id	sid	tid	token	lemma	pos
fcis-14690	1	1	frontiers	frontier	NOUN
fcis-14690	1	2	in	in	ADP
fcis-14690	1	3	computing	computing	NOUN
fcis-14690	1	4	and	and	CCONJ
fcis-14690	1	5	intelligent	intelligent	ADJ
fcis-14690	1	6	systems	system	NOUN
fcis-14690	1	7	issn	issn	VERB
fcis-14690	1	8	:	:	PUNCT
fcis-14690	1	9	2832	2832	NUM
fcis-14690	1	10	-	-	SYM
fcis-14690	1	11	6024	6024	NUM
fcis-14690	1	12	|	|	NOUN
fcis-14690	1	13	vol	vol	NOUN
fcis-14690	1	14	.	.	PROPN
fcis-14690	2	1	6	6	NUM
fcis-14690	2	2	,	,	PUNCT
fcis-14690	2	3	no	no	INTJ
fcis-14690	2	4	.	.	NOUN
fcis-14690	2	5	1	1	NUM
fcis-14690	2	6	,	,	PUNCT
fcis-14690	2	7	2023	2023	NUM
fcis-14690	2	8	60	60	NUM
fcis-14690	2	9	image	image	NOUN
fcis-14690	2	10	emotion	emotion	NOUN
fcis-14690	2	11	analysis	analysis	NOUN
fcis-14690	2	12	combining	combine	VERB
fcis-14690	2	13	attention	attention	NOUN
fcis-14690	2	14	mechanism	mechanism	NOUN
fcis-14690	2	15	and	and	CCONJ
fcis-14690	2	16	multi‐level	multi‐level	NOUN
fcis-14690	2	17	correlation	correlation	NOUN
fcis-14690	2	18	shuxia	shuxia	PROPN
fcis-14690	2	19	ren	ren	PROPN
fcis-14690	2	20	,	,	PUNCT
fcis-14690	2	21	simin	simin	PROPN
fcis-14690	2	22	li	li	PROPN
fcis-14690	2	23	school	school	PROPN
fcis-14690	2	24	of	of	ADP
fcis-14690	2	25	software	software	NOUN
fcis-14690	2	26	,	,	PUNCT
fcis-14690	2	27	tianjin	tianjin	PROPN
fcis-14690	2	28	polytechnic	polytechnic	PROPN
fcis-14690	2	29	university	university	PROPN
fcis-14690	2	30	,	,	PUNCT
fcis-14690	2	31	tianjin	tianjin	PROPN
fcis-14690	2	32	300387	300387	NUM
fcis-14690	2	33	,	,	PUNCT
fcis-14690	2	34	china	china	PROPN
fcis-14690	2	35	abstract	abstract	NOUN
fcis-14690	2	36	:	:	PUNCT
fcis-14690	2	37	the	the	DET
fcis-14690	2	38	development	development	NOUN
fcis-14690	2	39	of	of	ADP
fcis-14690	2	40	social	social	ADJ
fcis-14690	2	41	network	network	NOUN
fcis-14690	2	42	has	have	AUX
fcis-14690	2	43	brought	bring	VERB
fcis-14690	2	44	a	a	DET
fcis-14690	2	45	large	large	ADJ
fcis-14690	2	46	amount	amount	NOUN
fcis-14690	2	47	of	of	ADP
fcis-14690	2	48	image	image	NOUN
fcis-14690	2	49	information	information	NOUN
fcis-14690	2	50	,	,	PUNCT
fcis-14690	2	51	and	and	CCONJ
fcis-14690	2	52	the	the	DET
fcis-14690	2	53	research	research	NOUN
fcis-14690	2	54	on	on	ADP
fcis-14690	2	55	image	image	NOUN
fcis-14690	2	56	emotion	emotion	NOUN
fcis-14690	2	57	has	have	AUX
fcis-14690	2	58	gradually	gradually	ADV
fcis-14690	2	59	attracted	attract	VERB
fcis-14690	2	60	wide	wide	ADJ
fcis-14690	2	61	attention	attention	NOUN
fcis-14690	2	62	.	.	PUNCT
fcis-14690	3	1	the	the	DET
fcis-14690	3	2	current	current	ADJ
fcis-14690	3	3	image	image	NOUN
fcis-14690	3	4	emotion	emotion	NOUN
fcis-14690	3	5	analysis	analysis	NOUN
fcis-14690	3	6	methods	method	NOUN
fcis-14690	3	7	based	base	VERB
fcis-14690	3	8	on	on	ADP
fcis-14690	3	9	multi	multi	ADJ
fcis-14690	3	10	-	-	ADJ
fcis-14690	3	11	level	level	ADJ
fcis-14690	3	12	features	feature	NOUN
fcis-14690	3	13	simply	simply	ADV
fcis-14690	3	14	splice	splice	VERB
fcis-14690	3	15	the	the	DET
fcis-14690	3	16	features	feature	NOUN
fcis-14690	3	17	at	at	ADP
fcis-14690	3	18	each	each	DET
fcis-14690	3	19	level	level	NOUN
fcis-14690	3	20	and	and	CCONJ
fcis-14690	3	21	then	then	ADV
fcis-14690	3	22	classify	classify	VERB
fcis-14690	3	23	the	the	DET
fcis-14690	3	24	emotions	emotion	NOUN
fcis-14690	3	25	,	,	PUNCT
fcis-14690	3	26	which	which	PRON
fcis-14690	3	27	not	not	PART
fcis-14690	3	28	only	only	ADV
fcis-14690	3	29	ignores	ignore	VERB
fcis-14690	3	30	the	the	DET
fcis-14690	3	31	correlation	correlation	NOUN
fcis-14690	3	32	between	between	ADP
fcis-14690	3	33	features	feature	NOUN
fcis-14690	3	34	at	at	ADP
fcis-14690	3	35	different	different	ADJ
fcis-14690	3	36	levels	level	NOUN
fcis-14690	3	37	,	,	PUNCT
fcis-14690	3	38	but	but	CCONJ
fcis-14690	3	39	also	also	ADV
fcis-14690	3	40	ignores	ignore	VERB
fcis-14690	3	41	the	the	DET
fcis-14690	3	42	synergistic	synergistic	ADJ
fcis-14690	3	43	effect	effect	NOUN
fcis-14690	3	44	between	between	ADP
fcis-14690	3	45	global	global	ADJ
fcis-14690	3	46	features	feature	NOUN
fcis-14690	3	47	and	and	CCONJ
fcis-14690	3	48	local	local	ADJ
fcis-14690	3	49	features	feature	NOUN
fcis-14690	3	50	.	.	PUNCT
fcis-14690	4	1	therefore	therefore	ADV
fcis-14690	4	2	,	,	PUNCT
fcis-14690	4	3	this	this	DET
fcis-14690	4	4	paper	paper	NOUN
fcis-14690	4	5	proposes	propose	VERB
fcis-14690	4	6	an	an	DET
fcis-14690	4	7	emotion	emotion	NOUN
fcis-14690	4	8	model	model	NOUN
fcis-14690	4	9	(	(	PUNCT
fcis-14690	4	10	maml	maml	PROPN
fcis-14690	4	11	)	)	PUNCT
fcis-14690	4	12	based	base	VERB
fcis-14690	4	13	on	on	ADP
fcis-14690	4	14	mixed	mixed	ADJ
fcis-14690	4	15	attention	attention	NOUN
fcis-14690	4	16	and	and	CCONJ
fcis-14690	4	17	multi	multi	ADJ
fcis-14690	4	18	-	-	ADJ
fcis-14690	4	19	level	level	ADJ
fcis-14690	4	20	dependence	dependence	NOUN
fcis-14690	4	21	of	of	ADP
fcis-14690	4	22	images	image	NOUN
fcis-14690	4	23	,	,	PUNCT
fcis-14690	4	24	which	which	PRON
fcis-14690	4	25	uses	use	VERB
fcis-14690	4	26	spatial	spatial	ADJ
fcis-14690	4	27	and	and	CCONJ
fcis-14690	4	28	channel	channel	NOUN
fcis-14690	4	29	attention	attention	NOUN
fcis-14690	4	30	mechanisms	mechanism	NOUN
fcis-14690	4	31	to	to	PART
fcis-14690	4	32	extract	extract	VERB
fcis-14690	4	33	local	local	ADJ
fcis-14690	4	34	emotion	emotion	NOUN
fcis-14690	4	35	region	region	NOUN
fcis-14690	4	36	features	feature	NOUN
fcis-14690	4	37	of	of	ADP
fcis-14690	4	38	images	image	NOUN
fcis-14690	4	39	.	.	PUNCT
fcis-14690	5	1	bi	bi	ADJ
fcis-14690	5	2	-	-	ADJ
fcis-14690	5	3	directional	directional	ADJ
fcis-14690	5	4	long	long	ADJ
fcis-14690	5	5	short	short	ADJ
fcis-14690	5	6	term	term	NOUN
fcis-14690	5	7	memory	memory	NOUN
fcis-14690	5	8	network	network	NOUN
fcis-14690	5	9	(	(	PUNCT
fcis-14690	5	10	bilstm	bilstm	NOUN
fcis-14690	5	11	)	)	PUNCT
fcis-14690	5	12	is	be	AUX
fcis-14690	5	13	used	use	VERB
fcis-14690	5	14	to	to	PART
fcis-14690	5	15	establish	establish	VERB
fcis-14690	5	16	correlation	correlation	NOUN
fcis-14690	5	17	between	between	ADP
fcis-14690	5	18	multi	multi	ADJ
fcis-14690	5	19	-	-	ADJ
fcis-14690	5	20	level	level	ADJ
fcis-14690	5	21	image	image	NOUN
fcis-14690	5	22	global	global	ADJ
fcis-14690	5	23	features	feature	NOUN
fcis-14690	5	24	.	.	PUNCT
fcis-14690	6	1	the	the	DET
fcis-14690	6	2	experimental	experimental	ADJ
fcis-14690	6	3	results	result	NOUN
fcis-14690	6	4	of	of	ADP
fcis-14690	6	5	maml	maml	ADJ
fcis-14690	6	6	model	model	NOUN
fcis-14690	6	7	on	on	ADP
fcis-14690	6	8	artphoto	artphoto	NOUN
fcis-14690	6	9	and	and	CCONJ
fcis-14690	6	10	abstract	abstract	ADJ
fcis-14690	6	11	data	data	NOUN
fcis-14690	6	12	sets	set	NOUN
fcis-14690	6	13	prove	prove	VERB
fcis-14690	6	14	the	the	DET
fcis-14690	6	15	validity	validity	NOUN
fcis-14690	6	16	of	of	ADP
fcis-14690	6	17	maml	maml	PROPN
fcis-14690	6	18	model	model	NOUN
fcis-14690	6	19	.	.	PUNCT
fcis-14690	7	1	keywords	keyword	NOUN
fcis-14690	7	2	:	:	PUNCT
fcis-14690	7	3	multi	multi	ADJ
fcis-14690	7	4	-	-	ADJ
fcis-14690	7	5	level	level	ADJ
fcis-14690	7	6	features	feature	NOUN
fcis-14690	7	7	;	;	PUNCT
fcis-14690	7	8	mixed	mixed	ADJ
fcis-14690	7	9	attention	attention	NOUN
fcis-14690	7	10	;	;	PUNCT
fcis-14690	7	11	global	global	ADJ
fcis-14690	7	12	and	and	CCONJ
fcis-14690	7	13	local	local	ADJ
fcis-14690	7	14	;	;	PUNCT
fcis-14690	7	15	bilstm	bilstm	NOUN
fcis-14690	7	16	.	.	PUNCT
fcis-14690	8	1	1	1	X
fcis-14690	8	2	.	.	X
fcis-14690	8	3	introduction	introduction	NOUN
fcis-14690	8	4	with	with	ADP
fcis-14690	8	5	the	the	DET
fcis-14690	8	6	increasing	increase	VERB
fcis-14690	8	7	popularity	popularity	NOUN
fcis-14690	8	8	of	of	ADP
fcis-14690	8	9	social	social	ADJ
fcis-14690	8	10	media	medium	NOUN
fcis-14690	8	11	,	,	PUNCT
fcis-14690	8	12	more	more	ADJ
fcis-14690	8	13	and	and	CCONJ
fcis-14690	8	14	more	more	ADJ
fcis-14690	8	15	users	user	NOUN
fcis-14690	8	16	post	post	VERB
fcis-14690	8	17	images	image	NOUN
fcis-14690	8	18	on	on	ADP
fcis-14690	8	19	social	social	ADJ
fcis-14690	8	20	networks	network	NOUN
fcis-14690	8	21	to	to	PART
fcis-14690	8	22	express	express	VERB
fcis-14690	8	23	their	their	PRON
fcis-14690	8	24	views	view	NOUN
fcis-14690	8	25	and	and	CCONJ
fcis-14690	8	26	emotions	emotion	NOUN
fcis-14690	8	27	,	,	PUNCT
fcis-14690	8	28	and	and	CCONJ
fcis-14690	8	29	images	image	NOUN
fcis-14690	8	30	have	have	AUX
fcis-14690	8	31	gradually	gradually	ADV
fcis-14690	8	32	become	become	VERB
fcis-14690	8	33	an	an	DET
fcis-14690	8	34	important	important	ADJ
fcis-14690	8	35	carrier	carrier	NOUN
fcis-14690	8	36	for	for	SCONJ
fcis-14690	8	37	people	people	NOUN
fcis-14690	8	38	to	to	PART
fcis-14690	8	39	express	express	VERB
fcis-14690	8	40	their	their	PRON
fcis-14690	8	41	emotions	emotion	NOUN
fcis-14690	8	42	.	.	PUNCT
fcis-14690	9	1	emotion	emotion	NOUN
fcis-14690	9	2	research	research	NOUN
fcis-14690	9	3	plays	play	VERB
fcis-14690	9	4	an	an	DET
fcis-14690	9	5	indispensable	indispensable	ADJ
fcis-14690	9	6	role	role	NOUN
fcis-14690	9	7	in	in	ADP
fcis-14690	9	8	education	education	NOUN
fcis-14690	9	9	,	,	PUNCT
fcis-14690	9	10	advertising	advertising	NOUN
fcis-14690	9	11	,	,	PUNCT
fcis-14690	9	12	decision	decision	NOUN
fcis-14690	9	13	-	-	PUNCT
fcis-14690	9	14	making	making	NOUN
fcis-14690	9	15	,	,	PUNCT
fcis-14690	9	16	planning	planning	NOUN
fcis-14690	9	17	and	and	CCONJ
fcis-14690	9	18	many	many	ADJ
fcis-14690	9	19	other	other	ADJ
fcis-14690	9	20	activities	activity	NOUN
fcis-14690	9	21	.	.	PUNCT
fcis-14690	10	1	therefore	therefore	ADV
fcis-14690	10	2	,	,	PUNCT
fcis-14690	10	3	image	image	NOUN
fcis-14690	10	4	sentiment	sentiment	NOUN
fcis-14690	10	5	analysis	analysis	NOUN
fcis-14690	10	6	has	have	AUX
fcis-14690	10	7	attracted	attract	VERB
fcis-14690	10	8	much	much	ADJ
fcis-14690	10	9	attention	attention	NOUN
fcis-14690	10	10	from	from	ADP
fcis-14690	10	11	researchers	researcher	NOUN
fcis-14690	10	12	.	.	PUNCT
fcis-14690	11	1	however	however	ADV
fcis-14690	11	2	,	,	PUNCT
fcis-14690	11	3	because	because	SCONJ
fcis-14690	11	4	of	of	ADP
fcis-14690	11	5	the	the	DET
fcis-14690	11	6	subjectivity	subjectivity	NOUN
fcis-14690	11	7	and	and	CCONJ
fcis-14690	11	8	complexity	complexity	NOUN
fcis-14690	11	9	of	of	ADP
fcis-14690	11	10	emotion	emotion	NOUN
fcis-14690	11	11	,	,	PUNCT
fcis-14690	11	12	the	the	DET
fcis-14690	11	13	task	task	NOUN
fcis-14690	11	14	of	of	ADP
fcis-14690	11	15	emotion	emotion	NOUN
fcis-14690	11	16	analysis	analysis	NOUN
fcis-14690	11	17	for	for	ADP
fcis-14690	11	18	images	image	NOUN
fcis-14690	11	19	is	be	AUX
fcis-14690	11	20	still	still	ADV
fcis-14690	11	21	very	very	ADV
fcis-14690	11	22	challenging	challenging	ADJ
fcis-14690	11	23	.	.	PUNCT
fcis-14690	12	1	as	as	SCONJ
fcis-14690	12	2	shown	show	VERB
fcis-14690	12	3	in	in	ADP
fcis-14690	12	4	figure	figure	NOUN
fcis-14690	12	5	1	1	NUM
fcis-14690	12	6	,	,	PUNCT
fcis-14690	12	7	image	image	NOUN
fcis-14690	12	8	emotion	emotion	NOUN
fcis-14690	12	9	is	be	AUX
fcis-14690	12	10	closely	closely	ADV
fcis-14690	12	11	related	relate	VERB
fcis-14690	12	12	to	to	ADP
fcis-14690	12	13	image	image	NOUN
fcis-14690	12	14	color	color	NOUN
fcis-14690	12	15	,	,	PUNCT
fcis-14690	12	16	texture	texture	NOUN
fcis-14690	12	17	,	,	PUNCT
fcis-14690	12	18	line	line	NOUN
fcis-14690	12	19	and	and	CCONJ
fcis-14690	12	20	other	other	ADJ
fcis-14690	12	21	underlying	underlie	VERB
fcis-14690	12	22	features	feature	NOUN
fcis-14690	12	23	.	.	PUNCT
fcis-14690	13	1	at	at	ADP
fcis-14690	13	2	the	the	DET
fcis-14690	13	3	beginning	beginning	NOUN
fcis-14690	13	4	of	of	ADP
fcis-14690	13	5	the	the	DET
fcis-14690	13	6	study	study	NOUN
fcis-14690	13	7	,	,	PUNCT
fcis-14690	13	8	the	the	DET
fcis-14690	13	9	researchers	researcher	NOUN
fcis-14690	13	10	predicted	predict	VERB
fcis-14690	13	11	emotion	emotion	NOUN
fcis-14690	13	12	by	by	ADP
fcis-14690	13	13	designing	design	VERB
fcis-14690	13	14	various	various	ADJ
fcis-14690	13	15	manual	manual	ADJ
fcis-14690	13	16	features	feature	NOUN
fcis-14690	13	17	(	(	PUNCT
fcis-14690	13	18	such	such	ADJ
fcis-14690	13	19	as	as	ADP
fcis-14690	13	20	color	color	NOUN
fcis-14690	13	21	,	,	PUNCT
fcis-14690	13	22	texture	texture	NOUN
fcis-14690	13	23	,	,	PUNCT
fcis-14690	13	24	composition	composition	NOUN
fcis-14690	13	25	,	,	PUNCT
fcis-14690	13	26	etc	etc	X
fcis-14690	13	27	.	.	X
fcis-14690	13	28	)	)	PUNCT
fcis-14690	14	1	[	[	X
fcis-14690	14	2	1,2,3	1,2,3	NUM
fcis-14690	14	3	]	]	PUNCT
fcis-14690	14	4	.	.	PUNCT
fcis-14690	15	1	with	with	ADP
fcis-14690	15	2	the	the	DET
fcis-14690	15	3	development	development	NOUN
fcis-14690	15	4	of	of	ADP
fcis-14690	15	5	convolutional	convolutional	ADJ
fcis-14690	15	6	neural	neural	ADJ
fcis-14690	15	7	network	network	NOUN
fcis-14690	15	8	(	(	PUNCT
fcis-14690	15	9	cnn	cnn	PROPN
fcis-14690	15	10	)	)	PUNCT
fcis-14690	15	11	,	,	PUNCT
fcis-14690	15	12	image	image	NOUN
fcis-14690	15	13	emotion	emotion	NOUN
fcis-14690	15	14	analysis	analysis	NOUN
fcis-14690	15	15	methods	method	NOUN
fcis-14690	15	16	based	base	VERB
fcis-14690	15	17	on	on	ADP
fcis-14690	15	18	deep	deep	ADJ
fcis-14690	15	19	learning	learning	NOUN
fcis-14690	15	20	show	show	VERB
fcis-14690	15	21	excellent	excellent	ADJ
fcis-14690	15	22	performance	performance	NOUN
fcis-14690	15	23	[	[	X
fcis-14690	15	24	4,5,6	4,5,6	NUM
fcis-14690	15	25	]	]	PUNCT
fcis-14690	15	26	.	.	PUNCT
fcis-14690	16	1	at	at	ADP
fcis-14690	16	2	present	present	ADJ
fcis-14690	16	3	,	,	PUNCT
fcis-14690	16	4	researchers	researcher	NOUN
fcis-14690	16	5	have	have	AUX
fcis-14690	16	6	designed	design	VERB
fcis-14690	16	7	various	various	ADJ
fcis-14690	16	8	neural	neural	ADJ
fcis-14690	16	9	networks	network	NOUN
fcis-14690	16	10	for	for	ADP
fcis-14690	16	11	image	image	NOUN
fcis-14690	16	12	emotion	emotion	NOUN
fcis-14690	16	13	analysis	analysis	NOUN
fcis-14690	16	14	.	.	PUNCT
fcis-14690	17	1	peng	peng	PROPN
fcis-14690	17	2	et	et	PROPN
fcis-14690	17	3	al	al	PROPN
fcis-14690	17	4	.	.	PUNCT
fcis-14690	18	1	[	[	X
fcis-14690	18	2	7	7	NUM
fcis-14690	18	3	]	]	PUNCT
fcis-14690	18	4	extracted	extract	VERB
fcis-14690	18	5	global	global	ADJ
fcis-14690	18	6	features	feature	NOUN
fcis-14690	18	7	of	of	ADP
fcis-14690	18	8	images	image	NOUN
fcis-14690	18	9	for	for	ADP
fcis-14690	18	10	emotion	emotion	NOUN
fcis-14690	18	11	classification	classification	NOUN
fcis-14690	18	12	by	by	ADP
fcis-14690	18	13	fine	fine	ADV
fcis-14690	18	14	-	-	PUNCT
fcis-14690	18	15	tuning	tune	VERB
fcis-14690	18	16	the	the	DET
fcis-14690	18	17	cnn	cnn	PROPN
fcis-14690	18	18	model	model	NOUN
fcis-14690	18	19	pre	pre	VERB
fcis-14690	18	20	-	-	VERB
fcis-14690	18	21	trained	train	VERB
fcis-14690	18	22	in	in	ADP
fcis-14690	18	23	imagenet	imagenet	NOUN
fcis-14690	18	24	.	.	PUNCT
fcis-14690	19	1	research	research	NOUN
fcis-14690	20	1	[	[	X
fcis-14690	20	2	8	8	X
fcis-14690	20	3	]	]	PUNCT
fcis-14690	20	4	shows	show	VERB
fcis-14690	20	5	that	that	SCONJ
fcis-14690	20	6	when	when	SCONJ
fcis-14690	20	7	humans	human	NOUN
fcis-14690	20	8	observe	observe	VERB
fcis-14690	20	9	images	image	NOUN
fcis-14690	20	10	,	,	PUNCT
fcis-14690	20	11	emotion	emotion	NOUN
fcis-14690	20	12	is	be	AUX
fcis-14690	20	13	mainly	mainly	ADV
fcis-14690	20	14	caused	cause	VERB
fcis-14690	20	15	by	by	ADP
fcis-14690	20	16	local	local	ADJ
fcis-14690	20	17	emotional	emotional	ADJ
fcis-14690	20	18	areas	area	NOUN
fcis-14690	20	19	.	.	PUNCT
fcis-14690	21	1	based	base	VERB
fcis-14690	21	2	on	on	ADP
fcis-14690	21	3	descriptive	descriptive	ADJ
fcis-14690	21	4	visual	visual	ADJ
fcis-14690	21	5	attributes	attribute	NOUN
fcis-14690	21	6	,	,	PUNCT
fcis-14690	21	7	you	you	PRON
fcis-14690	21	8	et	et	VERB
fcis-14690	21	9	al	al	PROPN
fcis-14690	21	10	.	.	PUNCT
fcis-14690	22	1	[	[	X
fcis-14690	22	2	9	9	NUM
fcis-14690	22	3	]	]	PUNCT
fcis-14690	22	4	used	use	VERB
fcis-14690	22	5	the	the	DET
fcis-14690	22	6	attention	attention	NOUN
fcis-14690	22	7	model	model	NOUN
fcis-14690	22	8	to	to	PART
fcis-14690	22	9	find	find	VERB
fcis-14690	22	10	the	the	DET
fcis-14690	22	11	local	local	ADJ
fcis-14690	22	12	areas	area	NOUN
fcis-14690	22	13	that	that	PRON
fcis-14690	22	14	evoke	evoke	VERB
fcis-14690	22	15	the	the	DET
fcis-14690	22	16	audience	audience	NOUN
fcis-14690	22	17	's	's	PART
fcis-14690	22	18	emotions	emotion	NOUN
fcis-14690	22	19	,	,	PUNCT
fcis-14690	22	20	and	and	CCONJ
fcis-14690	22	21	then	then	ADV
fcis-14690	22	22	extracted	extract	VERB
fcis-14690	22	23	their	their	PRON
fcis-14690	22	24	features	feature	NOUN
fcis-14690	22	25	to	to	PART
fcis-14690	22	26	improve	improve	VERB
fcis-14690	22	27	the	the	DET
fcis-14690	22	28	performance	performance	NOUN
fcis-14690	22	29	of	of	ADP
fcis-14690	22	30	emotion	emotion	NOUN
fcis-14690	22	31	analysis	analysis	NOUN
fcis-14690	22	32	.	.	PUNCT
fcis-14690	23	1	zhao	zhao	PROPN
fcis-14690	23	2	et	et	PROPN
fcis-14690	23	3	al	al	PROPN
fcis-14690	23	4	.	.	PUNCT
fcis-14690	24	1	[	[	X
fcis-14690	24	2	10	10	NUM
fcis-14690	24	3	]	]	PUNCT
fcis-14690	24	4	showed	show	VERB
fcis-14690	24	5	through	through	ADP
fcis-14690	24	6	experiments	experiment	NOUN
fcis-14690	24	7	that	that	SCONJ
fcis-14690	24	8	the	the	DET
fcis-14690	24	9	effect	effect	NOUN
fcis-14690	24	10	of	of	ADP
fcis-14690	24	11	using	use	VERB
fcis-14690	24	12	both	both	CCONJ
fcis-14690	24	13	local	local	ADJ
fcis-14690	24	14	and	and	CCONJ
fcis-14690	24	15	global	global	ADJ
fcis-14690	24	16	features	feature	NOUN
fcis-14690	24	17	is	be	AUX
fcis-14690	24	18	better	well	ADJ
fcis-14690	24	19	than	than	ADP
fcis-14690	24	20	that	that	PRON
fcis-14690	24	21	of	of	ADP
fcis-14690	24	22	using	use	VERB
fcis-14690	24	23	one	one	NUM
fcis-14690	24	24	of	of	ADP
fcis-14690	24	25	them	they	PRON
fcis-14690	24	26	alone	alone	ADV
fcis-14690	24	27	.	.	PUNCT
fcis-14690	25	1	however	however	ADV
fcis-14690	25	2	,	,	PUNCT
fcis-14690	25	3	most	most	ADJ
fcis-14690	25	4	current	current	ADJ
fcis-14690	25	5	studies	study	NOUN
fcis-14690	25	6	ignore	ignore	VERB
fcis-14690	25	7	the	the	DET
fcis-14690	25	8	synergistic	synergistic	ADJ
fcis-14690	25	9	effect	effect	NOUN
fcis-14690	25	10	of	of	ADP
fcis-14690	25	11	local	local	ADJ
fcis-14690	25	12	and	and	CCONJ
fcis-14690	25	13	global	global	ADJ
fcis-14690	25	14	features	feature	NOUN
fcis-14690	25	15	,	,	PUNCT
fcis-14690	25	16	and	and	CCONJ
fcis-14690	25	17	the	the	DET
fcis-14690	25	18	attention	attention	NOUN
fcis-14690	25	19	model	model	NOUN
fcis-14690	25	20	used	use	VERB
fcis-14690	25	21	in	in	ADP
fcis-14690	25	22	image	image	NOUN
fcis-14690	25	23	emotion	emotion	NOUN
fcis-14690	25	24	analysis	analysis	NOUN
fcis-14690	25	25	tasks	task	NOUN
fcis-14690	25	26	usually	usually	ADV
fcis-14690	25	27	only	only	ADV
fcis-14690	25	28	considers	consider	VERB
fcis-14690	25	29	spatial	spatial	ADJ
fcis-14690	25	30	attention	attention	NOUN
fcis-14690	25	31	when	when	SCONJ
fcis-14690	25	32	focusing	focus	VERB
fcis-14690	25	33	on	on	ADP
fcis-14690	25	34	local	local	ADJ
fcis-14690	25	35	emotion	emotion	NOUN
fcis-14690	25	36	features	feature	NOUN
fcis-14690	25	37	.	.	PUNCT
fcis-14690	26	1	although	although	SCONJ
fcis-14690	26	2	spatial	spatial	ADJ
fcis-14690	26	3	attention	attention	NOUN
fcis-14690	26	4	regulates	regulate	VERB
fcis-14690	26	5	the	the	DET
fcis-14690	26	6	local	local	ADJ
fcis-14690	26	7	spatial	spatial	ADJ
fcis-14690	26	8	connectivity	connectivity	NOUN
fcis-14690	26	9	pattern	pattern	NOUN
fcis-14690	26	10	on	on	ADP
fcis-14690	26	11	each	each	DET
fcis-14690	26	12	channel	channel	NOUN
fcis-14690	26	13	through	through	ADP
fcis-14690	26	14	spatial	spatial	ADJ
fcis-14690	26	15	attention	attention	NOUN
fcis-14690	26	16	weights	weight	NOUN
fcis-14690	26	17	[	[	X
fcis-14690	26	18	9,11,12	9,11,12	NUM
fcis-14690	26	19	]	]	PUNCT
fcis-14690	26	20	,	,	PUNCT
fcis-14690	26	21	it	it	PRON
fcis-14690	26	22	ignores	ignore	VERB
fcis-14690	26	23	the	the	DET
fcis-14690	26	24	interdependencies	interdependency	NOUN
fcis-14690	26	25	between	between	ADP
fcis-14690	26	26	different	different	ADJ
fcis-14690	26	27	channels	channel	NOUN
fcis-14690	26	28	.	.	PUNCT
fcis-14690	27	1	however	however	ADV
fcis-14690	27	2	,	,	PUNCT
fcis-14690	27	3	it	it	PRON
fcis-14690	27	4	is	be	AUX
fcis-14690	27	5	very	very	ADV
fcis-14690	27	6	important	important	ADJ
fcis-14690	27	7	to	to	PART
fcis-14690	27	8	pay	pay	VERB
fcis-14690	27	9	attention	attention	NOUN
fcis-14690	27	10	to	to	ADP
fcis-14690	27	11	the	the	DET
fcis-14690	27	12	channel	channel	NOUN
fcis-14690	27	13	aspect	aspect	NOUN
fcis-14690	27	14	,	,	PUNCT
fcis-14690	27	15	which	which	PRON
fcis-14690	27	16	can	can	AUX
fcis-14690	27	17	be	be	AUX
fcis-14690	27	18	seen	see	VERB
fcis-14690	27	19	as	as	ADP
fcis-14690	27	20	a	a	DET
fcis-14690	27	21	process	process	NOUN
fcis-14690	27	22	of	of	ADP
fcis-14690	27	23	selecting	select	VERB
fcis-14690	27	24	semantic	semantic	ADJ
fcis-14690	27	25	attributes	attribute	NOUN
fcis-14690	27	26	and	and	CCONJ
fcis-14690	27	27	is	be	AUX
fcis-14690	27	28	essentially	essentially	ADV
fcis-14690	27	29	consistent	consistent	ADJ
fcis-14690	27	30	with	with	ADP
fcis-14690	27	31	the	the	DET
fcis-14690	27	32	characteristics	characteristic	NOUN
fcis-14690	27	33	of	of	ADP
fcis-14690	27	34	cnn	cnn	PROPN
fcis-14690	27	35	[	[	X
fcis-14690	27	36	13	13	NUM
fcis-14690	27	37	]	]	PUNCT
fcis-14690	27	38	.	.	PUNCT
fcis-14690	28	1	previous	previous	ADJ
fcis-14690	28	2	studies	study	NOUN
fcis-14690	28	3	have	have	AUX
fcis-14690	28	4	shown	show	VERB
fcis-14690	28	5	that	that	SCONJ
fcis-14690	28	6	image	image	NOUN
fcis-14690	28	7	emotion	emotion	NOUN
fcis-14690	28	8	is	be	AUX
fcis-14690	28	9	related	relate	VERB
fcis-14690	28	10	to	to	ADP
fcis-14690	28	11	features	feature	NOUN
fcis-14690	28	12	ranging	range	VERB
fcis-14690	28	13	from	from	ADP
fcis-14690	28	14	low	low	ADJ
fcis-14690	28	15	level	level	NOUN
fcis-14690	28	16	to	to	ADP
fcis-14690	28	17	high	high	ADJ
fcis-14690	28	18	level	level	NOUN
fcis-14690	28	19	[	[	X
fcis-14690	28	20	14	14	NUM
fcis-14690	28	21	]	]	PUNCT
fcis-14690	28	22	.	.	PUNCT
fcis-14690	29	1	there	there	PRON
fcis-14690	29	2	is	be	VERB
fcis-14690	29	3	correlation	correlation	NOUN
fcis-14690	29	4	between	between	ADP
fcis-14690	29	5	features	feature	NOUN
fcis-14690	29	6	at	at	ADP
fcis-14690	29	7	different	different	ADJ
fcis-14690	29	8	levels	level	NOUN
fcis-14690	29	9	.	.	PUNCT
fcis-14690	30	1	zhu	zhu	PROPN
fcis-14690	30	2	et	et	PROPN
fcis-14690	30	3	al	al	PROPN
fcis-14690	30	4	.	.	PUNCT
fcis-14690	31	1	[	[	X
fcis-14690	31	2	14	14	NUM
fcis-14690	31	3	]	]	PUNCT
fcis-14690	31	4	and	and	CCONJ
fcis-14690	31	5	zhang	zhang	PROPN
fcis-14690	31	6	et	et	PROPN
fcis-14690	31	7	al	al	PROPN
fcis-14690	31	8	.	.	PUNCT
fcis-14690	32	1	[	[	X
fcis-14690	32	2	15	15	NUM
fcis-14690	32	3	]	]	PUNCT
fcis-14690	32	4	used	use	VERB
fcis-14690	32	5	bidirectional	bidirectional	PROPN
fcis-14690	32	6	gru	gru	NOUN
fcis-14690	32	7	method	method	NOUN
fcis-14690	32	8	and	and	CCONJ
fcis-14690	32	9	gram	gram	NOUN
fcis-14690	32	10	method	method	NOUN
fcis-14690	32	11	respectively	respectively	ADV
fcis-14690	32	12	to	to	PART
fcis-14690	32	13	establish	establish	VERB
fcis-14690	32	14	correlation	correlation	NOUN
fcis-14690	32	15	between	between	ADP
fcis-14690	32	16	features	feature	NOUN
fcis-14690	32	17	at	at	ADP
fcis-14690	32	18	different	different	ADJ
fcis-14690	32	19	levels	level	NOUN
fcis-14690	32	20	.	.	PUNCT
fcis-14690	33	1	however	however	ADV
fcis-14690	33	2	,	,	PUNCT
fcis-14690	33	3	these	these	DET
fcis-14690	33	4	methods	method	NOUN
fcis-14690	33	5	pay	pay	VERB
fcis-14690	33	6	too	too	ADV
fcis-14690	33	7	much	much	ADJ
fcis-14690	33	8	attention	attention	NOUN
fcis-14690	33	9	to	to	ADP
fcis-14690	33	10	the	the	DET
fcis-14690	33	11	global	global	ADJ
fcis-14690	33	12	features	feature	NOUN
fcis-14690	33	13	of	of	ADP
fcis-14690	33	14	images	image	NOUN
fcis-14690	33	15	and	and	CCONJ
fcis-14690	33	16	neglect	neglect	VERB
fcis-14690	33	17	the	the	DET
fcis-14690	33	18	local	local	ADJ
fcis-14690	33	19	emotion	emotion	NOUN
fcis-14690	33	20	region	region	NOUN
fcis-14690	33	21	features	feature	VERB
fcis-14690	33	22	closely	closely	ADV
fcis-14690	33	23	related	relate	VERB
fcis-14690	33	24	to	to	ADP
fcis-14690	33	25	image	image	NOUN
fcis-14690	33	26	emotion	emotion	NOUN
fcis-14690	33	27	.	.	PUNCT
fcis-14690	34	1	fig	fig	NOUN
fcis-14690	34	2	1	1	NUM
fcis-14690	34	3	.	.	PUNCT
fcis-14690	35	1	emotional	emotional	ADJ
fcis-14690	35	2	image	image	NOUN
fcis-14690	35	3	sample	sample	NOUN
fcis-14690	35	4	to	to	PART
fcis-14690	35	5	sum	sum	VERB
fcis-14690	35	6	up	up	ADP
fcis-14690	35	7	,	,	PUNCT
fcis-14690	35	8	although	although	SCONJ
fcis-14690	35	9	researchers	researcher	NOUN
fcis-14690	35	10	have	have	AUX
fcis-14690	35	11	improved	improve	VERB
fcis-14690	35	12	the	the	DET
fcis-14690	35	13	effect	effect	NOUN
fcis-14690	35	14	of	of	ADP
fcis-14690	35	15	image	image	NOUN
fcis-14690	35	16	sentiment	sentiment	NOUN
fcis-14690	35	17	analysis	analysis	NOUN
fcis-14690	35	18	from	from	ADP
fcis-14690	35	19	multiple	multiple	ADJ
fcis-14690	35	20	perspectives	perspective	NOUN
fcis-14690	35	21	,	,	PUNCT
fcis-14690	35	22	how	how	SCONJ
fcis-14690	35	23	to	to	PART
fcis-14690	35	24	use	use	VERB
fcis-14690	35	25	the	the	DET
fcis-14690	35	26	spatial	spatial	ADJ
fcis-14690	35	27	and	and	CCONJ
fcis-14690	35	28	channel	channel	NOUN
fcis-14690	35	29	attention	attention	NOUN
fcis-14690	35	30	mechanisms	mechanism	NOUN
fcis-14690	35	31	as	as	ADV
fcis-14690	35	32	well	well	ADV
fcis-14690	35	33	as	as	ADP
fcis-14690	35	34	the	the	DET
fcis-14690	35	35	correlation	correlation	NOUN
fcis-14690	35	36	between	between	ADP
fcis-14690	35	37	features	feature	NOUN
fcis-14690	35	38	at	at	ADP
fcis-14690	35	39	different	different	ADJ
fcis-14690	35	40	levels	level	NOUN
fcis-14690	35	41	of	of	ADP
fcis-14690	35	42	images	image	NOUN
fcis-14690	35	43	to	to	PART
fcis-14690	35	44	extract	extract	VERB
fcis-14690	35	45	more	more	ADJ
fcis-14690	35	46	discriminative	discriminative	NOUN
fcis-14690	35	47	features	feature	NOUN
fcis-14690	35	48	is	be	AUX
fcis-14690	35	49	still	still	ADV
fcis-14690	35	50	a	a	DET
fcis-14690	35	51	problem	problem	NOUN
fcis-14690	35	52	to	to	PART
fcis-14690	35	53	be	be	AUX
fcis-14690	35	54	solved	solve	VERB
fcis-14690	35	55	.	.	PUNCT
fcis-14690	36	1	therefore	therefore	ADV
fcis-14690	36	2	,	,	PUNCT
fcis-14690	36	3	this	this	DET
fcis-14690	36	4	paper	paper	NOUN
fcis-14690	36	5	proposes	propose	VERB
fcis-14690	36	6	an	an	DET
fcis-14690	36	7	image	image	NOUN
fcis-14690	36	8	feature	feature	NOUN
fcis-14690	36	9	extraction	extraction	NOUN
fcis-14690	36	10	method	method	NOUN
fcis-14690	36	11	that	that	PRON
fcis-14690	36	12	combines	combine	VERB
fcis-14690	36	13	mixed	mixed	ADJ
fcis-14690	36	14	attention	attention	NOUN
fcis-14690	36	15	and	and	CCONJ
fcis-14690	36	16	multilevel	multilevel	ADJ
fcis-14690	36	17	feature	feature	NOUN
fcis-14690	36	18	dependence	dependence	NOUN
fcis-14690	36	19	.	.	PUNCT
fcis-14690	37	1	multiple	multiple	ADJ
fcis-14690	37	2	branches	branch	NOUN
fcis-14690	37	3	are	be	AUX
fcis-14690	37	4	used	use	VERB
fcis-14690	37	5	to	to	PART
fcis-14690	37	6	extract	extract	VERB
fcis-14690	37	7	multi	multi	ADJ
fcis-14690	37	8	-	-	ADJ
fcis-14690	37	9	level	level	ADJ
fcis-14690	37	10	features	feature	NOUN
fcis-14690	37	11	from	from	ADP
fcis-14690	37	12	low	low	ADJ
fcis-14690	37	13	to	to	ADP
fcis-14690	37	14	high	high	ADJ
fcis-14690	37	15	,	,	PUNCT
fcis-14690	37	16	and	and	CCONJ
fcis-14690	37	17	bilstm	bilstm	NOUN
fcis-14690	37	18	is	be	AUX
fcis-14690	37	19	used	use	VERB
fcis-14690	37	20	to	to	PART
fcis-14690	37	21	achieve	achieve	VERB
fcis-14690	37	22	multi	multi	ADJ
fcis-14690	37	23	-	-	ADJ
fcis-14690	37	24	level	level	ADJ
fcis-14690	37	25	feature	feature	NOUN
fcis-14690	37	26	fusion	fusion	NOUN
fcis-14690	37	27	.	.	PUNCT
fcis-14690	38	1	in	in	ADP
fcis-14690	38	2	addition	addition	NOUN
fcis-14690	38	3	,	,	PUNCT
fcis-14690	38	4	after	after	ADP
fcis-14690	38	5	the	the	DET
fcis-14690	38	6	highest	high	ADJ
fcis-14690	38	7	level	level	NOUN
fcis-14690	38	8	of	of	ADP
fcis-14690	38	9	convolutional	convolutional	ADJ
fcis-14690	38	10	blocks	block	NOUN
fcis-14690	38	11	,	,	PUNCT
fcis-14690	38	12	the	the	DET
fcis-14690	38	13	lightweight	lightweight	ADJ
fcis-14690	38	14	spatial	spatial	ADJ
fcis-14690	38	15	and	and	CCONJ
fcis-14690	38	16	channel	channel	NOUN
fcis-14690	38	17	attention	attention	NOUN
fcis-14690	38	18	module	module	NOUN
fcis-14690	38	19	cbam	cbam	NOUN
fcis-14690	38	20	is	be	AUX
fcis-14690	38	21	used	use	VERB
fcis-14690	38	22	to	to	PART
fcis-14690	38	23	focus	focus	VERB
fcis-14690	38	24	on	on	ADP
fcis-14690	38	25	the	the	DET
fcis-14690	38	26	local	local	ADJ
fcis-14690	38	27	emotion	emotion	NOUN
fcis-14690	38	28	area	area	NOUN
fcis-14690	38	29	.	.	PUNCT
fcis-14690	39	1	therefore	therefore	ADV
fcis-14690	39	2	,	,	PUNCT
fcis-14690	39	3	the	the	DET
fcis-14690	39	4	maml	maml	PROPN
fcis-14690	39	5	model	model	NOUN
fcis-14690	39	6	proposed	propose	VERB
fcis-14690	39	7	in	in	ADP
fcis-14690	39	8	this	this	DET
fcis-14690	39	9	paper	paper	NOUN
fcis-14690	39	10	can	can	AUX
fcis-14690	39	11	not	not	PART
fcis-14690	39	12	only	only	ADV
fcis-14690	39	13	extract	extract	VERB
fcis-14690	39	14	the	the	DET
fcis-14690	39	15	local	local	ADJ
fcis-14690	39	16	features	feature	NOUN
fcis-14690	39	17	of	of	ADP
fcis-14690	39	18	the	the	DET
fcis-14690	39	19	image	image	NOUN
fcis-14690	39	20	that	that	PRON
fcis-14690	39	21	cause	cause	VERB
fcis-14690	39	22	the	the	DET
fcis-14690	39	23	viewer	viewer	NOUN
fcis-14690	39	24	's	's	PART
fcis-14690	39	25	emotion	emotion	NOUN
fcis-14690	39	26	,	,	PUNCT
fcis-14690	39	27	but	but	CCONJ
fcis-14690	39	28	also	also	ADV
fcis-14690	39	29	extract	extract	VERB
fcis-14690	39	30	the	the	DET
fcis-14690	39	31	complete	complete	ADJ
fcis-14690	39	32	feature	feature	NOUN
fcis-14690	39	33	information	information	NOUN
fcis-14690	39	34	of	of	ADP
fcis-14690	39	35	the	the	DET
fcis-14690	39	36	image	image	NOUN
fcis-14690	39	37	.	.	PUNCT
fcis-14690	40	1	2	2	X
fcis-14690	40	2	.	.	X
fcis-14690	40	3	related	relate	VERB
fcis-14690	40	4	work	work	NOUN
fcis-14690	40	5	2.1	2.1	NUM
fcis-14690	40	6	.	.	PUNCT
fcis-14690	41	1	multi	multi	ADJ
fcis-14690	41	2	-	-	ADJ
fcis-14690	41	3	level	level	ADJ
fcis-14690	41	4	features	feature	NOUN
fcis-14690	41	5	closing	close	VERB
fcis-14690	41	6	the	the	DET
fcis-14690	41	7	affective	affective	NOUN
fcis-14690	41	8	gap	gap	NOUN
fcis-14690	41	9	is	be	AUX
fcis-14690	41	10	a	a	DET
fcis-14690	41	11	major	major	ADJ
fcis-14690	41	12	challenge	challenge	NOUN
fcis-14690	41	13	in	in	ADP
fcis-14690	41	14	affective	affective	ADJ
fcis-14690	41	15	prediction	prediction	NOUN
fcis-14690	41	16	.	.	PUNCT
fcis-14690	42	1	in	in	ADP
fcis-14690	42	2	the	the	DET
fcis-14690	42	3	past	past	NOUN
fcis-14690	42	4	,	,	PUNCT
fcis-14690	42	5	a	a	DET
fcis-14690	42	6	lot	lot	NOUN
fcis-14690	42	7	of	of	ADP
fcis-14690	42	8	efforts	effort	NOUN
fcis-14690	42	9	have	have	AUX
fcis-14690	42	10	been	be	AUX
fcis-14690	42	11	focused	focus	VERB
fcis-14690	42	12	on	on	ADP
fcis-14690	42	13	feature	feature	NOUN
fcis-14690	42	14	extraction	extraction	NOUN
fcis-14690	42	15	.	.	PUNCT
fcis-14690	43	1	in	in	ADP
fcis-14690	43	2	earlier	early	ADJ
fcis-14690	43	3	studies	study	NOUN
fcis-14690	43	4	,	,	PUNCT
fcis-14690	43	5	image	image	NOUN
fcis-14690	43	6	sentiment	sentiment	NOUN
fcis-14690	43	7	analysis	analysis	NOUN
fcis-14690	43	8	methods	method	NOUN
fcis-14690	43	9	mainly	mainly	ADV
fcis-14690	43	10	classified	classify	VERB
fcis-14690	43	11	emotions	emotion	NOUN
fcis-14690	43	12	according	accord	VERB
fcis-14690	43	13	to	to	ADP
fcis-14690	43	14	manual	manual	ADJ
fcis-14690	43	15	61	61	NUM
fcis-14690	43	16	features	feature	NOUN
fcis-14690	43	17	of	of	ADP
fcis-14690	43	18	images	image	NOUN
fcis-14690	43	19	,	,	PUNCT
fcis-14690	43	20	such	such	ADJ
fcis-14690	43	21	as	as	ADP
fcis-14690	43	22	low	low	ADJ
fcis-14690	43	23	-	-	PUNCT
fcis-14690	43	24	level	level	NOUN
fcis-14690	43	25	features	feature	NOUN
fcis-14690	43	26	such	such	ADJ
fcis-14690	43	27	as	as	ADP
fcis-14690	43	28	color	color	NOUN
fcis-14690	43	29	and	and	CCONJ
fcis-14690	43	30	texture	texture	NOUN
fcis-14690	43	31	,	,	PUNCT
fcis-14690	43	32	intermediate	intermediate	ADJ
fcis-14690	43	33	features	feature	NOUN
fcis-14690	43	34	such	such	ADJ
fcis-14690	43	35	as	as	ADP
fcis-14690	43	36	image	image	NOUN
fcis-14690	43	37	composition	composition	NOUN
fcis-14690	43	38	and	and	CCONJ
fcis-14690	43	39	aesthetics	aesthetic	NOUN
fcis-14690	43	40	,	,	PUNCT
fcis-14690	43	41	and	and	CCONJ
fcis-14690	43	42	high	high	ADJ
fcis-14690	43	43	-	-	PUNCT
fcis-14690	43	44	level	level	NOUN
fcis-14690	43	45	features	feature	NOUN
fcis-14690	43	46	that	that	PRON
fcis-14690	43	47	contain	contain	VERB
fcis-14690	43	48	semantic	semantic	ADJ
fcis-14690	43	49	information	information	NOUN
fcis-14690	43	50	.	.	PUNCT
fcis-14690	44	1	inspired	inspire	VERB
fcis-14690	44	2	by	by	ADP
fcis-14690	44	3	the	the	DET
fcis-14690	44	4	fact	fact	NOUN
fcis-14690	44	5	that	that	SCONJ
fcis-14690	44	6	cnn	cnn	PROPN
fcis-14690	44	7	methods	method	NOUN
fcis-14690	44	8	work	work	VERB
fcis-14690	44	9	well	well	ADV
fcis-14690	44	10	in	in	ADP
fcis-14690	44	11	other	other	ADJ
fcis-14690	44	12	visual	visual	ADJ
fcis-14690	44	13	recognition	recognition	NOUN
fcis-14690	44	14	tasks	task	NOUN
fcis-14690	44	15	with	with	ADP
fcis-14690	44	16	the	the	DET
fcis-14690	44	17	development	development	NOUN
fcis-14690	44	18	of	of	ADP
fcis-14690	44	19	deep	deep	ADJ
fcis-14690	44	20	learning	learning	NOUN
fcis-14690	44	21	,	,	PUNCT
fcis-14690	44	22	researchers	researcher	NOUN
fcis-14690	44	23	began	begin	VERB
fcis-14690	44	24	to	to	PART
fcis-14690	44	25	apply	apply	VERB
fcis-14690	44	26	cnn	cnn	PROPN
fcis-14690	44	27	-	-	PUNCT
fcis-14690	44	28	based	base	VERB
fcis-14690	44	29	methods	method	NOUN
fcis-14690	44	30	to	to	PART
fcis-14690	44	31	image	image	VERB
fcis-14690	44	32	sentiment	sentiment	NOUN
fcis-14690	44	33	analysis	analysis	NOUN
fcis-14690	44	34	.	.	PUNCT
fcis-14690	45	1	the	the	DET
fcis-14690	45	2	cnn	cnn	PROPN
fcis-14690	45	3	-	-	PUNCT
fcis-14690	45	4	based	base	VERB
fcis-14690	45	5	approach	approach	NOUN
fcis-14690	45	6	can	can	AUX
fcis-14690	45	7	learn	learn	VERB
fcis-14690	45	8	features	feature	NOUN
fcis-14690	45	9	automatically	automatically	ADV
fcis-14690	45	10	through	through	ADP
fcis-14690	45	11	its	its	PRON
fcis-14690	45	12	multi	multi	ADJ
fcis-14690	45	13	-	-	ADJ
fcis-14690	45	14	level	level	ADJ
fcis-14690	45	15	deep	deep	ADJ
fcis-14690	45	16	learning	learning	NOUN
fcis-14690	45	17	architecture	architecture	NOUN
fcis-14690	45	18	,	,	PUNCT
fcis-14690	45	19	rather	rather	ADV
fcis-14690	45	20	than	than	ADP
fcis-14690	45	21	manually	manually	ADV
fcis-14690	45	22	designing	design	VERB
fcis-14690	45	23	image	image	NOUN
fcis-14690	45	24	features	feature	NOUN
fcis-14690	45	25	.	.	PUNCT
fcis-14690	46	1	peng	peng	PROPN
fcis-14690	46	2	et	et	PROPN
fcis-14690	46	3	al	al	PROPN
fcis-14690	46	4	.	.	PUNCT
fcis-14690	47	1	[	[	X
fcis-14690	47	2	7	7	X
fcis-14690	47	3	]	]	PUNCT
fcis-14690	47	4	tried	try	VERB
fcis-14690	47	5	to	to	PART
fcis-14690	47	6	apply	apply	VERB
fcis-14690	47	7	cnn	cnn	PROPN
fcis-14690	47	8	model	model	NOUN
fcis-14690	47	9	to	to	PART
fcis-14690	47	10	image	image	VERB
fcis-14690	47	11	emotion	emotion	NOUN
fcis-14690	47	12	recognition	recognition	NOUN
fcis-14690	47	13	for	for	ADP
fcis-14690	47	14	the	the	DET
fcis-14690	47	15	first	first	ADJ
fcis-14690	47	16	time	time	NOUN
fcis-14690	47	17	.	.	PUNCT
fcis-14690	48	1	they	they	PRON
fcis-14690	48	2	fine	fine	ADV
fcis-14690	48	3	-	-	PUNCT
fcis-14690	48	4	tuned	tune	VERB
fcis-14690	48	5	the	the	DET
fcis-14690	48	6	pre	pre	ADJ
fcis-14690	48	7	-	-	ADJ
fcis-14690	48	8	trained	train	VERB
fcis-14690	48	9	cnn	cnn	PROPN
fcis-14690	48	10	on	on	ADP
fcis-14690	48	11	imagenet[16	imagenet[16	PROPN
fcis-14690	48	12	]	]	PUNCT
fcis-14690	48	13	,	,	PUNCT
fcis-14690	48	14	showing	show	VERB
fcis-14690	48	15	that	that	SCONJ
fcis-14690	48	16	the	the	DET
fcis-14690	48	17	cnn	cnn	PROPN
fcis-14690	48	18	model	model	NOUN
fcis-14690	48	19	is	be	AUX
fcis-14690	48	20	superior	superior	ADJ
fcis-14690	48	21	to	to	ADP
fcis-14690	48	22	the	the	DET
fcis-14690	48	23	manual	manual	ADJ
fcis-14690	48	24	method	method	NOUN
fcis-14690	48	25	on	on	ADP
fcis-14690	48	26	the	the	DET
fcis-14690	48	27	emotion6	emotion6	NOUN
fcis-14690	48	28	dataset	dataset	NOUN
fcis-14690	48	29	.	.	PUNCT
fcis-14690	49	1	you	you	PRON
fcis-14690	49	2	et	et	VERB
fcis-14690	49	3	al	al	PROPN
fcis-14690	49	4	.	.	PUNCT
fcis-14690	50	1	[	[	X
fcis-14690	50	2	17	17	NUM
fcis-14690	50	3	]	]	PUNCT
fcis-14690	50	4	combined	combine	VERB
fcis-14690	50	5	the	the	DET
fcis-14690	50	6	cnn	cnn	PROPN
fcis-14690	50	7	model	model	NOUN
fcis-14690	50	8	in	in	ADP
fcis-14690	50	9	literature	literature	NOUN
fcis-14690	50	10	[	[	X
fcis-14690	50	11	18	18	NUM
fcis-14690	50	12	]	]	PUNCT
fcis-14690	50	13	with	with	ADP
fcis-14690	50	14	support	support	NOUN
fcis-14690	50	15	vector	vector	NOUN
fcis-14690	50	16	machine	machine	NOUN
fcis-14690	50	17	(	(	PUNCT
fcis-14690	50	18	svm	svm	PROPN
fcis-14690	50	19	)	)	PUNCT
fcis-14690	50	20	to	to	PART
fcis-14690	50	21	detect	detect	VERB
fcis-14690	50	22	image	image	NOUN
fcis-14690	50	23	emotion	emotion	NOUN
fcis-14690	50	24	on	on	ADP
fcis-14690	50	25	a	a	DET
fcis-14690	50	26	web	web	NOUN
fcis-14690	50	27	image	image	NOUN
fcis-14690	50	28	dataset	dataset	NOUN
fcis-14690	50	29	.	.	PUNCT
fcis-14690	51	1	they	they	PRON
fcis-14690	51	2	demonstrated	demonstrate	VERB
fcis-14690	51	3	that	that	SCONJ
fcis-14690	51	4	the	the	DET
fcis-14690	51	5	cnn	cnn	PROPN
fcis-14690	51	6	method	method	NOUN
fcis-14690	51	7	can	can	AUX
fcis-14690	51	8	capture	capture	VERB
fcis-14690	51	9	more	more	ADV
fcis-14690	51	10	advanced	advanced	ADJ
fcis-14690	51	11	emotional	emotional	ADJ
fcis-14690	51	12	features	feature	NOUN
fcis-14690	51	13	than	than	ADP
fcis-14690	51	14	manual	manual	ADJ
fcis-14690	51	15	methods	method	NOUN
fcis-14690	51	16	.	.	PUNCT
fcis-14690	52	1	image	image	NOUN
fcis-14690	52	2	emotion	emotion	NOUN
fcis-14690	52	3	representation	representation	NOUN
fcis-14690	52	4	is	be	AUX
fcis-14690	52	5	related	relate	VERB
fcis-14690	52	6	to	to	ADP
fcis-14690	52	7	both	both	PRON
fcis-14690	52	8	lowlevel	lowlevel	VERB
fcis-14690	52	9	and	and	CCONJ
fcis-14690	52	10	high	high	ADJ
fcis-14690	52	11	-	-	PUNCT
fcis-14690	52	12	level	level	NOUN
fcis-14690	52	13	features	feature	NOUN
fcis-14690	52	14	,	,	PUNCT
fcis-14690	52	15	but	but	CCONJ
fcis-14690	52	16	these	these	DET
fcis-14690	52	17	methods	method	NOUN
fcis-14690	52	18	only	only	ADV
fcis-14690	52	19	use	use	VERB
fcis-14690	52	20	the	the	DET
fcis-14690	52	21	last	last	ADJ
fcis-14690	52	22	level	level	NOUN
fcis-14690	52	23	of	of	ADP
fcis-14690	52	24	cnn	cnn	PROPN
fcis-14690	52	25	high	high	ADJ
fcis-14690	52	26	-	-	PUNCT
fcis-14690	52	27	level	level	NOUN
fcis-14690	52	28	semantic	semantic	ADJ
fcis-14690	52	29	feature	feature	NOUN
fcis-14690	52	30	vector	vector	NOUN
fcis-14690	52	31	,	,	PUNCT
fcis-14690	52	32	ignoring	ignore	VERB
fcis-14690	52	33	the	the	DET
fcis-14690	52	34	importance	importance	NOUN
fcis-14690	52	35	of	of	ADP
fcis-14690	52	36	low	low	ADJ
fcis-14690	52	37	-	-	PUNCT
fcis-14690	52	38	level	level	NOUN
fcis-14690	52	39	features	feature	NOUN
fcis-14690	52	40	.	.	PUNCT
fcis-14690	53	1	some	some	DET
fcis-14690	53	2	researchers	researcher	NOUN
fcis-14690	53	3	combine	combine	VERB
fcis-14690	53	4	higher	high	ADJ
fcis-14690	53	5	-	-	PUNCT
fcis-14690	53	6	level	level	NOUN
fcis-14690	53	7	semantic	semantic	ADJ
fcis-14690	53	8	information	information	NOUN
fcis-14690	53	9	with	with	ADP
fcis-14690	53	10	lower	low	ADJ
fcis-14690	53	11	-	-	PUNCT
fcis-14690	53	12	level	level	NOUN
fcis-14690	53	13	visual	visual	ADJ
fcis-14690	53	14	features	feature	NOUN
fcis-14690	53	15	in	in	ADP
fcis-14690	53	16	different	different	ADJ
fcis-14690	53	17	ways	way	NOUN
fcis-14690	53	18	to	to	PART
fcis-14690	53	19	guide	guide	VERB
fcis-14690	53	20	emotion	emotion	NOUN
fcis-14690	53	21	classification	classification	NOUN
fcis-14690	53	22	.	.	PUNCT
fcis-14690	54	1	zhu	zhu	PROPN
fcis-14690	54	2	et	et	PROPN
fcis-14690	54	3	al	al	PROPN
fcis-14690	54	4	.	.	PUNCT
fcis-14690	55	1	[	[	X
fcis-14690	55	2	14	14	NUM
fcis-14690	55	3	]	]	PUNCT
fcis-14690	55	4	designed	design	VERB
fcis-14690	55	5	a	a	DET
fcis-14690	55	6	cnn	cnn	PROPN
fcis-14690	55	7	-	-	PUNCT
fcis-14690	55	8	rnn	rnn	PROPN
fcis-14690	55	9	model	model	NOUN
fcis-14690	55	10	to	to	PART
fcis-14690	55	11	extract	extract	VERB
fcis-14690	55	12	visual	visual	ADJ
fcis-14690	55	13	and	and	CCONJ
fcis-14690	55	14	semantic	semantic	ADJ
fcis-14690	55	15	features	feature	NOUN
fcis-14690	55	16	through	through	ADP
fcis-14690	55	17	underlying	underlie	VERB
fcis-14690	55	18	convolution	convolution	NOUN
fcis-14690	55	19	,	,	PUNCT
fcis-14690	55	20	and	and	CCONJ
fcis-14690	55	21	then	then	ADV
fcis-14690	55	22	aggregate	aggregate	VERB
fcis-14690	55	23	them	they	PRON
fcis-14690	55	24	using	use	VERB
fcis-14690	55	25	bidirectional	bidirectional	ADJ
fcis-14690	55	26	cyclic	cyclic	ADJ
fcis-14690	55	27	convolutional	convolutional	ADJ
fcis-14690	55	28	neural	neural	ADJ
fcis-14690	55	29	networks	network	NOUN
fcis-14690	55	30	(	(	PUNCT
fcis-14690	55	31	birnn	birnn	NOUN
fcis-14690	55	32	)	)	PUNCT
fcis-14690	55	33	.	.	PUNCT
fcis-14690	56	1	rao	rao	PROPN
fcis-14690	56	2	et	et	PROPN
fcis-14690	56	3	al	al	PROPN
fcis-14690	56	4	.	.	PUNCT
fcis-14690	57	1	[	[	X
fcis-14690	57	2	19	19	NUM
fcis-14690	57	3	]	]	PUNCT
fcis-14690	57	4	used	use	VERB
fcis-14690	57	5	three	three	NUM
fcis-14690	57	6	kinds	kind	NOUN
fcis-14690	57	7	of	of	ADP
fcis-14690	57	8	convolutional	convolutional	ADJ
fcis-14690	57	9	neural	neural	ADJ
fcis-14690	57	10	networks	network	NOUN
fcis-14690	57	11	to	to	PART
fcis-14690	57	12	obtain	obtain	VERB
fcis-14690	57	13	the	the	DET
fcis-14690	57	14	three	three	NUM
fcis-14690	57	15	hierarchical	hierarchical	ADJ
fcis-14690	57	16	features	feature	NOUN
fcis-14690	57	17	of	of	ADP
fcis-14690	57	18	the	the	DET
fcis-14690	57	19	original	original	ADJ
fcis-14690	57	20	image	image	NOUN
fcis-14690	57	21	,	,	PUNCT
fcis-14690	57	22	the	the	DET
fcis-14690	57	23	prominent	prominent	ADJ
fcis-14690	57	24	theme	theme	NOUN
fcis-14690	57	25	and	and	CCONJ
fcis-14690	57	26	the	the	DET
fcis-14690	57	27	color	color	NOUN
fcis-14690	57	28	respectively	respectively	ADV
fcis-14690	57	29	,	,	PUNCT
fcis-14690	57	30	and	and	CCONJ
fcis-14690	57	31	conducted	conduct	VERB
fcis-14690	57	32	a	a	DET
fcis-14690	57	33	comprehensive	comprehensive	ADJ
fcis-14690	57	34	analysis	analysis	NOUN
fcis-14690	57	35	of	of	ADP
fcis-14690	57	36	the	the	DET
fcis-14690	57	37	emotion	emotion	NOUN
fcis-14690	57	38	categories	category	NOUN
fcis-14690	57	39	,	,	PUNCT
fcis-14690	57	40	achieving	achieve	VERB
fcis-14690	57	41	good	good	ADJ
fcis-14690	57	42	results	result	NOUN
fcis-14690	57	43	.	.	PUNCT
fcis-14690	58	1	however	however	ADV
fcis-14690	58	2	,	,	PUNCT
fcis-14690	58	3	these	these	DET
fcis-14690	58	4	methods	method	NOUN
fcis-14690	58	5	often	often	ADV
fcis-14690	58	6	use	use	VERB
fcis-14690	58	7	aggregation	aggregation	NOUN
fcis-14690	58	8	functions	function	NOUN
fcis-14690	58	9	to	to	PART
fcis-14690	58	10	fuse	fuse	VERB
fcis-14690	58	11	the	the	DET
fcis-14690	58	12	features	feature	NOUN
fcis-14690	58	13	of	of	ADP
fcis-14690	58	14	each	each	DET
fcis-14690	58	15	level	level	NOUN
fcis-14690	58	16	of	of	ADP
fcis-14690	58	17	the	the	DET
fcis-14690	58	18	image	image	NOUN
fcis-14690	58	19	for	for	ADP
fcis-14690	58	20	emotion	emotion	NOUN
fcis-14690	58	21	classification	classification	NOUN
fcis-14690	58	22	.	.	PUNCT
fcis-14690	59	1	sowmyayani	sowmyayani	NOUN
fcis-14690	59	2	and	and	CCONJ
fcis-14690	59	3	rani	rani	PROPN
fcis-14690	59	4	[	[	X
fcis-14690	59	5	20	20	NUM
fcis-14690	59	6	]	]	PUNCT
fcis-14690	59	7	point	point	NOUN
fcis-14690	59	8	out	out	ADP
fcis-14690	59	9	that	that	SCONJ
fcis-14690	59	10	prominent	prominent	ADJ
fcis-14690	59	11	objects	object	NOUN
fcis-14690	59	12	in	in	ADP
fcis-14690	59	13	images	image	NOUN
fcis-14690	59	14	play	play	VERB
fcis-14690	59	15	an	an	DET
fcis-14690	59	16	important	important	ADJ
fcis-14690	59	17	role	role	NOUN
fcis-14690	59	18	in	in	ADP
fcis-14690	59	19	determining	determine	VERB
fcis-14690	59	20	emotion	emotion	NOUN
fcis-14690	59	21	.	.	PUNCT
fcis-14690	60	1	first	first	ADV
fcis-14690	60	2	,	,	PUNCT
fcis-14690	60	3	a	a	DET
fcis-14690	60	4	spectral	spectral	ADJ
fcis-14690	60	5	significance	significance	NOUN
fcis-14690	60	6	detection	detection	NOUN
fcis-14690	60	7	model	model	NOUN
fcis-14690	60	8	is	be	AUX
fcis-14690	60	9	used	use	VERB
fcis-14690	60	10	to	to	PART
fcis-14690	60	11	detect	detect	VERB
fcis-14690	60	12	significant	significant	ADJ
fcis-14690	60	13	objects	object	NOUN
fcis-14690	60	14	from	from	ADP
fcis-14690	60	15	the	the	DET
fcis-14690	60	16	entire	entire	ADJ
fcis-14690	60	17	image	image	NOUN
fcis-14690	60	18	,	,	PUNCT
fcis-14690	60	19	and	and	CCONJ
fcis-14690	60	20	then	then	ADV
fcis-14690	60	21	depth	depth	NOUN
fcis-14690	60	22	features	feature	NOUN
fcis-14690	60	23	and	and	CCONJ
fcis-14690	60	24	manual	manual	ADJ
fcis-14690	60	25	features	feature	NOUN
fcis-14690	60	26	of	of	ADP
fcis-14690	60	27	significant	significant	ADJ
fcis-14690	60	28	objects	object	NOUN
fcis-14690	60	29	are	be	AUX
fcis-14690	60	30	extracted	extract	VERB
fcis-14690	60	31	.	.	PUNCT
fcis-14690	61	1	however	however	ADV
fcis-14690	61	2	,	,	PUNCT
fcis-14690	61	3	prominent	prominent	ADJ
fcis-14690	61	4	objects	object	NOUN
fcis-14690	61	5	are	be	AUX
fcis-14690	61	6	important	important	ADJ
fcis-14690	61	7	for	for	ADP
fcis-14690	61	8	emotion	emotion	NOUN
fcis-14690	61	9	classification	classification	NOUN
fcis-14690	61	10	research	research	NOUN
fcis-14690	61	11	,	,	PUNCT
fcis-14690	61	12	but	but	CCONJ
fcis-14690	61	13	other	other	ADJ
fcis-14690	61	14	parts	part	NOUN
fcis-14690	61	15	of	of	ADP
fcis-14690	61	16	the	the	DET
fcis-14690	61	17	image	image	NOUN
fcis-14690	61	18	can	can	AUX
fcis-14690	61	19	not	not	PART
fcis-14690	61	20	be	be	AUX
fcis-14690	61	21	ignored	ignore	VERB
fcis-14690	61	22	,	,	PUNCT
fcis-14690	61	23	and	and	CCONJ
fcis-14690	61	24	some	some	DET
fcis-14690	61	25	images	image	NOUN
fcis-14690	61	26	are	be	AUX
fcis-14690	61	27	difficult	difficult	ADJ
fcis-14690	61	28	to	to	PART
fcis-14690	61	29	extract	extract	VERB
fcis-14690	61	30	prominent	prominent	ADJ
fcis-14690	61	31	objects	object	NOUN
fcis-14690	61	32	.	.	PUNCT
fcis-14690	62	1	the	the	DET
fcis-14690	62	2	above	above	ADJ
fcis-14690	62	3	methods	method	NOUN
fcis-14690	62	4	based	base	VERB
fcis-14690	62	5	on	on	ADP
fcis-14690	62	6	manual	manual	ADJ
fcis-14690	62	7	features	feature	NOUN
fcis-14690	62	8	and	and	CCONJ
fcis-14690	62	9	cnn	cnn	PROPN
fcis-14690	62	10	model	model	NOUN
fcis-14690	62	11	independently	independently	ADV
fcis-14690	62	12	consider	consider	VERB
fcis-14690	62	13	the	the	DET
fcis-14690	62	14	global	global	ADJ
fcis-14690	62	15	features	feature	NOUN
fcis-14690	62	16	of	of	ADP
fcis-14690	62	17	images	image	NOUN
fcis-14690	62	18	or	or	CCONJ
fcis-14690	62	19	the	the	DET
fcis-14690	62	20	features	feature	NOUN
fcis-14690	62	21	of	of	ADP
fcis-14690	62	22	significant	significant	ADJ
fcis-14690	62	23	objects	object	NOUN
fcis-14690	62	24	to	to	PART
fcis-14690	62	25	predict	predict	VERB
fcis-14690	62	26	emotion	emotion	NOUN
fcis-14690	62	27	.	.	PUNCT
fcis-14690	63	1	however	however	ADV
fcis-14690	63	2	,	,	PUNCT
fcis-14690	63	3	zhao	zhao	PROPN
fcis-14690	63	4	et	et	PROPN
fcis-14690	63	5	al	al	PROPN
fcis-14690	63	6	.	.	PUNCT
fcis-14690	64	1	[	[	X
fcis-14690	64	2	10	10	NUM
fcis-14690	64	3	]	]	PUNCT
fcis-14690	64	4	show	show	NOUN
fcis-14690	64	5	through	through	ADP
fcis-14690	64	6	experiments	experiment	NOUN
fcis-14690	64	7	that	that	SCONJ
fcis-14690	64	8	it	it	PRON
fcis-14690	64	9	is	be	AUX
fcis-14690	64	10	better	well	ADJ
fcis-14690	64	11	to	to	PART
fcis-14690	64	12	use	use	VERB
fcis-14690	64	13	local	local	ADJ
fcis-14690	64	14	and	and	CCONJ
fcis-14690	64	15	global	global	ADJ
fcis-14690	64	16	features	feature	NOUN
fcis-14690	64	17	at	at	ADP
fcis-14690	64	18	the	the	DET
fcis-14690	64	19	same	same	ADJ
fcis-14690	64	20	time	time	NOUN
fcis-14690	64	21	than	than	SCONJ
fcis-14690	64	22	to	to	PART
fcis-14690	64	23	use	use	VERB
fcis-14690	64	24	one	one	NUM
fcis-14690	64	25	of	of	ADP
fcis-14690	64	26	them	they	PRON
fcis-14690	64	27	alone	alone	ADV
fcis-14690	64	28	.	.	PUNCT
fcis-14690	65	1	therefore	therefore	ADV
fcis-14690	65	2	,	,	PUNCT
fcis-14690	65	3	multi	multi	ADJ
fcis-14690	65	4	-	-	ADJ
fcis-14690	65	5	level	level	ADJ
fcis-14690	65	6	global	global	ADJ
fcis-14690	65	7	features	feature	NOUN
fcis-14690	65	8	and	and	CCONJ
fcis-14690	65	9	local	local	ADJ
fcis-14690	65	10	regional	regional	ADJ
fcis-14690	65	11	features	feature	NOUN
fcis-14690	65	12	can	can	AUX
fcis-14690	65	13	be	be	AUX
fcis-14690	65	14	combined	combine	VERB
fcis-14690	65	15	to	to	PART
fcis-14690	65	16	classify	classify	VERB
fcis-14690	65	17	image	image	NOUN
fcis-14690	65	18	emotion	emotion	NOUN
fcis-14690	65	19	.	.	PUNCT
fcis-14690	66	1	2.2	2.2	NUM
fcis-14690	66	2	.	.	PUNCT
fcis-14690	67	1	mixed	mix	VERB
fcis-14690	67	2	attention	attention	NOUN
fcis-14690	67	3	mechanism	mechanism	NOUN
fcis-14690	67	4	the	the	DET
fcis-14690	67	5	basic	basic	ADJ
fcis-14690	67	6	idea	idea	NOUN
fcis-14690	67	7	of	of	ADP
fcis-14690	67	8	the	the	DET
fcis-14690	67	9	attention	attention	NOUN
fcis-14690	67	10	mechanism	mechanism	NOUN
fcis-14690	67	11	is	be	AUX
fcis-14690	67	12	to	to	PART
fcis-14690	67	13	make	make	VERB
fcis-14690	67	14	the	the	DET
fcis-14690	67	15	system	system	NOUN
fcis-14690	67	16	ignore	ignore	VERB
fcis-14690	67	17	irrelevant	irrelevant	ADJ
fcis-14690	67	18	information	information	NOUN
fcis-14690	67	19	and	and	CCONJ
fcis-14690	67	20	focus	focus	VERB
fcis-14690	67	21	on	on	ADP
fcis-14690	67	22	important	important	ADJ
fcis-14690	67	23	information	information	NOUN
fcis-14690	67	24	.	.	PUNCT
fcis-14690	68	1	with	with	ADP
fcis-14690	68	2	the	the	DET
fcis-14690	68	3	development	development	NOUN
fcis-14690	68	4	of	of	ADP
fcis-14690	68	5	deep	deep	ADJ
fcis-14690	68	6	learning	learning	NOUN
fcis-14690	68	7	,	,	PUNCT
fcis-14690	68	8	attention	attention	NOUN
fcis-14690	68	9	mechanism	mechanism	NOUN
fcis-14690	68	10	has	have	AUX
fcis-14690	68	11	been	be	AUX
fcis-14690	68	12	widely	widely	ADV
fcis-14690	68	13	applied	apply	VERB
fcis-14690	68	14	in	in	ADP
fcis-14690	68	15	the	the	DET
fcis-14690	68	16	field	field	NOUN
fcis-14690	68	17	of	of	ADP
fcis-14690	68	18	computer	computer	NOUN
fcis-14690	68	19	vision	vision	NOUN
fcis-14690	68	20	,	,	PUNCT
fcis-14690	68	21	such	such	ADJ
fcis-14690	68	22	as	as	ADP
fcis-14690	68	23	object	object	NOUN
fcis-14690	68	24	detection	detection	NOUN
fcis-14690	68	25	[	[	X
fcis-14690	68	26	21	21	NUM
fcis-14690	68	27	]	]	PUNCT
fcis-14690	68	28	,	,	PUNCT
fcis-14690	68	29	image	image	NOUN
fcis-14690	68	30	classification	classification	NOUN
fcis-14690	69	1	[	[	X
fcis-14690	69	2	22	22	NUM
fcis-14690	69	3	]	]	PUNCT
fcis-14690	69	4	,	,	PUNCT
fcis-14690	69	5	image	image	NOUN
fcis-14690	69	6	captions	caption	NOUN
fcis-14690	69	7	generation	generation	NOUN
fcis-14690	70	1	[	[	X
fcis-14690	70	2	23	23	NUM
fcis-14690	70	3	]	]	PUNCT
fcis-14690	70	4	,	,	PUNCT
fcis-14690	70	5	etc	etc	X
fcis-14690	70	6	.	.	X
fcis-14690	71	1	there	there	PRON
fcis-14690	71	2	have	have	AUX
fcis-14690	71	3	been	be	AUX
fcis-14690	71	4	some	some	DET
fcis-14690	71	5	studies	study	NOUN
fcis-14690	71	6	on	on	ADP
fcis-14690	71	7	integrated	integrate	VERB
fcis-14690	71	8	visual	visual	ADJ
fcis-14690	71	9	attention	attention	NOUN
fcis-14690	71	10	to	to	ADP
fcis-14690	71	11	the	the	DET
fcis-14690	71	12	cnn	cnn	PROPN
fcis-14690	71	13	emotion	emotion	NOUN
fcis-14690	71	14	classification	classification	NOUN
fcis-14690	71	15	framework	framework	NOUN
fcis-14690	71	16	.	.	PUNCT
fcis-14690	72	1	song	song	NOUN
fcis-14690	72	2	et	et	PROPN
fcis-14690	72	3	al	al	PROPN
fcis-14690	72	4	.	.	PUNCT
fcis-14690	73	1	[	[	X
fcis-14690	73	2	24	24	NUM
fcis-14690	73	3	]	]	PUNCT
fcis-14690	73	4	proposed	propose	VERB
fcis-14690	73	5	an	an	DET
fcis-14690	73	6	emotion	emotion	NOUN
fcis-14690	73	7	network	network	NOUN
fcis-14690	73	8	with	with	ADP
fcis-14690	73	9	visual	visual	ADJ
fcis-14690	73	10	attention	attention	NOUN
fcis-14690	73	11	,	,	PUNCT
fcis-14690	73	12	integrating	integrate	VERB
fcis-14690	73	13	visual	visual	ADJ
fcis-14690	73	14	attention	attention	NOUN
fcis-14690	73	15	into	into	ADP
fcis-14690	73	16	the	the	DET
fcis-14690	73	17	emotion	emotion	NOUN
fcis-14690	73	18	classification	classification	NOUN
fcis-14690	73	19	framework	framework	NOUN
fcis-14690	73	20	to	to	PART
fcis-14690	73	21	locate	locate	VERB
fcis-14690	73	22	local	local	ADJ
fcis-14690	73	23	areas	area	NOUN
fcis-14690	73	24	related	relate	VERB
fcis-14690	73	25	to	to	ADP
fcis-14690	73	26	emotion	emotion	NOUN
fcis-14690	73	27	.	.	PUNCT
fcis-14690	74	1	yang	yang	PROPN
fcis-14690	74	2	et	et	PROPN
fcis-14690	74	3	al	al	PROPN
fcis-14690	74	4	.	.	PUNCT
fcis-14690	75	1	[	[	X
fcis-14690	75	2	25	25	NUM
fcis-14690	75	3	]	]	PUNCT
fcis-14690	75	4	proposed	propose	VERB
fcis-14690	75	5	a	a	DET
fcis-14690	75	6	weakly	weakly	ADV
fcis-14690	75	7	supervised	supervised	ADJ
fcis-14690	75	8	model	model	NOUN
fcis-14690	75	9	coupled	couple	VERB
fcis-14690	75	10	with	with	ADP
fcis-14690	75	11	attention	attention	NOUN
fcis-14690	75	12	mechanism	mechanism	NOUN
fcis-14690	75	13	,	,	PUNCT
fcis-14690	75	14	which	which	PRON
fcis-14690	75	15	combined	combine	VERB
fcis-14690	75	16	visual	visual	ADJ
fcis-14690	75	17	emotion	emotion	NOUN
fcis-14690	75	18	detection	detection	NOUN
fcis-14690	75	19	and	and	CCONJ
fcis-14690	75	20	classification	classification	NOUN
fcis-14690	75	21	within	within	ADP
fcis-14690	75	22	a	a	DET
fcis-14690	75	23	unified	unified	ADJ
fcis-14690	75	24	cnn	cnn	PROPN
fcis-14690	75	25	framework	framework	NOUN
fcis-14690	75	26	.	.	PUNCT
fcis-14690	76	1	these	these	DET
fcis-14690	76	2	attention	attention	NOUN
fcis-14690	76	3	mechanisms	mechanism	NOUN
fcis-14690	76	4	mainly	mainly	ADV
fcis-14690	76	5	focus	focus	VERB
fcis-14690	76	6	on	on	ADP
fcis-14690	76	7	spatial	spatial	ADJ
fcis-14690	76	8	attention	attention	NOUN
fcis-14690	76	9	and	and	CCONJ
fcis-14690	76	10	ignore	ignore	VERB
fcis-14690	76	11	channel	channel	NOUN
fcis-14690	76	12	attention	attention	NOUN
fcis-14690	76	13	,	,	PUNCT
fcis-14690	76	14	but	but	CCONJ
fcis-14690	76	15	channel	channel	NOUN
fcis-14690	76	16	attention	attention	NOUN
fcis-14690	76	17	is	be	AUX
fcis-14690	76	18	also	also	ADV
fcis-14690	76	19	a	a	DET
fcis-14690	76	20	process	process	NOUN
fcis-14690	76	21	of	of	ADP
fcis-14690	76	22	selecting	select	VERB
fcis-14690	76	23	semantic	semantic	ADJ
fcis-14690	76	24	attributes	attribute	NOUN
fcis-14690	76	25	.	.	PUNCT
fcis-14690	77	1	zhao	zhao	PROPN
fcis-14690	77	2	et	et	PROPN
fcis-14690	77	3	al	al	PROPN
fcis-14690	77	4	.	.	PUNCT
fcis-14690	78	1	[	[	X
fcis-14690	78	2	26	26	NUM
fcis-14690	78	3	]	]	PUNCT
fcis-14690	78	4	established	establish	VERB
fcis-14690	78	5	a	a	DET
fcis-14690	78	6	deep	deep	ADJ
fcis-14690	78	7	attention	attention	NOUN
fcis-14690	78	8	network	network	NOUN
fcis-14690	78	9	that	that	PRON
fcis-14690	78	10	integrates	integrate	VERB
fcis-14690	78	11	spatial	spatial	ADJ
fcis-14690	78	12	attention	attention	NOUN
fcis-14690	78	13	and	and	CCONJ
fcis-14690	78	14	channel	channel	NOUN
fcis-14690	78	15	attention	attention	NOUN
fcis-14690	78	16	with	with	ADP
fcis-14690	78	17	the	the	DET
fcis-14690	78	18	same	same	ADJ
fcis-14690	78	19	polarity	polarity	NOUN
fcis-14690	78	20	for	for	ADP
fcis-14690	78	21	image	image	NOUN
fcis-14690	78	22	emotion	emotion	NOUN
fcis-14690	78	23	regression	regression	NOUN
fcis-14690	78	24	.	.	PUNCT
fcis-14690	79	1	in	in	ADP
fcis-14690	79	2	order	order	NOUN
fcis-14690	79	3	to	to	PART
fcis-14690	79	4	make	make	VERB
fcis-14690	79	5	full	full	ADJ
fcis-14690	79	6	use	use	NOUN
fcis-14690	79	7	of	of	ADP
fcis-14690	79	8	the	the	DET
fcis-14690	79	9	multi	multi	ADJ
fcis-14690	79	10	-	-	NOUN
fcis-14690	79	11	level	level	ADJ
fcis-14690	79	12	,	,	PUNCT
fcis-14690	79	13	spatial	spatial	ADJ
fcis-14690	79	14	and	and	CCONJ
fcis-14690	79	15	channel	channel	NOUN
fcis-14690	79	16	features	feature	NOUN
fcis-14690	79	17	of	of	ADP
fcis-14690	79	18	cnn	cnn	PROPN
fcis-14690	79	19	,	,	PUNCT
fcis-14690	79	20	li	li	PROPN
fcis-14690	79	21	et	et	PROPN
fcis-14690	79	22	al	al	PROPN
fcis-14690	79	23	.	.	PUNCT
fcis-14690	80	1	[	[	X
fcis-14690	80	2	6	6	NUM
fcis-14690	80	3	]	]	PUNCT
fcis-14690	80	4	proposed	propose	VERB
fcis-14690	80	5	a	a	DET
fcis-14690	80	6	scep	scep	NOUN
fcis-14690	80	7	model	model	NOUN
fcis-14690	80	8	,	,	PUNCT
fcis-14690	80	9	integrating	integrate	VERB
fcis-14690	80	10	spatial	spatial	ADJ
fcis-14690	80	11	attention	attention	NOUN
fcis-14690	80	12	and	and	CCONJ
fcis-14690	80	13	channel	channel	NOUN
fcis-14690	80	14	attention	attention	NOUN
fcis-14690	80	15	mechanisms	mechanism	NOUN
fcis-14690	80	16	into	into	ADP
fcis-14690	80	17	a	a	DET
fcis-14690	80	18	classical	classical	ADJ
fcis-14690	80	19	convolutional	convolutional	ADJ
fcis-14690	80	20	neural	neural	ADJ
fcis-14690	80	21	network	network	NOUN
fcis-14690	80	22	layer	layer	NOUN
fcis-14690	80	23	structure	structure	NOUN
fcis-14690	80	24	to	to	PART
fcis-14690	80	25	predict	predict	VERB
fcis-14690	80	26	image	image	NOUN
fcis-14690	80	27	emotion	emotion	NOUN
fcis-14690	80	28	.	.	PUNCT
fcis-14690	81	1	in	in	ADP
fcis-14690	81	2	2018	2018	NUM
fcis-14690	81	3	,	,	PUNCT
fcis-14690	81	4	woo	woo	VERB
fcis-14690	81	5	s	s	PRON
fcis-14690	81	6	et	et	NOUN
fcis-14690	81	7	al	al	PROPN
fcis-14690	81	8	.	.	PUNCT
fcis-14690	82	1	[	[	X
fcis-14690	82	2	27	27	NUM
fcis-14690	82	3	]	]	PUNCT
fcis-14690	82	4	proved	prove	VERB
fcis-14690	82	5	through	through	ADP
fcis-14690	82	6	experiments	experiment	NOUN
fcis-14690	82	7	that	that	SCONJ
fcis-14690	82	8	the	the	DET
fcis-14690	82	9	mode	mode	NOUN
fcis-14690	82	10	of	of	ADP
fcis-14690	82	11	channeling	channel	VERB
fcis-14690	82	12	attention	attention	NOUN
fcis-14690	82	13	module	module	NOUN
fcis-14690	82	14	first	first	ADV
fcis-14690	82	15	and	and	CCONJ
fcis-14690	82	16	then	then	ADV
fcis-14690	82	17	spatial	spatial	ADJ
fcis-14690	82	18	attention	attention	NOUN
fcis-14690	82	19	module	module	NOUN
fcis-14690	82	20	is	be	AUX
fcis-14690	82	21	more	more	ADV
fcis-14690	82	22	effective	effective	ADJ
fcis-14690	82	23	.	.	PUNCT
fcis-14690	83	1	therefore	therefore	ADV
fcis-14690	83	2	,	,	PUNCT
fcis-14690	83	3	combining	combine	VERB
fcis-14690	83	4	the	the	DET
fcis-14690	83	5	attention	attention	NOUN
fcis-14690	83	6	mechanism	mechanism	NOUN
fcis-14690	83	7	in	in	ADP
fcis-14690	83	8	channel	channel	NOUN
fcis-14690	83	9	and	and	CCONJ
fcis-14690	83	10	space	space	NOUN
fcis-14690	83	11	,	,	PUNCT
fcis-14690	83	12	they	they	PRON
fcis-14690	83	13	proposed	propose	VERB
fcis-14690	83	14	a	a	DET
fcis-14690	83	15	lightweight	lightweight	ADJ
fcis-14690	83	16	and	and	CCONJ
fcis-14690	83	17	universal	universal	ADJ
fcis-14690	83	18	modular	modular	ADJ
fcis-14690	83	19	hybrid	hybrid	ADJ
fcis-14690	83	20	attention	attention	NOUN
fcis-14690	83	21	mechanism	mechanism	NOUN
fcis-14690	83	22	model	model	NOUN
fcis-14690	83	23	cbam	cbam	NOUN
fcis-14690	83	24	.	.	PUNCT
fcis-14690	84	1	the	the	DET
fcis-14690	84	2	existing	exist	VERB
fcis-14690	84	3	researches	research	NOUN
fcis-14690	84	4	pay	pay	VERB
fcis-14690	84	5	too	too	ADV
fcis-14690	84	6	much	much	ADJ
fcis-14690	84	7	attention	attention	NOUN
fcis-14690	84	8	to	to	ADP
fcis-14690	84	9	the	the	DET
fcis-14690	84	10	spatial	spatial	ADJ
fcis-14690	84	11	attention	attention	NOUN
fcis-14690	84	12	but	but	CCONJ
fcis-14690	84	13	neglect	neglect	VERB
fcis-14690	84	14	the	the	DET
fcis-14690	84	15	channel	channel	NOUN
fcis-14690	84	16	attention	attention	NOUN
fcis-14690	84	17	,	,	PUNCT
fcis-14690	84	18	or	or	CCONJ
fcis-14690	84	19	ignore	ignore	VERB
fcis-14690	84	20	the	the	DET
fcis-14690	84	21	importance	importance	NOUN
fcis-14690	84	22	of	of	ADP
fcis-14690	84	23	the	the	DET
fcis-14690	84	24	global	global	ADJ
fcis-14690	84	25	feature	feature	NOUN
fcis-14690	84	26	when	when	SCONJ
fcis-14690	84	27	focusing	focus	VERB
fcis-14690	84	28	on	on	ADP
fcis-14690	84	29	the	the	DET
fcis-14690	84	30	local	local	ADJ
fcis-14690	84	31	emotion	emotion	NOUN
fcis-14690	84	32	region	region	NOUN
fcis-14690	84	33	of	of	ADP
fcis-14690	84	34	the	the	DET
fcis-14690	84	35	image	image	NOUN
fcis-14690	84	36	.	.	PUNCT
fcis-14690	85	1	therefore	therefore	ADV
fcis-14690	85	2	,	,	PUNCT
fcis-14690	85	3	in	in	ADP
fcis-14690	85	4	this	this	DET
fcis-14690	85	5	paper	paper	NOUN
fcis-14690	85	6	,	,	PUNCT
fcis-14690	85	7	when	when	SCONJ
fcis-14690	85	8	extracting	extract	VERB
fcis-14690	85	9	multi	multi	ADJ
fcis-14690	85	10	-	-	ADJ
fcis-14690	85	11	level	level	ADJ
fcis-14690	85	12	image	image	NOUN
fcis-14690	85	13	features	feature	NOUN
fcis-14690	85	14	,	,	PUNCT
fcis-14690	85	15	features	feature	NOUN
fcis-14690	85	16	of	of	ADP
fcis-14690	85	17	different	different	ADJ
fcis-14690	85	18	levels	level	NOUN
fcis-14690	85	19	are	be	AUX
fcis-14690	85	20	fused	fuse	VERB
fcis-14690	85	21	by	by	ADP
fcis-14690	85	22	bilstm	bilstm	NOUN
fcis-14690	85	23	to	to	PART
fcis-14690	85	24	form	form	VERB
fcis-14690	85	25	global	global	ADJ
fcis-14690	85	26	features	feature	NOUN
fcis-14690	85	27	,	,	PUNCT
fcis-14690	85	28	and	and	CCONJ
fcis-14690	85	29	a	a	DET
fcis-14690	85	30	branch	branch	NOUN
fcis-14690	85	31	is	be	AUX
fcis-14690	85	32	used	use	VERB
fcis-14690	85	33	separately	separately	ADV
fcis-14690	85	34	to	to	PART
fcis-14690	85	35	obtain	obtain	VERB
fcis-14690	85	36	local	local	ADJ
fcis-14690	85	37	features	feature	NOUN
fcis-14690	85	38	by	by	ADP
fcis-14690	85	39	integrating	integrate	VERB
fcis-14690	85	40	cbam	cbam	NOUN
fcis-14690	85	41	module	module	NOUN
fcis-14690	85	42	.	.	PUNCT
fcis-14690	86	1	in	in	ADP
fcis-14690	86	2	this	this	DET
fcis-14690	86	3	way	way	NOUN
fcis-14690	86	4	,	,	PUNCT
fcis-14690	86	5	not	not	PART
fcis-14690	86	6	only	only	ADV
fcis-14690	86	7	the	the	DET
fcis-14690	86	8	spatial	spatial	ADJ
fcis-14690	86	9	and	and	CCONJ
fcis-14690	86	10	channel	channel	NOUN
fcis-14690	86	11	attention	attention	NOUN
fcis-14690	86	12	mechanisms	mechanism	NOUN
fcis-14690	86	13	are	be	AUX
fcis-14690	86	14	used	use	VERB
fcis-14690	86	15	to	to	PART
fcis-14690	86	16	focus	focus	VERB
fcis-14690	86	17	on	on	ADP
fcis-14690	86	18	the	the	DET
fcis-14690	86	19	important	important	ADJ
fcis-14690	86	20	local	local	ADJ
fcis-14690	86	21	emotional	emotional	ADJ
fcis-14690	86	22	features	feature	NOUN
fcis-14690	86	23	,	,	PUNCT
fcis-14690	86	24	but	but	CCONJ
fcis-14690	86	25	also	also	ADV
fcis-14690	86	26	the	the	DET
fcis-14690	86	27	global	global	ADJ
fcis-14690	86	28	features	feature	NOUN
fcis-14690	86	29	are	be	AUX
fcis-14690	86	30	guaranteed	guarantee	VERB
fcis-14690	86	31	to	to	PART
fcis-14690	86	32	participate	participate	VERB
fcis-14690	86	33	in	in	ADP
fcis-14690	86	34	the	the	DET
fcis-14690	86	35	feature	feature	NOUN
fcis-14690	86	36	fusion	fusion	NOUN
fcis-14690	86	37	.	.	PUNCT
fcis-14690	87	1	2.3	2.3	NUM
fcis-14690	87	2	.	.	PUNCT
fcis-14690	87	3	bilstm	bilstm	NOUN
fcis-14690	87	4	after	after	ADP
fcis-14690	87	5	extracting	extract	VERB
fcis-14690	87	6	the	the	DET
fcis-14690	87	7	multi	multi	ADJ
fcis-14690	87	8	-	-	ADJ
fcis-14690	87	9	level	level	ADJ
fcis-14690	87	10	features	feature	NOUN
fcis-14690	87	11	of	of	ADP
fcis-14690	87	12	the	the	DET
fcis-14690	87	13	image	image	NOUN
fcis-14690	87	14	,	,	PUNCT
fcis-14690	87	15	it	it	PRON
fcis-14690	87	16	is	be	AUX
fcis-14690	87	17	necessary	necessary	ADJ
fcis-14690	87	18	to	to	PART
fcis-14690	87	19	fuse	fuse	VERB
fcis-14690	87	20	them	they	PRON
fcis-14690	87	21	and	and	CCONJ
fcis-14690	87	22	then	then	ADV
fcis-14690	87	23	carry	carry	VERB
fcis-14690	87	24	out	out	ADP
fcis-14690	87	25	the	the	DET
fcis-14690	87	26	emotion	emotion	NOUN
fcis-14690	87	27	classification	classification	NOUN
fcis-14690	87	28	.	.	PUNCT
fcis-14690	88	1	rao	rao	NOUN
fcis-14690	88	2	et	et	PROPN
fcis-14690	88	3	al	al	PROPN
fcis-14690	88	4	.	.	PUNCT
fcis-14690	89	1	[	[	X
fcis-14690	89	2	19	19	NUM
fcis-14690	89	3	]	]	PUNCT
fcis-14690	89	4	proposed	propose	VERB
fcis-14690	89	5	a	a	DET
fcis-14690	89	6	multi	multi	ADJ
fcis-14690	89	7	-	-	ADJ
fcis-14690	89	8	level	level	ADJ
fcis-14690	89	9	deep	deep	ADJ
fcis-14690	89	10	network	network	NOUN
fcis-14690	89	11	mldrnet	mldrnet	NOUN
fcis-14690	89	12	,	,	PUNCT
fcis-14690	89	13	which	which	PRON
fcis-14690	89	14	uses	use	VERB
fcis-14690	89	15	max	max	PROPN
fcis-14690	89	16	and	and	CCONJ
fcis-14690	89	17	avg	avg	PROPN
fcis-14690	89	18	aggregation	aggregation	NOUN
fcis-14690	89	19	functions	function	NOUN
fcis-14690	89	20	to	to	PART
fcis-14690	89	21	fuse	fuse	VERB
fcis-14690	89	22	features	feature	NOUN
fcis-14690	89	23	of	of	ADP
fcis-14690	89	24	different	different	ADJ
fcis-14690	89	25	levels	level	NOUN
fcis-14690	89	26	.	.	PUNCT
fcis-14690	90	1	but	but	CCONJ
fcis-14690	90	2	this	this	DET
fcis-14690	90	3	approach	approach	NOUN
fcis-14690	90	4	ignores	ignore	VERB
fcis-14690	90	5	the	the	DET
fcis-14690	90	6	dependencies	dependency	NOUN
fcis-14690	90	7	between	between	ADP
fcis-14690	90	8	the	the	DET
fcis-14690	90	9	multi	multi	ADJ
fcis-14690	90	10	-	-	ADJ
fcis-14690	90	11	level	level	ADJ
fcis-14690	90	12	features	feature	NOUN
fcis-14690	90	13	.	.	PUNCT
fcis-14690	91	1	zhu	zhu	PROPN
fcis-14690	91	2	et	et	PROPN
fcis-14690	91	3	al	al	PROPN
fcis-14690	91	4	.	.	PUNCT
fcis-14690	92	1	[	[	X
fcis-14690	92	2	14	14	NUM
fcis-14690	92	3	]	]	PUNCT
fcis-14690	92	4	proposed	propose	VERB
fcis-14690	92	5	a	a	DET
fcis-14690	92	6	unified	unified	ADJ
fcis-14690	92	7	cnn	cnn	PROPN
fcis-14690	92	8	-	-	PUNCT
fcis-14690	92	9	rnn	rnn	PROPN
fcis-14690	92	10	visual	visual	ADJ
fcis-14690	92	11	emotion	emotion	NOUN
fcis-14690	92	12	recognition	recognition	NOUN
fcis-14690	92	13	model	model	NOUN
fcis-14690	92	14	,	,	PUNCT
fcis-14690	92	15	which	which	PRON
fcis-14690	92	16	utilized	utilize	VERB
fcis-14690	92	17	the	the	DET
fcis-14690	92	18	features	feature	NOUN
fcis-14690	92	19	of	of	ADP
fcis-14690	92	20	multiple	multiple	ADJ
fcis-14690	92	21	branches	branch	NOUN
fcis-14690	92	22	in	in	ADP
fcis-14690	92	23	cnn	cnn	PROPN
fcis-14690	92	24	at	at	ADP
fcis-14690	92	25	different	different	ADJ
fcis-14690	92	26	levels	level	NOUN
fcis-14690	92	27	,	,	PUNCT
fcis-14690	92	28	and	and	CCONJ
fcis-14690	92	29	effectively	effectively	ADV
fcis-14690	92	30	integrated	integrate	VERB
fcis-14690	92	31	these	these	DET
fcis-14690	92	32	features	feature	NOUN
fcis-14690	92	33	by	by	ADP
fcis-14690	92	34	using	use	VERB
fcis-14690	92	35	the	the	DET
fcis-14690	92	36	dependency	dependency	NOUN
fcis-14690	92	37	relationship	relationship	NOUN
fcis-14690	92	38	between	between	ADP
fcis-14690	92	39	them	they	PRON
fcis-14690	92	40	through	through	ADP
fcis-14690	92	41	the	the	DET
fcis-14690	92	42	bidirectional	bidirectional	PROPN
fcis-14690	92	43	gru	gru	PROPN
fcis-14690	92	44	method	method	NOUN
fcis-14690	92	45	.	.	PUNCT
fcis-14690	93	1	zhang	zhang	PROPN
fcis-14690	93	2	et	et	PROPN
fcis-14690	93	3	al	al	PROPN
fcis-14690	93	4	.	.	PUNCT
fcis-14690	94	1	[	[	X
fcis-14690	94	2	28	28	NUM
fcis-14690	94	3	]	]	PUNCT
fcis-14690	94	4	proposed	propose	VERB
fcis-14690	94	5	a	a	DET
fcis-14690	94	6	multi	multi	ADJ
fcis-14690	94	7	-	-	ADJ
fcis-14690	94	8	level	level	ADJ
fcis-14690	94	9	hybrid	hybrid	ADJ
fcis-14690	94	10	model	model	NOUN
fcis-14690	94	11	,	,	PUNCT
fcis-14690	94	12	which	which	PRON
fcis-14690	94	13	learns	learn	VERB
fcis-14690	94	14	and	and	CCONJ
fcis-14690	94	15	integrates	integrate	VERB
fcis-14690	94	16	deep	deep	ADJ
fcis-14690	94	17	semantics	semantic	NOUN
fcis-14690	94	18	and	and	CCONJ
fcis-14690	94	19	shallow	shallow	ADJ
fcis-14690	94	20	visual	visual	ADJ
fcis-14690	94	21	representations	representation	NOUN
fcis-14690	94	22	.	.	PUNCT
fcis-14690	95	1	considering	consider	VERB
fcis-14690	95	2	that	that	SCONJ
fcis-14690	95	3	images	image	NOUN
fcis-14690	95	4	have	have	VERB
fcis-14690	95	5	strong	strong	ADJ
fcis-14690	95	6	texture	texture	ADJ
fcis-14690	95	7	features	feature	NOUN
fcis-14690	95	8	,	,	PUNCT
fcis-14690	95	9	gram	gram	NOUN
fcis-14690	95	10	matrix	matrix	NOUN
fcis-14690	95	11	is	be	AUX
fcis-14690	95	12	used	use	VERB
fcis-14690	95	13	to	to	PART
fcis-14690	95	14	establish	establish	VERB
fcis-14690	95	15	the	the	DET
fcis-14690	95	16	correlation	correlation	NOUN
fcis-14690	95	17	between	between	ADP
fcis-14690	95	18	multi	multi	ADJ
fcis-14690	95	19	-	-	ADJ
fcis-14690	95	20	level	level	ADJ
fcis-14690	95	21	features	feature	NOUN
fcis-14690	95	22	and	and	CCONJ
fcis-14690	95	23	form	form	VERB
fcis-14690	95	24	gram	gram	NOUN
fcis-14690	95	25	matrix	matrix	NOUN
fcis-14690	95	26	for	for	ADP
fcis-14690	95	27	emotion	emotion	NOUN
fcis-14690	95	28	classification	classification	NOUN
fcis-14690	95	29	.	.	PUNCT
fcis-14690	96	1	in	in	ADP
fcis-14690	96	2	the	the	DET
fcis-14690	96	3	above	above	ADJ
fcis-14690	96	4	methods	method	NOUN
fcis-14690	96	5	to	to	PART
fcis-14690	96	6	establish	establish	VERB
fcis-14690	96	7	multi	multi	ADJ
fcis-14690	96	8	-	-	ADJ
fcis-14690	96	9	level	level	ADJ
fcis-14690	96	10	feature	feature	NOUN
fcis-14690	96	11	correlation	correlation	NOUN
fcis-14690	96	12	,	,	PUNCT
fcis-14690	96	13	the	the	DET
fcis-14690	96	14	bidirectional	bidirectional	PROPN
fcis-14690	96	15	gru	gru	NOUN
fcis-14690	96	16	method	method	NOUN
fcis-14690	96	17	can	can	AUX
fcis-14690	96	18	retain	retain	VERB
fcis-14690	96	19	fewer	few	ADJ
fcis-14690	96	20	features	feature	NOUN
fcis-14690	96	21	per	per	ADP
fcis-14690	96	22	gru	gru	NOUN
fcis-14690	96	23	,	,	PUNCT
fcis-14690	96	24	so	so	ADV
fcis-14690	96	25	using	use	VERB
fcis-14690	96	26	bidirectional	bidirectional	ADJ
fcis-14690	96	27	gru	gru	NOUN
fcis-14690	96	28	to	to	PART
fcis-14690	96	29	establish	establish	VERB
fcis-14690	96	30	multi	multi	ADJ
fcis-14690	96	31	-	-	ADJ
fcis-14690	96	32	level	level	ADJ
fcis-14690	96	33	feature	feature	NOUN
fcis-14690	96	34	correlation	correlation	NOUN
fcis-14690	96	35	will	will	AUX
fcis-14690	96	36	affect	affect	VERB
fcis-14690	96	37	the	the	DET
fcis-14690	96	38	performance	performance	NOUN
fcis-14690	96	39	of	of	ADP
fcis-14690	96	40	image	image	NOUN
fcis-14690	96	41	emotion	emotion	NOUN
fcis-14690	96	42	classification	classification	NOUN
fcis-14690	96	43	.	.	PUNCT
fcis-14690	97	1	gram	gram	NOUN
fcis-14690	97	2	matrix	matrix	NOUN
fcis-14690	97	3	establishes	establish	VERB
fcis-14690	97	4	correlation	correlation	NOUN
fcis-14690	97	5	according	accord	VERB
fcis-14690	97	6	to	to	ADP
fcis-14690	97	7	the	the	DET
fcis-14690	97	8	texture	texture	NOUN
fcis-14690	97	9	features	feature	NOUN
fcis-14690	97	10	of	of	ADP
fcis-14690	97	11	image	image	NOUN
fcis-14690	97	12	,	,	PUNCT
fcis-14690	97	13	and	and	CCONJ
fcis-14690	97	14	the	the	DET
fcis-14690	97	15	weight	weight	NOUN
fcis-14690	97	16	of	of	ADP
fcis-14690	97	17	higher	high	ADJ
fcis-14690	97	18	semantic	semantic	ADJ
fcis-14690	97	19	features	feature	NOUN
fcis-14690	97	20	is	be	AUX
fcis-14690	97	21	reduced	reduce	VERB
fcis-14690	97	22	when	when	SCONJ
fcis-14690	97	23	the	the	DET
fcis-14690	97	24	lowerlevel	lowerlevel	NOUN
fcis-14690	97	25	to	to	ADP
fcis-14690	97	26	higher	high	ADJ
fcis-14690	97	27	-	-	PUNCT
fcis-14690	97	28	level	level	NOUN
fcis-14690	97	29	features	feature	NOUN
fcis-14690	97	30	is	be	AUX
fcis-14690	97	31	established	establish	VERB
fcis-14690	97	32	.	.	PUNCT
fcis-14690	98	1	since	since	SCONJ
fcis-14690	98	2	long	long	ADJ
fcis-14690	98	3	shortterm	shortterm	PROPN
fcis-14690	98	4	memory	memory	NOUN
fcis-14690	98	5	network	network	NOUN
fcis-14690	98	6	(	(	PUNCT
fcis-14690	98	7	lstm	lstm	NOUN
fcis-14690	98	8	)	)	PUNCT
fcis-14690	98	9	has	have	VERB
fcis-14690	98	10	more	more	ADJ
fcis-14690	98	11	memory	memory	NOUN
fcis-14690	98	12	units	unit	NOUN
fcis-14690	98	13	,	,	PUNCT
fcis-14690	98	14	it	it	PRON
fcis-14690	98	15	can	can	AUX
fcis-14690	98	16	retain	retain	VERB
fcis-14690	98	17	more	more	ADJ
fcis-14690	98	18	image	image	NOUN
fcis-14690	98	19	features	feature	NOUN
fcis-14690	98	20	,	,	PUNCT
fcis-14690	98	21	so	so	SCONJ
fcis-14690	98	22	this	this	DET
fcis-14690	98	23	paper	paper	NOUN
fcis-14690	98	24	uses	use	VERB
fcis-14690	98	25	bilstm	bilstm	NOUN
fcis-14690	98	26	network	network	NOUN
fcis-14690	98	27	to	to	PART
fcis-14690	98	28	mine	mine	VERB
fcis-14690	98	29	the	the	DET
fcis-14690	98	30	correlation	correlation	NOUN
fcis-14690	98	31	between	between	ADP
fcis-14690	98	32	features	feature	NOUN
fcis-14690	98	33	at	at	ADP
fcis-14690	98	34	different	different	ADJ
fcis-14690	98	35	levels	level	NOUN
fcis-14690	98	36	.	.	PUNCT
fcis-14690	99	1	therefore	therefore	ADV
fcis-14690	99	2	,	,	PUNCT
fcis-14690	99	3	in	in	ADP
fcis-14690	99	4	this	this	DET
fcis-14690	99	5	paper	paper	NOUN
fcis-14690	99	6	,	,	PUNCT
fcis-14690	99	7	we	we	PRON
fcis-14690	99	8	propose	propose	VERB
fcis-14690	99	9	an	an	DET
fcis-14690	99	10	image	image	NOUN
fcis-14690	99	11	sentiment	sentiment	NOUN
fcis-14690	99	12	analysis	analysis	NOUN
fcis-14690	99	13	model	model	NOUN
fcis-14690	99	14	maml	maml	PROPN
fcis-14690	99	15	based	base	VERB
fcis-14690	99	16	on	on	ADP
fcis-14690	99	17	mixed	mixed	ADJ
fcis-14690	99	18	attention	attention	NOUN
fcis-14690	99	19	and	and	CCONJ
fcis-14690	99	20	multilevel	multilevel	NOUN
fcis-14690	99	21	dependency	dependency	NOUN
fcis-14690	99	22	,	,	PUNCT
fcis-14690	99	23	which	which	PRON
fcis-14690	99	24	combines	combine	VERB
fcis-14690	99	25	the	the	DET
fcis-14690	99	26	local	local	ADJ
fcis-14690	99	27	features	feature	NOUN
fcis-14690	99	28	obtained	obtain	VERB
fcis-14690	99	29	by	by	ADP
fcis-14690	99	30	cbam	cbam	NOUN
fcis-14690	99	31	module	module	NOUN
fcis-14690	99	32	with	with	ADP
fcis-14690	99	33	the	the	DET
fcis-14690	99	34	multi	multi	ADJ
fcis-14690	99	35	-	-	ADJ
fcis-14690	99	36	level	level	ADJ
fcis-14690	99	37	global	global	ADJ
fcis-14690	99	38	features	feature	NOUN
fcis-14690	99	39	62	62	NUM
fcis-14690	99	40	integrated	integrate	VERB
fcis-14690	99	41	by	by	ADP
fcis-14690	99	42	bilstm	bilstm	NOUN
fcis-14690	99	43	,	,	PUNCT
fcis-14690	99	44	so	so	SCONJ
fcis-14690	99	45	as	as	SCONJ
fcis-14690	99	46	to	to	PART
fcis-14690	99	47	improve	improve	VERB
fcis-14690	99	48	the	the	DET
fcis-14690	99	49	performance	performance	NOUN
fcis-14690	99	50	of	of	ADP
fcis-14690	99	51	image	image	NOUN
fcis-14690	99	52	sentiment	sentiment	NOUN
fcis-14690	99	53	analysis	analysis	NOUN
fcis-14690	99	54	.	.	PUNCT
fcis-14690	100	1	3	3	X
fcis-14690	100	2	.	.	X
fcis-14690	100	3	the	the	DET
fcis-14690	100	4	proposed	propose	VERB
fcis-14690	100	5	method	method	NOUN
fcis-14690	100	6	the	the	DET
fcis-14690	100	7	emotion	emotion	NOUN
fcis-14690	100	8	analysis	analysis	NOUN
fcis-14690	100	9	model	model	NOUN
fcis-14690	100	10	maml	maml	PROPN
fcis-14690	100	11	proposed	propose	VERB
fcis-14690	100	12	in	in	ADP
fcis-14690	100	13	this	this	DET
fcis-14690	100	14	paper	paper	NOUN
fcis-14690	100	15	is	be	AUX
fcis-14690	100	16	shown	show	VERB
fcis-14690	100	17	in	in	ADP
fcis-14690	100	18	figure	figure	NOUN
fcis-14690	100	19	2	2	NUM
fcis-14690	100	20	,	,	PUNCT
fcis-14690	100	21	which	which	PRON
fcis-14690	100	22	mainly	mainly	ADV
fcis-14690	100	23	consists	consist	VERB
fcis-14690	100	24	of	of	ADP
fcis-14690	100	25	three	three	NUM
fcis-14690	100	26	parts	part	NOUN
fcis-14690	100	27	:	:	PUNCT
fcis-14690	100	28	multi	multi	ADJ
fcis-14690	100	29	-	-	ADJ
fcis-14690	100	30	level	level	ADJ
fcis-14690	100	31	feature	feature	NOUN
fcis-14690	100	32	extraction	extraction	NOUN
fcis-14690	100	33	,	,	PUNCT
fcis-14690	100	34	local	local	ADJ
fcis-14690	100	35	emotion	emotion	NOUN
fcis-14690	100	36	feature	feature	NOUN
fcis-14690	100	37	recognition	recognition	NOUN
fcis-14690	100	38	,	,	PUNCT
fcis-14690	100	39	and	and	CCONJ
fcis-14690	100	40	multi	multi	ADJ
fcis-14690	100	41	-	-	ADJ
fcis-14690	100	42	level	level	ADJ
fcis-14690	100	43	feature	feature	NOUN
fcis-14690	100	44	dependency	dependency	NOUN
fcis-14690	100	45	establishment	establishment	NOUN
fcis-14690	100	46	.	.	PUNCT
fcis-14690	101	1	many	many	ADJ
fcis-14690	101	2	studies	study	NOUN
fcis-14690	101	3	have	have	AUX
fcis-14690	101	4	shown	show	VERB
fcis-14690	101	5	that	that	SCONJ
fcis-14690	101	6	image	image	NOUN
fcis-14690	101	7	emotion	emotion	NOUN
fcis-14690	101	8	is	be	AUX
fcis-14690	101	9	related	relate	VERB
fcis-14690	101	10	to	to	ADP
fcis-14690	101	11	lower	low	ADJ
fcis-14690	101	12	-	-	PUNCT
fcis-14690	101	13	level	level	NOUN
fcis-14690	101	14	to	to	ADP
fcis-14690	101	15	higher	high	ADJ
fcis-14690	101	16	-	-	PUNCT
fcis-14690	101	17	level	level	NOUN
fcis-14690	101	18	features	feature	NOUN
fcis-14690	101	19	,	,	PUNCT
fcis-14690	101	20	so	so	SCONJ
fcis-14690	101	21	this	this	DET
fcis-14690	101	22	paper	paper	NOUN
fcis-14690	101	23	uses	use	VERB
fcis-14690	101	24	the	the	DET
fcis-14690	101	25	layered	layered	ADJ
fcis-14690	101	26	stacking	stacking	NOUN
fcis-14690	101	27	structure	structure	NOUN
fcis-14690	101	28	of	of	ADP
fcis-14690	101	29	cnn	cnn	PROPN
fcis-14690	101	30	to	to	PART
fcis-14690	101	31	extract	extract	VERB
fcis-14690	101	32	multilevel	multilevel	NOUN
fcis-14690	101	33	features	feature	NOUN
fcis-14690	101	34	of	of	ADP
fcis-14690	101	35	images	image	NOUN
fcis-14690	101	36	.	.	PUNCT
fcis-14690	102	1	considering	consider	VERB
fcis-14690	102	2	the	the	DET
fcis-14690	102	3	importance	importance	NOUN
fcis-14690	102	4	of	of	ADP
fcis-14690	102	5	local	local	ADJ
fcis-14690	102	6	and	and	CCONJ
fcis-14690	102	7	global	global	ADJ
fcis-14690	102	8	features	feature	NOUN
fcis-14690	102	9	to	to	PART
fcis-14690	102	10	image	image	VERB
fcis-14690	102	11	emotion	emotion	NOUN
fcis-14690	102	12	analysis	analysis	NOUN
fcis-14690	102	13	,	,	PUNCT
fcis-14690	102	14	cbam	cbam	NOUN
fcis-14690	102	15	module	module	NOUN
fcis-14690	102	16	is	be	AUX
fcis-14690	102	17	introduced	introduce	VERB
fcis-14690	102	18	in	in	ADP
fcis-14690	102	19	this	this	DET
fcis-14690	102	20	paper	paper	NOUN
fcis-14690	102	21	,	,	PUNCT
fcis-14690	102	22	and	and	CCONJ
fcis-14690	102	23	spatial	spatial	ADJ
fcis-14690	102	24	and	and	CCONJ
fcis-14690	102	25	channel	channel	NOUN
fcis-14690	102	26	attention	attention	NOUN
fcis-14690	102	27	mechanisms	mechanism	NOUN
fcis-14690	102	28	are	be	AUX
fcis-14690	102	29	used	use	VERB
fcis-14690	102	30	to	to	ADP
fcis-14690	102	31	mine	mine	ADJ
fcis-14690	102	32	regions	region	NOUN
fcis-14690	102	33	and	and	CCONJ
fcis-14690	102	34	features	feature	NOUN
fcis-14690	102	35	with	with	ADP
fcis-14690	102	36	richer	rich	ADJ
fcis-14690	102	37	image	image	NOUN
fcis-14690	102	38	emotion	emotion	NOUN
fcis-14690	102	39	information	information	NOUN
fcis-14690	102	40	.	.	PUNCT
fcis-14690	103	1	since	since	SCONJ
fcis-14690	103	2	there	there	PRON
fcis-14690	103	3	is	be	VERB
fcis-14690	103	4	a	a	DET
fcis-14690	103	5	dependency	dependency	NOUN
fcis-14690	103	6	between	between	ADP
fcis-14690	103	7	features	feature	NOUN
fcis-14690	103	8	at	at	ADP
fcis-14690	103	9	different	different	ADJ
fcis-14690	103	10	levels	level	NOUN
fcis-14690	103	11	,	,	PUNCT
fcis-14690	103	12	this	this	DET
fcis-14690	103	13	paper	paper	NOUN
fcis-14690	103	14	uses	use	VERB
fcis-14690	103	15	bilstm	bilstm	NOUN
fcis-14690	103	16	to	to	PART
fcis-14690	103	17	integrate	integrate	VERB
fcis-14690	103	18	features	feature	NOUN
fcis-14690	103	19	at	at	ADP
fcis-14690	103	20	5	5	NUM
fcis-14690	103	21	different	different	ADJ
fcis-14690	103	22	levels	level	NOUN
fcis-14690	103	23	of	of	ADP
fcis-14690	103	24	images	image	NOUN
fcis-14690	103	25	,	,	PUNCT
fcis-14690	103	26	establish	establish	VERB
fcis-14690	103	27	the	the	DET
fcis-14690	103	28	correlation	correlation	NOUN
fcis-14690	103	29	between	between	ADP
fcis-14690	103	30	them	they	PRON
fcis-14690	103	31	,	,	PUNCT
fcis-14690	103	32	and	and	CCONJ
fcis-14690	103	33	refine	refine	VERB
fcis-14690	103	34	the	the	DET
fcis-14690	103	35	feature	feature	NOUN
fcis-14690	103	36	information	information	NOUN
fcis-14690	103	37	at	at	ADP
fcis-14690	103	38	each	each	DET
fcis-14690	103	39	level	level	NOUN
fcis-14690	103	40	.	.	PUNCT
fcis-14690	104	1	fig	fig	NOUN
fcis-14690	104	2	2	2	NUM
fcis-14690	104	3	.	.	PUNCT
fcis-14690	104	4	maml	maml	PROPN
fcis-14690	104	5	model	model	NOUN
fcis-14690	104	6	3.1	3.1	NUM
fcis-14690	104	7	.	.	PUNCT
fcis-14690	105	1	multi	multi	ADJ
fcis-14690	105	2	-	-	ADJ
fcis-14690	105	3	level	level	ADJ
fcis-14690	105	4	feature	feature	NOUN
fcis-14690	105	5	extraction	extraction	NOUN
fcis-14690	105	6	at	at	ADP
fcis-14690	105	7	present	present	ADJ
fcis-14690	105	8	,	,	PUNCT
fcis-14690	105	9	convolutional	convolutional	ADJ
fcis-14690	105	10	neural	neural	ADJ
fcis-14690	105	11	networks	network	NOUN
fcis-14690	105	12	have	have	AUX
fcis-14690	105	13	been	be	AUX
fcis-14690	105	14	widely	widely	ADV
fcis-14690	105	15	used	use	VERB
fcis-14690	105	16	in	in	ADP
fcis-14690	105	17	image	image	NOUN
fcis-14690	105	18	emotion	emotion	NOUN
fcis-14690	105	19	analysis	analysis	NOUN
fcis-14690	105	20	tasks	task	NOUN
fcis-14690	105	21	,	,	PUNCT
fcis-14690	105	22	and	and	CCONJ
fcis-14690	105	23	good	good	ADJ
fcis-14690	105	24	results	result	NOUN
fcis-14690	105	25	have	have	AUX
fcis-14690	105	26	been	be	AUX
fcis-14690	105	27	obtained	obtain	VERB
fcis-14690	105	28	by	by	ADP
fcis-14690	105	29	using	use	VERB
fcis-14690	105	30	the	the	DET
fcis-14690	105	31	last	last	ADJ
fcis-14690	105	32	layer	layer	NOUN
fcis-14690	105	33	of	of	ADP
fcis-14690	105	34	high	high	ADJ
fcis-14690	105	35	-	-	PUNCT
fcis-14690	105	36	level	level	NOUN
fcis-14690	105	37	semantic	semantic	ADJ
fcis-14690	105	38	features	feature	NOUN
fcis-14690	105	39	extracted	extract	VERB
fcis-14690	105	40	by	by	ADP
fcis-14690	105	41	neural	neural	ADJ
fcis-14690	105	42	networks	network	NOUN
fcis-14690	105	43	for	for	ADP
fcis-14690	105	44	emotion	emotion	NOUN
fcis-14690	105	45	classification	classification	NOUN
fcis-14690	105	46	.	.	PUNCT
fcis-14690	106	1	because	because	SCONJ
fcis-14690	106	2	of	of	ADP
fcis-14690	106	3	the	the	DET
fcis-14690	106	4	hierarchical	hierarchical	ADJ
fcis-14690	106	5	structure	structure	NOUN
fcis-14690	106	6	of	of	ADP
fcis-14690	106	7	convolutional	convolutional	ADJ
fcis-14690	106	8	neural	neural	ADJ
fcis-14690	106	9	networks	network	NOUN
fcis-14690	106	10	,	,	PUNCT
fcis-14690	106	11	cnns	cnn	NOUN
fcis-14690	106	12	can	can	AUX
fcis-14690	106	13	capture	capture	VERB
fcis-14690	106	14	features	feature	NOUN
fcis-14690	106	15	at	at	ADP
fcis-14690	106	16	different	different	ADJ
fcis-14690	106	17	levels	level	NOUN
fcis-14690	106	18	.	.	PUNCT
fcis-14690	107	1	the	the	DET
fcis-14690	107	2	features	feature	NOUN
fcis-14690	107	3	extracted	extract	VERB
fcis-14690	107	4	by	by	ADP
fcis-14690	107	5	shallow	shallow	ADJ
fcis-14690	107	6	network	network	NOUN
fcis-14690	107	7	contain	contain	VERB
fcis-14690	107	8	more	more	ADJ
fcis-14690	107	9	pixel	pixel	PROPN
fcis-14690	107	10	information	information	NOUN
fcis-14690	107	11	,	,	PUNCT
fcis-14690	107	12	such	such	ADJ
fcis-14690	107	13	as	as	ADP
fcis-14690	107	14	color	color	NOUN
fcis-14690	107	15	,	,	PUNCT
fcis-14690	107	16	texture	texture	NOUN
fcis-14690	107	17	,	,	PUNCT
fcis-14690	107	18	edge	edge	NOUN
fcis-14690	107	19	and	and	CCONJ
fcis-14690	107	20	so	so	ADV
fcis-14690	107	21	on	on	ADV
fcis-14690	107	22	.	.	PUNCT
fcis-14690	108	1	the	the	DET
fcis-14690	108	2	deep	deep	ADJ
fcis-14690	108	3	network	network	NOUN
fcis-14690	108	4	receptive	receptive	ADJ
fcis-14690	108	5	field	field	NOUN
fcis-14690	108	6	is	be	AUX
fcis-14690	108	7	increased	increase	VERB
fcis-14690	108	8	,	,	PUNCT
fcis-14690	108	9	and	and	CCONJ
fcis-14690	108	10	the	the	DET
fcis-14690	108	11	extracted	extract	VERB
fcis-14690	108	12	features	feature	NOUN
fcis-14690	108	13	contain	contain	VERB
fcis-14690	108	14	more	more	ADJ
fcis-14690	108	15	abstract	abstract	ADJ
fcis-14690	108	16	semantic	semantic	ADJ
fcis-14690	108	17	information	information	NOUN
fcis-14690	108	18	,	,	PUNCT
fcis-14690	108	19	but	but	CCONJ
fcis-14690	108	20	the	the	DET
fcis-14690	108	21	resolution	resolution	NOUN
fcis-14690	108	22	is	be	AUX
fcis-14690	108	23	lower	low	ADJ
fcis-14690	108	24	and	and	CCONJ
fcis-14690	108	25	the	the	DET
fcis-14690	108	26	perception	perception	NOUN
fcis-14690	108	27	of	of	ADP
fcis-14690	108	28	detail	detail	NOUN
fcis-14690	108	29	is	be	AUX
fcis-14690	108	30	poor	poor	ADJ
fcis-14690	108	31	.	.	PUNCT
fcis-14690	109	1	consider	consider	VERB
fcis-14690	109	2	that	that	DET
fcis-14690	109	3	image	image	NOUN
fcis-14690	109	4	emotion	emotion	NOUN
fcis-14690	109	5	is	be	AUX
fcis-14690	109	6	not	not	PART
fcis-14690	109	7	only	only	ADV
fcis-14690	109	8	related	relate	VERB
fcis-14690	109	9	to	to	ADP
fcis-14690	109	10	higher	high	ADJ
fcis-14690	109	11	-	-	PUNCT
fcis-14690	109	12	level	level	NOUN
fcis-14690	109	13	features	feature	NOUN
fcis-14690	109	14	,	,	PUNCT
fcis-14690	109	15	but	but	CCONJ
fcis-14690	109	16	also	also	ADV
fcis-14690	109	17	to	to	ADP
fcis-14690	109	18	lower	low	ADJ
fcis-14690	109	19	-	-	PUNCT
fcis-14690	109	20	level	level	NOUN
fcis-14690	109	21	features	feature	NOUN
fcis-14690	109	22	.	.	PUNCT
fcis-14690	110	1	in	in	ADP
fcis-14690	110	2	this	this	DET
fcis-14690	110	3	paper	paper	NOUN
fcis-14690	110	4	,	,	PUNCT
fcis-14690	110	5	resnet18	resnet18	NOUN
fcis-14690	110	6	is	be	AUX
fcis-14690	110	7	used	use	VERB
fcis-14690	110	8	to	to	PART
fcis-14690	110	9	extract	extract	VERB
fcis-14690	110	10	multi	multi	ADJ
fcis-14690	110	11	-	-	ADJ
fcis-14690	110	12	layer	layer	ADJ
fcis-14690	110	13	features	feature	NOUN
fcis-14690	110	14	of	of	ADP
fcis-14690	110	15	images	image	NOUN
fcis-14690	110	16	.	.	PUNCT
fcis-14690	111	1	transforms	transform	VERB
fcis-14690	111	2	an	an	DET
fcis-14690	111	3	image	image	NOUN
fcis-14690	111	4	to	to	ADP
fcis-14690	111	5	224×224×3	224×224×3	NUM
fcis-14690	111	6	pixels	pixel	NOUN
fcis-14690	111	7	is	be	AUX
fcis-14690	111	8	first	first	ADV
fcis-14690	111	9	preprocessed	preprocesse	VERB
fcis-14690	111	10	using	using	NOUN
fcis-14690	111	11	transforms	transform	NOUN
fcis-14690	111	12	,	,	PUNCT
fcis-14690	111	13	and	and	CCONJ
fcis-14690	111	14	the	the	DET
fcis-14690	111	15	image	image	NOUN
fcis-14690	111	16	is	be	AUX
fcis-14690	111	17	then	then	ADV
fcis-14690	111	18	input	input	ADJ
fcis-14690	111	19	into	into	ADP
fcis-14690	111	20	the	the	DET
fcis-14690	111	21	resnet18	resnet18	NOUN
fcis-14690	111	22	network	network	NOUN
fcis-14690	111	23	,	,	PUNCT
fcis-14690	111	24	which	which	PRON
fcis-14690	111	25	consists	consist	VERB
fcis-14690	111	26	of	of	ADP
fcis-14690	111	27	5	5	NUM
fcis-14690	111	28	parts	part	NOUN
fcis-14690	111	29	.	.	PUNCT
fcis-14690	112	1	the	the	DET
fcis-14690	112	2	convolution	convolution	NOUN
fcis-14690	112	3	layer	layer	NOUN
fcis-14690	112	4	in	in	ADP
fcis-14690	112	5	the	the	DET
fcis-14690	112	6	first	first	ADJ
fcis-14690	112	7	part	part	NOUN
fcis-14690	112	8	,	,	PUNCT
fcis-14690	112	9	level1	level1	PROPN
fcis-14690	112	10	,	,	PUNCT
fcis-14690	112	11	is	be	AUX
fcis-14690	112	12	7×7×64	7×7×64	NUM
fcis-14690	112	13	.	.	PUNCT
fcis-14690	113	1	the	the	DET
fcis-14690	113	2	lowest	low	ADJ
fcis-14690	113	3	layer	layer	NOUN
fcis-14690	113	4	feature	feature	NOUN
fcis-14690	113	5	f0	f0	PROPN
fcis-14690	113	6	of	of	ADP
fcis-14690	113	7	the	the	DET
fcis-14690	113	8	image	image	NOUN
fcis-14690	113	9	is	be	AUX
fcis-14690	113	10	extracted	extract	VERB
fcis-14690	113	11	by	by	ADP
fcis-14690	113	12	the	the	DET
fcis-14690	113	13	first	first	ADJ
fcis-14690	113	14	convolution	convolution	NOUN
fcis-14690	113	15	block	block	NOUN
fcis-14690	113	16	and	and	CCONJ
fcis-14690	113	17	a	a	DET
fcis-14690	113	18	3×3×64	3×3×64	NUM
fcis-14690	113	19	maximum	maximum	ADJ
fcis-14690	113	20	pooling	pool	VERB
fcis-14690	113	21	layer	layer	NOUN
fcis-14690	113	22	.	.	PUNCT
fcis-14690	114	1	the	the	DET
fcis-14690	114	2	convolution	convolution	NOUN
fcis-14690	114	3	layers	layer	NOUN
fcis-14690	114	4	contained	contain	VERB
fcis-14690	114	5	in	in	ADP
fcis-14690	114	6	the	the	DET
fcis-14690	114	7	other	other	ADJ
fcis-14690	114	8	four	four	NUM
fcis-14690	114	9	parts	part	NOUN
fcis-14690	114	10	level2	level2	PROPN
fcis-14690	114	11	,	,	PUNCT
fcis-14690	114	12	level3	level3	PROPN
fcis-14690	114	13	,	,	PUNCT
fcis-14690	114	14	level4	level4	PROPN
fcis-14690	114	15	and	and	CCONJ
fcis-14690	114	16	level5	level5	PROPN
fcis-14690	114	17	are	be	AUX
fcis-14690	114	18	3×3×64	3×3×64	NUM
fcis-14690	114	19	,	,	PUNCT
fcis-14690	114	20	3×3×128	3×3×128	NUM
fcis-14690	114	21	,	,	PUNCT
fcis-14690	114	22	3×3×256	3×3×256	NUM
fcis-14690	114	23	and	and	CCONJ
fcis-14690	114	24	3×3×512	3×3×512	NUM
fcis-14690	114	25	respectively	respectively	ADV
fcis-14690	114	26	.	.	PUNCT
fcis-14690	115	1	each	each	DET
fcis-14690	115	2	part	part	NOUN
fcis-14690	115	3	corresponds	correspond	VERB
fcis-14690	115	4	to	to	ADP
fcis-14690	115	5	the	the	DET
fcis-14690	115	6	features	feature	NOUN
fcis-14690	115	7	of	of	ADP
fcis-14690	115	8	a	a	DET
fcis-14690	115	9	level	level	NOUN
fcis-14690	115	10	,	,	PUNCT
fcis-14690	115	11	so	so	CCONJ
fcis-14690	115	12	the	the	DET
fcis-14690	115	13	features	feature	NOUN
fcis-14690	115	14	of	of	ADP
fcis-14690	115	15	the	the	DET
fcis-14690	115	16	other	other	ADJ
fcis-14690	115	17	four	four	NUM
fcis-14690	115	18	levels	level	NOUN
fcis-14690	115	19	are	be	AUX
fcis-14690	115	20	f1	f1	ADJ
fcis-14690	115	21	,	,	PUNCT
fcis-14690	115	22	f2	f2	PROPN
fcis-14690	115	23	,	,	PUNCT
fcis-14690	115	24	f3	f3	PROPN
fcis-14690	115	25	and	and	CCONJ
fcis-14690	115	26	f4	f4	PROPN
fcis-14690	115	27	.	.	NOUN
fcis-14690	115	28	3.2	3.2	NUM
fcis-14690	115	29	.	.	PUNCT
fcis-14690	116	1	local	local	ADJ
fcis-14690	116	2	emotion	emotion	NOUN
fcis-14690	116	3	feature	feature	NOUN
fcis-14690	116	4	recognition	recognition	NOUN
fcis-14690	116	5	attention	attention	NOUN
fcis-14690	116	6	mechanism	mechanism	NOUN
fcis-14690	116	7	is	be	AUX
fcis-14690	116	8	widely	widely	ADV
fcis-14690	116	9	used	use	VERB
fcis-14690	116	10	in	in	ADP
fcis-14690	116	11	the	the	DET
fcis-14690	116	12	field	field	NOUN
fcis-14690	116	13	of	of	ADP
fcis-14690	116	14	computer	computer	NOUN
fcis-14690	116	15	vision	vision	NOUN
fcis-14690	116	16	,	,	PUNCT
fcis-14690	116	17	but	but	CCONJ
fcis-14690	116	18	the	the	DET
fcis-14690	116	19	existing	exist	VERB
fcis-14690	116	20	researches	research	NOUN
fcis-14690	116	21	pay	pay	VERB
fcis-14690	116	22	too	too	ADV
fcis-14690	116	23	much	much	ADJ
fcis-14690	116	24	attention	attention	NOUN
fcis-14690	116	25	to	to	ADP
fcis-14690	116	26	the	the	DET
fcis-14690	116	27	spatial	spatial	ADJ
fcis-14690	116	28	attention	attention	NOUN
fcis-14690	116	29	of	of	ADP
fcis-14690	116	30	images	image	NOUN
fcis-14690	116	31	and	and	CCONJ
fcis-14690	116	32	ignore	ignore	VERB
fcis-14690	116	33	the	the	DET
fcis-14690	116	34	channel	channel	NOUN
fcis-14690	116	35	attention	attention	NOUN
fcis-14690	116	36	,	,	PUNCT
fcis-14690	116	37	or	or	CCONJ
fcis-14690	116	38	use	use	VERB
fcis-14690	116	39	the	the	DET
fcis-14690	116	40	local	local	ADJ
fcis-14690	116	41	features	feature	NOUN
fcis-14690	116	42	extracted	extract	VERB
fcis-14690	116	43	by	by	ADP
fcis-14690	116	44	attention	attention	NOUN
fcis-14690	116	45	mechanism	mechanism	NOUN
fcis-14690	116	46	and	and	CCONJ
fcis-14690	116	47	ignore	ignore	VERB
fcis-14690	116	48	the	the	DET
fcis-14690	116	49	global	global	ADJ
fcis-14690	116	50	features	feature	NOUN
fcis-14690	116	51	.	.	PUNCT
fcis-14690	117	1	therefore	therefore	ADV
fcis-14690	117	2	,	,	PUNCT
fcis-14690	117	3	in	in	ADP
fcis-14690	117	4	this	this	DET
fcis-14690	117	5	paper	paper	NOUN
fcis-14690	117	6	,	,	PUNCT
fcis-14690	117	7	a	a	DET
fcis-14690	117	8	hybrid	hybrid	ADJ
fcis-14690	117	9	attention	attention	NOUN
fcis-14690	117	10	mechanism	mechanism	NOUN
fcis-14690	117	11	cbam	cbam	NOUN
fcis-14690	117	12	is	be	AUX
fcis-14690	117	13	used	use	VERB
fcis-14690	117	14	to	to	PART
fcis-14690	117	15	extract	extract	VERB
fcis-14690	117	16	local	local	ADJ
fcis-14690	117	17	features	feature	NOUN
fcis-14690	117	18	,	,	PUNCT
fcis-14690	117	19	which	which	PRON
fcis-14690	117	20	can	can	AUX
fcis-14690	117	21	generate	generate	VERB
fcis-14690	117	22	attention	attention	NOUN
fcis-14690	117	23	feature	feature	NOUN
fcis-14690	117	24	map	map	NOUN
fcis-14690	117	25	information	information	NOUN
fcis-14690	117	26	in	in	ADP
fcis-14690	117	27	two	two	NUM
fcis-14690	117	28	dimensions	dimension	NOUN
fcis-14690	117	29	of	of	ADP
fcis-14690	117	30	channel	channel	NOUN
fcis-14690	117	31	and	and	CCONJ
fcis-14690	117	32	space	space	NOUN
fcis-14690	117	33	,	,	PUNCT
fcis-14690	117	34	and	and	CCONJ
fcis-14690	117	35	then	then	ADV
fcis-14690	117	36	multiply	multiply	VERB
fcis-14690	117	37	the	the	DET
fcis-14690	117	38	two	two	NUM
fcis-14690	117	39	-	-	PUNCT
fcis-14690	117	40	feature	feature	NOUN
fcis-14690	117	41	map	map	NOUN
fcis-14690	117	42	information	information	NOUN
fcis-14690	117	43	with	with	ADP
fcis-14690	117	44	the	the	DET
fcis-14690	117	45	original	original	ADJ
fcis-14690	117	46	input	input	NOUN
fcis-14690	117	47	feature	feature	NOUN
fcis-14690	117	48	map	map	NOUN
fcis-14690	117	49	for	for	ADP
fcis-14690	117	50	adaptive	adaptive	ADJ
fcis-14690	117	51	feature	feature	NOUN
fcis-14690	117	52	correction	correction	NOUN
fcis-14690	117	53	to	to	PART
fcis-14690	117	54	produce	produce	VERB
fcis-14690	117	55	the	the	DET
fcis-14690	117	56	final	final	ADJ
fcis-14690	117	57	feature	feature	NOUN
fcis-14690	117	58	map	map	NOUN
fcis-14690	117	59	.	.	PUNCT
fcis-14690	118	1	cbam	cbam	NOUN
fcis-14690	118	2	is	be	AUX
fcis-14690	118	3	mainly	mainly	ADV
fcis-14690	118	4	composed	compose	VERB
fcis-14690	118	5	of	of	ADP
fcis-14690	118	6	channel	channel	NOUN
fcis-14690	118	7	attention	attention	NOUN
fcis-14690	118	8	module	module	NOUN
fcis-14690	118	9	and	and	CCONJ
fcis-14690	118	10	spatial	spatial	ADJ
fcis-14690	118	11	attention	attention	NOUN
fcis-14690	118	12	module	module	NOUN
fcis-14690	118	13	.	.	PUNCT
fcis-14690	119	1	the	the	DET
fcis-14690	119	2	complete	complete	ADJ
fcis-14690	119	3	cbam	cbam	NOUN
fcis-14690	119	4	module	module	NOUN
fcis-14690	119	5	is	be	AUX
fcis-14690	119	6	shown	show	VERB
fcis-14690	119	7	in	in	ADP
fcis-14690	119	8	figure	figure	NOUN
fcis-14690	119	9	3	3	NUM
fcis-14690	119	10	.	.	PUNCT
fcis-14690	119	11	fig	fig	NOUN
fcis-14690	119	12	3	3	NUM
fcis-14690	119	13	.	.	PUNCT
fcis-14690	120	1	cbam	cbam	NOUN
fcis-14690	120	2	model	model	NOUN
fcis-14690	120	3	the	the	DET
fcis-14690	120	4	channel	channel	NOUN
fcis-14690	120	5	attention	attention	NOUN
fcis-14690	120	6	module	module	NOUN
fcis-14690	120	7	uses	use	VERB
fcis-14690	120	8	average	average	ADJ
fcis-14690	120	9	pooling	pooling	NOUN
fcis-14690	120	10	and	and	CCONJ
fcis-14690	120	11	maximum	maximum	ADJ
fcis-14690	120	12	pooling	pooling	NOUN
fcis-14690	120	13	to	to	PART
fcis-14690	120	14	learn	learn	VERB
fcis-14690	120	15	discriminant	discriminant	NOUN
fcis-14690	120	16	features	feature	NOUN
fcis-14690	120	17	,	,	PUNCT
fcis-14690	120	18	c	c	PROPN
fcis-14690	120	19	avgf	avgf	PROPN
fcis-14690	120	20	and	and	CCONJ
fcis-14690	120	21	cfmax	cfmax	NOUN
fcis-14690	120	22	represents	represent	VERB
fcis-14690	120	23	the	the	DET
fcis-14690	120	24	features	feature	NOUN
fcis-14690	120	25	obtained	obtain	VERB
fcis-14690	120	26	after	after	ADP
fcis-14690	120	27	average	average	ADJ
fcis-14690	120	28	pooling	pooling	NOUN
fcis-14690	120	29	and	and	CCONJ
fcis-14690	120	30	maximum	maximum	ADJ
fcis-14690	120	31	pooling	pooling	NOUN
fcis-14690	120	32	,	,	PUNCT
fcis-14690	120	33	respectively	respectively	ADV
fcis-14690	120	34	.	.	PUNCT
fcis-14690	121	1	this	this	DET
fcis-14690	121	2	feature	feature	NOUN
fcis-14690	121	3	is	be	AUX
fcis-14690	121	4	then	then	ADV
fcis-14690	121	5	fed	feed	VERB
fcis-14690	121	6	into	into	ADP
fcis-14690	121	7	a	a	DET
fcis-14690	121	8	shared	share	VERB
fcis-14690	121	9	multi	multi	ADJ
fcis-14690	121	10	-	-	ADJ
fcis-14690	121	11	layer	layer	ADJ
fcis-14690	121	12	perceptron	perceptron	NOUN
fcis-14690	121	13	(	(	PUNCT
fcis-14690	121	14	mlp	mlp	PROPN
fcis-14690	121	15	)	)	PUNCT
fcis-14690	121	16	network	network	NOUN
fcis-14690	121	17	to	to	PART
fcis-14690	121	18	generate	generate	VERB
fcis-14690	121	19	the	the	DET
fcis-14690	121	20	final	final	ADJ
fcis-14690	121	21	channel	channel	NOUN
fcis-14690	121	22	attention	attention	NOUN
fcis-14690	121	23	feature	feature	NOUN
fcis-14690	121	24	map	map	NOUN
fcis-14690	121	25	11	11	NUM
fcis-14690	121	26	c	c	NOUN
fcis-14690	121	27	c	c	PROPN
fcis-14690	121	28	rm	rm	PROPN
fcis-14690	121	29	.	.	PUNCT
fcis-14690	122	1	in	in	ADP
fcis-14690	122	2	order	order	NOUN
fcis-14690	122	3	to	to	PART
fcis-14690	122	4	reduce	reduce	VERB
fcis-14690	122	5	the	the	DET
fcis-14690	122	6	calculation	calculation	NOUN
fcis-14690	122	7	parameters	parameter	NOUN
fcis-14690	122	8	,	,	PUNCT
fcis-14690	122	9	a	a	DET
fcis-14690	122	10	dimensionality	dimensionality	NOUN
fcis-14690	122	11	reduction	reduction	NOUN
fcis-14690	122	12	parameter	parameter	NOUN
fcis-14690	122	13	r	r	NOUN
fcis-14690	122	14	is	be	AUX
fcis-14690	122	15	used	use	VERB
fcis-14690	122	16	in	in	ADP
fcis-14690	122	17	mlp	mlp	PROPN
fcis-14690	122	18	,	,	PUNCT
fcis-14690	122	19	11/	11/	PROPN
fcis-14690	122	20			PROPN
fcis-14690	122	21	rc	rc	PROPN
fcis-14690	122	22	c	c	PROPN
fcis-14690	122	23	rm	rm	PROPN
fcis-14690	122	24	,	,	PUNCT
fcis-14690	122	25	so	so	ADV
fcis-14690	122	26	the	the	DET
fcis-14690	122	27	calculation	calculation	NOUN
fcis-14690	122	28	formula	formula	NOUN
fcis-14690	122	29	of	of	ADP
fcis-14690	122	30	the	the	DET
fcis-14690	122	31	channel	channel	NOUN
fcis-14690	122	32	attention	attention	NOUN
fcis-14690	122	33	module	module	NOUN
fcis-14690	122	34	is	be	AUX
fcis-14690	122	35	)	)	PUNCT
fcis-14690	122	36	)	)	PUNCT
fcis-14690	122	37	)	)	PUNCT
fcis-14690	123	1	(	(	PUNCT
fcis-14690	123	2	(	(	PUNCT
fcis-14690	123	3	)	)	PUNCT
fcis-14690	123	4	)	)	PUNCT
fcis-14690	123	5	(	(	PUNCT
fcis-14690	123	6	(	(	PUNCT
fcis-14690	123	7	(	(	PUNCT
fcis-14690	123	8	)	)	PUNCT
fcis-14690	123	9	)	)	PUNCT
fcis-14690	123	10	)	)	PUNCT
fcis-14690	124	1	(	(	PUNCT
fcis-14690	124	2	(	(	PUNCT
fcis-14690	124	3	)	)	PUNCT
fcis-14690	124	4	)	)	PUNCT
fcis-14690	124	5	(	(	PUNCT
fcis-14690	124	6	(	(	PUNCT
fcis-14690	124	7	(	(	PUNCT
fcis-14690	124	8	)	)	PUNCT
fcis-14690	124	9	(	(	PUNCT
fcis-14690	124	10	max0101	max0101	NOUN
fcis-14690	124	11	cc	cc	PROPN
fcis-14690	124	12	avg	avg	PROPN
fcis-14690	124	13	c	c	PROPN
fcis-14690	124	14	fwwfww	fwwfww	PROPN
fcis-14690	124	15	fmaxpoolmlpfavgpoolmlpfm	fmaxpoolmlpfavgpoolmlpfm	PROPN
fcis-14690	124	16			PROPN
fcis-14690	124	17			ADJ
fcis-14690	124	18			X
fcis-14690	124	19			X
fcis-14690	124	20	(	(	PUNCT
fcis-14690	124	21	1	1	X
fcis-14690	124	22	)	)	PUNCT
fcis-14690	124	23	the	the	DET
fcis-14690	124	24	spatial	spatial	ADJ
fcis-14690	124	25	attention	attention	NOUN
fcis-14690	124	26	module	module	NOUN
fcis-14690	124	27	generates	generate	VERB
fcis-14690	124	28	the	the	DET
fcis-14690	124	29	spatial	spatial	ADJ
fcis-14690	124	30	attention	attention	NOUN
fcis-14690	124	31	feature	feature	NOUN
fcis-14690	124	32	map	map	NOUN
fcis-14690	125	1	wh	wh	NOUN
fcis-14690	125	2	s	s	NOUN
fcis-14690	125	3	rfm	rfm	PROPN
fcis-14690	125	4	,	,	PUNCT
fcis-14690	125	5	)	)	PUNCT
fcis-14690	125	6	(	(	PUNCT
fcis-14690	125	7			NOUN
fcis-14690	125	8	based	base	VERB
fcis-14690	125	9	on	on	ADP
fcis-14690	125	10	the	the	DET
fcis-14690	125	11	channel	channel	NOUN
fcis-14690	125	12	attention	attention	NOUN
fcis-14690	125	13	feature	feature	NOUN
fcis-14690	125	14	map	map	NOUN
fcis-14690	125	15	,	,	PUNCT
fcis-14690	125	16	which	which	PRON
fcis-14690	125	17	is	be	AUX
fcis-14690	125	18	the	the	DET
fcis-14690	125	19	same	same	ADJ
fcis-14690	125	20	as	as	ADP
fcis-14690	125	21	the	the	DET
fcis-14690	125	22	channel	channel	NOUN
fcis-14690	125	23	attention	attention	NOUN
fcis-14690	125	24	mechanism	mechanism	NOUN
fcis-14690	125	25	,	,	PUNCT
fcis-14690	125	26	and	and	CCONJ
fcis-14690	125	27	simultaneously	simultaneously	ADV
fcis-14690	125	28	uses	use	VERB
fcis-14690	125	29	average	average	ADJ
fcis-14690	125	30	pooling	pooling	NOUN
fcis-14690	125	31	and	and	CCONJ
fcis-14690	125	32	maximum	maximum	ADJ
fcis-14690	125	33	pooling	pooling	NOUN
fcis-14690	125	34	to	to	PART
fcis-14690	125	35	generate	generate	VERB
fcis-14690	125	36	discriminant	discriminant	NOUN
fcis-14690	125	37	features	feature	NOUN
fcis-14690	125	38	.	.	PUNCT
fcis-14690	126	1	however	however	ADV
fcis-14690	126	2	,	,	PUNCT
fcis-14690	126	3	the	the	DET
fcis-14690	126	4	spatial	spatial	ADJ
fcis-14690	126	5	attention	attention	NOUN
fcis-14690	126	6	mechanism	mechanism	NOUN
fcis-14690	126	7	generates	generate	VERB
fcis-14690	126	8	2d	2d	NUM
fcis-14690	126	9	feature	feature	NOUN
fcis-14690	126	10	maps	map	NOUN
fcis-14690	126	11	whs	whs	VERB
fcis-14690	126	12	avg	avg	PROPN
fcis-14690	126	13	rf	rf	PROPN
fcis-14690	126	14			PROPN
fcis-14690	126	15	1	1	NUM
fcis-14690	126	16	and	and	CCONJ
fcis-14690	126	17	whs	whs	VERB
fcis-14690	126	18	rf	rf	VERB
fcis-14690	126	19			PROPN
fcis-14690	126	20	1	1	NUM
fcis-14690	126	21	max	max	NOUN
fcis-14690	126	22	,	,	PUNCT
fcis-14690	126	23	so	so	CCONJ
fcis-14690	126	24	the	the	DET
fcis-14690	126	25	calculation	calculation	NOUN
fcis-14690	126	26	formula	formula	NOUN
fcis-14690	126	27	of	of	ADP
fcis-14690	126	28	the	the	DET
fcis-14690	126	29	spatial	spatial	ADJ
fcis-14690	126	30	attention	attention	NOUN
fcis-14690	126	31	module	module	NOUN
fcis-14690	126	32	is	be	AUX
fcis-14690	126	33	)	)	PUNCT
fcis-14690	126	34	)	)	PUNCT
fcis-14690	126	35	;	;	PUNCT
fcis-14690	126	36	(	(	PUNCT
fcis-14690	126	37	(	(	PUNCT
fcis-14690	126	38	)	)	PUNCT
fcis-14690	126	39	]	]	PUNCT
fcis-14690	126	40	)	)	PUNCT
fcis-14690	126	41	)	)	PUNCT
fcis-14690	126	42	(	(	PUNCT
fcis-14690	126	43	)	)	PUNCT
fcis-14690	126	44	;	;	PUNCT
fcis-14690	126	45	(	(	PUNCT
fcis-14690	126	46	(	(	PUNCT
fcis-14690	126	47	[	[	X
fcis-14690	126	48	(	(	PUNCT
fcis-14690	126	49	)	)	PUNCT
fcis-14690	126	50	(	(	PUNCT
fcis-14690	126	51	max	max	PROPN
fcis-14690	126	52	77	77	NUM
fcis-14690	126	53	77	77	NUM
fcis-14690	126	54	ss	ss	PROPN
fcis-14690	126	55	avg	avg	PROPN
fcis-14690	126	56	s	s	PROPN
fcis-14690	126	57	fff	fff	PROPN
fcis-14690	126	58	fmaxpoolfavgpoolffm	fmaxpoolfavgpoolffm	PROPN
fcis-14690	126	59			PROPN
fcis-14690	126	60			PROPN
fcis-14690	126	61			PROPN
fcis-14690	126	62			PROPN
fcis-14690	126	63			X
fcis-14690	126	64			X
fcis-14690	126	65	(	(	PUNCT
fcis-14690	126	66	2	2	NUM
fcis-14690	126	67	)	)	PUNCT
fcis-14690	126	68	based	base	VERB
fcis-14690	126	69	on	on	ADP
fcis-14690	126	70	resnet18	resnet18	NOUN
fcis-14690	126	71	,	,	PUNCT
fcis-14690	126	72	a	a	DET
fcis-14690	126	73	total	total	NOUN
fcis-14690	126	74	of	of	ADP
fcis-14690	126	75	5	5	NUM
fcis-14690	126	76	features	feature	NOUN
fcis-14690	126	77	of	of	ADP
fcis-14690	126	78	different	different	ADJ
fcis-14690	126	79	levels	level	NOUN
fcis-14690	126	80	are	be	AUX
fcis-14690	126	81	generated	generate	VERB
fcis-14690	126	82	in	in	ADP
fcis-14690	126	83	this	this	DET
fcis-14690	126	84	paper	paper	NOUN
fcis-14690	126	85	.	.	PUNCT
fcis-14690	127	1	considering	consider	VERB
fcis-14690	127	2	that	that	SCONJ
fcis-14690	127	3	it	it	PRON
fcis-14690	127	4	is	be	AUX
fcis-14690	127	5	better	well	ADJ
fcis-14690	127	6	to	to	PART
fcis-14690	127	7	use	use	VERB
fcis-14690	127	8	both	both	DET
fcis-14690	127	9	local	local	ADJ
fcis-14690	127	10	features	feature	NOUN
fcis-14690	127	11	and	and	CCONJ
fcis-14690	127	12	global	global	ADJ
fcis-14690	127	13	features	feature	NOUN
fcis-14690	127	14	for	for	ADP
fcis-14690	127	15	emotion	emotion	NOUN
fcis-14690	127	16	analysis	analysis	NOUN
fcis-14690	127	17	than	than	SCONJ
fcis-14690	127	18	to	to	PART
fcis-14690	127	19	use	use	VERB
fcis-14690	127	20	one	one	NUM
fcis-14690	127	21	of	of	ADP
fcis-14690	127	22	them	they	PRON
fcis-14690	127	23	alone	alone	ADV
fcis-14690	127	24	,	,	PUNCT
fcis-14690	127	25	this	this	DET
fcis-14690	127	26	paper	paper	NOUN
fcis-14690	127	27	inputs	input	VERB
fcis-14690	127	28	the	the	DET
fcis-14690	127	29	features	feature	NOUN
fcis-14690	127	30	of	of	ADP
fcis-14690	127	31	the	the	DET
fcis-14690	127	32	highest	high	ADJ
fcis-14690	127	33	level	level	NOUN
fcis-14690	127	34	with	with	ADP
fcis-14690	127	35	the	the	DET
fcis-14690	127	36	richest	rich	ADJ
fcis-14690	127	37	semantic	semantic	ADJ
fcis-14690	127	38	information	information	NOUN
fcis-14690	127	39	into	into	ADP
fcis-14690	127	40	the	the	DET
fcis-14690	127	41	cbam	cbam	NOUN
fcis-14690	127	42	module	module	NOUN
fcis-14690	127	43	to	to	PART
fcis-14690	127	44	generate	generate	VERB
fcis-14690	127	45	features	feature	NOUN
fcis-14690	127	46	in	in	ADP
fcis-14690	127	47	the	the	DET
fcis-14690	127	48	local	local	ADJ
fcis-14690	127	49	emotion	emotion	NOUN
fcis-14690	127	50	region	region	NOUN
fcis-14690	127	51	.	.	PUNCT
fcis-14690	128	1	cbam	cbam	NOUN
fcis-14690	128	2	first	first	ADV
fcis-14690	128	3	passes	pass	VERB
fcis-14690	128	4	the	the	DET
fcis-14690	128	5	input	input	NOUN
fcis-14690	128	6	feature	feature	NOUN
fcis-14690	128	7	whcrf	whcrf	VERB
fcis-14690	128	8			ADV
fcis-14690	128	9	through	through	ADP
fcis-14690	128	10	the	the	DET
fcis-14690	128	11	channel	channel	NOUN
fcis-14690	128	12	attention	attention	NOUN
fcis-14690	128	13	module	module	NOUN
fcis-14690	128	14	to	to	PART
fcis-14690	128	15	obtain	obtain	VERB
fcis-14690	128	16	1d	1d	NUM
fcis-14690	128	17	channel	channel	NOUN
fcis-14690	128	18	attention	attention	NOUN
fcis-14690	128	19	feature	feature	NOUN
fcis-14690	128	20	diagram	diagram	NOUN
fcis-14690	128	21	11	11	NUM
fcis-14690	128	22	c	c	PROPN
fcis-14690	128	23	c	c	PROPN
fcis-14690	128	24	rm	rm	PROPN
fcis-14690	128	25	,	,	PUNCT
fcis-14690	128	26	and	and	CCONJ
fcis-14690	128	27	then	then	ADV
fcis-14690	128	28	the	the	DET
fcis-14690	128	29	spatial	spatial	ADJ
fcis-14690	128	30	attention	attention	NOUN
fcis-14690	128	31	module	module	NOUN
fcis-14690	128	32	obtains	obtain	VERB
fcis-14690	128	33	2d	2d	NUM
fcis-14690	128	34	space	space	NOUN
fcis-14690	128	35	attention	attention	NOUN
fcis-14690	128	36	feature	feature	NOUN
fcis-14690	128	37	diagram	diagram	NOUN
fcis-14690	128	38	wh	wh	PROPN
fcis-14690	128	39	s	s	PROPN
fcis-14690	128	40	rm	rm	NOUN
fcis-14690	128	41			PROPN
fcis-14690	128	42	1	1	NUM
fcis-14690	128	43	according	accord	VERB
fcis-14690	128	44	to	to	ADP
fcis-14690	128	45	the	the	DET
fcis-14690	128	46	channel	channel	NOUN
fcis-14690	128	47	attention	attention	NOUN
fcis-14690	128	48	feature	feature	NOUN
fcis-14690	128	49	diagram	diagram	NOUN
fcis-14690	128	50	.	.	PUNCT
fcis-14690	129	1	the	the	DET
fcis-14690	129	2	general	general	ADJ
fcis-14690	129	3	process	process	NOUN
fcis-14690	129	4	is	be	AUX
fcis-14690	129	5	as	as	SCONJ
fcis-14690	129	6	follows	follow	VERB
fcis-14690	129	7	:	:	PUNCT
fcis-14690	129	8	44c	44c	NUM
fcis-14690	129	9	ffmf	ffmf	PROPN
fcis-14690	129	10			PROPN
fcis-14690	129	11	)	)	PUNCT
fcis-14690	129	12	(	(	PUNCT
fcis-14690	129	13	4	4	NUM
fcis-14690	129	14	(	(	PUNCT
fcis-14690	129	15	3	3	NUM
fcis-14690	129	16	)	)	PUNCT
fcis-14690	129	17	4s	4s	NUM
fcis-14690	129	18	ffmf	ffmf	PROPN
fcis-14690	129	19			PROPN
fcis-14690	129	20	)	)	PUNCT
fcis-14690	129	21	(	(	PUNCT
fcis-14690	129	22	44	44	NUM
fcis-14690	129	23	(	(	PUNCT
fcis-14690	129	24	4	4	NUM
fcis-14690	129	25	)	)	PUNCT
fcis-14690	129	26	63	63	NUM
fcis-14690	129	27	where	where	PROPN
fcis-14690	129	28	represents	represent	VERB
fcis-14690	129	29	element	element	ADJ
fcis-14690	129	30	level	level	NOUN
fcis-14690	129	31	multiplication	multiplication	NOUN
fcis-14690	129	32	.	.	PUNCT
fcis-14690	130	1	4f	4f	NUM
fcis-14690	130	2	is	be	AUX
fcis-14690	130	3	the	the	DET
fcis-14690	130	4	local	local	ADJ
fcis-14690	130	5	emotion	emotion	NOUN
fcis-14690	130	6	feature	feature	NOUN
fcis-14690	130	7	map	map	NOUN
fcis-14690	130	8	after	after	ADP
fcis-14690	130	9	channel	channel	NOUN
fcis-14690	130	10	attention	attention	NOUN
fcis-14690	130	11	and	and	CCONJ
fcis-14690	130	12	spatial	spatial	ADJ
fcis-14690	130	13	attention	attention	NOUN
fcis-14690	130	14	.	.	PUNCT
fcis-14690	131	1	in	in	ADP
fcis-14690	131	2	order	order	NOUN
fcis-14690	131	3	to	to	PART
fcis-14690	131	4	reduce	reduce	VERB
fcis-14690	131	5	overfitting	overfitting	NOUN
fcis-14690	131	6	,	,	PUNCT
fcis-14690	131	7	two	two	NUM
fcis-14690	131	8	2×2	2×2	NUM
fcis-14690	131	9	maximum	maximum	ADJ
fcis-14690	131	10	pooling	pool	VERB
fcis-14690	131	11	layers	layer	NOUN
fcis-14690	131	12	are	be	AUX
fcis-14690	131	13	continuously	continuously	ADV
fcis-14690	131	14	used	use	VERB
fcis-14690	131	15	to	to	PART
fcis-14690	131	16	filter	filter	VERB
fcis-14690	131	17	the	the	DET
fcis-14690	131	18	features	feature	NOUN
fcis-14690	131	19	output	output	NOUN
fcis-14690	131	20	by	by	ADP
fcis-14690	131	21	cbam	cbam	NOUN
fcis-14690	131	22	module	module	NOUN
fcis-14690	131	23	and	and	CCONJ
fcis-14690	131	24	the	the	DET
fcis-14690	131	25	features	feature	NOUN
fcis-14690	131	26	output	output	NOUN
fcis-14690	131	27	by	by	ADP
fcis-14690	131	28	resnet18	resnet18	NOUN
fcis-14690	131	29	at	at	ADP
fcis-14690	131	30	five	five	NUM
fcis-14690	131	31	different	different	ADJ
fcis-14690	131	32	levels	level	NOUN
fcis-14690	131	33	to	to	PART
fcis-14690	131	34	reduce	reduce	VERB
fcis-14690	131	35	the	the	DET
fcis-14690	131	36	feature	feature	NOUN
fcis-14690	131	37	dimension	dimension	NOUN
fcis-14690	131	38	.	.	PUNCT
fcis-14690	132	1	the	the	DET
fcis-14690	132	2	specific	specific	ADJ
fcis-14690	132	3	implementation	implementation	NOUN
fcis-14690	132	4	method	method	NOUN
fcis-14690	132	5	is	be	AUX
fcis-14690	132	6	as	as	SCONJ
fcis-14690	132	7	follows	follow	VERB
fcis-14690	132	8	:	:	PUNCT
fcis-14690	132	9	)	)	PUNCT
fcis-14690	132	10	(	(	PUNCT
fcis-14690	132	11	ii	ii	PROPN
fcis-14690	132	12	fmaxpoolp	fmaxpoolp	NOUN
fcis-14690	132	13			PROPN
fcis-14690	132	14	(	(	PUNCT
fcis-14690	132	15	5	5	NUM
fcis-14690	132	16	)	)	PUNCT
fcis-14690	132	17	)	)	PUNCT
fcis-14690	133	1	(	(	PUNCT
fcis-14690	133	2	ii	ii	PROPN
fcis-14690	133	3	pmaxpoolp	pmaxpoolp	VERB
fcis-14690	133	4			PROPN
fcis-14690	133	5	(	(	PUNCT
fcis-14690	133	6	6	6	NUM
fcis-14690	133	7	)	)	PUNCT
fcis-14690	133	8	)	)	PUNCT
fcis-14690	133	9	(	(	PUNCT
fcis-14690	133	10	ii	ii	PROPN
fcis-14690	133	11	pflattenx	pflattenx	VERB
fcis-14690	133	12			PROPN
fcis-14690	133	13	(	(	PUNCT
fcis-14690	133	14	7	7	NUM
fcis-14690	133	15	)	)	PUNCT
fcis-14690	133	16	)	)	PUNCT
fcis-14690	133	17	(	(	PUNCT
fcis-14690	133	18	iiii	iiii	PROPN
fcis-14690	133	19	bxwreluf	bxwreluf	X
fcis-14690	133	20			PROPN
fcis-14690	133	21	(	(	PUNCT
fcis-14690	133	22	8)	8)	NUM
fcis-14690	133	23	where	where	SCONJ
fcis-14690	133	24	i=0,1,2,3,4	i=0,1,2,3,4	ADJ
fcis-14690	133	25	,	,	PUNCT
fcis-14690	133	26	wi	wi	PROPN
fcis-14690	133	27	,	,	PUNCT
fcis-14690	133	28	bi	bi	NOUN
fcis-14690	133	29	are	be	AUX
fcis-14690	133	30	the	the	DET
fcis-14690	133	31	parameters	parameter	NOUN
fcis-14690	133	32	of	of	ADP
fcis-14690	133	33	the	the	DET
fcis-14690	133	34	full	full	ADJ
fcis-14690	133	35	connection	connection	NOUN
fcis-14690	133	36	layer	layer	NOUN
fcis-14690	133	37	.	.	PUNCT
fcis-14690	134	1	3.3	3.3	NUM
fcis-14690	134	2	.	.	PUNCT
fcis-14690	135	1	multi	multi	ADJ
fcis-14690	135	2	-	-	ADJ
fcis-14690	135	3	level	level	ADJ
fcis-14690	135	4	feature	feature	NOUN
fcis-14690	135	5	dependency	dependency	NOUN
fcis-14690	135	6	establishment	establishment	NOUN
fcis-14690	135	7	fig	fig	NOUN
fcis-14690	135	8	4	4	NUM
fcis-14690	135	9	.	.	PUNCT
fcis-14690	135	10	bilstm	bilstm	NOUN
fcis-14690	135	11	model	model	NOUN
fcis-14690	135	12	five	five	NUM
fcis-14690	135	13	different	different	ADJ
fcis-14690	135	14	levels	level	NOUN
fcis-14690	135	15	of	of	ADP
fcis-14690	135	16	features	feature	NOUN
fcis-14690	135	17	are	be	AUX
fcis-14690	135	18	obtained	obtain	VERB
fcis-14690	135	19	through	through	ADP
fcis-14690	135	20	resnet18	resnet18	NOUN
fcis-14690	135	21	,	,	PUNCT
fcis-14690	135	22	and	and	CCONJ
fcis-14690	135	23	there	there	PRON
fcis-14690	135	24	is	be	VERB
fcis-14690	135	25	a	a	DET
fcis-14690	135	26	certain	certain	ADJ
fcis-14690	135	27	dependence	dependence	NOUN
fcis-14690	135	28	among	among	ADP
fcis-14690	135	29	them	they	PRON
fcis-14690	135	30	,	,	PUNCT
fcis-14690	135	31	which	which	PRON
fcis-14690	135	32	is	be	AUX
fcis-14690	135	33	often	often	ADV
fcis-14690	135	34	ignored	ignore	VERB
fcis-14690	135	35	by	by	ADP
fcis-14690	135	36	existing	exist	VERB
fcis-14690	135	37	studies	study	NOUN
fcis-14690	135	38	.	.	PUNCT
fcis-14690	136	1	therefore	therefore	ADV
fcis-14690	136	2	,	,	PUNCT
fcis-14690	136	3	bilstm	bilstm	NOUN
fcis-14690	136	4	is	be	AUX
fcis-14690	136	5	used	use	VERB
fcis-14690	136	6	in	in	ADP
fcis-14690	136	7	this	this	DET
fcis-14690	136	8	paper	paper	NOUN
fcis-14690	136	9	to	to	PART
fcis-14690	136	10	explore	explore	VERB
fcis-14690	136	11	the	the	DET
fcis-14690	136	12	dependency	dependency	NOUN
fcis-14690	136	13	relationship	relationship	NOUN
fcis-14690	136	14	between	between	ADP
fcis-14690	136	15	them	they	PRON
fcis-14690	136	16	,	,	PUNCT
fcis-14690	136	17	and	and	CCONJ
fcis-14690	136	18	the	the	DET
fcis-14690	136	19	general	general	ADJ
fcis-14690	136	20	flow	flow	NOUN
fcis-14690	136	21	of	of	ADP
fcis-14690	136	22	bilstm	bilstm	NOUN
fcis-14690	136	23	is	be	AUX
fcis-14690	136	24	shown	show	VERB
fcis-14690	136	25	in	in	ADP
fcis-14690	136	26	figure	figure	NOUN
fcis-14690	136	27	4	4	NUM
fcis-14690	136	28	.	.	PUNCT
fcis-14690	137	1	the	the	DET
fcis-14690	137	2	lstm	lstm	NOUN
fcis-14690	137	3	mainly	mainly	ADV
fcis-14690	137	4	includes	include	VERB
fcis-14690	137	5	the	the	DET
fcis-14690	137	6	forgetting	forget	VERB
fcis-14690	137	7	gate	gate	NOUN
fcis-14690	137	8	,	,	PUNCT
fcis-14690	137	9	the	the	DET
fcis-14690	137	10	input	input	NOUN
fcis-14690	137	11	gate	gate	NOUN
fcis-14690	137	12	,	,	PUNCT
fcis-14690	137	13	the	the	DET
fcis-14690	137	14	output	output	NOUN
fcis-14690	137	15	gate	gate	NOUN
fcis-14690	137	16	and	and	CCONJ
fcis-14690	137	17	the	the	DET
fcis-14690	137	18	recording	recording	NOUN
fcis-14690	137	19	unit	unit	NOUN
fcis-14690	137	20	of	of	ADP
fcis-14690	137	21	the	the	DET
fcis-14690	137	22	last	last	ADJ
fcis-14690	137	23	moment	moment	NOUN
fcis-14690	137	24	.	.	PUNCT
fcis-14690	138	1	the	the	DET
fcis-14690	138	2	multi	multi	ADJ
fcis-14690	138	3	-	-	ADJ
fcis-14690	138	4	level	level	ADJ
fcis-14690	138	5	image	image	NOUN
fcis-14690	138	6	features	feature	NOUN
fcis-14690	138	7	are	be	AUX
fcis-14690	138	8	obtained	obtain	VERB
fcis-14690	138	9	in	in	ADP
fcis-14690	138	10	the	the	DET
fcis-14690	138	11	multi	multi	ADJ
fcis-14690	138	12	-	-	ADJ
fcis-14690	138	13	level	level	ADJ
fcis-14690	138	14	feature	feature	NOUN
fcis-14690	138	15	extraction	extraction	NOUN
fcis-14690	138	16	part	part	NOUN
fcis-14690	138	17	,	,	PUNCT
fcis-14690	138	18	and	and	CCONJ
fcis-14690	138	19	they	they	PRON
fcis-14690	138	20	are	be	AUX
fcis-14690	138	21	used	use	VERB
fcis-14690	138	22	as	as	ADP
fcis-14690	138	23	the	the	DET
fcis-14690	138	24	input	input	NOUN
fcis-14690	138	25	of	of	ADP
fcis-14690	138	26	each	each	DET
fcis-14690	138	27	moment	moment	NOUN
fcis-14690	138	28	of	of	ADP
fcis-14690	138	29	lstm	lstm	PROPN
fcis-14690	138	30	unit	unit	NOUN
fcis-14690	138	31	.	.	PUNCT
fcis-14690	139	1	the	the	DET
fcis-14690	139	2	output	output	NOUN
fcis-14690	139	3	process	process	NOUN
fcis-14690	139	4	of	of	ADP
fcis-14690	139	5	the	the	DET
fcis-14690	139	6	entire	entire	ADJ
fcis-14690	139	7	lstm	lstm	ADJ
fcis-14690	139	8	unit	unit	NOUN
fcis-14690	139	9	is	be	AUX
fcis-14690	139	10	as	as	SCONJ
fcis-14690	139	11	follows	follow	VERB
fcis-14690	139	12	:	:	PUNCT
fcis-14690	139	13	)	)	PUNCT
fcis-14690	139	14	)	)	PUNCT
fcis-14690	139	15	,	,	PUNCT
fcis-14690	139	16	(	(	PUNCT
fcis-14690	139	17	(	(	PUNCT
fcis-14690	139	18	1	1	NUM
fcis-14690	139	19	fitft	fitft	NOUN
fcis-14690	139	20	bphwf	bphwf	NOUN
fcis-14690	139	21			PROPN
fcis-14690	139	22			PROPN
fcis-14690	139	23	(	(	PUNCT
fcis-14690	139	24	9	9	NUM
fcis-14690	139	25	)	)	PUNCT
fcis-14690	139	26	)	)	PUNCT
fcis-14690	139	27	)	)	PUNCT
fcis-14690	139	28	,	,	PUNCT
fcis-14690	139	29	(	(	PUNCT
fcis-14690	139	30	(	(	PUNCT
fcis-14690	139	31	1	1	NUM
fcis-14690	139	32	iitit	iitit	NOUN
fcis-14690	139	33	bphwi	bphwi	PROPN
fcis-14690	139	34			PROPN
fcis-14690	139	35			PROPN
fcis-14690	139	36	(	(	PUNCT
fcis-14690	139	37	10	10	NUM
fcis-14690	139	38	)	)	PUNCT
fcis-14690	139	39	)	)	PUNCT
fcis-14690	139	40	)	)	PUNCT
fcis-14690	139	41	,	,	PUNCT
fcis-14690	139	42	(	(	PUNCT
fcis-14690	139	43	tanh	tanh	NOUN
fcis-14690	139	44	(	(	PUNCT
fcis-14690	139	45	~	~	PUNCT
fcis-14690	139	46	1	1	NUM
fcis-14690	139	47	citct	citct	NOUN
fcis-14690	139	48	bphwc	bphwc	PROPN
fcis-14690	139	49			PROPN
fcis-14690	139	50			PROPN
fcis-14690	139	51	(	(	PUNCT
fcis-14690	139	52	11	11	NUM
fcis-14690	139	53	)	)	PUNCT
fcis-14690	139	54	ttttt	ttttt	NOUN
fcis-14690	139	55	cicfc	cicfc	NOUN
fcis-14690	139	56	~	~	PUNCT
fcis-14690	139	57	1	1	NUM
fcis-14690	139	58			NUM
fcis-14690	139	59			X
fcis-14690	139	60	(	(	PUNCT
fcis-14690	139	61	12	12	NUM
fcis-14690	139	62	)	)	PUNCT
fcis-14690	139	63	)	)	PUNCT
fcis-14690	139	64	)	)	PUNCT
fcis-14690	139	65	,	,	PUNCT
fcis-14690	139	66	(	(	PUNCT
fcis-14690	139	67	(	(	PUNCT
fcis-14690	139	68	1	1	NUM
fcis-14690	139	69	oitot	oitot	ADJ
fcis-14690	139	70	bphwo	bphwo	NOUN
fcis-14690	139	71			PROPN
fcis-14690	139	72			PROPN
fcis-14690	139	73	(	(	PUNCT
fcis-14690	139	74	13	13	NUM
fcis-14690	139	75	)	)	PUNCT
fcis-14690	139	76	)	)	PUNCT
fcis-14690	139	77	tanh	tanh	NOUN
fcis-14690	139	78	(	(	PUNCT
fcis-14690	139	79	tit	tit	NOUN
fcis-14690	139	80	coh	coh	NOUN
fcis-14690	139	81			NOUN
fcis-14690	139	82	(	(	PUNCT
fcis-14690	139	83	14	14	NUM
fcis-14690	139	84	)	)	PUNCT
fcis-14690	139	85	where	where	SCONJ
fcis-14690	139	86	tf	tf	INTJ
fcis-14690	139	87	,	,	PUNCT
fcis-14690	139	88	ti	ti	PROPN
fcis-14690	139	89	,	,	PUNCT
fcis-14690	139	90	tc	tc	X
fcis-14690	139	91	~	~	PUNCT
fcis-14690	139	92	,	,	PUNCT
fcis-14690	139	93	tc	tc	X
fcis-14690	139	94	and	and	CCONJ
fcis-14690	139	95	to	to	PART
fcis-14690	139	96	are	are	VERB
fcis-14690	139	97	respectively	respectively	ADV
fcis-14690	139	98	the	the	DET
fcis-14690	139	99	forgetting	forget	VERB
fcis-14690	139	100	gate	gate	NOUN
fcis-14690	139	101	,	,	PUNCT
fcis-14690	139	102	the	the	DET
fcis-14690	139	103	input	input	NOUN
fcis-14690	139	104	gate	gate	NOUN
fcis-14690	139	105	,	,	PUNCT
fcis-14690	139	106	the	the	DET
fcis-14690	139	107	candidate	candidate	NOUN
fcis-14690	139	108	storage	storage	NOUN
fcis-14690	139	109	unit	unit	NOUN
fcis-14690	139	110	value	value	NOUN
fcis-14690	139	111	,	,	PUNCT
fcis-14690	139	112	the	the	DET
fcis-14690	139	113	memory	memory	NOUN
fcis-14690	139	114	unit	unit	NOUN
fcis-14690	139	115	value	value	NOUN
fcis-14690	139	116	of	of	ADP
fcis-14690	139	117	the	the	DET
fcis-14690	139	118	current	current	ADJ
fcis-14690	139	119	moment	moment	NOUN
fcis-14690	139	120	,	,	PUNCT
fcis-14690	139	121	and	and	CCONJ
fcis-14690	139	122	the	the	DET
fcis-14690	139	123	output	output	NOUN
fcis-14690	139	124	gate	gate	NOUN
fcis-14690	139	125	.	.	PUNCT
fcis-14690	140	1	th	th	X
fcis-14690	140	2	is	be	AUX
fcis-14690	140	3	the	the	DET
fcis-14690	140	4	output	output	NOUN
fcis-14690	140	5	information	information	NOUN
fcis-14690	140	6	of	of	ADP
fcis-14690	140	7	the	the	DET
fcis-14690	140	8	lstm	lstm	PROPN
fcis-14690	140	9	unit	unit	NOUN
fcis-14690	140	10	.	.	PUNCT
fcis-14690	141	1	after	after	SCONJ
fcis-14690	141	2	the	the	DET
fcis-14690	141	3	forward	forward	ADJ
fcis-14690	141	4	lstm	lstm	NOUN
fcis-14690	141	5	model	model	NOUN
fcis-14690	141	6	and	and	CCONJ
fcis-14690	141	7	the	the	DET
fcis-14690	141	8	reverse	reverse	ADJ
fcis-14690	141	9	lstm	lstm	PROPN
fcis-14690	141	10	model	model	NOUN
fcis-14690	141	11	,	,	PUNCT
fcis-14690	141	12	the	the	DET
fcis-14690	141	13	forward	forward	ADV
fcis-14690	141	14	hidden	hide	VERB
fcis-14690	141	15	feature	feature	NOUN
fcis-14690	141	16	th	th	X
fcis-14690	141	17			NOUN
fcis-14690	141	18	and	and	CCONJ
fcis-14690	141	19	the	the	DET
fcis-14690	141	20	reverse	reverse	ADJ
fcis-14690	141	21	hidden	hide	VERB
fcis-14690	141	22	feature	feature	NOUN
fcis-14690	141	23	th	th	X
fcis-14690	141	24			NUM
fcis-14690	141	25	are	be	AUX
fcis-14690	141	26	obtained	obtain	VERB
fcis-14690	141	27	for	for	ADP
fcis-14690	141	28	each	each	DET
fcis-14690	141	29	layer	layer	NOUN
fcis-14690	141	30	of	of	ADP
fcis-14690	141	31	image	image	NOUN
fcis-14690	141	32	features	feature	NOUN
fcis-14690	141	33	.	.	PUNCT
fcis-14690	142	1	the	the	DET
fcis-14690	142	2	features	feature	NOUN
fcis-14690	142	3	of	of	ADP
fcis-14690	142	4	five	five	NUM
fcis-14690	142	5	different	different	ADJ
fcis-14690	142	6	levels	level	NOUN
fcis-14690	142	7	of	of	ADP
fcis-14690	142	8	the	the	DET
fcis-14690	142	9	image	image	NOUN
fcis-14690	142	10	are	be	AUX
fcis-14690	142	11	fused	fuse	VERB
fcis-14690	142	12	by	by	ADP
fcis-14690	142	13	bilstm	bilstm	NOUN
fcis-14690	142	14	and	and	CCONJ
fcis-14690	142	15	input	input	NOUN
fcis-14690	142	16	into	into	ADP
fcis-14690	142	17	layernorm	layernorm	NOUN
fcis-14690	142	18	layer	layer	NOUN
fcis-14690	142	19	to	to	PART
fcis-14690	142	20	obtain	obtain	VERB
fcis-14690	142	21	the	the	DET
fcis-14690	142	22	final	final	ADJ
fcis-14690	142	23	global	global	ADJ
fcis-14690	142	24	feature	feature	NOUN
fcis-14690	142	25	f.	f.	PROPN
fcis-14690	142	26	after	after	ADP
fcis-14690	142	27	obtaining	obtain	VERB
fcis-14690	142	28	the	the	DET
fcis-14690	142	29	local	local	ADJ
fcis-14690	142	30	features	feature	NOUN
fcis-14690	142	31	extracted	extract	VERB
fcis-14690	142	32	by	by	ADP
fcis-14690	142	33	cbam	cbam	NOUN
fcis-14690	142	34	module	module	NOUN
fcis-14690	142	35	and	and	CCONJ
fcis-14690	142	36	the	the	DET
fcis-14690	142	37	global	global	ADJ
fcis-14690	142	38	features	feature	NOUN
fcis-14690	142	39	after	after	ADP
fcis-14690	142	40	bilstm	bilstm	NOUN
fcis-14690	142	41	fusion	fusion	NOUN
fcis-14690	142	42	of	of	ADP
fcis-14690	142	43	multilevel	multilevel	ADJ
fcis-14690	142	44	features	feature	NOUN
fcis-14690	142	45	,	,	PUNCT
fcis-14690	142	46	this	this	DET
fcis-14690	142	47	paper	paper	NOUN
fcis-14690	142	48	uses	use	VERB
fcis-14690	142	49	cat	cat	NOUN
fcis-14690	142	50	(	(	PUNCT
fcis-14690	142	51	)	)	PUNCT
fcis-14690	142	52	function	function	NOUN
fcis-14690	142	53	to	to	PART
fcis-14690	142	54	concatenate	concatenate	VERB
fcis-14690	142	55	these	these	DET
fcis-14690	142	56	two	two	NUM
fcis-14690	142	57	features	feature	NOUN
fcis-14690	142	58	to	to	PART
fcis-14690	142	59	obtain	obtain	VERB
fcis-14690	142	60	feature	feature	NOUN
fcis-14690	142	61	f	f	PROPN
fcis-14690	143	1			ADJ
fcis-14690	143	2	.	.	PUNCT
fcis-14690	144	1	finally	finally	ADV
fcis-14690	144	2	,	,	PUNCT
fcis-14690	144	3	the	the	DET
fcis-14690	144	4	dimension	dimension	NOUN
fcis-14690	144	5	is	be	AUX
fcis-14690	144	6	further	far	ADV
fcis-14690	144	7	reduced	reduce	VERB
fcis-14690	144	8	through	through	ADP
fcis-14690	144	9	the	the	DET
fcis-14690	144	10	full	full	ADJ
fcis-14690	144	11	connection	connection	NOUN
fcis-14690	144	12	layer	layer	NOUN
fcis-14690	144	13	and	and	CCONJ
fcis-14690	144	14	sent	send	VERB
fcis-14690	144	15	to	to	ADP
fcis-14690	144	16	the	the	DET
fcis-14690	144	17	softmax	softmax	NOUN
fcis-14690	144	18	layer	layer	NOUN
fcis-14690	144	19	for	for	ADP
fcis-14690	144	20	classification	classification	NOUN
fcis-14690	144	21	:	:	PUNCT
fcis-14690	144	22	)	)	PUNCT
fcis-14690	144	23	)	)	PUNCT
fcis-14690	144	24	)	)	PUNCT
fcis-14690	144	25	)	)	PUNCT
fcis-14690	145	1	(	(	PUNCT
fcis-14690	145	2	(	(	PUNCT
fcis-14690	145	3	(	(	PUNCT
fcis-14690	145	4	remax	remax	X
fcis-14690	145	5	(	(	PUNCT
fcis-14690	145	6	2112	2112	NUM
fcis-14690	145	7	bbfflattenwluwsoftc	bbfflattenwluwsoftc	NOUN
fcis-14690	145	8			NOUN
fcis-14690	145	9	(	(	PUNCT
fcis-14690	145	10	15	15	NUM
fcis-14690	145	11	)	)	PUNCT
fcis-14690	145	12	4	4	NUM
fcis-14690	145	13	.	.	PUNCT
fcis-14690	146	1	the	the	DET
fcis-14690	146	2	experiments	experiment	NOUN
fcis-14690	146	3	4.1	4.1	NUM
fcis-14690	146	4	.	.	PUNCT
fcis-14690	146	5	dataset	dataset	VERB
fcis-14690	146	6	this	this	DET
fcis-14690	146	7	paper	paper	NOUN
fcis-14690	146	8	evaluates	evaluate	VERB
fcis-14690	146	9	our	our	PRON
fcis-14690	146	10	approach	approach	NOUN
fcis-14690	146	11	using	use	VERB
fcis-14690	146	12	two	two	NUM
fcis-14690	146	13	datasets	dataset	NOUN
fcis-14690	146	14	,	,	PUNCT
fcis-14690	146	15	artphoto	artphoto	NOUN
fcis-14690	146	16	and	and	CCONJ
fcis-14690	146	17	abstract	abstract	ADJ
fcis-14690	147	1	[	[	X
fcis-14690	147	2	34	34	NUM
fcis-14690	147	3	]	]	PUNCT
fcis-14690	147	4	,	,	PUNCT
fcis-14690	147	5	the	the	DET
fcis-14690	147	6	details	detail	NOUN
fcis-14690	147	7	of	of	ADP
fcis-14690	147	8	which	which	PRON
fcis-14690	147	9	are	be	AUX
fcis-14690	147	10	shown	show	VERB
fcis-14690	147	11	in	in	ADP
fcis-14690	147	12	table	table	NOUN
fcis-14690	147	13	1	1	NUM
fcis-14690	147	14	.	.	X
fcis-14690	147	15	artphoto	artphoto	NOUN
fcis-14690	147	16	:	:	PUNCT
fcis-14690	147	17	it	it	PRON
fcis-14690	147	18	is	be	AUX
fcis-14690	147	19	a	a	DET
fcis-14690	147	20	selection	selection	NOUN
fcis-14690	147	21	of	of	ADP
fcis-14690	147	22	806	806	NUM
fcis-14690	147	23	photos	photo	NOUN
fcis-14690	147	24	from	from	ADP
fcis-14690	147	25	an	an	DET
fcis-14690	147	26	art	art	NOUN
fcis-14690	147	27	sharing	sharing	NOUN
fcis-14690	147	28	website	website	NOUN
fcis-14690	147	29	,	,	PUNCT
fcis-14690	147	30	which	which	PRON
fcis-14690	147	31	are	be	AUX
fcis-14690	147	32	taken	take	VERB
fcis-14690	147	33	by	by	ADP
fcis-14690	147	34	artists	artist	NOUN
fcis-14690	147	35	who	who	PRON
fcis-14690	147	36	consciously	consciously	ADV
fcis-14690	147	37	manipulate	manipulate	VERB
fcis-14690	147	38	the	the	DET
fcis-14690	147	39	position	position	NOUN
fcis-14690	147	40	,	,	PUNCT
fcis-14690	147	41	brightness	brightness	NOUN
fcis-14690	147	42	,	,	PUNCT
fcis-14690	147	43	color	color	NOUN
fcis-14690	147	44	,	,	PUNCT
fcis-14690	147	45	etc	etc	X
fcis-14690	147	46	.	.	X
fcis-14690	147	47	of	of	ADP
fcis-14690	147	48	the	the	DET
fcis-14690	147	49	image	image	NOUN
fcis-14690	147	50	to	to	PART
fcis-14690	147	51	evoke	evoke	VERB
fcis-14690	147	52	a	a	DET
fcis-14690	147	53	certain	certain	ADJ
fcis-14690	147	54	emotion	emotion	NOUN
fcis-14690	147	55	in	in	ADP
fcis-14690	147	56	the	the	DET
fcis-14690	147	57	viewer	viewer	NOUN
fcis-14690	147	58	.	.	PUNCT
fcis-14690	148	1	abstract	abstract	ADJ
fcis-14690	148	2	:	:	PUNCT
fcis-14690	148	3	this	this	DET
fcis-14690	148	4	dataset	dataset	NOUN
fcis-14690	148	5	consists	consist	VERB
fcis-14690	148	6	of	of	ADP
fcis-14690	148	7	228	228	NUM
fcis-14690	148	8	abstract	abstract	ADJ
fcis-14690	148	9	paintings	painting	NOUN
fcis-14690	148	10	.	.	PUNCT
fcis-14690	149	1	unlike	unlike	ADP
fcis-14690	149	2	images	image	NOUN
fcis-14690	149	3	in	in	ADP
fcis-14690	149	4	the	the	DET
fcis-14690	149	5	artphoto	artphoto	NOUN
fcis-14690	149	6	dataset	dataset	NOUN
fcis-14690	149	7	,	,	PUNCT
fcis-14690	149	8	images	image	NOUN
fcis-14690	149	9	in	in	ADP
fcis-14690	149	10	the	the	DET
fcis-14690	149	11	abstract	abstract	ADJ
fcis-14690	149	12	dataset	dataset	NOUN
fcis-14690	149	13	represent	represent	VERB
fcis-14690	149	14	emotions	emotion	NOUN
fcis-14690	149	15	through	through	ADP
fcis-14690	149	16	overall	overall	ADJ
fcis-14690	149	17	colors	color	NOUN
fcis-14690	149	18	and	and	CCONJ
fcis-14690	149	19	textures	texture	NOUN
fcis-14690	149	20	,	,	PUNCT
fcis-14690	149	21	rather	rather	ADV
fcis-14690	149	22	than	than	ADP
fcis-14690	149	23	some	some	DET
fcis-14690	149	24	emotional	emotional	ADJ
fcis-14690	149	25	object	object	NOUN
fcis-14690	149	26	.	.	PUNCT
fcis-14690	150	1	in	in	ADP
fcis-14690	150	2	this	this	DET
fcis-14690	150	3	dataset	dataset	NOUN
fcis-14690	150	4	,	,	PUNCT
fcis-14690	150	5	each	each	DET
fcis-14690	150	6	painting	painting	NOUN
fcis-14690	150	7	was	be	AUX
fcis-14690	150	8	voted	vote	VERB
fcis-14690	150	9	on	on	ADP
fcis-14690	150	10	by	by	ADP
fcis-14690	150	11	14	14	NUM
fcis-14690	150	12	different	different	ADJ
fcis-14690	150	13	people	people	NOUN
fcis-14690	150	14	to	to	PART
fcis-14690	150	15	determine	determine	VERB
fcis-14690	150	16	its	its	PRON
fcis-14690	150	17	emotional	emotional	ADJ
fcis-14690	150	18	category	category	NOUN
fcis-14690	150	19	.	.	PUNCT
fcis-14690	151	1	select	select	VERB
fcis-14690	151	2	the	the	DET
fcis-14690	151	3	emotion	emotion	NOUN
fcis-14690	151	4	category	category	NOUN
fcis-14690	151	5	with	with	ADP
fcis-14690	151	6	the	the	DET
fcis-14690	151	7	most	most	ADJ
fcis-14690	151	8	votes	vote	NOUN
fcis-14690	151	9	as	as	ADP
fcis-14690	151	10	the	the	DET
fcis-14690	151	11	emotion	emotion	NOUN
fcis-14690	151	12	category	category	NOUN
fcis-14690	151	13	for	for	ADP
fcis-14690	151	14	the	the	DET
fcis-14690	151	15	image	image	NOUN
fcis-14690	151	16	.	.	PUNCT
fcis-14690	152	1	table	table	NOUN
fcis-14690	152	2	1	1	NUM
fcis-14690	152	3	.	.	PUNCT
fcis-14690	153	1	statistical	statistical	ADJ
fcis-14690	153	2	data	datum	NOUN
fcis-14690	153	3	of	of	ADP
fcis-14690	153	4	image	image	NOUN
fcis-14690	153	5	emotion	emotion	NOUN
fcis-14690	153	6	dataset	dataset	VERB
fcis-14690	153	7	dataset	dataset	VERB
fcis-14690	153	8	positive	positive	ADJ
fcis-14690	153	9	negative	negative	ADJ
fcis-14690	153	10	sum	sum	NOUN
fcis-14690	153	11	artphoto	artphoto	NOUN
fcis-14690	153	12	378	378	NUM
fcis-14690	153	13	428	428	NUM
fcis-14690	153	14	806	806	NUM
fcis-14690	153	15	abstract	abstract	ADJ
fcis-14690	153	16	139	139	NUM
fcis-14690	153	17	89	89	NUM
fcis-14690	153	18	228	228	NUM
fcis-14690	153	19	4.2	4.2	NUM
fcis-14690	153	20	.	.	PUNCT
fcis-14690	154	1	experiment	experiment	NOUN
fcis-14690	154	2	settings	setting	NOUN
fcis-14690	154	3	the	the	DET
fcis-14690	154	4	model	model	NOUN
fcis-14690	154	5	in	in	ADP
fcis-14690	154	6	this	this	DET
fcis-14690	154	7	article	article	NOUN
fcis-14690	154	8	is	be	AUX
fcis-14690	154	9	based	base	VERB
fcis-14690	154	10	on	on	ADP
fcis-14690	154	11	a	a	DET
fcis-14690	154	12	pre	pre	ADJ
fcis-14690	154	13	-	-	ADJ
fcis-14690	154	14	trained	train	VERB
fcis-14690	154	15	resnet18	resnet18	NOUN
fcis-14690	154	16	network	network	NOUN
fcis-14690	154	17	and	and	CCONJ
fcis-14690	154	18	implemented	implement	VERB
fcis-14690	154	19	using	use	VERB
fcis-14690	154	20	the	the	DET
fcis-14690	154	21	pytorch	pytorch	NOUN
fcis-14690	154	22	framework	framework	NOUN
fcis-14690	154	23	,	,	PUNCT
fcis-14690	154	24	with	with	ADP
fcis-14690	154	25	40	40	NUM
fcis-14690	154	26	epochs	epoch	NOUN
fcis-14690	154	27	trained	train	VERB
fcis-14690	154	28	on	on	ADP
fcis-14690	154	29	nvidia	nvidia	PROPN
fcis-14690	154	30	geforce	geforce	NOUN
fcis-14690	154	31	mx450	mx450	PROPN
fcis-14690	154	32	.	.	PUNCT
fcis-14690	155	1	the	the	DET
fcis-14690	155	2	batch	batch	NOUN
fcis-14690	155	3	size	size	NOUN
fcis-14690	155	4	and	and	CCONJ
fcis-14690	155	5	learning	learning	NOUN
fcis-14690	155	6	rate	rate	NOUN
fcis-14690	155	7	are	be	AUX
fcis-14690	155	8	set	set	VERB
fcis-14690	155	9	to	to	ADP
fcis-14690	155	10	4	4	NUM
fcis-14690	155	11	and	and	CCONJ
fcis-14690	155	12	0.0001	0.0001	NUM
fcis-14690	155	13	,	,	PUNCT
fcis-14690	155	14	respectively	respectively	ADV
fcis-14690	155	15	.	.	PUNCT
fcis-14690	156	1	during	during	ADP
fcis-14690	156	2	model	model	NOUN
fcis-14690	156	3	training	training	NOUN
fcis-14690	156	4	,	,	PUNCT
fcis-14690	156	5	images	image	NOUN
fcis-14690	156	6	were	be	AUX
fcis-14690	156	7	randomly	randomly	ADV
fcis-14690	156	8	cropped	crop	VERB
fcis-14690	156	9	to	to	ADP
fcis-14690	156	10	different	different	ADJ
fcis-14690	156	11	sizes	size	NOUN
fcis-14690	156	12	and	and	CCONJ
fcis-14690	156	13	aspect	aspect	NOUN
fcis-14690	156	14	ratios	ratio	NOUN
fcis-14690	156	15	and	and	CCONJ
fcis-14690	156	16	scaled	scale	VERB
fcis-14690	156	17	to	to	ADP
fcis-14690	156	18	224×224	224×224	NUM
fcis-14690	156	19	pixels	pixel	NOUN
fcis-14690	156	20	,	,	PUNCT
fcis-14690	156	21	and	and	CCONJ
fcis-14690	156	22	then	then	ADV
fcis-14690	156	23	randomly	randomly	ADV
fcis-14690	156	24	flipped	flip	VERB
fcis-14690	156	25	horizontally	horizontally	ADV
fcis-14690	156	26	with	with	ADP
fcis-14690	156	27	a	a	DET
fcis-14690	156	28	probability	probability	NOUN
fcis-14690	156	29	of	of	ADP
fcis-14690	156	30	0.5	0.5	NUM
fcis-14690	156	31	to	to	PART
fcis-14690	156	32	expand	expand	VERB
fcis-14690	156	33	the	the	DET
fcis-14690	156	34	data	datum	NOUN
fcis-14690	156	35	and	and	CCONJ
fcis-14690	156	36	prevent	prevent	VERB
fcis-14690	156	37	overfitting	overfitting	NOUN
fcis-14690	156	38	.	.	PUNCT
fcis-14690	157	1	finally	finally	ADV
fcis-14690	157	2	,	,	PUNCT
fcis-14690	157	3	the	the	DET
fcis-14690	157	4	image	image	NOUN
fcis-14690	157	5	is	be	AUX
fcis-14690	157	6	standardized	standardize	VERB
fcis-14690	157	7	according	accord	VERB
fcis-14690	157	8	to	to	ADP
fcis-14690	157	9	the	the	DET
fcis-14690	157	10	channel	channel	NOUN
fcis-14690	157	11	,	,	PUNCT
fcis-14690	157	12	so	so	SCONJ
fcis-14690	157	13	as	as	SCONJ
fcis-14690	157	14	to	to	PART
fcis-14690	157	15	accelerate	accelerate	VERB
fcis-14690	157	16	the	the	DET
fcis-14690	157	17	convergence	convergence	NOUN
fcis-14690	157	18	speed	speed	NOUN
fcis-14690	157	19	of	of	ADP
fcis-14690	157	20	the	the	DET
fcis-14690	157	21	model	model	NOUN
fcis-14690	157	22	.	.	PUNCT
fcis-14690	158	1	artphoto	artphoto	NOUN
fcis-14690	158	2	and	and	CCONJ
fcis-14690	158	3	abstract	abstract	ADJ
fcis-14690	158	4	were	be	AUX
fcis-14690	158	5	randomly	randomly	ADV
fcis-14690	158	6	divided	divide	VERB
fcis-14690	158	7	into	into	ADP
fcis-14690	158	8	80	80	NUM
fcis-14690	158	9	%	%	NOUN
fcis-14690	158	10	training	training	NOUN
fcis-14690	158	11	datasets	dataset	NOUN
fcis-14690	158	12	and	and	CCONJ
fcis-14690	158	13	20	20	NUM
fcis-14690	158	14	%	%	NOUN
fcis-14690	158	15	test	test	NOUN
fcis-14690	158	16	datasets	dataset	NOUN
fcis-14690	158	17	.	.	PUNCT
fcis-14690	159	1	4.3	4.3	NUM
fcis-14690	159	2	.	.	PUNCT
fcis-14690	159	3	compare	compare	VERB
fcis-14690	159	4	with	with	ADP
fcis-14690	159	5	the	the	DET
fcis-14690	159	6	previous	previous	ADJ
fcis-14690	159	7	methods	method	NOUN
fcis-14690	159	8	the	the	DET
fcis-14690	159	9	method	method	NOUN
fcis-14690	159	10	presented	present	VERB
fcis-14690	159	11	in	in	ADP
fcis-14690	159	12	this	this	DET
fcis-14690	159	13	paper	paper	NOUN
fcis-14690	159	14	is	be	AUX
fcis-14690	159	15	compared	compare	VERB
fcis-14690	159	16	with	with	ADP
fcis-14690	159	17	the	the	DET
fcis-14690	159	18	following	follow	VERB
fcis-14690	159	19	different	different	ADJ
fcis-14690	159	20	baselines	baseline	NOUN
fcis-14690	159	21	:	:	PUNCT
fcis-14690	159	22	gch	gch	NOUN
fcis-14690	160	1	[	[	X
fcis-14690	160	2	15	15	NUM
fcis-14690	160	3	]	]	X
fcis-14690	160	4	:	:	PUNCT
fcis-14690	160	5	the	the	DET
fcis-14690	160	6	global	global	ADJ
fcis-14690	160	7	view	view	NOUN
fcis-14690	160	8	of	of	ADP
fcis-14690	160	9	the	the	DET
fcis-14690	160	10	image	image	NOUN
fcis-14690	160	11	is	be	AUX
fcis-14690	160	12	composed	compose	VERB
fcis-14690	160	13	using	use	VERB
fcis-14690	160	14	a	a	DET
fcis-14690	160	15	64	64	NUM
fcis-14690	160	16	-	-	PUNCT
fcis-14690	160	17	bit	bit	NOUN
fcis-14690	160	18	binary	binary	ADJ
fcis-14690	160	19	color	color	NOUN
fcis-14690	160	20	histogram	histogram	NOUN
fcis-14690	160	21	feature	feature	NOUN
fcis-14690	160	22	.	.	PUNCT
fcis-14690	161	1	sentibank	sentibank	NOUN
fcis-14690	162	1	[	[	X
fcis-14690	162	2	29	29	NUM
fcis-14690	162	3	]	]	SYM
fcis-14690	162	4	:	:	PUNCT
fcis-14690	162	5	a	a	DET
fcis-14690	162	6	1200	1200	NUM
fcis-14690	162	7	-	-	PUNCT
fcis-14690	162	8	dimensional	dimensional	ADJ
fcis-14690	162	9	intermediate	intermediate	ADJ
fcis-14690	162	10	feature	feature	NOUN
fcis-14690	162	11	called	call	VERB
fcis-14690	162	12	adjective	adjective	ADJ
fcis-14690	162	13	-	-	PUNCT
fcis-14690	162	14	noun	noun	NOUN
fcis-14690	162	15	pair	pair	NOUN
fcis-14690	162	16	(	(	PUNCT
fcis-14690	162	17	anps	anps	NOUN
fcis-14690	162	18	)	)	PUNCT
fcis-14690	162	19	was	be	AUX
fcis-14690	162	20	proposed	propose	VERB
fcis-14690	162	21	to	to	PART
fcis-14690	162	22	describe	describe	VERB
fcis-14690	162	23	the	the	DET
fcis-14690	162	24	relationship	relationship	NOUN
fcis-14690	162	25	between	between	ADP
fcis-14690	162	26	image	image	NOUN
fcis-14690	162	27	content	content	NOUN
fcis-14690	162	28	and	and	CCONJ
fcis-14690	162	29	emotion	emotion	NOUN
fcis-14690	162	30	.	.	PUNCT
fcis-14690	163	1	this	this	DET
fcis-14690	163	2	study	study	NOUN
fcis-14690	163	3	is	be	AUX
fcis-14690	163	4	an	an	DET
fcis-14690	163	5	important	important	ADJ
fcis-14690	163	6	work	work	NOUN
fcis-14690	163	7	to	to	PART
fcis-14690	163	8	explore	explore	VERB
fcis-14690	163	9	the	the	DET
fcis-14690	163	10	correspondence	correspondence	NOUN
fcis-14690	163	11	between	between	ADP
fcis-14690	163	12	early	early	ADJ
fcis-14690	163	13	semantic	semantic	ADJ
fcis-14690	163	14	information	information	NOUN
fcis-14690	163	15	and	and	CCONJ
fcis-14690	163	16	emotion	emotion	NOUN
fcis-14690	163	17	.	.	PUNCT
fcis-14690	164	1	rao	rao	NOUN
fcis-14690	165	1	[	[	X
fcis-14690	165	2	30	30	NUM
fcis-14690	165	3	]	]	SYM
fcis-14690	165	4	:	:	PUNCT
fcis-14690	165	5	early	early	ADJ
fcis-14690	165	6	explorations	exploration	NOUN
fcis-14690	165	7	in	in	ADP
fcis-14690	165	8	the	the	DET
fcis-14690	165	9	analysis	analysis	NOUN
fcis-14690	165	10	of	of	ADP
fcis-14690	165	11	local	local	ADJ
fcis-14690	165	12	areas	area	NOUN
fcis-14690	165	13	related	relate	VERB
fcis-14690	165	14	to	to	ADP
fcis-14690	165	15	emotion	emotion	NOUN
fcis-14690	165	16	.	.	PUNCT
fcis-14690	166	1	the	the	DET
fcis-14690	166	2	image	image	NOUN
fcis-14690	166	3	is	be	AUX
fcis-14690	166	4	segmented	segment	VERB
fcis-14690	166	5	into	into	ADP
fcis-14690	166	6	different	different	ADJ
fcis-14690	166	7	blocks	block	NOUN
fcis-14690	166	8	by	by	ADP
fcis-14690	166	9	image	image	NOUN
fcis-14690	166	10	segmentation	segmentation	NOUN
fcis-14690	166	11	,	,	PUNCT
fcis-14690	166	12	called	call	VERB
fcis-14690	166	13	multi	multi	ADJ
fcis-14690	166	14	-	-	ADJ
fcis-14690	166	15	scale	scale	ADJ
fcis-14690	166	16	blocks	block	NOUN
fcis-14690	166	17	,	,	PUNCT
fcis-14690	166	18	and	and	CCONJ
fcis-14690	166	19	the	the	DET
fcis-14690	166	20	vision	vision	NOUN
fcis-14690	166	21	bag	bag	NOUN
fcis-14690	166	22	features	feature	NOUN
fcis-14690	166	23	based	base	VERB
fcis-14690	166	24	on	on	ADP
fcis-14690	166	25	sift	sift	ADJ
fcis-14690	166	26	contain	contain	VERB
fcis-14690	166	27	both	both	DET
fcis-14690	166	28	local	local	ADJ
fcis-14690	166	29	and	and	CCONJ
fcis-14690	166	30	global	global	ADJ
fcis-14690	166	31	information	information	NOUN
fcis-14690	166	32	extracted	extract	VERB
fcis-14690	166	33	from	from	ADP
fcis-14690	166	34	the	the	DET
fcis-14690	166	35	image	image	NOUN
fcis-14690	166	36	blocks	block	NOUN
fcis-14690	166	37	.	.	PUNCT
fcis-14690	167	1	pcnn	pcnn	PROPN
fcis-14690	168	1	[	[	X
fcis-14690	168	2	31	31	NUM
fcis-14690	168	3	]	]	PUNCT
fcis-14690	168	4	:	:	PUNCT
fcis-14690	168	5	a	a	DET
fcis-14690	168	6	progressive	progressive	ADJ
fcis-14690	168	7	training	training	NOUN
fcis-14690	168	8	framework	framework	NOUN
fcis-14690	168	9	based	base	VERB
fcis-14690	168	10	on	on	ADP
fcis-14690	168	11	vggnet	vggnet	NOUN
fcis-14690	168	12	.	.	PUNCT
fcis-14690	169	1	they	they	PRON
fcis-14690	169	2	used	use	VERB
fcis-14690	169	3	large	large	ADJ
fcis-14690	169	4	amounts	amount	NOUN
fcis-14690	169	5	of	of	ADP
fcis-14690	169	6	weakly	weakly	ADJ
fcis-14690	169	7	supervised	supervised	ADJ
fcis-14690	169	8	data	datum	NOUN
fcis-14690	169	9	to	to	PART
fcis-14690	169	10	make	make	VERB
fcis-14690	169	11	the	the	DET
fcis-14690	169	12	model	model	NOUN
fcis-14690	169	13	learn	learn	VERB
fcis-14690	169	14	some	some	DET
fcis-14690	169	15	common	common	ADJ
fcis-14690	169	16	visual	visual	ADJ
fcis-14690	169	17	features	feature	NOUN
fcis-14690	169	18	to	to	PART
fcis-14690	169	19	reduce	reduce	VERB
fcis-14690	169	20	the	the	DET
fcis-14690	169	21	difficulty	difficulty	NOUN
fcis-14690	169	22	of	of	ADP
fcis-14690	169	23	training	train	VERB
fcis-14690	169	24	visual	visual	ADJ
fcis-14690	169	25	emotion	emotion	NOUN
fcis-14690	169	26	datasets	dataset	NOUN
fcis-14690	169	27	.	.	PUNCT
fcis-14690	170	1	ar	ar	PROPN
fcis-14690	171	1	[	[	X
fcis-14690	171	2	32	32	NUM
fcis-14690	171	3	]	]	PUNCT
fcis-14690	171	4	:	:	PUNCT
fcis-14690	171	5	this	this	DET
fcis-14690	171	6	study	study	NOUN
fcis-14690	171	7	proposes	propose	VERB
fcis-14690	171	8	a	a	DET
fcis-14690	171	9	new	new	ADJ
fcis-14690	171	10	concept	concept	NOUN
fcis-14690	171	11	of	of	ADP
fcis-14690	171	12	emotion	emotion	NOUN
fcis-14690	171	13	region	region	NOUN
fcis-14690	171	14	to	to	PART
fcis-14690	171	15	explore	explore	VERB
fcis-14690	171	16	local	local	ADJ
fcis-14690	171	17	region	region	NOUN
fcis-14690	171	18	and	and	CCONJ
fcis-14690	171	19	emotion	emotion	NOUN
fcis-14690	171	20	arousal	arousal	NOUN
fcis-14690	171	21	,	,	PUNCT
fcis-14690	171	22	and	and	CCONJ
fcis-14690	171	23	uses	use	VERB
fcis-14690	171	24	ready	ready	ADV
fcis-14690	171	25	-	-	PUNCT
fcis-14690	171	26	made	make	VERB
fcis-14690	171	27	object	object	NOUN
fcis-14690	171	28	detection	detection	NOUN
fcis-14690	171	29	technology	technology	NOUN
fcis-14690	171	30	as	as	ADP
fcis-14690	171	31	local	local	ADJ
fcis-14690	171	32	information	information	NOUN
fcis-14690	171	33	for	for	ADP
fcis-14690	171	34	analysis	analysis	NOUN
fcis-14690	171	35	in	in	ADP
fcis-14690	171	36	combination	combination	NOUN
fcis-14690	171	37	with	with	ADP
fcis-14690	171	38	vgg	vgg	PROPN
fcis-14690	171	39	model	model	NOUN
fcis-14690	171	40	.	.	PUNCT
fcis-14690	172	1	64	64	NUM
fcis-14690	172	2	cnngsr	cnngsr	VERB
fcis-14690	172	3	[	[	X
fcis-14690	172	4	33	33	NUM
fcis-14690	172	5	]	]	X
fcis-14690	172	6	:	:	PUNCT
fcis-14690	172	7	xiong	xiong	PROPN
fcis-14690	172	8	et	et	PROPN
fcis-14690	172	9	al	al	PROPN
fcis-14690	172	10	proposed	propose	VERB
fcis-14690	172	11	r	r	X
fcis-14690	172	12	-	-	PUNCT
fcis-14690	172	13	cnngsr	cnngsr	PROPN
fcis-14690	172	14	obtained	obtain	VERB
fcis-14690	172	15	the	the	DET
fcis-14690	172	16	initial	initial	ADJ
fcis-14690	172	17	emotion	emotion	NOUN
fcis-14690	172	18	prediction	prediction	NOUN
fcis-14690	172	19	model	model	NOUN
fcis-14690	172	20	by	by	ADP
fcis-14690	172	21	using	use	VERB
fcis-14690	172	22	group	group	NOUN
fcis-14690	172	23	sparse	sparse	ADJ
fcis-14690	172	24	regularization	regularization	NOUN
fcis-14690	172	25	through	through	ADP
fcis-14690	172	26	cnn	cnn	PROPN
fcis-14690	172	27	,	,	PUNCT
fcis-14690	172	28	and	and	CCONJ
fcis-14690	172	29	then	then	ADV
fcis-14690	172	30	obtained	obtain	VERB
fcis-14690	172	31	a	a	DET
fcis-14690	172	32	compact	compact	ADJ
fcis-14690	172	33	neural	neural	ADJ
fcis-14690	172	34	network	network	NOUN
fcis-14690	172	35	,	,	PUNCT
fcis-14690	172	36	then	then	ADV
fcis-14690	172	37	combined	combine	VERB
fcis-14690	172	38	the	the	DET
fcis-14690	172	39	underlying	underlie	VERB
fcis-14690	172	40	features	feature	NOUN
fcis-14690	172	41	and	and	CCONJ
fcis-14690	172	42	emotion	emotion	NOUN
fcis-14690	172	43	features	feature	NOUN
fcis-14690	172	44	to	to	PART
fcis-14690	172	45	automatically	automatically	ADV
fcis-14690	172	46	detect	detect	VERB
fcis-14690	172	47	the	the	DET
fcis-14690	172	48	emotion	emotion	NOUN
fcis-14690	172	49	region	region	NOUN
fcis-14690	172	50	,	,	PUNCT
fcis-14690	172	51	and	and	CCONJ
fcis-14690	172	52	finally	finally	ADV
fcis-14690	172	53	integrated	integrate	VERB
fcis-14690	172	54	the	the	DET
fcis-14690	172	55	whole	whole	ADJ
fcis-14690	172	56	image	image	NOUN
fcis-14690	172	57	and	and	CCONJ
fcis-14690	172	58	emotion	emotion	NOUN
fcis-14690	172	59	region	region	NOUN
fcis-14690	172	60	to	to	PART
fcis-14690	172	61	predict	predict	VERB
fcis-14690	172	62	the	the	DET
fcis-14690	172	63	overall	overall	ADJ
fcis-14690	172	64	emotion	emotion	NOUN
fcis-14690	172	65	of	of	ADP
fcis-14690	172	66	the	the	DET
fcis-14690	172	67	image	image	NOUN
fcis-14690	172	68	.	.	PUNCT
fcis-14690	173	1	zhang	zhang	PROPN
fcis-14690	174	1	[	[	X
fcis-14690	174	2	28	28	NUM
fcis-14690	174	3	]	]	X
fcis-14690	174	4	:	:	PUNCT
fcis-14690	174	5	zhang	zhang	PROPN
fcis-14690	174	6	et	et	PROPN
fcis-14690	174	7	al	al	PROPN
fcis-14690	174	8	.	.	PROPN
fcis-14690	174	9	proposed	propose	VERB
fcis-14690	174	10	a	a	DET
fcis-14690	174	11	multi	multi	ADJ
fcis-14690	174	12	-	-	ADJ
fcis-14690	174	13	level	level	ADJ
fcis-14690	174	14	hybrid	hybrid	NOUN
fcis-14690	174	15	model	model	NOUN
fcis-14690	174	16	that	that	PRON
fcis-14690	174	17	learns	learn	VERB
fcis-14690	174	18	and	and	CCONJ
fcis-14690	174	19	integrates	integrate	VERB
fcis-14690	174	20	deep	deep	ADJ
fcis-14690	174	21	semantics	semantic	NOUN
fcis-14690	174	22	and	and	CCONJ
fcis-14690	174	23	shallow	shallow	ADJ
fcis-14690	174	24	visual	visual	ADJ
fcis-14690	174	25	representations	representation	NOUN
fcis-14690	174	26	for	for	ADP
fcis-14690	174	27	emotion	emotion	NOUN
fcis-14690	174	28	classification	classification	NOUN
fcis-14690	174	29	.	.	PUNCT
fcis-14690	175	1	4.4	4.4	NUM
fcis-14690	175	2	.	.	PUNCT
fcis-14690	176	1	experimental	experimental	ADJ
fcis-14690	176	2	result	result	NOUN
fcis-14690	176	3	in	in	ADP
fcis-14690	176	4	the	the	DET
fcis-14690	176	5	experiment	experiment	NOUN
fcis-14690	176	6	,	,	PUNCT
fcis-14690	176	7	two	two	NUM
fcis-14690	176	8	datasets	dataset	NOUN
fcis-14690	176	9	,	,	PUNCT
fcis-14690	176	10	artphoto	artphoto	NOUN
fcis-14690	176	11	and	and	CCONJ
fcis-14690	176	12	abstract	abstract	ADJ
fcis-14690	176	13	,	,	PUNCT
fcis-14690	176	14	were	be	AUX
fcis-14690	176	15	randomly	randomly	ADV
fcis-14690	176	16	divided	divide	VERB
fcis-14690	176	17	into	into	ADP
fcis-14690	176	18	a	a	DET
fcis-14690	176	19	training	training	NOUN
fcis-14690	176	20	set	set	NOUN
fcis-14690	176	21	and	and	CCONJ
fcis-14690	176	22	a	a	DET
fcis-14690	176	23	test	test	NOUN
fcis-14690	176	24	set	set	VERB
fcis-14690	176	25	at	at	ADP
fcis-14690	176	26	a	a	DET
fcis-14690	176	27	ratio	ratio	NOUN
fcis-14690	176	28	of	of	ADP
fcis-14690	176	29	8:2	8:2	NUM
fcis-14690	176	30	.	.	PUNCT
fcis-14690	177	1	in	in	ADP
fcis-14690	177	2	the	the	DET
fcis-14690	177	3	above	above	ADJ
fcis-14690	177	4	experimental	experimental	ADJ
fcis-14690	177	5	environment	environment	NOUN
fcis-14690	177	6	,	,	PUNCT
fcis-14690	177	7	the	the	DET
fcis-14690	177	8	performance	performance	NOUN
fcis-14690	177	9	effects	effect	NOUN
fcis-14690	177	10	of	of	ADP
fcis-14690	177	11	maml	maml	ADJ
fcis-14690	177	12	emotion	emotion	NOUN
fcis-14690	177	13	model	model	NOUN
fcis-14690	177	14	on	on	ADP
fcis-14690	177	15	the	the	DET
fcis-14690	177	16	two	two	NUM
fcis-14690	177	17	data	data	NOUN
fcis-14690	177	18	sets	set	NOUN
fcis-14690	177	19	are	be	AUX
fcis-14690	177	20	shown	show	VERB
fcis-14690	177	21	in	in	ADP
fcis-14690	177	22	table	table	NOUN
fcis-14690	177	23	3	3	NUM
fcis-14690	177	24	.	.	PUNCT
fcis-14690	178	1	as	as	SCONJ
fcis-14690	178	2	can	can	AUX
fcis-14690	178	3	be	be	AUX
fcis-14690	178	4	seen	see	VERB
fcis-14690	178	5	from	from	ADP
fcis-14690	178	6	table	table	NOUN
fcis-14690	178	7	3	3	NUM
fcis-14690	178	8	,	,	PUNCT
fcis-14690	178	9	the	the	DET
fcis-14690	178	10	accuracy	accuracy	NOUN
fcis-14690	178	11	of	of	ADP
fcis-14690	178	12	maml	maml	NOUN
fcis-14690	178	13	model	model	NOUN
fcis-14690	178	14	on	on	ADP
fcis-14690	178	15	artphoto	artphoto	NOUN
fcis-14690	178	16	and	and	CCONJ
fcis-14690	178	17	abstract	abstract	ADJ
fcis-14690	178	18	datasets	dataset	NOUN
fcis-14690	178	19	reached	reach	VERB
fcis-14690	178	20	79.38	79.38	NUM
fcis-14690	178	21	%	%	NOUN
fcis-14690	178	22	and	and	CCONJ
fcis-14690	178	23	82.14	82.14	NUM
fcis-14690	178	24	%	%	NOUN
fcis-14690	178	25	respectively	respectively	ADV
fcis-14690	178	26	,	,	PUNCT
fcis-14690	178	27	higher	high	ADJ
fcis-14690	178	28	than	than	ADP
fcis-14690	178	29	the	the	DET
fcis-14690	178	30	baseline	baseline	NOUN
fcis-14690	178	31	method	method	NOUN
fcis-14690	178	32	,	,	PUNCT
fcis-14690	178	33	thus	thus	ADV
fcis-14690	178	34	verifying	verify	VERB
fcis-14690	178	35	the	the	DET
fcis-14690	178	36	effectiveness	effectiveness	NOUN
fcis-14690	178	37	of	of	ADP
fcis-14690	178	38	the	the	DET
fcis-14690	178	39	proposed	propose	VERB
fcis-14690	178	40	method	method	NOUN
fcis-14690	178	41	.	.	PUNCT
fcis-14690	179	1	table	table	NOUN
fcis-14690	179	2	2	2	NUM
fcis-14690	179	3	.	.	PUNCT
fcis-14690	180	1	classification	classification	NOUN
fcis-14690	180	2	results	result	NOUN
fcis-14690	180	3	of	of	ADP
fcis-14690	180	4	different	different	ADJ
fcis-14690	180	5	methods	method	NOUN
fcis-14690	180	6	method	method	VERB
fcis-14690	180	7	dataset	dataset	NOUN
fcis-14690	180	8	artphoto	artphoto	NOUN
fcis-14690	180	9	abstract	abstract	ADJ
fcis-14690	180	10	gch	gch	NOUN
fcis-14690	180	11	66.53	66.53	NUM
fcis-14690	180	12	67.33	67.33	NUM
fcis-14690	180	13	sentibank	sentibank	NOUN
fcis-14690	180	14	67.33	67.33	NUM
fcis-14690	180	15	64.30	64.30	NUM
fcis-14690	180	16	rao	rao	NOUN
fcis-14690	180	17	71.53	71.53	NUM
fcis-14690	180	18	67.82	67.82	NUM
fcis-14690	180	19	pcnn	pcnn	NOUN
fcis-14690	180	20	70.96	70.96	NUM
fcis-14690	180	21	70.84	70.84	NUM
fcis-14690	180	22	ar	ar	NOUN
fcis-14690	180	23	74.80	74.80	NUM
fcis-14690	180	24	76.03	76.03	NUM
fcis-14690	180	25	r	r	NOUN
fcis-14690	180	26	-	-	PUNCT
fcis-14690	180	27	cnngsr	cnngsr	ADJ
fcis-14690	180	28	75.02	75.02	NUM
fcis-14690	180	29	75.89	75.89	NUM
fcis-14690	180	30	zhang	zhang	X
fcis-14690	180	31	75.63	75.63	NUM
fcis-14690	180	32	77.85	77.85	NUM
fcis-14690	180	33	ours	ours	PRON
fcis-14690	180	34	79.38	79.38	NUM
fcis-14690	180	35	82.14	82.14	NUM
fcis-14690	180	36	4.5	4.5	NUM
fcis-14690	180	37	.	.	PUNCT
fcis-14690	181	1	contrast	contrast	NOUN
fcis-14690	181	2	experiment	experiment	NOUN
fcis-14690	181	3	our	our	PRON
fcis-14690	181	4	method	method	NOUN
fcis-14690	181	5	is	be	AUX
fcis-14690	181	6	compared	compare	VERB
fcis-14690	181	7	with	with	ADP
fcis-14690	181	8	several	several	ADJ
fcis-14690	181	9	previous	previous	ADJ
fcis-14690	181	10	methods	method	NOUN
fcis-14690	181	11	on	on	ADP
fcis-14690	181	12	two	two	NUM
fcis-14690	181	13	datasets	dataset	NOUN
fcis-14690	181	14	,	,	PUNCT
fcis-14690	181	15	artphoto	artphoto	NOUN
fcis-14690	181	16	and	and	CCONJ
fcis-14690	181	17	abstract	abstract	ADJ
fcis-14690	181	18	.	.	PUNCT
fcis-14690	182	1	as	as	SCONJ
fcis-14690	182	2	can	can	AUX
fcis-14690	182	3	be	be	AUX
fcis-14690	182	4	seen	see	VERB
fcis-14690	182	5	from	from	ADP
fcis-14690	182	6	table	table	NOUN
fcis-14690	182	7	3	3	NUM
fcis-14690	182	8	,	,	PUNCT
fcis-14690	182	9	the	the	DET
fcis-14690	182	10	method	method	NOUN
fcis-14690	182	11	based	base	VERB
fcis-14690	182	12	on	on	ADP
fcis-14690	182	13	deep	deep	ADJ
fcis-14690	182	14	learning	learning	NOUN
fcis-14690	182	15	has	have	VERB
fcis-14690	182	16	better	well	ADJ
fcis-14690	182	17	performance	performance	NOUN
fcis-14690	182	18	than	than	ADP
fcis-14690	182	19	the	the	DET
fcis-14690	182	20	method	method	NOUN
fcis-14690	182	21	based	base	VERB
fcis-14690	182	22	on	on	ADP
fcis-14690	182	23	traditional	traditional	ADJ
fcis-14690	182	24	manual	manual	ADJ
fcis-14690	182	25	features	feature	NOUN
fcis-14690	182	26	.	.	PUNCT
fcis-14690	183	1	however	however	ADV
fcis-14690	183	2	,	,	PUNCT
fcis-14690	183	3	the	the	DET
fcis-14690	183	4	method	method	NOUN
fcis-14690	183	5	based	base	VERB
fcis-14690	183	6	on	on	ADP
fcis-14690	183	7	sift	sift	ADJ
fcis-14690	183	8	visual	visual	ADJ
fcis-14690	183	9	bag	bag	NOUN
fcis-14690	183	10	features	feature	NOUN
fcis-14690	183	11	of	of	ADP
fcis-14690	183	12	rao	rao	PROPN
fcis-14690	183	13	et	et	PROPN
fcis-14690	183	14	al	al	PROPN
fcis-14690	183	15	.	.	PUNCT
fcis-14690	184	1	[	[	X
fcis-14690	184	2	30	30	NUM
fcis-14690	184	3	]	]	PUNCT
fcis-14690	184	4	,	,	PUNCT
fcis-14690	184	5	which	which	PRON
fcis-14690	184	6	extracts	extract	VERB
fcis-14690	184	7	local	local	ADJ
fcis-14690	184	8	and	and	CCONJ
fcis-14690	184	9	global	global	ADJ
fcis-14690	184	10	information	information	NOUN
fcis-14690	184	11	from	from	ADP
fcis-14690	184	12	image	image	NOUN
fcis-14690	184	13	blocks	block	NOUN
fcis-14690	184	14	,	,	PUNCT
fcis-14690	184	15	has	have	VERB
fcis-14690	184	16	0.57	0.57	NUM
fcis-14690	184	17	%	%	NOUN
fcis-14690	184	18	higher	high	ADJ
fcis-14690	184	19	accuracy	accuracy	NOUN
fcis-14690	184	20	in	in	ADP
fcis-14690	184	21	emotion	emotion	NOUN
fcis-14690	184	22	prediction	prediction	NOUN
fcis-14690	184	23	on	on	ADP
fcis-14690	184	24	artphoto	artphoto	NOUN
fcis-14690	184	25	data	datum	NOUN
fcis-14690	184	26	set	set	VERB
fcis-14690	184	27	than	than	ADP
fcis-14690	184	28	that	that	PRON
fcis-14690	184	29	of	of	ADP
fcis-14690	184	30	pcnn	pcnn	PROPN
fcis-14690	184	31	method	method	NOUN
fcis-14690	184	32	[	[	X
fcis-14690	184	33	31	31	NUM
fcis-14690	184	34	]	]	PUNCT
fcis-14690	184	35	.	.	PUNCT
fcis-14690	185	1	it	it	PRON
fcis-14690	185	2	shows	show	VERB
fcis-14690	185	3	that	that	SCONJ
fcis-14690	185	4	manual	manual	ADJ
fcis-14690	185	5	features	feature	NOUN
fcis-14690	185	6	play	play	VERB
fcis-14690	185	7	an	an	DET
fcis-14690	185	8	important	important	ADJ
fcis-14690	185	9	role	role	NOUN
fcis-14690	185	10	in	in	ADP
fcis-14690	185	11	image	image	NOUN
fcis-14690	185	12	sentiment	sentiment	NOUN
fcis-14690	185	13	analysis	analysis	NOUN
fcis-14690	185	14	.	.	PUNCT
fcis-14690	186	1	ar	ar	NOUN
fcis-14690	186	2	method	method	NOUN
fcis-14690	186	3	[	[	X
fcis-14690	186	4	32	32	NUM
fcis-14690	186	5	]	]	PUNCT
fcis-14690	186	6	and	and	CCONJ
fcis-14690	186	7	rcnngsr	rcnngsr	ADJ
fcis-14690	186	8	method	method	NOUN
fcis-14690	186	9	[	[	X
fcis-14690	186	10	33	33	NUM
fcis-14690	186	11	]	]	PUNCT
fcis-14690	186	12	automatically	automatically	ADV
fcis-14690	186	13	detect	detect	VERB
fcis-14690	186	14	the	the	DET
fcis-14690	186	15	local	local	ADJ
fcis-14690	186	16	emotion	emotion	NOUN
fcis-14690	186	17	region	region	NOUN
fcis-14690	186	18	,	,	PUNCT
fcis-14690	186	19	and	and	CCONJ
fcis-14690	186	20	then	then	ADV
fcis-14690	186	21	combine	combine	VERB
fcis-14690	186	22	the	the	DET
fcis-14690	186	23	features	feature	NOUN
fcis-14690	186	24	of	of	ADP
fcis-14690	186	25	the	the	DET
fcis-14690	186	26	local	local	ADJ
fcis-14690	186	27	emotion	emotion	NOUN
fcis-14690	186	28	region	region	NOUN
fcis-14690	186	29	with	with	ADP
fcis-14690	186	30	the	the	DET
fcis-14690	186	31	overall	overall	ADJ
fcis-14690	186	32	image	image	NOUN
fcis-14690	186	33	features	feature	VERB
fcis-14690	186	34	to	to	PART
fcis-14690	186	35	produce	produce	VERB
fcis-14690	186	36	the	the	DET
fcis-14690	186	37	final	final	ADJ
fcis-14690	186	38	emotion	emotion	NOUN
fcis-14690	186	39	prediction	prediction	NOUN
fcis-14690	186	40	.	.	PUNCT
fcis-14690	187	1	different	different	ADJ
fcis-14690	187	2	from	from	ADP
fcis-14690	187	3	the	the	DET
fcis-14690	187	4	previous	previous	ADJ
fcis-14690	187	5	methods	method	NOUN
fcis-14690	187	6	,	,	PUNCT
fcis-14690	187	7	zhang	zhang	PROPN
fcis-14690	187	8	et	et	PROPN
fcis-14690	187	9	al	al	PROPN
fcis-14690	187	10	.	.	PUNCT
fcis-14690	188	1	[	[	X
fcis-14690	188	2	28	28	NUM
fcis-14690	188	3	]	]	X
fcis-14690	188	4	used	use	VERB
fcis-14690	188	5	a	a	DET
fcis-14690	188	6	multi	multi	ADJ
fcis-14690	188	7	-	-	ADJ
fcis-14690	188	8	level	level	ADJ
fcis-14690	188	9	mixed	mixed	ADJ
fcis-14690	188	10	model	model	NOUN
fcis-14690	188	11	to	to	PART
fcis-14690	188	12	achieve	achieve	VERB
fcis-14690	188	13	better	well	ADJ
fcis-14690	188	14	effect	effect	NOUN
fcis-14690	188	15	in	in	ADP
fcis-14690	188	16	emotion	emotion	NOUN
fcis-14690	188	17	classification	classification	NOUN
fcis-14690	188	18	by	by	ADP
fcis-14690	188	19	combining	combine	VERB
fcis-14690	188	20	features	feature	NOUN
fcis-14690	188	21	of	of	ADP
fcis-14690	188	22	different	different	ADJ
fcis-14690	188	23	levels	level	NOUN
fcis-14690	188	24	from	from	ADP
fcis-14690	188	25	low	low	ADJ
fcis-14690	188	26	to	to	ADP
fcis-14690	188	27	high	high	ADJ
fcis-14690	188	28	.	.	PUNCT
fcis-14690	189	1	in	in	ADP
fcis-14690	189	2	this	this	DET
fcis-14690	189	3	paper	paper	NOUN
fcis-14690	189	4	,	,	PUNCT
fcis-14690	189	5	we	we	PRON
fcis-14690	189	6	make	make	VERB
fcis-14690	189	7	full	full	ADJ
fcis-14690	189	8	use	use	NOUN
fcis-14690	189	9	of	of	ADP
fcis-14690	189	10	the	the	DET
fcis-14690	189	11	multi	multi	ADJ
fcis-14690	189	12	-	-	ADJ
fcis-14690	189	13	level	level	ADJ
fcis-14690	189	14	features	feature	NOUN
fcis-14690	189	15	of	of	ADP
fcis-14690	189	16	convolutional	convolutional	ADJ
fcis-14690	189	17	neural	neural	ADJ
fcis-14690	189	18	network	network	NOUN
fcis-14690	189	19	to	to	PART
fcis-14690	189	20	extract	extract	VERB
fcis-14690	189	21	features	feature	NOUN
fcis-14690	189	22	at	at	ADP
fcis-14690	189	23	different	different	ADJ
fcis-14690	189	24	levels	level	NOUN
fcis-14690	189	25	from	from	ADP
fcis-14690	189	26	low	low	ADJ
fcis-14690	189	27	to	to	ADP
fcis-14690	189	28	high	high	ADJ
fcis-14690	189	29	,	,	PUNCT
fcis-14690	189	30	then	then	ADV
fcis-14690	189	31	use	use	VERB
fcis-14690	189	32	bilstm	bilstm	NOUN
fcis-14690	189	33	to	to	PART
fcis-14690	189	34	establish	establish	VERB
fcis-14690	189	35	the	the	DET
fcis-14690	189	36	correlation	correlation	NOUN
fcis-14690	189	37	between	between	ADP
fcis-14690	189	38	features	feature	NOUN
fcis-14690	189	39	at	at	ADP
fcis-14690	189	40	different	different	ADJ
fcis-14690	189	41	levels	level	NOUN
fcis-14690	189	42	,	,	PUNCT
fcis-14690	189	43	and	and	CCONJ
fcis-14690	189	44	use	use	VERB
fcis-14690	189	45	cbam	cbam	NOUN
fcis-14690	189	46	module	module	NOUN
fcis-14690	189	47	to	to	PART
fcis-14690	189	48	extract	extract	VERB
fcis-14690	189	49	local	local	ADJ
fcis-14690	189	50	features	feature	NOUN
fcis-14690	189	51	related	relate	VERB
fcis-14690	189	52	to	to	ADP
fcis-14690	189	53	image	image	NOUN
fcis-14690	189	54	and	and	CCONJ
fcis-14690	189	55	emotion	emotion	NOUN
fcis-14690	189	56	.	.	PUNCT
fcis-14690	190	1	the	the	DET
fcis-14690	190	2	accuracy	accuracy	NOUN
fcis-14690	190	3	of	of	ADP
fcis-14690	190	4	the	the	DET
fcis-14690	190	5	proposed	propose	VERB
fcis-14690	190	6	method	method	NOUN
fcis-14690	190	7	on	on	ADP
fcis-14690	190	8	artphoto	artphoto	NOUN
fcis-14690	190	9	and	and	CCONJ
fcis-14690	190	10	abstract	abstract	ADJ
fcis-14690	190	11	data	data	NOUN
fcis-14690	190	12	sets	set	NOUN
fcis-14690	190	13	is	be	AUX
fcis-14690	190	14	improved	improve	VERB
fcis-14690	190	15	compared	compare	VERB
fcis-14690	190	16	with	with	ADP
fcis-14690	190	17	previous	previous	ADJ
fcis-14690	190	18	methods	method	NOUN
fcis-14690	190	19	,	,	PUNCT
fcis-14690	190	20	which	which	PRON
fcis-14690	190	21	proves	prove	VERB
fcis-14690	190	22	that	that	SCONJ
fcis-14690	190	23	the	the	DET
fcis-14690	190	24	proposed	propose	VERB
fcis-14690	190	25	method	method	NOUN
fcis-14690	190	26	is	be	AUX
fcis-14690	190	27	effective	effective	ADJ
fcis-14690	190	28	in	in	ADP
fcis-14690	190	29	image	image	NOUN
fcis-14690	190	30	sentiment	sentiment	NOUN
fcis-14690	190	31	analysis	analysis	NOUN
fcis-14690	190	32	.	.	PUNCT
fcis-14690	191	1	4.6	4.6	NUM
fcis-14690	191	2	.	.	PUNCT
fcis-14690	191	3	ablation	ablation	NOUN
fcis-14690	191	4	experiment	experiment	NOUN
fcis-14690	191	5	in	in	ADP
fcis-14690	191	6	order	order	NOUN
fcis-14690	191	7	to	to	PART
fcis-14690	191	8	verify	verify	VERB
fcis-14690	191	9	the	the	DET
fcis-14690	191	10	effectiveness	effectiveness	NOUN
fcis-14690	191	11	of	of	ADP
fcis-14690	191	12	the	the	DET
fcis-14690	191	13	mixed	mixed	ADJ
fcis-14690	191	14	attention	attention	NOUN
fcis-14690	191	15	mechanism	mechanism	NOUN
fcis-14690	191	16	and	and	CCONJ
fcis-14690	191	17	bilstm	bilstm	NOUN
fcis-14690	191	18	module	module	NOUN
fcis-14690	191	19	in	in	ADP
fcis-14690	191	20	the	the	DET
fcis-14690	191	21	maml	maml	PROPN
fcis-14690	191	22	model	model	NOUN
fcis-14690	191	23	,	,	PUNCT
fcis-14690	191	24	an	an	DET
fcis-14690	191	25	ablation	ablation	NOUN
fcis-14690	191	26	experiment	experiment	NOUN
fcis-14690	191	27	was	be	AUX
fcis-14690	191	28	conducted	conduct	VERB
fcis-14690	191	29	on	on	ADP
fcis-14690	191	30	the	the	DET
fcis-14690	191	31	abstract	abstract	ADJ
fcis-14690	191	32	dataset	dataset	NOUN
fcis-14690	191	33	,	,	PUNCT
fcis-14690	191	34	and	and	CCONJ
fcis-14690	191	35	the	the	DET
fcis-14690	191	36	experimental	experimental	ADJ
fcis-14690	191	37	results	result	NOUN
fcis-14690	191	38	are	be	AUX
fcis-14690	191	39	shown	show	VERB
fcis-14690	191	40	in	in	ADP
fcis-14690	191	41	table	table	NOUN
fcis-14690	191	42	4.it	4.it	NUM
fcis-14690	191	43	can	can	AUX
fcis-14690	191	44	be	be	AUX
fcis-14690	191	45	seen	see	VERB
fcis-14690	191	46	from	from	ADP
fcis-14690	191	47	table	table	NOUN
fcis-14690	191	48	4	4	NUM
fcis-14690	191	49	that	that	SCONJ
fcis-14690	191	50	the	the	DET
fcis-14690	191	51	performance	performance	NOUN
fcis-14690	191	52	of	of	ADP
fcis-14690	191	53	the	the	DET
fcis-14690	191	54	model	model	NOUN
fcis-14690	191	55	is	be	AUX
fcis-14690	191	56	improved	improve	VERB
fcis-14690	191	57	after	after	SCONJ
fcis-14690	191	58	the	the	DET
fcis-14690	191	59	cbam	cbam	NOUN
fcis-14690	191	60	and	and	CCONJ
fcis-14690	191	61	bilstm	bilstm	NOUN
fcis-14690	191	62	modules	module	NOUN
fcis-14690	191	63	are	be	AUX
fcis-14690	191	64	added	add	VERB
fcis-14690	191	65	to	to	PART
fcis-14690	191	66	resnet18	resnet18	VERB
fcis-14690	191	67	respectively	respectively	ADV
fcis-14690	191	68	,	,	PUNCT
fcis-14690	191	69	and	and	CCONJ
fcis-14690	191	70	the	the	DET
fcis-14690	191	71	performance	performance	NOUN
fcis-14690	191	72	is	be	AUX
fcis-14690	191	73	the	the	DET
fcis-14690	191	74	best	good	ADJ
fcis-14690	191	75	after	after	SCONJ
fcis-14690	191	76	the	the	DET
fcis-14690	191	77	two	two	NUM
fcis-14690	191	78	modules	module	NOUN
fcis-14690	191	79	are	be	AUX
fcis-14690	191	80	added	add	VERB
fcis-14690	191	81	at	at	ADP
fcis-14690	191	82	the	the	DET
fcis-14690	191	83	same	same	ADJ
fcis-14690	191	84	time	time	NOUN
fcis-14690	191	85	,	,	PUNCT
fcis-14690	191	86	thus	thus	ADV
fcis-14690	191	87	verifying	verify	VERB
fcis-14690	191	88	the	the	DET
fcis-14690	191	89	effectiveness	effectiveness	NOUN
fcis-14690	191	90	of	of	ADP
fcis-14690	191	91	the	the	DET
fcis-14690	191	92	proposed	propose	VERB
fcis-14690	191	93	method	method	NOUN
fcis-14690	191	94	in	in	ADP
fcis-14690	191	95	this	this	DET
fcis-14690	191	96	paper	paper	NOUN
fcis-14690	191	97	.	.	PUNCT
fcis-14690	192	1	table	table	NOUN
fcis-14690	192	2	3	3	NUM
fcis-14690	192	3	.	.	PUNCT
fcis-14690	193	1	results	result	NOUN
fcis-14690	193	2	of	of	ADP
fcis-14690	193	3	ablation	ablation	NOUN
fcis-14690	193	4	experiments	experiment	NOUN
fcis-14690	193	5	on	on	ADP
fcis-14690	193	6	an	an	DET
fcis-14690	193	7	abstract	abstract	ADJ
fcis-14690	193	8	dataset	dataset	NOUN
fcis-14690	193	9	method	method	NOUN
fcis-14690	193	10	accuracy	accuracy	NOUN
fcis-14690	193	11	resnet18	resnet18	NOUN
fcis-14690	193	12	73.47	73.47	NUM
fcis-14690	193	13	cbam	cbam	NOUN
fcis-14690	193	14	78.57	78.57	NUM
fcis-14690	193	15	bilstm	bilstm	NOUN
fcis-14690	193	16	80.63	80.63	NUM
fcis-14690	193	17	cbam+bilstm	cbam+bilstm	PROPN
fcis-14690	193	18	82.14	82.14	NUM
fcis-14690	193	19	5	5	NUM
fcis-14690	193	20	.	.	PUNCT
fcis-14690	194	1	summary	summary	NOUN
fcis-14690	194	2	in	in	ADP
fcis-14690	194	3	this	this	DET
fcis-14690	194	4	paper	paper	NOUN
fcis-14690	194	5	,	,	PUNCT
fcis-14690	194	6	we	we	PRON
fcis-14690	194	7	propose	propose	VERB
fcis-14690	194	8	an	an	DET
fcis-14690	194	9	image	image	NOUN
fcis-14690	194	10	sentiment	sentiment	NOUN
fcis-14690	194	11	analysis	analysis	NOUN
fcis-14690	194	12	model	model	NOUN
fcis-14690	194	13	,	,	PUNCT
fcis-14690	194	14	maml	maml	PROPN
fcis-14690	194	15	.	.	PUNCT
fcis-14690	195	1	it	it	PRON
fcis-14690	195	2	uses	use	VERB
fcis-14690	195	3	multiple	multiple	ADJ
fcis-14690	195	4	branches	branch	NOUN
fcis-14690	195	5	in	in	ADP
fcis-14690	195	6	resnet18	resnet18	NOUN
fcis-14690	195	7	to	to	PART
fcis-14690	195	8	extract	extract	VERB
fcis-14690	195	9	multi	multi	ADJ
fcis-14690	195	10	-	-	ADJ
fcis-14690	195	11	level	level	ADJ
fcis-14690	195	12	features	feature	NOUN
fcis-14690	195	13	,	,	PUNCT
fcis-14690	195	14	and	and	CCONJ
fcis-14690	195	15	uses	use	VERB
fcis-14690	195	16	the	the	DET
fcis-14690	195	17	dependency	dependency	NOUN
fcis-14690	195	18	relationship	relationship	NOUN
fcis-14690	195	19	between	between	ADP
fcis-14690	195	20	features	feature	NOUN
fcis-14690	195	21	at	at	ADP
fcis-14690	195	22	different	different	ADJ
fcis-14690	195	23	levels	level	NOUN
fcis-14690	195	24	through	through	ADP
fcis-14690	195	25	bilstm	bilstm	NOUN
fcis-14690	195	26	method	method	NOUN
fcis-14690	195	27	to	to	PART
fcis-14690	195	28	effectively	effectively	ADV
fcis-14690	195	29	integrate	integrate	VERB
fcis-14690	195	30	these	these	DET
fcis-14690	195	31	features	feature	NOUN
fcis-14690	195	32	,	,	PUNCT
fcis-14690	195	33	and	and	CCONJ
fcis-14690	195	34	focuses	focus	VERB
fcis-14690	195	35	on	on	ADP
fcis-14690	195	36	the	the	DET
fcis-14690	195	37	main	main	ADJ
fcis-14690	195	38	features	feature	NOUN
fcis-14690	195	39	in	in	ADP
fcis-14690	195	40	the	the	DET
fcis-14690	195	41	emotion	emotion	NOUN
fcis-14690	195	42	region	region	NOUN
fcis-14690	195	43	through	through	ADP
fcis-14690	195	44	the	the	DET
fcis-14690	195	45	mixed	mixed	ADJ
fcis-14690	195	46	attention	attention	NOUN
fcis-14690	195	47	mechanism	mechanism	NOUN
fcis-14690	195	48	of	of	ADP
fcis-14690	195	49	space	space	NOUN
fcis-14690	195	50	and	and	CCONJ
fcis-14690	195	51	channel	channel	NOUN
fcis-14690	195	52	,	,	PUNCT
fcis-14690	195	53	so	so	SCONJ
fcis-14690	195	54	as	as	SCONJ
fcis-14690	195	55	to	to	PART
fcis-14690	195	56	make	make	VERB
fcis-14690	195	57	the	the	DET
fcis-14690	195	58	final	final	ADJ
fcis-14690	195	59	emotion	emotion	NOUN
fcis-14690	195	60	features	feature	VERB
fcis-14690	195	61	more	more	ADJ
fcis-14690	195	62	discriminative	discriminative	NOUN
fcis-14690	195	63	.	.	PUNCT
fcis-14690	196	1	this	this	DET
fcis-14690	196	2	method	method	NOUN
fcis-14690	196	3	shows	show	VERB
fcis-14690	196	4	excellent	excellent	ADJ
fcis-14690	196	5	performance	performance	NOUN
fcis-14690	196	6	on	on	ADP
fcis-14690	196	7	both	both	CCONJ
fcis-14690	196	8	artphoto	artphoto	NOUN
fcis-14690	196	9	and	and	CCONJ
fcis-14690	196	10	abstract	abstract	ADJ
fcis-14690	196	11	datasets	dataset	NOUN
fcis-14690	196	12	.	.	PUNCT
fcis-14690	197	1	in	in	ADP
fcis-14690	197	2	the	the	DET
fcis-14690	197	3	future	future	NOUN
fcis-14690	197	4	,	,	PUNCT
fcis-14690	197	5	we	we	PRON
fcis-14690	197	6	will	will	AUX
fcis-14690	197	7	consider	consider	VERB
fcis-14690	197	8	designing	design	VERB
fcis-14690	197	9	a	a	DET
fcis-14690	197	10	more	more	ADV
fcis-14690	197	11	reasonable	reasonable	ADJ
fcis-14690	197	12	feature	feature	NOUN
fcis-14690	197	13	extraction	extraction	NOUN
fcis-14690	197	14	network	network	NOUN
fcis-14690	197	15	to	to	PART
fcis-14690	197	16	solve	solve	VERB
fcis-14690	197	17	the	the	DET
fcis-14690	197	18	problem	problem	NOUN
fcis-14690	197	19	of	of	ADP
fcis-14690	197	20	sample	sample	NOUN
fcis-14690	197	21	imbalance	imbalance	NOUN
fcis-14690	197	22	and	and	CCONJ
fcis-14690	197	23	the	the	DET
fcis-14690	197	24	global	global	ADJ
fcis-14690	197	25	and	and	CCONJ
fcis-14690	197	26	local	local	ADJ
fcis-14690	197	27	emotional	emotional	ADJ
fcis-14690	197	28	features	feature	NOUN
fcis-14690	197	29	in	in	ADP
fcis-14690	197	30	images	image	NOUN
fcis-14690	197	31	can	can	AUX
fcis-14690	197	32	be	be	AUX
fcis-14690	197	33	mined	mine	VERB
fcis-14690	197	34	more	more	ADV
fcis-14690	197	35	accurately	accurately	ADV
fcis-14690	197	36	through	through	ADP
fcis-14690	197	37	deep	deep	ADJ
fcis-14690	197	38	learning	learning	NOUN
fcis-14690	197	39	methods	method	NOUN
fcis-14690	197	40	,	,	PUNCT
fcis-14690	197	41	so	so	SCONJ
fcis-14690	197	42	as	as	SCONJ
fcis-14690	197	43	to	to	PART
fcis-14690	197	44	further	far	ADV
fcis-14690	197	45	improve	improve	VERB
fcis-14690	197	46	the	the	DET
fcis-14690	197	47	effect	effect	NOUN
fcis-14690	197	48	of	of	ADP
fcis-14690	197	49	image	image	NOUN
fcis-14690	197	50	emotion	emotion	NOUN
fcis-14690	197	51	analysis	analysis	NOUN
fcis-14690	197	52	.	.	PUNCT
fcis-14690	198	1	acknowledgments	acknowledgment	NOUN
fcis-14690	198	2	this	this	DET
fcis-14690	198	3	work	work	NOUN
fcis-14690	198	4	was	be	AUX
fcis-14690	198	5	financially	financially	ADV
fcis-14690	198	6	supported	support	VERB
fcis-14690	198	7	by	by	ADP
fcis-14690	198	8	tianjin	tianjin	PROPN
fcis-14690	198	9	natural	natural	PROPN
fcis-14690	198	10	science	science	PROPN
fcis-14690	198	11	foundation	foundation	PROPN
fcis-14690	198	12	of	of	ADP
fcis-14690	198	13	china	china	PROPN
fcis-14690	198	14	(	(	PUNCT
fcis-14690	198	15	19jcybjc18700	19jcybjc18700	NUM
fcis-14690	198	16	)	)	PUNCT
fcis-14690	198	17	.	.	PUNCT
fcis-14690	199	1	references	reference	NOUN
fcis-14690	199	2	[	[	X
fcis-14690	199	3	1	1	NUM
fcis-14690	199	4	]	]	X
fcis-14690	199	5	machajdik	machajdik	PROPN
fcis-14690	199	6	j	j	PROPN
fcis-14690	199	7	,	,	PUNCT
fcis-14690	199	8	hanbury	hanbury	PROPN
fcis-14690	199	9	a.	a.	NOUN
fcis-14690	199	10	affective	affective	PROPN
fcis-14690	199	11	image	image	NOUN
fcis-14690	199	12	classification	classification	NOUN
fcis-14690	199	13	using	use	VERB
fcis-14690	199	14	features	feature	NOUN
fcis-14690	199	15	inspired	inspire	VERB
fcis-14690	199	16	by	by	ADP
fcis-14690	199	17	psychology	psychology	NOUN
fcis-14690	199	18	and	and	CCONJ
fcis-14690	199	19	art	art	NOUN
fcis-14690	199	20	theory[c]//proceedings	theory[c]//proceeding	NOUN
fcis-14690	199	21	of	of	ADP
fcis-14690	199	22	the	the	DET
fcis-14690	199	23	18th	18th	ADJ
fcis-14690	199	24	acm	acm	PROPN
fcis-14690	199	25	international	international	ADJ
fcis-14690	199	26	conference	conference	NOUN
fcis-14690	199	27	on	on	ADP
fcis-14690	199	28	multimedia	multimedia	NOUN
fcis-14690	199	29	.	.	PUNCT
fcis-14690	200	1	2010	2010	NUM
fcis-14690	200	2	:	:	PUNCT
fcis-14690	200	3	83	83	NUM
fcis-14690	200	4	-	-	SYM
fcis-14690	200	5	92	92	NUM
fcis-14690	200	6	.	.	PUNCT
fcis-14690	201	1	[	[	X
fcis-14690	201	2	2	2	X
fcis-14690	201	3	]	]	X
fcis-14690	201	4	yanulevskaya	yanulevskaya	PROPN
fcis-14690	201	5	v	v	NOUN
fcis-14690	201	6	,	,	PUNCT
fcis-14690	201	7	van	van	PROPN
fcis-14690	201	8	gemert	gemert	PROPN
fcis-14690	201	9	j	j	PROPN
fcis-14690	201	10	c	c	PROPN
fcis-14690	201	11	,	,	PUNCT
fcis-14690	201	12	rothk	rothk	NOUN
fcis-14690	201	13	,	,	PUNCT
fcis-14690	201	14	et	et	PROPN
fcis-14690	201	15	al	al	PROPN
fcis-14690	201	16	.	.	PUNCT
fcis-14690	202	1	emotional	emotional	ADJ
fcis-14690	202	2	valence	valence	NOUN
fcis-14690	202	3	categorization	categorization	NOUN
fcis-14690	202	4	using	use	VERB
fcis-14690	202	5	holistic	holistic	ADJ
fcis-14690	202	6	image	image	NOUN
fcis-14690	202	7	features[c]//2008	features[c]//2008	PROPN
fcis-14690	202	8	15th	15th	ADJ
fcis-14690	202	9	ieee	ieee	PROPN
fcis-14690	202	10	international	international	ADJ
fcis-14690	202	11	conference	conference	NOUN
fcis-14690	202	12	on	on	ADP
fcis-14690	202	13	image	image	NOUN
fcis-14690	202	14	processing	processing	NOUN
fcis-14690	202	15	.	.	PUNCT
fcis-14690	203	1	ieee	ieee	NOUN
fcis-14690	203	2	,	,	PUNCT
fcis-14690	203	3	2008:101	2008:101	NUM
fcis-14690	203	4	-	-	PUNCT
fcis-14690	203	5	104	104	NUM
fcis-14690	203	6	.	.	PUNCT
fcis-14690	204	1	[	[	X
fcis-14690	204	2	3	3	X
fcis-14690	204	3	]	]	X
fcis-14690	204	4	zhao	zhao	PROPN
fcis-14690	204	5	s	s	PROPN
fcis-14690	204	6	,	,	PUNCT
fcis-14690	204	7	gao	gao	PROPN
fcis-14690	204	8	y	y	PROPN
fcis-14690	204	9	,	,	PUNCT
fcis-14690	204	10	jiang	jiang	PROPN
fcis-14690	204	11	x	x	PROPN
fcis-14690	204	12	,	,	PUNCT
fcis-14690	204	13	et	et	PROPN
fcis-14690	204	14	al	al	PROPN
fcis-14690	204	15	.	.	PUNCT
fcis-14690	204	16	exploring	explore	VERB
fcis-14690	204	17	principles	principle	NOUN
fcis-14690	204	18	-	-	PUNCT
fcis-14690	204	19	of	of	ADP
fcis-14690	204	20	-	-	PUNCT
fcis-14690	204	21	art	art	NOUN
fcis-14690	204	22	features	feature	NOUN
fcis-14690	204	23	for	for	ADP
fcis-14690	204	24	image	image	NOUN
fcis-14690	204	25	emotion	emotion	NOUN
fcis-14690	204	26	recognition[c]//proceedings	recognition[c]//proceeding	NOUN
fcis-14690	204	27	of	of	ADP
fcis-14690	204	28	the	the	DET
fcis-14690	204	29	22nd	22nd	PROPN
fcis-14690	204	30	acm	acm	PROPN
fcis-14690	204	31	international	international	ADJ
fcis-14690	204	32	conference	conference	NOUN
fcis-14690	204	33	on	on	ADP
fcis-14690	204	34	multimedia	multimedia	NOUN
fcis-14690	204	35	.	.	PUNCT
fcis-14690	205	1	2014	2014	NUM
fcis-14690	205	2	:	:	PUNCT
fcis-14690	205	3	4756	4756	NUM
fcis-14690	205	4	.	.	PUNCT
fcis-14690	206	1	[	[	X
fcis-14690	206	2	4	4	NUM
fcis-14690	206	3	]	]	PUNCT
fcis-14690	206	4	xu	xu	PROPN
fcis-14690	206	5	l	l	PROPN
fcis-14690	206	6	,	,	PUNCT
fcis-14690	206	7	wang	wang	PROPN
fcis-14690	206	8	z	z	PROPN
fcis-14690	206	9	,	,	PUNCT
fcis-14690	206	10	wu	wu	PROPN
fcis-14690	206	11	b	b	PROPN
fcis-14690	206	12	,	,	PUNCT
fcis-14690	207	1	et	et	PROPN
fcis-14690	207	2	al	al	PROPN
fcis-14690	207	3	.	.	PROPN
fcis-14690	207	4	mdan	mdan	PROPN
fcis-14690	207	5	:	:	PUNCT
fcis-14690	207	6	multi	multi	ADJ
fcis-14690	207	7	-	-	ADJ
fcis-14690	207	8	level	level	ADJ
fcis-14690	207	9	dependent	dependent	ADJ
fcis-14690	207	10	attention	attention	NOUN
fcis-14690	207	11	network	network	NOUN
fcis-14690	207	12	for	for	ADP
fcis-14690	207	13	visual	visual	ADJ
fcis-14690	207	14	emotion	emotion	NOUN
fcis-14690	207	15	analysis[c]//proceedings	analysis[c]//proceedings	PROPN
fcis-14690	207	16	of	of	ADP
fcis-14690	207	17	the	the	DET
fcis-14690	207	18	ieee	ieee	NOUN
fcis-14690	207	19	/	/	SYM
fcis-14690	207	20	cvf	cvf	NOUN
fcis-14690	207	21	conference	conference	NOUN
fcis-14690	207	22	on	on	ADP
fcis-14690	207	23	computer	computer	NOUN
fcis-14690	207	24	vision	vision	NOUN
fcis-14690	207	25	and	and	CCONJ
fcis-14690	207	26	pattern	pattern	NOUN
fcis-14690	207	27	recognition	recognition	NOUN
fcis-14690	207	28	.	.	PUNCT
fcis-14690	208	1	2022	2022	NUM
fcis-14690	208	2	:	:	PUNCT
fcis-14690	208	3	9479	9479	NUM
fcis-14690	208	4	-	-	SYM
fcis-14690	208	5	9488	9488	NUM
fcis-14690	208	6	.	.	PUNCT
fcis-14690	209	1	[	[	X
fcis-14690	209	2	5	5	NUM
fcis-14690	209	3	]	]	PUNCT
fcis-14690	209	4	rao	rao	PROPN
fcis-14690	209	5	t	t	PROPN
fcis-14690	209	6	,	,	PUNCT
fcis-14690	209	7	li	li	PROPN
fcis-14690	209	8	x	x	PROPN
fcis-14690	209	9	,	,	PUNCT
fcis-14690	209	10	zhang	zhang	PROPN
fcis-14690	209	11	h	h	PROPN
fcis-14690	209	12	,	,	PUNCT
fcis-14690	209	13	et	et	PROPN
fcis-14690	209	14	al	al	PROPN
fcis-14690	209	15	.	.	PUNCT
fcis-14690	210	1	multi	multi	ADJ
fcis-14690	210	2	-	-	ADJ
fcis-14690	210	3	level	level	ADJ
fcis-14690	210	4	region	region	NOUN
fcis-14690	210	5	-	-	PUNCT
fcis-14690	210	6	based	base	VERB
fcis-14690	210	7	convolutional	convolutional	ADJ
fcis-14690	210	8	neural	neural	ADJ
fcis-14690	210	9	network	network	NOUN
fcis-14690	210	10	for	for	ADP
fcis-14690	210	11	image	image	NOUN
fcis-14690	210	12	emotion	emotion	NOUN
fcis-14690	210	13	classification	classification	NOUN
fcis-14690	210	14	[	[	X
fcis-14690	210	15	j	j	X
fcis-14690	210	16	]	]	X
fcis-14690	210	17	.	.	PUNCT
fcis-14690	211	1	neurocomputing	neurocomputing	NOUN
fcis-14690	211	2	,	,	PUNCT
fcis-14690	211	3	2019	2019	NUM
fcis-14690	211	4	,	,	PUNCT
fcis-14690	211	5	333	333	NUM
fcis-14690	211	6	:	:	PUNCT
fcis-14690	211	7	429	429	NUM
fcis-14690	211	8	-	-	SYM
fcis-14690	211	9	439	439	NUM
fcis-14690	211	10	.	.	PUNCT
fcis-14690	212	1	[	[	X
fcis-14690	212	2	6	6	NUM
fcis-14690	212	3	]	]	SYM
fcis-14690	212	4	li	li	PROPN
fcis-14690	212	5	b	b	PROPN
fcis-14690	212	6	,	,	PUNCT
fcis-14690	212	7	ren	ren	PROPN
fcis-14690	212	8	h	h	NOUN
fcis-14690	212	9	,	,	PUNCT
fcis-14690	212	10	jiang	jiang	PROPN
fcis-14690	212	11	x	x	PROPN
fcis-14690	212	12	,	,	PUNCT
fcis-14690	212	13	et	et	PROPN
fcis-14690	212	14	al	al	PROPN
fcis-14690	212	15	.	.	PROPN
fcis-14690	213	1	scep	scep	PROPN
fcis-14690	213	2	—	—	PUNCT
fcis-14690	213	3	a	a	DET
fcis-14690	213	4	new	new	ADJ
fcis-14690	213	5	image	image	NOUN
fcis-14690	213	6	dimensional	dimensional	ADJ
fcis-14690	213	7	emotion	emotion	NOUN
fcis-14690	213	8	recognition	recognition	NOUN
fcis-14690	213	9	model	model	NOUN
fcis-14690	213	10	based	base	VERB
fcis-14690	213	11	on	on	ADP
fcis-14690	213	12	spatial	spatial	ADJ
fcis-14690	213	13	and	and	CCONJ
fcis-14690	213	14	channel	channel	NOUN
fcis-14690	213	15	-	-	PUNCT
fcis-14690	213	16	wise	wise	ADJ
fcis-14690	213	17	attention	attention	NOUN
fcis-14690	213	18	mechanisms[j	mechanisms[j	NOUN
fcis-14690	213	19	]	]	PUNCT
fcis-14690	213	20	.	.	PUNCT
fcis-14690	214	1	ieee	ieee	PROPN
fcis-14690	214	2	access,2021	access,2021	PROPN
fcis-14690	214	3	,	,	PUNCT
fcis-14690	214	4	9	9	NUM
fcis-14690	214	5	:	:	SYM
fcis-14690	214	6	25278	25278	NUM
fcis-14690	214	7	-	-	SYM
fcis-14690	214	8	25290	25290	NUM
fcis-14690	214	9	.	.	PUNCT
fcis-14690	215	1	[	[	X
fcis-14690	215	2	7	7	X
fcis-14690	215	3	]	]	X
fcis-14690	215	4	peng	peng	PROPN
fcis-14690	215	5	k	k	PROPN
fcis-14690	215	6	c	c	PROPN
fcis-14690	215	7	,	,	PUNCT
fcis-14690	215	8	chen	chen	PROPN
fcis-14690	215	9	t	t	PROPN
fcis-14690	215	10	,	,	PUNCT
fcis-14690	215	11	sadovnik	sadovnik	X
fcis-14690	215	12	a	a	X
fcis-14690	215	13	,	,	PUNCT
fcis-14690	215	14	et	et	PROPN
fcis-14690	215	15	al	al	PROPN
fcis-14690	215	16	.	.	PROPN
fcis-14690	215	17	amixed	amixe	VERB
fcis-14690	215	18	bag	bag	NOUN
fcis-14690	215	19	of	of	ADP
fcis-14690	215	20	emotions	emotion	NOUN
fcis-14690	215	21	:	:	PUNCT
fcis-14690	215	22	model	model	NOUN
fcis-14690	215	23	,	,	PUNCT
fcis-14690	215	24	predict	predict	VERB
fcis-14690	215	25	,	,	PUNCT
fcis-14690	215	26	and	and	CCONJ
fcis-14690	215	27	transfer	transfer	VERB
fcis-14690	215	28	emotion	emotion	NOUN
fcis-14690	215	29	distributions	distribution	NOUN
fcis-14690	215	30	65	65	NUM
fcis-14690	216	1	[	[	PUNCT
fcis-14690	216	2	c]//	c]//	ADJ
fcis-14690	216	3	proceedings	proceeding	NOUN
fcis-14690	216	4	of	of	ADP
fcis-14690	216	5	the	the	DET
fcis-14690	216	6	ieee	ieee	NOUN
fcis-14690	216	7	conference	conference	NOUN
fcis-14690	216	8	on	on	ADP
fcis-14690	216	9	computer	computer	NOUN
fcis-14690	216	10	vision	vision	NOUN
fcis-14690	216	11	and	and	CCONJ
fcis-14690	216	12	pattern	pattern	NOUN
fcis-14690	216	13	recognition	recognition	NOUN
fcis-14690	216	14	.	.	PUNCT
fcis-14690	217	1	2015	2015	NUM
fcis-14690	217	2	:	:	PUNCT
fcis-14690	217	3	860	860	NUM
fcis-14690	217	4	-	-	SYM
fcis-14690	217	5	868	868	NUM
fcis-14690	217	6	.	.	PUNCT
fcis-14690	218	1	[	[	X
fcis-14690	218	2	8	8	NUM
fcis-14690	218	3	]	]	X
fcis-14690	218	4	compton	compton	PROPN
fcis-14690	218	5	r	r	PROPN
fcis-14690	218	6	j.	j.	PROPN
fcis-14690	218	7	the	the	DET
fcis-14690	218	8	interface	interface	NOUN
fcis-14690	218	9	between	between	ADP
fcis-14690	218	10	emotion	emotion	NOUN
fcis-14690	218	11	and	and	CCONJ
fcis-14690	218	12	attention	attention	NOUN
fcis-14690	218	13	:	:	PUNCT
fcis-14690	218	14	a	a	DET
fcis-14690	218	15	review	review	NOUN
fcis-14690	218	16	of	of	ADP
fcis-14690	218	17	evidence	evidence	NOUN
fcis-14690	218	18	from	from	ADP
fcis-14690	218	19	psychology	psychology	NOUN
fcis-14690	218	20	and	and	CCONJ
fcis-14690	218	21	neuroscience[j	neuroscience[j	NOUN
fcis-14690	218	22	]	]	PUNCT
fcis-14690	218	23	.	.	PUNCT
fcis-14690	219	1	behavioral	behavioral	ADJ
fcis-14690	219	2	and	and	CCONJ
fcis-14690	219	3	cognitive	cognitive	ADJ
fcis-14690	219	4	neuroscience	neuroscience	NOUN
fcis-14690	219	5	reviews	review	NOUN
fcis-14690	219	6	,	,	PUNCT
fcis-14690	219	7	2003	2003	NUM
fcis-14690	219	8	,	,	PUNCT
fcis-14690	219	9	2(2	2(2	NUM
fcis-14690	219	10	):	):	PUNCT
fcis-14690	219	11	115	115	NUM
fcis-14690	219	12	-	-	SYM
fcis-14690	219	13	129	129	NUM
fcis-14690	219	14	.	.	PUNCT
fcis-14690	220	1	[	[	X
fcis-14690	220	2	9	9	NUM
fcis-14690	220	3	]	]	PUNCT
fcis-14690	220	4	you	you	PRON
fcis-14690	220	5	q	q	NOUN
fcis-14690	220	6	,	,	PUNCT
fcis-14690	220	7	jin	jin	PROPN
fcis-14690	220	8	h	h	PROPN
fcis-14690	220	9	,	,	PUNCT
fcis-14690	220	10	luo	luo	PROPN
fcis-14690	220	11	j.	j.	PROPN
fcis-14690	220	12	visual	visual	PROPN
fcis-14690	220	13	sentiment	sentiment	NOUN
fcis-14690	220	14	analysis	analysis	NOUN
fcis-14690	220	15	by	by	ADP
fcis-14690	220	16	attending	attend	VERB
fcis-14690	220	17	on	on	ADP
fcis-14690	220	18	local	local	ADJ
fcis-14690	220	19	image	image	NOUN
fcis-14690	220	20	regions[c]//proceedings	regions[c]//proceeding	NOUN
fcis-14690	220	21	of	of	ADP
fcis-14690	220	22	the	the	DET
fcis-14690	220	23	aaai	aaai	PROPN
fcis-14690	220	24	conference	conference	NOUN
fcis-14690	220	25	on	on	ADP
fcis-14690	220	26	artificial	artificial	ADJ
fcis-14690	220	27	intelligence	intelligence	NOUN
fcis-14690	220	28	.	.	PUNCT
fcis-14690	221	1	2017	2017	NUM
fcis-14690	221	2	,	,	PUNCT
fcis-14690	221	3	31(1	31(1	NUM
fcis-14690	221	4	)	)	PUNCT
fcis-14690	221	5	.	.	PUNCT
fcis-14690	222	1	[	[	X
fcis-14690	222	2	10	10	NUM
fcis-14690	222	3	]	]	X
fcis-14690	222	4	zhao	zhao	PROPN
fcis-14690	222	5	s	s	PROPN
fcis-14690	222	6	,	,	PUNCT
fcis-14690	222	7	yao	yao	PROPN
fcis-14690	222	8	x	x	PRON
fcis-14690	222	9	,	,	PUNCT
fcis-14690	222	10	yang	yang	PROPN
fcis-14690	222	11	j	j	PROPN
fcis-14690	222	12	,	,	PUNCT
fcis-14690	222	13	et	et	PROPN
fcis-14690	222	14	al	al	PROPN
fcis-14690	222	15	.	.	PROPN
fcis-14690	222	16	affective	affective	PROPN
fcis-14690	222	17	image	image	NOUN
fcis-14690	222	18	content	content	NOUN
fcis-14690	222	19	analysis	analysis	NOUN
fcis-14690	222	20	:	:	PUNCT
fcis-14690	222	21	two	two	NUM
fcis-14690	222	22	decades	decade	NOUN
fcis-14690	222	23	review	review	VERB
fcis-14690	222	24	and	and	CCONJ
fcis-14690	222	25	new	new	ADJ
fcis-14690	222	26	perspectives[j	perspectives[j	PROPN
fcis-14690	222	27	]	]	PUNCT
fcis-14690	222	28	.	.	PUNCT
fcis-14690	223	1	ieee	ieee	NOUN
fcis-14690	223	2	transactions	transaction	NOUN
fcis-14690	223	3	on	on	ADP
fcis-14690	223	4	pattern	pattern	NOUN
fcis-14690	223	5	analysis	analysis	NOUN
fcis-14690	223	6	and	and	CCONJ
fcis-14690	223	7	machine	machine	NOUN
fcis-14690	223	8	intelligence	intelligence	NOUN
fcis-14690	223	9	,	,	PUNCT
fcis-14690	223	10	2021	2021	NUM
fcis-14690	223	11	,	,	PUNCT
fcis-14690	223	12	44(10	44(10	NUM
fcis-14690	223	13	):	):	PUNCT
fcis-14690	223	14	6729	6729	NUM
fcis-14690	223	15	-	-	SYM
fcis-14690	223	16	6751	6751	NUM
fcis-14690	223	17	.	.	PUNCT
fcis-14690	224	1	[	[	X
fcis-14690	224	2	11	11	NUM
fcis-14690	224	3	]	]	PUNCT
fcis-14690	224	4	song	song	NOUN
fcis-14690	224	5	k	k	PROPN
fcis-14690	224	6	,	,	PUNCT
fcis-14690	224	7	yao	yao	PROPN
fcis-14690	224	8	t	t	PROPN
fcis-14690	224	9	,	,	PUNCT
fcis-14690	224	10	ling	ling	PROPN
fcis-14690	224	11	q	q	PROPN
fcis-14690	224	12	,	,	PUNCT
fcis-14690	224	13	et	et	PROPN
fcis-14690	224	14	al	al	PROPN
fcis-14690	224	15	.	.	PUNCT
fcis-14690	224	16	boosting	boost	VERB
fcis-14690	224	17	image	image	NOUN
fcis-14690	224	18	sentiment	sentiment	NOUN
fcis-14690	224	19	analysis	analysis	NOUN
fcis-14690	224	20	with	with	ADP
fcis-14690	224	21	visual	visual	ADJ
fcis-14690	224	22	attention[j	attention[j	PROPN
fcis-14690	224	23	]	]	PUNCT
fcis-14690	224	24	.	.	PUNCT
fcis-14690	225	1	neurocomputing,2018	neurocomputing,2018	NOUN
fcis-14690	225	2	,	,	PUNCT
fcis-14690	225	3	312	312	NUM
fcis-14690	225	4	:	:	SYM
fcis-14690	225	5	218	218	NUM
fcis-14690	225	6	-	-	SYM
fcis-14690	225	7	228	228	NUM
fcis-14690	225	8	.	.	PUNCT
fcis-14690	226	1	[	[	X
fcis-14690	226	2	12	12	NUM
fcis-14690	226	3	]	]	X
fcis-14690	226	4	yang	yang	PROPN
fcis-14690	226	5	j	j	PROPN
fcis-14690	226	6	,	,	PUNCT
fcis-14690	226	7	she	she	PRON
fcis-14690	226	8	d	d	VERB
fcis-14690	226	9	,	,	PUNCT
fcis-14690	227	1	lai	lai	PROPN
fcis-14690	227	2	y	y	PROPN
fcis-14690	227	3	k	k	PROPN
fcis-14690	227	4	,	,	PUNCT
fcis-14690	227	5	et	et	PROPN
fcis-14690	227	6	al	al	PROPN
fcis-14690	227	7	.	.	PROPN
fcis-14690	227	8	weakly	weakly	ADV
fcis-14690	227	9	supervised	supervised	ADJ
fcis-14690	227	10	coupled	couple	VERB
fcis-14690	227	11	networks	network	NOUN
fcis-14690	227	12	for	for	ADP
fcis-14690	227	13	visual	visual	ADJ
fcis-14690	227	14	sentiment	sentiment	NOUN
fcis-14690	227	15	analysis[c]//proceedings	analysis[c]//proceedings	PROPN
fcis-14690	227	16	of	of	ADP
fcis-14690	227	17	the	the	DET
fcis-14690	227	18	ieee	ieee	NOUN
fcis-14690	227	19	conference	conference	NOUN
fcis-14690	227	20	on	on	ADP
fcis-14690	227	21	computer	computer	NOUN
fcis-14690	227	22	vision	vision	NOUN
fcis-14690	227	23	and	and	CCONJ
fcis-14690	227	24	pattern	pattern	NOUN
fcis-14690	227	25	recognition	recognition	NOUN
fcis-14690	227	26	.	.	PUNCT
fcis-14690	228	1	2018	2018	NUM
fcis-14690	228	2	:	:	PUNCT
fcis-14690	228	3	7584	7584	NUM
fcis-14690	228	4	-	-	SYM
fcis-14690	228	5	7592	7592	NUM
fcis-14690	228	6	.	.	PUNCT
fcis-14690	229	1	[	[	X
fcis-14690	229	2	13	13	NUM
fcis-14690	229	3	]	]	X
fcis-14690	229	4	chen	chen	PROPN
fcis-14690	229	5	l	l	PROPN
fcis-14690	229	6	,	,	PUNCT
fcis-14690	229	7	zhang	zhang	PROPN
fcis-14690	229	8	h	h	PROPN
fcis-14690	229	9	,	,	PUNCT
fcis-14690	229	10	xiao	xiao	PROPN
fcis-14690	229	11	j	j	PROPN
fcis-14690	229	12	,	,	PUNCT
fcis-14690	229	13	et	et	PROPN
fcis-14690	229	14	al	al	PROPN
fcis-14690	229	15	.	.	PUNCT
fcis-14690	229	16	sca	sca	PROPN
fcis-14690	229	17	-	-	PUNCT
fcis-14690	229	18	cnn	cnn	PROPN
fcis-14690	229	19	:	:	PUNCT
fcis-14690	229	20	spatial	spatial	ADJ
fcis-14690	229	21	and	and	CCONJ
fcis-14690	229	22	channelwise	channelwise	VERB
fcis-14690	229	23	attention	attention	NOUN
fcis-14690	229	24	in	in	ADP
fcis-14690	229	25	convolutional	convolutional	ADJ
fcis-14690	229	26	networks	network	NOUN
fcis-14690	229	27	for	for	ADP
fcis-14690	229	28	image	image	NOUN
fcis-14690	229	29	captioning	captioning	NOUN
fcis-14690	229	30	[	[	X
fcis-14690	229	31	c]//proceedings	c]//proceeding	NOUN
fcis-14690	229	32	of	of	ADP
fcis-14690	229	33	the	the	DET
fcis-14690	229	34	ieee	ieee	NOUN
fcis-14690	229	35	conference	conference	NOUN
fcis-14690	229	36	on	on	ADP
fcis-14690	229	37	computer	computer	NOUN
fcis-14690	229	38	vision	vision	NOUN
fcis-14690	229	39	and	and	CCONJ
fcis-14690	229	40	pattern	pattern	NOUN
fcis-14690	229	41	recognition	recognition	NOUN
fcis-14690	229	42	.	.	PUNCT
fcis-14690	230	1	2017	2017	NUM
fcis-14690	230	2	:	:	PUNCT
fcis-14690	230	3	5659	5659	NUM
fcis-14690	230	4	-	-	SYM
fcis-14690	230	5	5667	5667	NUM
fcis-14690	230	6	.	.	PUNCT
fcis-14690	231	1	[	[	X
fcis-14690	231	2	14	14	NUM
fcis-14690	231	3	]	]	X
fcis-14690	231	4	zhu	zhu	PROPN
fcis-14690	231	5	x	x	SYM
fcis-14690	231	6	,	,	PUNCT
fcis-14690	231	7	li	li	PROPN
fcis-14690	231	8	l	l	PROPN
fcis-14690	231	9	,	,	PUNCT
fcis-14690	231	10	zhang	zhang	PROPN
fcis-14690	231	11	w	w	PROPN
fcis-14690	231	12	,	,	PUNCT
fcis-14690	231	13	et	et	PROPN
fcis-14690	231	14	al	al	PROPN
fcis-14690	231	15	.	.	PUNCT
fcis-14690	231	16	dependency	dependency	NOUN
fcis-14690	231	17	exploitation	exploitation	NOUN
fcis-14690	231	18	:	:	PUNCT
fcis-14690	231	19	a	a	DET
fcis-14690	231	20	unified	unified	ADJ
fcis-14690	231	21	cnn	cnn	PROPN
fcis-14690	231	22	-	-	PUNCT
fcis-14690	231	23	rnn	rnn	PROPN
fcis-14690	231	24	approach	approach	NOUN
fcis-14690	231	25	for	for	ADP
fcis-14690	231	26	visual	visual	ADJ
fcis-14690	231	27	emotion	emotion	NOUN
fcis-14690	231	28	recognition	recognition	NOUN
fcis-14690	231	29	[	[	X
fcis-14690	231	30	c]//	c]//	PROPN
fcis-14690	231	31	ijcai	ijcai	PROPN
fcis-14690	231	32	.	.	PUNCT
fcis-14690	232	1	2017	2017	NUM
fcis-14690	232	2	:	:	PUNCT
fcis-14690	232	3	3595	3595	NUM
fcis-14690	232	4	-	-	SYM
fcis-14690	232	5	3601	3601	NUM
fcis-14690	232	6	.	.	PUNCT
fcis-14690	233	1	[	[	X
fcis-14690	233	2	15	15	NUM
fcis-14690	233	3	]	]	X
fcis-14690	233	4	siersdorfer	siersdorfer	NOUN
fcis-14690	233	5	s	s	NOUN
fcis-14690	233	6	,	,	PUNCT
fcis-14690	233	7	minack	minack	ADJ
fcis-14690	233	8	e	e	NOUN
fcis-14690	233	9	,	,	PUNCT
fcis-14690	233	10	deng	deng	PROPN
fcis-14690	233	11	f	f	PROPN
fcis-14690	233	12	,	,	PUNCT
fcis-14690	233	13	et	et	PROPN
fcis-14690	233	14	al	al	PROPN
fcis-14690	233	15	.	.	PROPN
fcis-14690	233	16	analyzing	analyze	VERB
fcis-14690	233	17	and	and	CCONJ
fcis-14690	233	18	predicting	predict	VERB
fcis-14690	233	19	sentiment	sentiment	NOUN
fcis-14690	233	20	of	of	ADP
fcis-14690	233	21	images	image	NOUN
fcis-14690	233	22	on	on	ADP
fcis-14690	233	23	the	the	DET
fcis-14690	233	24	social	social	ADJ
fcis-14690	233	25	web[c]//	web[c]//	PROPN
fcis-14690	233	26	proceedings	proceeding	NOUN
fcis-14690	233	27	of	of	ADP
fcis-14690	233	28	the	the	DET
fcis-14690	233	29	18th	18th	ADJ
fcis-14690	233	30	acm	acm	PROPN
fcis-14690	233	31	international	international	ADJ
fcis-14690	233	32	conference	conference	NOUN
fcis-14690	233	33	on	on	ADP
fcis-14690	233	34	multimedia	multimedia	NOUN
fcis-14690	233	35	.	.	PUNCT
fcis-14690	234	1	2010	2010	NUM
fcis-14690	234	2	:	:	PUNCT
fcis-14690	235	1	71	71	NUM
fcis-14690	235	2	5	5	NUM
fcis-14690	235	3	-	-	SYM
fcis-14690	235	4	718	718	NUM
fcis-14690	235	5	.	.	PUNCT
fcis-14690	236	1	[	[	X
fcis-14690	236	2	16	16	NUM
fcis-14690	236	3	]	]	X
fcis-14690	236	4	deng	deng	PROPN
fcis-14690	236	5	j	j	PROPN
fcis-14690	236	6	,	,	PUNCT
fcis-14690	236	7	dong	dong	PROPN
fcis-14690	236	8	w	w	PROPN
fcis-14690	236	9	,	,	PUNCT
fcis-14690	236	10	socher	socher	NOUN
fcis-14690	236	11	r	r	NOUN
fcis-14690	236	12	,	,	PUNCT
fcis-14690	236	13	et	et	PROPN
fcis-14690	236	14	al	al	PROPN
fcis-14690	236	15	.	.	PROPN
fcis-14690	236	16	imagenet	imagenet	PROPN
fcis-14690	236	17	:	:	PUNCT
fcis-14690	236	18	a	a	DET
fcis-14690	236	19	large	large	ADJ
fcis-14690	236	20	-	-	PUNCT
fcis-14690	236	21	scale	scale	NOUN
fcis-14690	236	22	hierarchical	hierarchical	ADJ
fcis-14690	236	23	image	image	NOUN
fcis-14690	236	24	database[c]//2009	database[c]//2009	PROPN
fcis-14690	236	25	ieee	ieee	NOUN
fcis-14690	236	26	conference	conference	NOUN
fcis-14690	236	27	on	on	ADP
fcis-14690	236	28	computer	computer	NOUN
fcis-14690	236	29	vision	vision	NOUN
fcis-14690	236	30	and	and	CCONJ
fcis-14690	236	31	pattern	pattern	NOUN
fcis-14690	236	32	recognition	recognition	NOUN
fcis-14690	236	33	.	.	PUNCT
fcis-14690	237	1	ieee	ieee	PROPN
fcis-14690	237	2	,	,	PUNCT
fcis-14690	237	3	2009	2009	NUM
fcis-14690	237	4	:	:	PUNCT
fcis-14690	237	5	248	248	NUM
fcis-14690	237	6	-	-	SYM
fcis-14690	237	7	255	255	NUM
fcis-14690	237	8	.	.	PUNCT
fcis-14690	238	1	[	[	X
fcis-14690	238	2	17	17	NUM
fcis-14690	238	3	]	]	X
fcis-14690	238	4	you	you	PRON
fcis-14690	238	5	q	q	PROPN
fcis-14690	238	6	,	,	PUNCT
fcis-14690	238	7	luo	luo	PROPN
fcis-14690	238	8	j	j	PROPN
fcis-14690	238	9	,	,	PUNCT
fcis-14690	238	10	jin	jin	PROPN
fcis-14690	238	11	h	h	PROPN
fcis-14690	238	12	,	,	PUNCT
fcis-14690	238	13	et	et	PROPN
fcis-14690	238	14	al	al	PROPN
fcis-14690	238	15	.	.	PUNCT
fcis-14690	238	16	building	build	VERB
fcis-14690	238	17	a	a	DET
fcis-14690	238	18	large	large	ADJ
fcis-14690	238	19	scale	scale	NOUN
fcis-14690	238	20	dataset	dataset	NOUN
fcis-14690	238	21	for	for	ADP
fcis-14690	238	22	image	image	NOUN
fcis-14690	238	23	emotion	emotion	NOUN
fcis-14690	238	24	recognition	recognition	NOUN
fcis-14690	238	25	:	:	PUNCT
fcis-14690	238	26	the	the	DET
fcis-14690	238	27	fine	fine	ADJ
fcis-14690	238	28	print	print	NOUN
fcis-14690	238	29	and	and	CCONJ
fcis-14690	238	30	the	the	DET
fcis-14690	238	31	benchmark	benchmark	NOUN
fcis-14690	238	32	[	[	X
fcis-14690	238	33	c]//proceedings	c]//proceeding	NOUN
fcis-14690	238	34	of	of	ADP
fcis-14690	238	35	the	the	DET
fcis-14690	238	36	aaai	aaai	PROPN
fcis-14690	238	37	conference	conference	NOUN
fcis-14690	238	38	on	on	ADP
fcis-14690	238	39	artificial	artificial	ADJ
fcis-14690	238	40	intelligence	intelligence	NOUN
fcis-14690	238	41	.	.	PUNCT
fcis-14690	239	1	2016	2016	NUM
fcis-14690	239	2	,	,	PUNCT
fcis-14690	239	3	30(1	30(1	NUM
fcis-14690	239	4	)	)	PUNCT
fcis-14690	239	5	.	.	PUNCT
fcis-14690	240	1	[	[	X
fcis-14690	240	2	18	18	NUM
fcis-14690	240	3	]	]	PUNCT
fcis-14690	240	4	krizhevsky	krizhevsky	NOUN
fcis-14690	240	5	a	a	PROPN
fcis-14690	240	6	,	,	PUNCT
fcis-14690	240	7	sutskever	sutskever	VERB
fcis-14690	240	8	i	i	PROPN
fcis-14690	240	9	,	,	PUNCT
fcis-14690	240	10	hinton	hinton	PROPN
fcis-14690	240	11	ge	ge	PROPN
fcis-14690	240	12	.	.	PUNCT
fcis-14690	241	1	imagenet	imagenet	PROPN
fcis-14690	241	2	classification	classification	NOUN
fcis-14690	241	3	with	with	ADP
fcis-14690	241	4	deep	deep	ADJ
fcis-14690	241	5	convolutional	convolutional	ADJ
fcis-14690	241	6	neural	neural	ADJ
fcis-14690	241	7	networks[j	networks[j	NOUN
fcis-14690	241	8	]	]	X
fcis-14690	241	9	.	.	PUNCT
fcis-14690	242	1	advances	advance	NOUN
fcis-14690	242	2	in	in	ADP
fcis-14690	242	3	neural	neural	ADJ
fcis-14690	242	4	information	information	NOUN
fcis-14690	242	5	processing	processing	NOUN
fcis-14690	242	6	systems	system	NOUN
fcis-14690	242	7	,	,	PUNCT
fcis-14690	242	8	2012	2012	NUM
fcis-14690	242	9	,	,	PUNCT
fcis-14690	242	10	25	25	NUM
fcis-14690	242	11	.	.	PUNCT
fcis-14690	243	1	[	[	X
fcis-14690	243	2	19	19	NUM
fcis-14690	243	3	]	]	PUNCT
fcis-14690	243	4	rao	rao	PROPN
fcis-14690	243	5	t	t	PROPN
fcis-14690	243	6	,	,	PUNCT
fcis-14690	243	7	li	li	PROPN
fcis-14690	243	8	x	x	PROPN
fcis-14690	243	9	,	,	PUNCT
fcis-14690	243	10	xu	xu	PROPN
fcis-14690	243	11	m.	m.	NOUN
fcis-14690	243	12	learning	learn	VERB
fcis-14690	243	13	multi	multi	ADJ
fcis-14690	243	14	-	-	ADJ
fcis-14690	243	15	level	level	ADJ
fcis-14690	243	16	deep	deep	ADJ
fcis-14690	243	17	representations	representation	NOUN
fcis-14690	243	18	for	for	ADP
fcis-14690	243	19	image	image	NOUN
fcis-14690	243	20	emotion	emotion	NOUN
fcis-14690	243	21	classification[j].neural	classification[j].neural	ADJ
fcis-14690	243	22	processing	processing	NOUN
fcis-14690	243	23	letters	letter	NOUN
fcis-14690	243	24	,	,	PUNCT
fcis-14690	243	25	2020	2020	NUM
fcis-14690	243	26	,	,	PUNCT
fcis-14690	243	27	51	51	NUM
fcis-14690	243	28	:	:	SYM
fcis-14690	243	29	2043	2043	NUM
fcis-14690	243	30	-	-	SYM
fcis-14690	243	31	2061	2061	NUM
fcis-14690	243	32	.	.	PUNCT
fcis-14690	244	1	[	[	X
fcis-14690	244	2	20	20	NUM
fcis-14690	244	3	]	]	X
fcis-14690	244	4	sowmyayani	sowmyayani	PROPN
fcis-14690	244	5	s	s	NOUN
fcis-14690	244	6	,	,	PUNCT
fcis-14690	244	7	rani	rani	PROPN
fcis-14690	244	8	p	p	PROPN
fcis-14690	244	9	a	a	DET
fcis-14690	244	10	j.	j.	PROPN
fcis-14690	244	11	salient	salient	PROPN
fcis-14690	244	12	object	object	PROPN
fcis-14690	244	13	based	base	VERB
fcis-14690	244	14	visual	visual	ADJ
fcis-14690	244	15	sentiment	sentiment	NOUN
fcis-14690	244	16	analysis	analysis	NOUN
fcis-14690	244	17	by	by	ADP
fcis-14690	244	18	combining	combine	VERB
fcis-14690	244	19	deep	deep	ADJ
fcis-14690	244	20	features	feature	NOUN
fcis-14690	244	21	and	and	CCONJ
fcis-14690	244	22	handcrafted	handcrafted	ADJ
fcis-14690	244	23	features[j	features[j	NOUN
fcis-14690	244	24	]	]	PUNCT
fcis-14690	244	25	.	.	PUNCT
fcis-14690	245	1	multimedia	multimedia	NOUN
fcis-14690	245	2	tools	tool	NOUN
fcis-14690	245	3	and	and	CCONJ
fcis-14690	245	4	applications	application	NOUN
fcis-14690	245	5	,	,	PUNCT
fcis-14690	245	6	2022	2022	NUM
fcis-14690	245	7	,	,	PUNCT
fcis-14690	245	8	81(6	81(6	NUM
fcis-14690	245	9	):	):	PUNCT
fcis-14690	245	10	7941	7941	NUM
fcis-14690	245	11	-	-	SYM
fcis-14690	245	12	7955	7955	NUM
fcis-14690	245	13	.	.	PUNCT
fcis-14690	246	1	[	[	X
fcis-14690	246	2	21	21	NUM
fcis-14690	246	3	]	]	X
fcis-14690	246	4	li	li	PROPN
fcis-14690	246	5	w	w	PROPN
fcis-14690	246	6	,	,	PUNCT
fcis-14690	246	7	liu	liu	PROPN
fcis-14690	246	8	k	k	PROPN
fcis-14690	246	9	,	,	PUNCT
fcis-14690	246	10	zhang	zhang	PROPN
fcis-14690	246	11	l	l	PROPN
fcis-14690	246	12	,	,	PUNCT
fcis-14690	246	13	et	et	PROPN
fcis-14690	246	14	al	al	PROPN
fcis-14690	246	15	.	.	PROPN
fcis-14690	246	16	object	object	PROPN
fcis-14690	246	17	detection	detection	NOUN
fcis-14690	246	18	based	base	VERB
fcis-14690	246	19	on	on	ADP
fcis-14690	246	20	an	an	DET
fcis-14690	246	21	adaptive	adaptive	ADJ
fcis-14690	246	22	attention	attention	NOUN
fcis-14690	246	23	mechanism[j].scientific	mechanism[j].scientific	NOUN
fcis-14690	246	24	reports,2020	reports,2020	NOUN
fcis-14690	246	25	,	,	PUNCT
fcis-14690	246	26	10	10	NUM
fcis-14690	246	27	(	(	PUNCT
fcis-14690	246	28	1	1	NUM
fcis-14690	246	29	):	):	NUM
fcis-14690	246	30	11307	11307	NUM
fcis-14690	246	31	.	.	PUNCT
fcis-14690	247	1	[	[	X
fcis-14690	247	2	22	22	NUM
fcis-14690	247	3	]	]	SYM
fcis-14690	247	4	hu	hu	PROPN
fcis-14690	248	1	j	j	PROPN
fcis-14690	248	2	,	,	PUNCT
fcis-14690	248	3	shen	shen	PROPN
fcis-14690	248	4	l	l	PROPN
fcis-14690	248	5	,	,	PUNCT
fcis-14690	248	6	sun	sun	PROPN
fcis-14690	248	7	g.	g.	PROPN
fcis-14690	248	8	squeeze	squeeze	PROPN
fcis-14690	248	9	-	-	PUNCT
fcis-14690	248	10	and	and	CCONJ
fcis-14690	248	11	-	-	PUNCT
fcis-14690	248	12	excitation	excitation	NOUN
fcis-14690	248	13	networks	network	NOUN
fcis-14690	248	14	[	[	PUNCT
fcis-14690	248	15	c]//	c]//	ADJ
fcis-14690	248	16	proceedings	proceeding	NOUN
fcis-14690	248	17	of	of	ADP
fcis-14690	248	18	the	the	DET
fcis-14690	248	19	ieee	ieee	NOUN
fcis-14690	248	20	conference	conference	NOUN
fcis-14690	248	21	on	on	ADP
fcis-14690	248	22	computer	computer	NOUN
fcis-14690	248	23	vision	vision	NOUN
fcis-14690	248	24	and	and	CCONJ
fcis-14690	248	25	pattern	pattern	NOUN
fcis-14690	248	26	recognition	recognition	NOUN
fcis-14690	248	27	.	.	PUNCT
fcis-14690	249	1	2018	2018	NUM
fcis-14690	249	2	:	:	PUNCT
fcis-14690	249	3	7132	7132	NUM
fcis-14690	249	4	-	-	SYM
fcis-14690	249	5	7141	7141	NUM
fcis-14690	249	6	.	.	PUNCT
fcis-14690	250	1	[	[	X
fcis-14690	250	2	23	23	NUM
fcis-14690	250	3	]	]	X
fcis-14690	250	4	chen	chen	PROPN
fcis-14690	250	5	l	l	PROPN
fcis-14690	250	6	,	,	PUNCT
fcis-14690	250	7	zhang	zhang	PROPN
fcis-14690	250	8	h	h	PROPN
fcis-14690	250	9	,	,	PUNCT
fcis-14690	250	10	xiao	xiao	PROPN
fcis-14690	250	11	j	j	PROPN
fcis-14690	250	12	,	,	PUNCT
fcis-14690	250	13	et	et	PROPN
fcis-14690	250	14	al	al	PROPN
fcis-14690	250	15	.	.	PUNCT
fcis-14690	250	16	sca	sca	PROPN
fcis-14690	250	17	-	-	PUNCT
fcis-14690	250	18	cnn	cnn	PROPN
fcis-14690	250	19	:	:	PUNCT
fcis-14690	250	20	spatial	spatial	ADJ
fcis-14690	250	21	and	and	CCONJ
fcis-14690	250	22	channelwise	channelwise	VERB
fcis-14690	250	23	attention	attention	NOUN
fcis-14690	250	24	in	in	ADP
fcis-14690	250	25	convolutional	convolutional	ADJ
fcis-14690	250	26	networks	network	NOUN
fcis-14690	250	27	for	for	ADP
fcis-14690	250	28	image	image	NOUN
fcis-14690	250	29	captioning	captioning	NOUN
fcis-14690	250	30	[	[	X
fcis-14690	250	31	c]//proceedings	c]//proceeding	NOUN
fcis-14690	250	32	of	of	ADP
fcis-14690	250	33	the	the	DET
fcis-14690	250	34	ieee	ieee	NOUN
fcis-14690	250	35	conference	conference	NOUN
fcis-14690	250	36	on	on	ADP
fcis-14690	250	37	computer	computer	NOUN
fcis-14690	250	38	vision	vision	NOUN
fcis-14690	250	39	and	and	CCONJ
fcis-14690	250	40	pattern	pattern	NOUN
fcis-14690	250	41	recognition	recognition	NOUN
fcis-14690	250	42	.	.	PUNCT
fcis-14690	251	1	2017	2017	NUM
fcis-14690	251	2	:	:	PUNCT
fcis-14690	251	3	5659	5659	NUM
fcis-14690	251	4	-	-	SYM
fcis-14690	251	5	5667	5667	NUM
fcis-14690	251	6	.	.	PUNCT
fcis-14690	252	1	[	[	X
fcis-14690	252	2	24	24	NUM
fcis-14690	252	3	]	]	PUNCT
fcis-14690	252	4	song	song	NOUN
fcis-14690	252	5	k	k	PROPN
fcis-14690	252	6	,	,	PUNCT
fcis-14690	252	7	yao	yao	PROPN
fcis-14690	252	8	t	t	PROPN
fcis-14690	252	9	,	,	PUNCT
fcis-14690	252	10	ling	ling	PROPN
fcis-14690	252	11	q	q	PROPN
fcis-14690	252	12	,	,	PUNCT
fcis-14690	252	13	et	et	PROPN
fcis-14690	252	14	al	al	PROPN
fcis-14690	252	15	.	.	PUNCT
fcis-14690	252	16	boosting	boost	VERB
fcis-14690	252	17	image	image	NOUN
fcis-14690	252	18	sentiment	sentiment	NOUN
fcis-14690	252	19	analysis	analysis	NOUN
fcis-14690	252	20	with	with	ADP
fcis-14690	252	21	visual	visual	ADJ
fcis-14690	252	22	attention[j	attention[j	PROPN
fcis-14690	252	23	]	]	PUNCT
fcis-14690	252	24	.	.	PUNCT
fcis-14690	253	1	neurocomputing,2018	neurocomputing,2018	NOUN
fcis-14690	253	2	,	,	PUNCT
fcis-14690	253	3	312	312	NUM
fcis-14690	253	4	:	:	SYM
fcis-14690	253	5	218	218	NUM
fcis-14690	253	6	-	-	SYM
fcis-14690	253	7	228	228	NUM
fcis-14690	253	8	.	.	PUNCT
fcis-14690	254	1	[	[	X
fcis-14690	254	2	25	25	NUM
fcis-14690	254	3	]	]	X
fcis-14690	254	4	yang	yang	PROPN
fcis-14690	254	5	j	j	PROPN
fcis-14690	254	6	,	,	PUNCT
fcis-14690	254	7	she	she	PRON
fcis-14690	254	8	d	d	VERB
fcis-14690	254	9	,	,	PUNCT
fcis-14690	255	1	lai	lai	PROPN
fcis-14690	255	2	y	y	PROPN
fcis-14690	255	3	k	k	PROPN
fcis-14690	255	4	,	,	PUNCT
fcis-14690	255	5	et	et	PROPN
fcis-14690	255	6	al	al	PROPN
fcis-14690	255	7	.	.	PROPN
fcis-14690	255	8	weakly	weakly	ADV
fcis-14690	255	9	supervised	supervised	ADJ
fcis-14690	255	10	coupled	couple	VERB
fcis-14690	255	11	networks	network	NOUN
fcis-14690	255	12	for	for	ADP
fcis-14690	255	13	visual	visual	ADJ
fcis-14690	255	14	sentiment	sentiment	NOUN
fcis-14690	255	15	analysis[c]//proceedings	analysis[c]//proceedings	PROPN
fcis-14690	255	16	of	of	ADP
fcis-14690	255	17	the	the	DET
fcis-14690	255	18	ieee	ieee	NOUN
fcis-14690	255	19	conference	conference	NOUN
fcis-14690	255	20	on	on	ADP
fcis-14690	255	21	computer	computer	NOUN
fcis-14690	255	22	vision	vision	NOUN
fcis-14690	255	23	and	and	CCONJ
fcis-14690	255	24	pattern	pattern	NOUN
fcis-14690	255	25	recognition	recognition	NOUN
fcis-14690	255	26	.	.	PUNCT
fcis-14690	256	1	2018	2018	NUM
fcis-14690	256	2	:	:	PUNCT
fcis-14690	256	3	7584	7584	NUM
fcis-14690	256	4	-	-	SYM
fcis-14690	256	5	7592	7592	NUM
fcis-14690	256	6	.	.	PUNCT
fcis-14690	257	1	[	[	X
fcis-14690	257	2	26	26	NUM
fcis-14690	257	3	]	]	X
fcis-14690	257	4	zhao	zhao	PROPN
fcis-14690	257	5	s	s	PROPN
fcis-14690	257	6	,	,	PUNCT
fcis-14690	257	7	jia	jia	PROPN
fcis-14690	257	8	z	z	PROPN
fcis-14690	257	9	,	,	PUNCT
fcis-14690	257	10	chen	chen	PROPN
fcis-14690	257	11	h	h	PROPN
fcis-14690	257	12	,	,	PUNCT
fcis-14690	258	1	et	et	PROPN
fcis-14690	258	2	al	al	PROPN
fcis-14690	258	3	.	.	PUNCT
fcis-14690	258	4	pdanet	pdanet	PROPN
fcis-14690	258	5	:	:	PUNCT
fcis-14690	258	6	polarity	polarity	NOUN
fcis-14690	258	7	-	-	PUNCT
fcis-14690	258	8	consistent	consistent	ADJ
fcis-14690	258	9	deep	deep	ADJ
fcis-14690	258	10	attention	attention	NOUN
fcis-14690	258	11	network	network	NOUN
fcis-14690	258	12	for	for	ADP
fcis-14690	258	13	fine	fine	ADV
fcis-14690	258	14	-	-	PUNCT
fcis-14690	258	15	grained	grain	VERB
fcis-14690	258	16	visual	visual	ADJ
fcis-14690	258	17	emotion	emotion	NOUN
fcis-14690	258	18	regression	regression	NOUN
fcis-14690	259	1	[	[	X
fcis-14690	259	2	c]//	c]//	ADJ
fcis-14690	259	3	proceedings	proceeding	NOUN
fcis-14690	259	4	of	of	ADP
fcis-14690	259	5	the	the	DET
fcis-14690	259	6	27th	27th	ADJ
fcis-14690	259	7	acm	acm	PROPN
fcis-14690	259	8	international	international	PROPN
fcis-14690	259	9	conference	conference	NOUN
fcis-14690	259	10	onmultimedia	onmultimedia	NOUN
fcis-14690	259	11	.	.	PUNCT
fcis-14690	260	1	2019	2019	NUM
fcis-14690	260	2	:	:	PUNCT
fcis-14690	260	3	192	192	NUM
fcis-14690	260	4	-	-	SYM
fcis-14690	260	5	201	201	NUM
fcis-14690	260	6	.	.	PUNCT
fcis-14690	260	7	　 	　 	SPACE
fcis-14690	261	1	[	[	X
fcis-14690	261	2	27	27	NUM
fcis-14690	261	3	]	]	X
fcis-14690	261	4	woo	woo	PROPN
fcis-14690	261	5	s	s	PROPN
fcis-14690	261	6	,	,	PUNCT
fcis-14690	261	7	park	park	PROPN
fcis-14690	261	8	j	j	PROPN
fcis-14690	261	9	,	,	PUNCT
fcis-14690	261	10	lee	lee	PROPN
fcis-14690	261	11	j	j	PROPN
fcis-14690	261	12	y	y	PROPN
fcis-14690	261	13	,	,	PUNCT
fcis-14690	261	14	et	et	PROPN
fcis-14690	261	15	al	al	PROPN
fcis-14690	261	16	.	.	PUNCT
fcis-14690	261	17	cbam	cbam	PROPN
fcis-14690	261	18	:	:	PUNCT
fcis-14690	261	19	convolutionalblock	convolutionalblock	VERB
fcis-14690	261	20	attention	attention	NOUN
fcis-14690	261	21	module[c]//proceedings	module[c]//proceeding	NOUN
fcis-14690	261	22	of	of	ADP
fcis-14690	261	23	the	the	DET
fcis-14690	261	24	european	european	PROPN
fcis-14690	261	25	conference	conference	PROPN
fcis-14690	261	26	on	on	ADP
fcis-14690	261	27	computer	computer	NOUN
fcis-14690	261	28	vision	vision	NOUN
fcis-14690	261	29	(	(	PUNCT
fcis-14690	261	30	eccv	eccv	ADV
fcis-14690	261	31	)	)	PUNCT
fcis-14690	261	32	.	.	PUNCT
fcis-14690	262	1	2018	2018	NUM
fcis-14690	262	2	:	:	PUNCT
fcis-14690	262	3	3	3	NUM
fcis-14690	262	4	-	-	SYM
fcis-14690	262	5	19	19	NUM
fcis-14690	262	6	.	.	PUNCT
fcis-14690	263	1	[	[	X
fcis-14690	263	2	28	28	NUM
fcis-14690	263	3	]	]	X
fcis-14690	263	4	zhang	zhang	PROPN
fcis-14690	263	5	h	h	PROPN
fcis-14690	263	6	,	,	PUNCT
fcis-14690	263	7	xu	xu	PROPN
fcis-14690	264	1	d	d	PROPN
fcis-14690	264	2	,	,	PUNCT
fcis-14690	264	3	luo	luo	PROPN
fcis-14690	264	4	g	g	PROPN
fcis-14690	264	5	,	,	PUNCT
fcis-14690	264	6	et	et	PROPN
fcis-14690	264	7	al	al	PROPN
fcis-14690	264	8	.	.	PUNCT
fcis-14690	264	9	learning	learn	VERB
fcis-14690	264	10	multi	multi	ADJ
fcis-14690	264	11	-	-	ADJ
fcis-14690	264	12	level	level	ADJ
fcis-14690	264	13	representations	representation	NOUN
fcis-14690	264	14	for	for	ADP
fcis-14690	264	15	affective	affective	ADJ
fcis-14690	264	16	image	image	NOUN
fcis-14690	264	17	recognition[j	recognition[j	NOUN
fcis-14690	264	18	]	]	PUNCT
fcis-14690	264	19	.	.	PUNCT
fcis-14690	265	1	neural	neural	ADJ
fcis-14690	265	2	computing	computing	NOUN
fcis-14690	265	3	and	and	CCONJ
fcis-14690	265	4	applications	application	NOUN
fcis-14690	265	5	,	,	PUNCT
fcis-14690	265	6	2022	2022	NUM
fcis-14690	265	7	,	,	PUNCT
fcis-14690	265	8	34(16	34(16	NUM
fcis-14690	265	9	):	):	PUNCT
fcis-14690	265	10	14107	14107	NUM
fcis-14690	265	11	-	-	SYM
fcis-14690	265	12	14120	14120	NUM
fcis-14690	265	13	.	.	PUNCT
fcis-14690	266	1	[	[	X
fcis-14690	266	2	29	29	NUM
fcis-14690	266	3	]	]	X
fcis-14690	266	4	borth	borth	PROPN
fcis-14690	266	5	d	d	PROPN
fcis-14690	266	6	,	,	PUNCT
fcis-14690	266	7	ji	ji	PROPN
fcis-14690	266	8	r	r	PROPN
fcis-14690	266	9	,	,	PUNCT
fcis-14690	266	10	chen	chen	PROPN
fcis-14690	266	11	t	t	PROPN
fcis-14690	266	12	,	,	PUNCT
fcis-14690	266	13	et	et	PROPN
fcis-14690	266	14	al	al	PROPN
fcis-14690	266	15	.	.	PUNCT
fcis-14690	267	1	large	large	ADJ
fcis-14690	267	2	-	-	PUNCT
fcis-14690	267	3	scale	scale	NOUN
fcis-14690	267	4	visual	visual	ADJ
fcis-14690	267	5	sentiment	sentiment	NOUN
fcis-14690	267	6	ontology	ontology	NOUN
fcis-14690	267	7	and	and	CCONJ
fcis-14690	267	8	detectors	detector	NOUN
fcis-14690	267	9	using	use	VERB
fcis-14690	267	10	adjective	adjective	ADJ
fcis-14690	267	11	noun	noun	NOUN
fcis-14690	267	12	pairs[c]//	pairs[c]//	ADJ
fcis-14690	267	13	proceedings	proceeding	NOUN
fcis-14690	267	14	of	of	ADP
fcis-14690	267	15	the	the	DET
fcis-14690	267	16	21st	21st	PROPN
fcis-14690	267	17	acm	acm	PROPN
fcis-14690	267	18	international	international	ADJ
fcis-14690	267	19	conference	conference	NOUN
fcis-14690	267	20	on	on	ADP
fcis-14690	267	21	multimedia	multimedia	NOUN
fcis-14690	267	22	.	.	PUNCT
fcis-14690	268	1	2013	2013	NUM
fcis-14690	268	2	:	:	PUNCT
fcis-14690	268	3	223	223	NUM
fcis-14690	268	4	-	-	SYM
fcis-14690	268	5	232	232	NUM
fcis-14690	268	6	.	.	PUNCT
fcis-14690	269	1	[	[	X
fcis-14690	269	2	30	30	NUM
fcis-14690	269	3	]	]	X
fcis-14690	269	4	rao	rao	PROPN
fcis-14690	269	5	t	t	PROPN
fcis-14690	269	6	,	,	PUNCT
fcis-14690	269	7	xu	xu	PROPN
fcis-14690	269	8	m	m	PROPN
fcis-14690	269	9	,	,	PUNCT
fcis-14690	269	10	liu	liu	PROPN
fcis-14690	269	11	h	h	PROPN
fcis-14690	269	12	,	,	PUNCT
fcis-14690	269	13	et	et	PROPN
fcis-14690	269	14	al	al	PROPN
fcis-14690	269	15	.	.	PUNCT
fcis-14690	270	1	multi	multi	ADJ
fcis-14690	270	2	-	-	ADJ
fcis-14690	270	3	scale	scale	ADJ
fcis-14690	270	4	blocks	block	NOUN
fcis-14690	270	5	based	base	VERB
fcis-14690	270	6	image	image	NOUN
fcis-14690	270	7	emotion	emotion	NOUN
fcis-14690	270	8	classification	classification	NOUN
fcis-14690	270	9	using	use	VERB
fcis-14690	270	10	multiple	multiple	ADJ
fcis-14690	270	11	instance	instance	NOUN
fcis-14690	270	12	learning[c]//	learning[c]//	PROPN
fcis-14690	270	13	2016	2016	NUM
fcis-14690	270	14	ieee	ieee	NOUN
fcis-14690	270	15	international	international	ADJ
fcis-14690	270	16	conference	conference	NOUN
fcis-14690	270	17	on	on	ADP
fcis-14690	270	18	imageprocessing	imageprocesse	VERB
fcis-14690	270	19	(	(	PUNCT
fcis-14690	270	20	icip	icip	PROPN
fcis-14690	270	21	)	)	PUNCT
fcis-14690	270	22	.	.	PUNCT
fcis-14690	271	1	ieee	ieee	PROPN
fcis-14690	271	2	,	,	PUNCT
fcis-14690	271	3	2016	2016	NUM
fcis-14690	271	4	:	:	PUNCT
fcis-14690	271	5	634	634	NUM
fcis-14690	271	6	-	-	NUM
fcis-14690	271	7	638	638	NUM
fcis-14690	271	8	.	.	PUNCT
fcis-14690	272	1	[	[	X
fcis-14690	272	2	31	31	NUM
fcis-14690	272	3	]	]	PUNCT
fcis-14690	272	4	you	you	PRON
fcis-14690	272	5	q	q	PROPN
fcis-14690	272	6	,	,	PUNCT
fcis-14690	272	7	luo	luo	PROPN
fcis-14690	272	8	j	j	PROPN
fcis-14690	272	9	,	,	PUNCT
fcis-14690	272	10	jin	jin	PROPN
fcis-14690	272	11	h	h	PROPN
fcis-14690	272	12	,	,	PUNCT
fcis-14690	272	13	et	et	PROPN
fcis-14690	272	14	al	al	PROPN
fcis-14690	272	15	.	.	PUNCT
fcis-14690	272	16	robust	robust	ADJ
fcis-14690	272	17	image	image	NOUN
fcis-14690	272	18	sentiment	sentiment	NOUN
fcis-14690	272	19	analysis	analysis	NOUN
fcis-14690	272	20	using	use	VERB
fcis-14690	272	21	progressively	progressively	ADV
fcis-14690	272	22	trained	train	VERB
fcis-14690	272	23	and	and	CCONJ
fcis-14690	272	24	domain	domain	NOUN
fcis-14690	272	25	transferred	transfer	VERB
fcis-14690	272	26	deep	deep	ADJ
fcis-14690	272	27	networks	network	NOUN
fcis-14690	272	28	[	[	X
fcis-14690	272	29	c]//proceedings	c]//proceeding	NOUN
fcis-14690	272	30	of	of	ADP
fcis-14690	272	31	the	the	DET
fcis-14690	272	32	aaai	aaai	PROPN
fcis-14690	272	33	conference	conference	NOUN
fcis-14690	272	34	on	on	ADP
fcis-14690	272	35	artificial	artificial	ADJ
fcis-14690	272	36	intelligence	intelligence	NOUN
fcis-14690	272	37	.	.	PUNCT
fcis-14690	273	1	2015	2015	NUM
fcis-14690	273	2	,	,	PUNCT
fcis-14690	273	3	29(1	29(1	NUM
fcis-14690	273	4	)	)	PUNCT
fcis-14690	273	5	.	.	PUNCT
fcis-14690	274	1	[	[	X
fcis-14690	274	2	32	32	NUM
fcis-14690	274	3	]	]	PUNCT
fcis-14690	274	4	yang	yang	PROPN
fcis-14690	274	5	j	j	PROPN
fcis-14690	274	6	,	,	PUNCT
fcis-14690	274	7	she	she	PRON
fcis-14690	274	8	d	d	VERB
fcis-14690	274	9	,	,	PUNCT
fcis-14690	274	10	sun	sun	PROPN
fcis-14690	274	11	m	m	PROPN
fcis-14690	274	12	,	,	PUNCT
fcis-14690	274	13	et	et	PROPN
fcis-14690	274	14	al	al	PROPN
fcis-14690	274	15	.	.	PROPN
fcis-14690	274	16	visual	visual	ADJ
fcis-14690	274	17	sentiment	sentiment	NOUN
fcis-14690	274	18	prediction	prediction	NOUN
fcis-14690	274	19	based	base	VERB
fcis-14690	274	20	on	on	ADP
fcis-14690	274	21	automatic	automatic	ADJ
fcis-14690	274	22	discovery	discovery	NOUN
fcis-14690	274	23	of	of	ADP
fcis-14690	274	24	affective	affective	ADJ
fcis-14690	274	25	regions[j	regions[j	NOUN
fcis-14690	274	26	]	]	PUNCT
fcis-14690	274	27	.	.	PUNCT
fcis-14690	275	1	ieee	ieee	NOUN
fcis-14690	275	2	transactions	transaction	NOUN
fcis-14690	275	3	on	on	ADP
fcis-14690	275	4	multimedia	multimedia	NOUN
fcis-14690	275	5	,	,	PUNCT
fcis-14690	275	6	2018	2018	NUM
fcis-14690	275	7	,	,	PUNCT
fcis-14690	275	8	20(9	20(9	NUM
fcis-14690	275	9	):	):	PUNCT
fcis-14690	275	10	2513	2513	NUM
fcis-14690	275	11	-	-	SYM
fcis-14690	275	12	2525	2525	NUM
fcis-14690	275	13	.	.	PUNCT
fcis-14690	276	1	[	[	X
fcis-14690	276	2	33	33	NUM
fcis-14690	276	3	]	]	PUNCT
fcis-14690	276	4	xiong	xiong	PROPN
fcis-14690	276	5	h	h	PROPN
fcis-14690	276	6	,	,	PUNCT
fcis-14690	276	7	liu	liu	PROPN
fcis-14690	276	8	q	q	PROPN
fcis-14690	276	9	,	,	PUNCT
fcis-14690	276	10	song	song	NOUN
fcis-14690	276	11	s	s	PROPN
fcis-14690	276	12	,	,	PUNCT
fcis-14690	276	13	et	et	PROPN
fcis-14690	276	14	al	al	PROPN
fcis-14690	276	15	.	.	PUNCT
fcis-14690	276	16	region	region	NOUN
fcis-14690	276	17	-	-	PUNCT
fcis-14690	276	18	based	base	VERB
fcis-14690	276	19	convolutional	convolutional	ADJ
fcis-14690	276	20	neural	neural	ADJ
fcis-14690	276	21	network	network	NOUN
fcis-14690	276	22	using	use	VERB
fcis-14690	276	23	group	group	NOUN
fcis-14690	276	24	sparse	sparse	ADJ
fcis-14690	276	25	regularization	regularization	NOUN
fcis-14690	276	26	for	for	ADP
fcis-14690	276	27	image	image	NOUN
fcis-14690	276	28	sentiment	sentiment	NOUN
fcis-14690	276	29	classification[j	classification[j	PROPN
fcis-14690	276	30	]	]	PUNCT
fcis-14690	276	31	.	.	PUNCT
fcis-14690	277	1	eurasip	eurasip	PROPN
fcis-14690	277	2	journal	journal	PROPN
fcis-14690	277	3	on	on	ADP
fcis-14690	277	4	image	image	NOUN
fcis-14690	277	5	and	and	CCONJ
fcis-14690	277	6	video	video	NOUN
fcis-14690	277	7	processing	processing	NOUN
fcis-14690	277	8	,	,	PUNCT
fcis-14690	277	9	2019	2019	NUM
fcis-14690	277	10	,	,	PUNCT
fcis-14690	277	11	2019(1	2019(1	NUM
fcis-14690	277	12	):	):	PUNCT
fcis-14690	277	13	1	1	NUM
fcis-14690	277	14	-	-	SYM
fcis-14690	277	15	9	9	NUM
fcis-14690	277	16	.	.	PUNCT
fcis-14690	278	1	[	[	X
fcis-14690	278	2	34	34	NUM
fcis-14690	278	3	]	]	X
fcis-14690	278	4	machajdik	machajdik	PROPN
fcis-14690	278	5	j	j	PROPN
fcis-14690	278	6	,	,	PUNCT
fcis-14690	278	7	hanbury	hanbury	PROPN
fcis-14690	278	8	a.	a.	NOUN
fcis-14690	278	9	affective	affective	PROPN
fcis-14690	278	10	image	image	NOUN
fcis-14690	278	11	classification	classification	NOUN
fcis-14690	278	12	using	use	VERB
fcis-14690	278	13	features	feature	NOUN
fcis-14690	278	14	inspired	inspire	VERB
fcis-14690	278	15	by	by	ADP
fcis-14690	278	16	psychology	psychology	NOUN
fcis-14690	278	17	and	and	CCONJ
fcis-14690	278	18	art	art	NOUN
fcis-14690	278	19	theory	theory	NOUN
fcis-14690	278	20	.	.	PUNCT
fcis-14690	279	1	proceedings	proceeding	NOUN
fcis-14690	279	2	of	of	ADP
fcis-14690	279	3	the	the	DET
fcis-14690	279	4	18th	18th	ADJ
fcis-14690	279	5	acm	acm	PROPN
fcis-14690	279	6	international	international	ADJ
fcis-14690	279	7	conference	conference	NOUN
fcis-14690	279	8	on	on	ADP
fcis-14690	279	9	multimedia	multimedia	NOUN
fcis-14690	279	10	,	,	PUNCT
fcis-14690	279	11	2010	2010	NUM
fcis-14690	279	12	.	.	PUNCT
fcis-14690	280	1	1	1	NUM
fcis-14690	280	2	,	,	PUNCT
fcis-14690	280	3	5	5	NUM
fcis-14690	280	4	.	.	PUNCT