and	CCONJ
ajst-20212	108	13	hownet	hownet	PROPN
ajst-20212	108	14	lexicon	lexicon	NOUN
ajst-20212	108	15	;	;	PUNCT
ajst-20212	108	16	proceedings	proceeding	NOUN
ajst-20212	108	17	of	of	ADP
ajst-20212	108	18	the	the	DET
ajst-20212	108	19	international	international	ADJ
ajst-20212	108	20	conference	conference	NOUN
ajst-20212	108	21	on	on	ADP
ajst-20212	108	22	web	web	NOUN
ajst-20212	108	23	information	information	NOUN
ajst-20212	108	24	systems	system	NOUN
ajst-20212	108	25	and	and	CCONJ
ajst-20212	108	26	mining	mining	NOUN
ajst-20212	108	27	,	,	PUNCT
ajst-20212	108	28	f	f	X
ajst-20212	108	29	,	,	PUNCT
ajst-20212	108	30	2011	2011	NUM
ajst-20212	109	1	[	[	X
ajst-20212	109	2	c	c	X
ajst-20212	109	3	]	]	PUNCT
ajst-20212	109	4	.	.	PUNCT
ajst-20212	109	5	springer	springer	NOUN
ajst-20212	109	6	.	.	PUNCT
ajst-20212	110	1	[	[	X
ajst-20212	110	2	2	2	NUM
ajst-20212	110	3	]	]	X
ajst-20212	110	4	akhtar	akhtar	PROPN
ajst-20212	110	5	n	n	PROPN
ajst-20212	110	6	,	,	PUNCT
ajst-20212	110	7	zubair	zubair	PROPN
ajst-20212	110	8	n	n	CCONJ
ajst-20212	110	9	,	,	PUNCT
ajst-20212	110	10	kumar	kumar	PROPN
ajst-20212	110	11	a	a	PROPN
ajst-20212	110	12	,	,	PUNCT
ajst-20212	110	13	et	et	PROPN
ajst-20212	110	14	al	al	PROPN
ajst-20212	110	15	.	.	PROPN
ajst-20212	110	16	aspect	aspect	PROPN
ajst-20212	110	17	based	base	VERB
ajst-20212	110	18	sentiment	sentiment	NOUN
ajst-20212	110	19	oriented	orient	VERB
ajst-20212	110	20	summarization	summarization	NOUN
ajst-20212	110	21	of	of	ADP
ajst-20212	110	22	hotel	hotel	NOUN
ajst-20212	110	23	reviews	review	NOUN
ajst-20212	110	24	[	[	X
ajst-20212	110	25	j	j	X
ajst-20212	110	26	]	]	X
ajst-20212	110	27	.	.	PUNCT
ajst-20212	111	1	procedia	procedia	PROPN
ajst-20212	111	2	computer	computer	NOUN
ajst-20212	111	3	science	science	NOUN
ajst-20212	111	4	,	,	PUNCT
ajst-20212	111	5	2017	2017	NUM
ajst-20212	111	6	,	,	PUNCT
ajst-20212	111	7	115	115	NUM
ajst-20212	111	8	:	:	PUNCT
ajst-20212	111	9	563	563	NUM
ajst-20212	111	10	-	-	SYM
ajst-20212	111	11	71	71	NUM
ajst-20212	111	12	.	.	PUNCT
ajst-20212	112	1	[	[	X
ajst-20212	112	2	3	3	X
ajst-20212	112	3	]	]	X
ajst-20212	112	4	luo	luo	PROPN
ajst-20212	112	5	h	h	PROPN
ajst-20212	112	6	,	,	PUNCT
ajst-20212	112	7	li	li	PROPN
ajst-20212	112	8	t	t	PROPN
ajst-20212	112	9	,	,	PUNCT
ajst-20212	112	10	liu	liu	PROPN
ajst-20212	112	11	b	b	PROPN
ajst-20212	112	12	,	,	PUNCT
ajst-20212	112	13	et	et	PROPN
ajst-20212	112	14	al	al	PROPN
ajst-20212	112	15	.	.	PUNCT
ajst-20212	112	16	improving	improve	VERB
ajst-20212	112	17	aspect	aspect	NOUN
ajst-20212	112	18	term	term	NOUN
ajst-20212	112	19	extraction	extraction	NOUN
ajst-20212	112	20	with	with	ADP
ajst-20212	112	21	bidirectional	bidirectional	ADJ
ajst-20212	112	22	dependency	dependency	NOUN
ajst-20212	112	23	tree	tree	NOUN
ajst-20212	112	24	representation	representation	NOUN
ajst-20212	112	25	[	[	X
ajst-20212	112	26	j	j	X
ajst-20212	112	27	]	]	X
ajst-20212	112	28	.	.	PUNCT
ajst-20212	113	1	ieee	ieee	PROPN
ajst-20212	113	2	/	/	SYM
ajst-20212	113	3	acm	acm	PROPN
ajst-20212	113	4	transactions	transaction	NOUN
ajst-20212	113	5	on	on	ADP
ajst-20212	113	6	audio	audio	NOUN
ajst-20212	113	7	,	,	PUNCT
ajst-20212	113	8	speech	speech	NOUN
ajst-20212	113	9	,	,	PUNCT
ajst-20212	113	10	and	and	CCONJ
ajst-20212	113	11	language	language	NOUN
ajst-20212	113	12	processing	processing	NOUN
ajst-20212	113	13	,	,	PUNCT
ajst-20212	113	14	2019	2019	NUM
ajst-20212	113	15	,	,	PUNCT
ajst-20212	113	16	27(7	27(7	NUM
ajst-20212	113	17	):	):	PUNCT
ajst-20212	113	18	1201	1201	NUM
ajst-20212	113	19	-	-	SYM
ajst-20212	113	20	12	12	NUM
ajst-20212	113	21	.	.	PUNCT
ajst-20212	114	1	[	[	X
ajst-20212	114	2	4	4	X
ajst-20212	114	3	]	]	PUNCT
ajst-20212	114	4	rubtsova	rubtsova	X
ajst-20212	114	5	y	y	PROPN
ajst-20212	114	6	,	,	PUNCT
ajst-20212	114	7	koshelnikov	koshelnikov	PROPN
ajst-20212	114	8	s.	s.	PROPN
ajst-20212	114	9	aspect	aspect	PROPN
ajst-20212	114	10	extraction	extraction	NOUN
ajst-20212	114	11	from	from	ADP
ajst-20212	114	12	reviews	review	NOUN
ajst-20212	114	13	using	use	VERB
ajst-20212	114	14	conditional	conditional	ADJ
ajst-20212	114	15	random	random	ADJ
ajst-20212	114	16	fields	field	NOUN
ajst-20212	114	17	;	;	PUNCT
ajst-20212	114	18	proceedings	proceeding	NOUN
ajst-20212	114	19	of	of	ADP
ajst-20212	114	20	the	the	DET
ajst-20212	114	21	international	international	ADJ
ajst-20212	114	22	conference	conference	NOUN
ajst-20212	114	23	on	on	ADP
ajst-20212	114	24	knowledge	knowledge	NOUN
ajst-20212	114	25	engineering	engineering	NOUN
ajst-20212	114	26	and	and	CCONJ
ajst-20212	114	27	the	the	DET
ajst-20212	114	28	semantic	semantic	ADJ
ajst-20212	114	29	web	web	NOUN
ajst-20212	114	30	,	,	PUNCT
ajst-20212	114	31	f	f	PROPN
ajst-20212	114	32	,	,	PUNCT
ajst-20212	114	33	2015	2015	NUM
ajst-20212	115	1	[	[	X
ajst-20212	115	2	c	c	X
ajst-20212	115	3	]	]	PUNCT
ajst-20212	115	4	.	.	PUNCT
ajst-20212	116	1	springer	springer	NOUN
ajst-20212	116	2	.	.	PUNCT
ajst-20212	117	1	[	[	X
ajst-20212	117	2	5	5	NUM
ajst-20212	117	3	]	]	PUNCT
ajst-20212	117	4	xu	xu	PROPN
ajst-20212	117	5	h	h	PROPN
ajst-20212	117	6	,	,	PUNCT
ajst-20212	117	7	liu	liu	PROPN
ajst-20212	117	8	b	b	PROPN
ajst-20212	117	9	,	,	PUNCT
ajst-20212	117	10	shu	shu	PROPN
ajst-20212	117	11	l	l	PROPN
ajst-20212	117	12	,	,	PUNCT
ajst-20212	117	13	et	et	PROPN
ajst-20212	117	14	al	al	PROPN
ajst-20212	117	15	.	.	PROPN
ajst-20212	117	16	double	double	ADJ
ajst-20212	117	17	embeddings	embedding	NOUN
ajst-20212	117	18	and	and	CCONJ
ajst-20212	117	19	cnn	cnn	PROPN
ajst-20212	117	20	-	-	PUNCT
ajst-20212	117	21	based	base	VERB
ajst-20212	117	22	sequence	sequence	NOUN
ajst-20212	117	23	labeling	labeling	NOUN
ajst-20212	117	24	for	for	ADP
ajst-20212	117	25	aspect	aspect	NOUN
ajst-20212	117	26	extraction	extraction	NOUN
ajst-20212	117	27	[	[	X
ajst-20212	117	28	j	j	X
ajst-20212	117	29	]	]	X
ajst-20212	117	30	.	.	PUNCT
ajst-20212	118	1	ar	ar	PROPN
ajst-20212	118	2	xiv	xiv	PROPN
ajst-20212	118	3	preprint	preprint	PROPN
ajst-20212	118	4	arxiv:180504601	arxiv:180504601	ADP
ajst-20212	118	5	,	,	PUNCT
ajst-20212	118	6	2018	2018	NUM
ajst-20212	118	7	.	.	PUNCT
ajst-20212	119	1	[	[	X
ajst-20212	119	2	6	6	NUM
ajst-20212	119	3	]	]	X
ajst-20212	119	4	chen	chen	PROPN
ajst-20212	119	5	p	p	X
ajst-20212	119	6	,	,	PUNCT
ajst-20212	119	7	sun	sun	PROPN
ajst-20212	119	8	z	z	PROPN
ajst-20212	119	9	,	,	PUNCT
ajst-20212	119	10	bing	bing	VERB
ajst-20212	119	11	l	l	NOUN
ajst-20212	119	12	,	,	PUNCT
ajst-20212	119	13	et	et	PROPN
ajst-20212	119	14	al	al	PROPN
ajst-20212	119	15	.	.	PROPN
ajst-20212	119	16	recurrent	recurrent	ADJ
ajst-20212	119	17	attention	attention	NOUN
ajst-20212	119	18	network	network	NOUN
ajst-20212	119	19	on	on	ADP
ajst-20212	119	20	memory	memory	NOUN
ajst-20212	119	21	for	for	ADP
ajst-20212	119	22	aspect	aspect	NOUN
ajst-20212	119	23	sentiment	sentiment	NOUN
ajst-20212	119	24	analysis	analysis	NOUN
ajst-20212	119	25	;	;	PUNCT
ajst-20212	119	26	proceedings	proceeding	NOUN
ajst-20212	119	27	of	of	ADP
ajst-20212	119	28	the	the	DET
ajst-20212	119	29	proceedings	proceeding	NOUN
ajst-20212	119	30	of	of	ADP
ajst-20212	119	31	the	the	DET
ajst-20212	119	32	2017	2017	NUM
ajst-20212	119	33	conference	conference	NOUN
ajst-20212	119	34	on	on	ADP
ajst-20212	119	35	empirical	empirical	ADJ
ajst-20212	119	36	methods	method	NOUN
ajst-20212	119	37	in	in	ADP
ajst-20212	119	38	natural	natural	ADJ
ajst-20212	119	39	language	language	NOUN
ajst-20212	119	40	processing	processing	NOUN
ajst-20212	119	41	,	,	PUNCT
ajst-20212	119	42	f	f	PROPN
ajst-20212	119	43	,	,	PUNCT
ajst-20212	119	44	2017	2017	NUM
ajst-20212	119	45	[	[	X
ajst-20212	119	46	c	c	X
ajst-20212	119	47	]	]	PUNCT
ajst-20212	119	48	.	.	PUNCT
ajst-20212	120	1	[	[	X
ajst-20212	120	2	7	7	NUM
ajst-20212	120	3	]	]	X
ajst-20212	120	4	phan	phan	PROPN
ajst-20212	120	5	h	h	PROPN
ajst-20212	120	6	t	t	PROPN
ajst-20212	120	7	,	,	PUNCT
ajst-20212	120	8	nguyen	nguyen	PROPN
ajst-20212	120	9	n	n	PROPN
ajst-20212	120	10	t	t	PROPN
ajst-20212	120	11	,	,	PUNCT
ajst-20212	120	12	hwang	hwang	PROPN
ajst-20212	120	13	d.	d.	PROPN
ajst-20212	120	14	convolutional	convolutional	ADJ
ajst-20212	120	15	attention	attention	NOUN
ajst-20212	120	16	neural	neural	ADJ
ajst-20212	120	17	network	network	NOUN
ajst-20212	120	18	over	over	ADP
ajst-20212	120	19	graph	graph	NOUN
ajst-20212	120	20	structures	structure	NOUN
ajst-20212	120	21	for	for	ADP
ajst-20212	120	22	improving	improve	VERB
ajst-20212	120	23	the	the	DET
ajst-20212	120	24	performance	performance	NOUN
ajst-20212	120	25	of	of	ADP
ajst-20212	120	26	aspect	aspect	NOUN
ajst-20212	120	27	-	-	PUNCT
ajst-20212	120	28	level	level	NOUN
ajst-20212	120	29	sentiment	sentiment	NOUN
ajst-20212	120	30	analysis	analysis	NOUN
ajst-20212	120	31	[	[	X
ajst-20212	120	32	j	j	X
ajst-20212	120	33	]	]	X
ajst-20212	120	34	.	.	PUNCT
ajst-20212	121	1	information	information	NOUN
ajst-20212	121	2	sciences	sciences	PROPN
ajst-20212	121	3	,	,	PUNCT
ajst-20212	121	4	2022	2022	NUM
ajst-20212	121	5	.	.	PUNCT
ajst-20212	122	1	[	[	X
ajst-20212	122	2	8	8	NUM
ajst-20212	122	3	]	]	X
ajst-20212	122	4	li	li	PROPN
ajst-20212	122	5	r	r	PROPN
ajst-20212	122	6	,	,	PUNCT
ajst-20212	122	7	chen	chen	PROPN
ajst-20212	122	8	h	h	PROPN
ajst-20212	122	9	,	,	PUNCT
ajst-20212	122	10	feng	feng	PROPN
ajst-20212	122	11	f	f	AUX
ajst-20212	122	12	,	,	PUNCT
ajst-20212	122	13	et	et	PROPN
ajst-20212	122	14	al	al	PROPN
ajst-20212	122	15	.	.	PUNCT
ajst-20212	123	1	dual	dual	ADJ
ajst-20212	123	2	graph	graph	NOUN
ajst-20212	123	3	convolutional	convolutional	ADJ
ajst-20212	123	4	networks	network	NOUN
ajst-20212	123	5	for	for	ADP
ajst-20212	123	6	aspect	aspect	NOUN
ajst-20212	123	7	-	-	PUNCT
ajst-20212	123	8	based	base	VERB
ajst-20212	123	9	sentiment	sentiment	NOUN
ajst-20212	123	10	analysis	analysis	NOUN
ajst-20212	123	11	;	;	PUNCT
ajst-20212	123	12	proceedings	proceeding	NOUN
ajst-20212	123	13	of	of	ADP
ajst-20212	123	14	the	the	DET
ajst-20212	123	15	proceedings	proceeding	NOUN
ajst-20212	123	16	of	of	ADP
ajst-20212	123	17	the	the	DET
ajst-20212	123	18	59th	59th	ADJ
ajst-20212	123	19	annual	annual	ADJ
ajst-20212	123	20	meeting	meeting	NOUN
ajst-20212	123	21	of	of	ADP
ajst-20212	123	22	the	the	DET
ajst-20212	123	23	association	association	NOUN
ajst-20212	123	24	for	for	ADP
ajst-20212	123	25	computational	computational	ADJ
ajst-20212	123	26	linguistics	linguistic	NOUN
ajst-20212	123	27	and	and	CCONJ
ajst-20212	123	28	the	the	DET
ajst-20212	123	29	11th	11th	ADJ
ajst-20212	123	30	international	international	ADJ
ajst-20212	123	31	joint	joint	ADJ
ajst-20212	123	32	conference	conference	NOUN
ajst-20212	123	33	on	on	ADP
ajst-20212	123	34	natural	natural	ADJ
ajst-20212	123	35	language	language	NOUN
ajst-20212	123	36	processing	processing	NOUN
ajst-20212	123	37	(	(	PUNCT
ajst-20212	123	38	volume	volume	NOUN
ajst-20212	123	39	1	1	NUM
ajst-20212	123	40	:	:	PUNCT
ajst-20212	123	41	long	long	ADJ
ajst-20212	123	42	papers	paper	NOUN
ajst-20212	123	43	)	)	PUNCT
ajst-20212	123	44	,	,	PUNCT
ajst-20212	123	45	f	f	X
ajst-20212	123	46	,	,	PUNCT
ajst-20212	123	47	2021	2021	NUM
ajst-20212	124	1	[	[	X
ajst-20212	124	2	c	c	X
ajst-20212	124	3	]	]	PUNCT
ajst-20212	124	4	.	.	PUNCT
ajst-20212	125	1	[	[	X
ajst-20212	125	2	9	9	NUM
ajst-20212	125	3	]	]	X
ajst-20212	125	4	xiao	xiao	PROPN
ajst-20212	125	5	z	z	PROPN
ajst-20212	125	6	,	,	PUNCT
ajst-20212	125	7	wu	wu	PROPN
ajst-20212	125	8	j	j	PROPN
ajst-20212	125	9	,	,	PUNCT
ajst-20212	125	10	chen	chen	PROPN
ajst-20212	126	1	q	q	PROPN
ajst-20212	126	2	,	,	PUNCT
ajst-20212	126	3	et	et	PROPN
ajst-20212	126	4	al	al	PROPN
ajst-20212	126	5	.	.	PROPN
ajst-20212	126	6	bert4gcn	bert4gcn	NUM
ajst-20212	126	7	:	:	PUNCT
ajst-20212	127	1	using	use	VERB
ajst-20212	127	2	bert	bert	PROPN
ajst-20212	127	3	intermediate	intermediate	ADJ
ajst-20212	127	4	layers	layer	NOUN
ajst-20212	127	5	to	to	PART
ajst-20212	127	6	augment	augment	VERB
ajst-20212	127	7	gcn	gcn	NOUN
ajst-20212	127	8	for	for	ADP
ajst-20212	127	9	aspect	aspect	NOUN
ajst-20212	127	10	-	-	PUNCT
ajst-20212	127	11	based	base	VERB