id	sid	tid	token	lemma	pos
ajst-31431	1	1	academic	academic	ADJ
ajst-31431	1	2	journal	journal	NOUN
ajst-31431	1	3	of	of	ADP
ajst-31431	1	4	science	science	NOUN
ajst-31431	1	5	and	and	CCONJ
ajst-31431	1	6	technology	technology	NOUN
ajst-31431	1	7	issn	issn	NOUN
ajst-31431	1	8	:	:	PUNCT
ajst-31431	1	9	2771	2771	NUM
ajst-31431	1	10	-	-	SYM
ajst-31431	1	11	3032	3032	NUM
ajst-31431	1	12	|	|	NOUN
ajst-31431	1	13	vol	vol	NOUN
ajst-31431	1	14	.	.	PROPN
ajst-31431	2	1	16	16	NUM
ajst-31431	2	2	,	,	PUNCT
ajst-31431	2	3	no	no	INTJ
ajst-31431	2	4	.	.	NOUN
ajst-31431	2	5	1	1	NUM
ajst-31431	2	6	,	,	PUNCT
ajst-31431	2	7	2025	2025	NUM
ajst-31431	2	8	40	40	NUM
ajst-31431	2	9	a	a	DET
ajst-31431	2	10	review	review	NOUN
ajst-31431	2	11	of	of	ADP
ajst-31431	2	12	key	key	ADJ
ajst-31431	2	13	technologies	technology	NOUN
ajst-31431	2	14	for	for	ADP
ajst-31431	2	15	the	the	DET
ajst-31431	2	16	analysis	analysis	NOUN
ajst-31431	2	17	and	and	CCONJ
ajst-31431	2	18	classification	classification	NOUN
ajst-31431	2	19	of	of	ADP
ajst-31431	2	20	social	social	ADJ
ajst-31431	2	21	media	medium	NOUN
ajst-31431	2	22	hate	hate	VERB
ajst-31431	2	23	speech	speech	NOUN
ajst-31431	2	24	content	content	NOUN
ajst-31431	2	25	.	.	PUNCT
ajst-31431	3	1	yingjie	yingjie	PROPN
ajst-31431	3	2	liu	liu	PROPN
ajst-31431	3	3	*	*	PROPN
ajst-31431	3	4	,	,	PUNCT
ajst-31431	3	5	chen	chen	PROPN
ajst-31431	3	6	gong	gong	PROPN
ajst-31431	3	7	and	and	CCONJ
ajst-31431	3	8	xingyue	xingyue	PROPN
ajst-31431	3	9	ma	ma	PROPN
ajst-31431	3	10	school	school	PROPN
ajst-31431	3	11	of	of	ADP
ajst-31431	3	12	information	information	NOUN
ajst-31431	3	13	network	network	NOUN
ajst-31431	3	14	security	security	NOUN
ajst-31431	3	15	people	people	NOUN
ajst-31431	3	16	's	's	PART
ajst-31431	3	17	public	public	ADJ
ajst-31431	3	18	security	security	NOUN
ajst-31431	3	19	university	university	PROPN
ajst-31431	3	20	of	of	ADP
ajst-31431	3	21	china	china	PROPN
ajst-31431	3	22	,	,	PUNCT
ajst-31431	3	23	beijng	beijng	VERB
ajst-31431	3	24	,	,	PUNCT
ajst-31431	3	25	china	china	PROPN
ajst-31431	3	26	*	*	PUNCT
ajst-31431	3	27	corresponding	correspond	VERB
ajst-31431	3	28	author	author	NOUN
ajst-31431	3	29	:	:	PUNCT
ajst-31431	3	30	1697666388@qq.com	1697666388@qq.com	NUM
ajst-31431	3	31	abstract	abstract	NOUN
ajst-31431	3	32	:	:	PUNCT
ajst-31431	3	33	with	with	ADP
ajst-31431	3	34	the	the	DET
ajst-31431	3	35	explosive	explosive	ADJ
ajst-31431	3	36	growth	growth	NOUN
ajst-31431	3	37	of	of	ADP
ajst-31431	3	38	social	social	ADJ
ajst-31431	3	39	media	medium	NOUN
ajst-31431	3	40	content	content	NOUN
ajst-31431	3	41	and	and	CCONJ
ajst-31431	3	42	the	the	DET
ajst-31431	3	43	enhanced	enhanced	ADJ
ajst-31431	3	44	information	information	NOUN
ajst-31431	3	45	-	-	PUNCT
ajst-31431	3	46	sharing	share	VERB
ajst-31431	3	47	capabilities	capability	NOUN
ajst-31431	3	48	of	of	ADP
ajst-31431	3	49	platforms	platform	NOUN
ajst-31431	3	50	,	,	PUNCT
ajst-31431	3	51	the	the	DET
ajst-31431	3	52	proliferation	proliferation	NOUN
ajst-31431	3	53	of	of	ADP
ajst-31431	3	54	online	online	ADJ
ajst-31431	3	55	hate	hate	NOUN
ajst-31431	3	56	speech	speech	NOUN
ajst-31431	3	57	has	have	AUX
ajst-31431	3	58	become	become	VERB
ajst-31431	3	59	a	a	DET
ajst-31431	3	60	global	global	ADJ
ajst-31431	3	61	governance	governance	NOUN
ajst-31431	3	62	challenge	challenge	NOUN
ajst-31431	3	63	.	.	PUNCT
ajst-31431	4	1	its	its	PRON
ajst-31431	4	2	dissemination	dissemination	NOUN
ajst-31431	4	3	patterns	pattern	NOUN
ajst-31431	4	4	are	be	AUX
ajst-31431	4	5	rapidly	rapidly	ADV
ajst-31431	4	6	evolving	evolve	VERB
ajst-31431	4	7	towards	towards	ADP
ajst-31431	4	8	multimodality	multimodality	NOUN
ajst-31431	4	9	(	(	PUNCT
ajst-31431	4	10	the	the	DET
ajst-31431	4	11	deep	deep	ADJ
ajst-31431	4	12	integration	integration	NOUN
ajst-31431	4	13	of	of	ADP
ajst-31431	4	14	text	text	NOUN
ajst-31431	4	15	and	and	CCONJ
ajst-31431	4	16	images	image	NOUN
ajst-31431	4	17	)	)	PUNCT
ajst-31431	4	18	,	,	PUNCT
ajst-31431	4	19	further	far	ADV
ajst-31431	4	20	complicating	complicate	VERB
ajst-31431	4	21	content	content	NOUN
ajst-31431	4	22	security	security	NOUN
ajst-31431	4	23	management	management	NOUN
ajst-31431	4	24	.	.	PUNCT
ajst-31431	5	1	on	on	ADP
ajst-31431	5	2	one	one	NUM
ajst-31431	5	3	hand	hand	NOUN
ajst-31431	5	4	,	,	PUNCT
ajst-31431	5	5	the	the	DET
ajst-31431	5	6	subtlety	subtlety	NOUN
ajst-31431	5	7	and	and	CCONJ
ajst-31431	5	8	strong	strong	ADJ
ajst-31431	5	9	contextual	contextual	ADJ
ajst-31431	5	10	dependence	dependence	NOUN
ajst-31431	5	11	of	of	ADP
ajst-31431	5	12	hate	hate	NOUN
ajst-31431	5	13	speech	speech	NOUN
ajst-31431	5	14	significantly	significantly	ADV
ajst-31431	5	15	increase	increase	VERB
ajst-31431	5	16	the	the	DET
ajst-31431	5	17	difficulty	difficulty	NOUN
ajst-31431	5	18	of	of	ADP
ajst-31431	5	19	detection	detection	NOUN
ajst-31431	5	20	.	.	PUNCT
ajst-31431	6	1	on	on	ADP
ajst-31431	6	2	the	the	DET
ajst-31431	6	3	other	other	ADJ
ajst-31431	6	4	hand	hand	NOUN
ajst-31431	6	5	,	,	PUNCT
ajst-31431	6	6	emerging	emerge	VERB
ajst-31431	6	7	forms	form	NOUN
ajst-31431	6	8	of	of	ADP
ajst-31431	6	9	dissemination	dissemination	NOUN
ajst-31431	6	10	,	,	PUNCT
ajst-31431	6	11	such	such	ADJ
ajst-31431	6	12	as	as	ADP
ajst-31431	6	13	meme	meme	NOUN
ajst-31431	6	14	images	image	NOUN
ajst-31431	6	15	,	,	PUNCT
ajst-31431	6	16	present	present	ADJ
ajst-31431	6	17	dual	dual	ADJ
ajst-31431	6	18	challenges	challenge	NOUN
ajst-31431	6	19	for	for	ADP
ajst-31431	6	20	classification	classification	NOUN
ajst-31431	6	21	tasks	task	NOUN
ajst-31431	6	22	due	due	ADP
ajst-31431	6	23	to	to	ADP
ajst-31431	6	24	their	their	PRON
ajst-31431	6	25	inherent	inherent	ADJ
ajst-31431	6	26	characteristics	characteristic	NOUN
ajst-31431	6	27	:	:	PUNCT
ajst-31431	6	28	a	a	DET
ajst-31431	6	29	seemingly	seemingly	ADV
ajst-31431	6	30	humorous	humorous	ADJ
ajst-31431	6	31	facade	facade	NOUN
ajst-31431	6	32	,	,	PUNCT
ajst-31431	6	33	reliance	reliance	NOUN
ajst-31431	6	34	on	on	ADP
ajst-31431	6	35	cultural	cultural	ADJ
ajst-31431	6	36	context	context	NOUN
ajst-31431	6	37	,	,	PUNCT
ajst-31431	6	38	and	and	CCONJ
ajst-31431	6	39	semantic	semantic	ADJ
ajst-31431	6	40	conflicts	conflict	NOUN
ajst-31431	6	41	between	between	ADP
ajst-31431	6	42	text	text	NOUN
ajst-31431	6	43	and	and	CCONJ
ajst-31431	6	44	image	image	NOUN
ajst-31431	6	45	.	.	PUNCT
ajst-31431	7	1	to	to	PART
ajst-31431	7	2	address	address	VERB
ajst-31431	7	3	these	these	DET
ajst-31431	7	4	issues	issue	NOUN
ajst-31431	7	5	,	,	PUNCT
ajst-31431	7	6	this	this	DET
ajst-31431	7	7	paper	paper	NOUN
ajst-31431	7	8	focuses	focus	VERB
ajst-31431	7	9	on	on	ADP
ajst-31431	7	10	two	two	NUM
ajst-31431	7	11	major	major	ADJ
ajst-31431	7	12	technical	technical	ADJ
ajst-31431	7	13	approaches	approach	NOUN
ajst-31431	7	14	:	:	PUNCT
ajst-31431	7	15	unimodal	unimodal	ADJ
ajst-31431	7	16	text	text	NOUN
ajst-31431	7	17	analysis	analysis	NOUN
ajst-31431	7	18	and	and	CCONJ
ajst-31431	7	19	multimodal	multimodal	NOUN
ajst-31431	7	20	content	content	NOUN
ajst-31431	7	21	classification	classification	NOUN
ajst-31431	7	22	.	.	PUNCT
ajst-31431	8	1	it	it	PRON
ajst-31431	8	2	provides	provide	VERB
ajst-31431	8	3	a	a	DET
ajst-31431	8	4	systematic	systematic	ADJ
ajst-31431	8	5	review	review	NOUN
ajst-31431	8	6	of	of	ADP
ajst-31431	8	7	the	the	DET
ajst-31431	8	8	research	research	NOUN
ajst-31431	8	9	progress	progress	NOUN
ajst-31431	8	10	in	in	ADP
ajst-31431	8	11	hate	hate	NOUN
ajst-31431	8	12	speech	speech	NOUN
ajst-31431	8	13	detection	detection	NOUN
ajst-31431	8	14	methods	method	NOUN
ajst-31431	8	15	based	base	VERB
ajst-31431	8	16	on	on	ADP
ajst-31431	8	17	text	text	NOUN
ajst-31431	8	18	and	and	CCONJ
ajst-31431	8	19	multimodal	multimodal	NOUN
ajst-31431	8	20	detection	detection	NOUN
ajst-31431	8	21	methods	method	NOUN
ajst-31431	8	22	,	,	PUNCT
ajst-31431	8	23	while	while	SCONJ
ajst-31431	8	24	also	also	ADV
ajst-31431	8	25	analyzing	analyze	VERB
ajst-31431	8	26	their	their	PRON
ajst-31431	8	27	limitations	limitation	NOUN
ajst-31431	8	28	.	.	PUNCT
ajst-31431	9	1	furthermore	furthermore	ADV
ajst-31431	9	2	,	,	PUNCT
ajst-31431	9	3	this	this	DET
ajst-31431	9	4	paper	paper	NOUN
ajst-31431	9	5	consolidates	consolidate	VERB
ajst-31431	9	6	the	the	DET
ajst-31431	9	7	characteristics	characteristic	NOUN
ajst-31431	9	8	and	and	CCONJ
ajst-31431	9	9	applicable	applicable	ADJ
ajst-31431	9	10	scenarios	scenario	NOUN
ajst-31431	9	11	of	of	ADP
ajst-31431	9	12	current	current	ADJ
ajst-31431	9	13	mainstream	mainstream	NOUN
ajst-31431	9	14	unimodal	unimodal	ADJ
ajst-31431	9	15	and	and	CCONJ
ajst-31431	9	16	multimodal	multimodal	NOUN
ajst-31431	9	17	hate	hate	NOUN
ajst-31431	9	18	speech	speech	NOUN
ajst-31431	9	19	datasets	dataset	NOUN
ajst-31431	9	20	,	,	PUNCT
ajst-31431	9	21	offering	offer	VERB
ajst-31431	9	22	reference	reference	NOUN
ajst-31431	9	23	directions	direction	NOUN
ajst-31431	9	24	for	for	ADP
ajst-31431	9	25	optimizing	optimize	VERB
ajst-31431	9	26	technical	technical	ADJ
ajst-31431	9	27	approaches	approach	NOUN
ajst-31431	9	28	and	and	CCONJ
ajst-31431	9	29	constructing	construct	VERB
ajst-31431	9	30	datasets	dataset	NOUN
ajst-31431	9	31	in	in	ADP
ajst-31431	9	32	subsequent	subsequent	ADJ
ajst-31431	9	33	research	research	NOUN
ajst-31431	9	34	.	.	PUNCT
ajst-31431	10	1	keywords	keyword	NOUN
ajst-31431	10	2	:	:	PUNCT
ajst-31431	10	3	hate	hate	VERB
ajst-31431	10	4	speech	speech	NOUN
ajst-31431	10	5	detection	detection	NOUN
ajst-31431	10	6	;	;	PUNCT
ajst-31431	10	7	machine	machine	NOUN
ajst-31431	10	8	learning	learning	NOUN
ajst-31431	10	9	;	;	PUNCT
ajst-31431	10	10	deep	deep	ADJ
ajst-31431	10	11	learning	learning	NOUN
ajst-31431	10	12	;	;	PUNCT
ajst-31431	10	13	multimodal	multimodal	NOUN
ajst-31431	10	14	analysis	analysis	NOUN
ajst-31431	10	15	.	.	PUNCT
ajst-31431	11	1	1	1	X
ajst-31431	11	2	.	.	X
ajst-31431	11	3	introduction	introduction	NOUN
ajst-31431	11	4	with	with	ADP
ajst-31431	11	5	the	the	DET
ajst-31431	11	6	explosive	explosive	ADJ
ajst-31431	11	7	proliferation	proliferation	NOUN
ajst-31431	11	8	of	of	ADP
ajst-31431	11	9	social	social	ADJ
ajst-31431	11	10	media	medium	NOUN
ajst-31431	11	11	platforms	platform	NOUN
ajst-31431	11	12	such	such	ADJ
ajst-31431	11	13	as	as	ADP
ajst-31431	11	14	twitter	twitter	NOUN
ajst-31431	11	15	,	,	PUNCT
ajst-31431	11	16	instagram	instagram	PROPN
ajst-31431	11	17	,	,	PUNCT
ajst-31431	11	18	and	and	CCONJ
ajst-31431	11	19	youtube	youtube	NOUN
ajst-31431	11	20	,	,	PUNCT
ajst-31431	11	21	users	user	NOUN
ajst-31431	11	22	generate	generate	VERB
ajst-31431	11	23	billions	billion	NOUN
ajst-31431	11	24	of	of	ADP
ajst-31431	11	25	multimodal	multimodal	NOUN
ajst-31431	11	26	content	content	NOUN
ajst-31431	11	27	items	item	NOUN
ajst-31431	11	28	daily	daily	ADV
ajst-31431	11	29	,	,	PUNCT
ajst-31431	11	30	encompassing	encompass	VERB
ajst-31431	11	31	text	text	NOUN
ajst-31431	11	32	,	,	PUNCT
ajst-31431	11	33	images	image	NOUN
ajst-31431	11	34	,	,	PUNCT
ajst-31431	11	35	audio	audio	NOUN
ajst-31431	11	36	,	,	PUNCT
ajst-31431	11	37	and	and	CCONJ
ajst-31431	11	38	video	video	NOUN
ajst-31431	11	39	.	.	PUNCT
ajst-31431	12	1	among	among	ADP
ajst-31431	12	2	this	this	DET
ajst-31431	12	3	content	content	NOUN
ajst-31431	12	4	,	,	PUNCT
ajst-31431	12	5	negative	negative	ADJ
ajst-31431	12	6	statements	statement	NOUN
ajst-31431	12	7	targeting	target	VERB
ajst-31431	12	8	specific	specific	ADJ
ajst-31431	12	9	individuals	individual	NOUN
ajst-31431	12	10	or	or	CCONJ
ajst-31431	12	11	groups	group	NOUN
ajst-31431	12	12	can	can	AUX
ajst-31431	12	13	evolve	evolve	VERB
ajst-31431	12	14	into	into	ADP
ajst-31431	12	15	incendiary	incendiary	ADJ
ajst-31431	12	16	hate	hate	NOUN
ajst-31431	12	17	speech	speech	NOUN
ajst-31431	12	18	.	.	PUNCT
ajst-31431	13	1	such	such	ADJ
ajst-31431	13	2	content	content	NOUN
ajst-31431	13	3	not	not	PART
ajst-31431	13	4	only	only	ADV
ajst-31431	13	5	exacerbates	exacerbate	VERB
ajst-31431	13	6	social	social	ADJ
ajst-31431	13	7	polarization	polarization	NOUN
ajst-31431	13	8	and	and	CCONJ
ajst-31431	13	9	group	group	NOUN
ajst-31431	13	10	fragmentation	fragmentation	NOUN
ajst-31431	13	11	but	but	CCONJ
ajst-31431	13	12	may	may	AUX
ajst-31431	13	13	also	also	ADV
ajst-31431	13	14	trigger	trigger	VERB
ajst-31431	13	15	mental	mental	ADJ
ajst-31431	13	16	health	health	NOUN
ajst-31431	13	17	crises	crisis	NOUN
ajst-31431	13	18	such	such	ADJ
ajst-31431	13	19	as	as	ADP
ajst-31431	13	20	depression	depression	NOUN
ajst-31431	13	21	and	and	CCONJ
ajst-31431	13	22	anxiety	anxiety	NOUN
ajst-31431	13	23	,	,	PUNCT
ajst-31431	13	24	the	the	DET
ajst-31431	13	25	detrimental	detrimental	ADJ
ajst-31431	13	26	effects	effect	NOUN
ajst-31431	13	27	of	of	ADP
ajst-31431	13	28	which	which	PRON
ajst-31431	13	29	have	have	AUX
ajst-31431	13	30	been	be	AUX
ajst-31431	13	31	substantiated	substantiate	VERB
ajst-31431	13	32	in	in	ADP
ajst-31431	13	33	numerous	numerous	ADJ
ajst-31431	13	34	sociological	sociological	ADJ
ajst-31431	13	35	studies	study	NOUN
ajst-31431	13	36	[	[	X
ajst-31431	13	37	1	1	NUM
ajst-31431	13	38	]	]	PUNCT
ajst-31431	13	39	.	.	PUNCT
ajst-31431	14	1	consequently	consequently	ADV
ajst-31431	14	2	,	,	PUNCT
ajst-31431	14	3	hate	hate	VERB
ajst-31431	14	4	speech	speech	NOUN
ajst-31431	14	5	detection	detection	NOUN
ajst-31431	14	6	holds	hold	VERB
ajst-31431	14	7	significant	significant	ADJ
ajst-31431	14	8	practical	practical	ADJ
ajst-31431	14	9	importance	importance	NOUN
ajst-31431	14	10	.	.	PUNCT
ajst-31431	15	1	in	in	ADP
ajst-31431	15	2	the	the	DET
ajst-31431	15	3	era	era	NOUN
ajst-31431	15	4	of	of	ADP
ajst-31431	15	5	big	big	ADJ
ajst-31431	15	6	data	datum	NOUN
ajst-31431	15	7	,	,	PUNCT
ajst-31431	15	8	manual	manual	ADJ
ajst-31431	15	9	processing	processing	NOUN
ajst-31431	15	10	and	and	CCONJ
ajst-31431	15	11	classification	classification	NOUN
ajst-31431	15	12	of	of	ADP
ajst-31431	15	13	vast	vast	ADJ
ajst-31431	15	14	amounts	amount	NOUN
ajst-31431	15	15	of	of	ADP
ajst-31431	15	16	textual	textual	ADJ
ajst-31431	15	17	data	datum	NOUN
ajst-31431	15	18	are	be	AUX
ajst-31431	15	19	both	both	DET
ajst-31431	15	20	time	time	NOUN
ajst-31431	15	21	-	-	PUNCT
ajst-31431	15	22	consuming	consume	VERB
ajst-31431	15	23	and	and	CCONJ
ajst-31431	15	24	susceptible	susceptible	ADJ
ajst-31431	15	25	to	to	ADP
ajst-31431	15	26	human	human	ADJ
ajst-31431	15	27	bias	bias	NOUN
ajst-31431	15	28	(	(	PUNCT
ajst-31431	15	29	mullah	mullah	PROPN
ajst-31431	15	30	&	&	CCONJ
ajst-31431	15	31	zainon	zainon	PROPN
ajst-31431	15	32	,	,	PUNCT
ajst-31431	15	33	2021	2021	NUM
ajst-31431	15	34	)	)	PUNCT
ajst-31431	15	35	.	.	PUNCT
ajst-31431	16	1	as	as	SCONJ
ajst-31431	16	2	research	research	NOUN
ajst-31431	16	3	utilizing	utilize	VERB
ajst-31431	16	4	natural	natural	ADJ
ajst-31431	16	5	language	language	NOUN
ajst-31431	16	6	processing	processing	NOUN
ajst-31431	16	7	(	(	PUNCT
ajst-31431	16	8	nlp	nlp	NOUN
ajst-31431	16	9	)	)	PUNCT
ajst-31431	16	10	for	for	ADP
ajst-31431	16	11	hate	hate	NOUN
ajst-31431	16	12	speech	speech	NOUN
ajst-31431	16	13	detection	detection	NOUN
ajst-31431	16	14	(	(	PUNCT
ajst-31431	16	15	hsd	hsd	NOUN
ajst-31431	16	16	)	)	PUNCT
ajst-31431	16	17	has	have	AUX
ajst-31431	16	18	grown	grow	VERB
ajst-31431	16	19	,	,	PUNCT
ajst-31431	16	20	traditional	traditional	ADJ
ajst-31431	16	21	methods	method	NOUN
ajst-31431	16	22	relying	rely	VERB
ajst-31431	16	23	on	on	ADP
ajst-31431	16	24	keyword	keyword	NOUN
ajst-31431	16	25	rules	rule	NOUN
ajst-31431	16	26	or	or	CCONJ
ajst-31431	16	27	shallow	shallow	ADJ
ajst-31431	16	28	machine	machine	NOUN
ajst-31431	16	29	learning	learn	VERB
ajst-31431	16	30	techniques	technique	NOUN
ajst-31431	16	31	(	(	PUNCT
ajst-31431	16	32	e.g.	e.g.	ADV
ajst-31431	16	33	,	,	PUNCT
ajst-31431	16	34	tf	tf	PROPN
ajst-31431	16	35	-	-	PUNCT
ajst-31431	16	36	idf	idf	NOUN
ajst-31431	16	37	+	+	CCONJ
ajst-31431	16	38	svm	svm	PROPN
ajst-31431	16	39	)	)	PUNCT
ajst-31431	16	40	have	have	AUX
ajst-31431	16	41	proven	prove	VERB
ajst-31431	16	42	inadequate	inadequate	ADJ
ajst-31431	16	43	.	.	PUNCT
ajst-31431	17	1	this	this	PRON
ajst-31431	17	2	is	be	AUX
ajst-31431	17	3	because	because	SCONJ
ajst-31431	17	4	they	they	PRON
ajst-31431	17	5	struggle	struggle	VERB
ajst-31431	17	6	to	to	PART
ajst-31431	17	7	capture	capture	VERB
ajst-31431	17	8	the	the	DET
ajst-31431	17	9	contextual	contextual	ADJ
ajst-31431	17	10	semantic	semantic	ADJ
ajst-31431	17	11	relationships	relationship	NOUN
ajst-31431	17	12	embedded	embed	VERB
ajst-31431	17	13	within	within	ADP
ajst-31431	17	14	the	the	DET
ajst-31431	17	15	highly	highly	ADV
ajst-31431	17	16	unstructured	unstructured	ADJ
ajst-31431	17	17	nature	nature	NOUN
ajst-31431	17	18	of	of	ADP
ajst-31431	17	19	social	social	ADJ
ajst-31431	17	20	media	medium	NOUN
ajst-31431	17	21	text	text	NOUN
ajst-31431	17	22	,	,	PUNCT
ajst-31431	17	23	which	which	PRON
ajst-31431	17	24	often	often	ADV
ajst-31431	17	25	employs	employ	VERB
ajst-31431	17	26	subtle	subtle	ADJ
ajst-31431	17	27	expressions	expression	NOUN
ajst-31431	17	28	of	of	ADP
ajst-31431	17	29	hostility	hostility	NOUN
ajst-31431	17	30	such	such	ADJ
ajst-31431	17	31	as	as	ADP
ajst-31431	17	32	metaphors	metaphor	NOUN
ajst-31431	17	33	and	and	CCONJ
ajst-31431	17	34	irony	irony	NOUN
ajst-31431	17	35	.	.	PUNCT
ajst-31431	18	1	the	the	DET
ajst-31431	18	2	contextual	contextual	ADJ
ajst-31431	18	3	dependency	dependency	NOUN
ajst-31431	18	4	,	,	PUNCT
ajst-31431	18	5	polysemy	polysemy	NOUN
ajst-31431	18	6	,	,	PUNCT
ajst-31431	18	7	and	and	CCONJ
ajst-31431	18	8	dynamic	dynamic	ADJ
ajst-31431	18	9	nature	nature	NOUN
ajst-31431	18	10	of	of	ADP
ajst-31431	18	11	internet	internet	NOUN
ajst-31431	18	12	slang	slang	PROPN
ajst-31431	18	13	pose	pose	VERB
ajst-31431	18	14	significant	significant	ADJ
ajst-31431	18	15	challenges	challenge	NOUN
ajst-31431	18	16	for	for	ADP
ajst-31431	18	17	semantic	semantic	ADJ
ajst-31431	18	18	parsing	parsing	NOUN
ajst-31431	18	19	in	in	ADP
ajst-31431	18	20	text	text	NOUN
ajst-31431	18	21	classification	classification	NOUN
ajst-31431	18	22	.	.	PUNCT
ajst-31431	19	1	consequently	consequently	ADV
ajst-31431	19	2	,	,	PUNCT
ajst-31431	19	3	conventional	conventional	ADJ
ajst-31431	19	4	rule	rule	NOUN
ajst-31431	19	5	-	-	PUNCT
ajst-31431	19	6	based	base	VERB
ajst-31431	19	7	hate	hate	NOUN
ajst-31431	19	8	speech	speech	NOUN
ajst-31431	19	9	detection	detection	NOUN
ajst-31431	19	10	approaches	approach	NOUN
ajst-31431	19	11	lack	lack	VERB
ajst-31431	19	12	scalability	scalability	NOUN
ajst-31431	19	13	and	and	CCONJ
ajst-31431	19	14	adaptability	adaptability	NOUN
ajst-31431	19	15	,	,	PUNCT
ajst-31431	19	16	making	make	VERB
ajst-31431	19	17	it	it	PRON
ajst-31431	19	18	difficult	difficult	ADJ
ajst-31431	19	19	to	to	PART
ajst-31431	19	20	handle	handle	VERB
ajst-31431	19	21	the	the	DET
ajst-31431	19	22	massive	massive	ADJ
ajst-31431	19	23	volume	volume	NOUN
ajst-31431	19	24	of	of	ADP
ajst-31431	19	25	user	user	NOUN
ajst-31431	19	26	-	-	PUNCT
ajst-31431	19	27	generated	generate	VERB
ajst-31431	19	28	content	content	NOUN
ajst-31431	19	29	on	on	ADP
ajst-31431	19	30	social	social	ADJ
ajst-31431	19	31	media	medium	NOUN
ajst-31431	19	32	platforms	platform	NOUN
ajst-31431	19	33	.	.	PUNCT
ajst-31431	20	1	in	in	ADP
ajst-31431	20	2	contrast	contrast	NOUN
ajst-31431	20	3	,	,	PUNCT
ajst-31431	20	4	machine	machine	NOUN
ajst-31431	20	5	learning	learning	NOUN
ajst-31431	20	6	and	and	CCONJ
ajst-31431	20	7	deep	deep	ADJ
ajst-31431	20	8	learning	learning	NOUN
ajst-31431	20	9	techniques	technique	NOUN
ajst-31431	20	10	have	have	AUX
ajst-31431	20	11	demonstrated	demonstrate	VERB
ajst-31431	20	12	promising	promising	ADJ
ajst-31431	20	13	results	result	NOUN
ajst-31431	20	14	in	in	ADP
ajst-31431	20	15	the	the	DET
ajst-31431	20	16	automatic	automatic	ADJ
ajst-31431	20	17	identification	identification	NOUN
ajst-31431	20	18	of	of	ADP
ajst-31431	20	19	hate	hate	NOUN
ajst-31431	20	20	speech	speech	NOUN
ajst-31431	20	21	.	.	PUNCT
ajst-31431	21	1	furthermore	furthermore	ADV
ajst-31431	21	2	,	,	PUNCT
ajst-31431	21	3	shifts	shift	NOUN
ajst-31431	21	4	in	in	ADP
ajst-31431	21	5	attention	attention	NOUN
ajst-31431	21	6	patterns	pattern	NOUN
ajst-31431	21	7	driven	drive	VERB
ajst-31431	21	8	by	by	ADP
ajst-31431	21	9	technological	technological	ADJ
ajst-31431	21	10	revolutions	revolution	NOUN
ajst-31431	21	11	have	have	AUX
ajst-31431	21	12	fueled	fuel	VERB
ajst-31431	21	13	the	the	DET
ajst-31431	21	14	popularity	popularity	NOUN
ajst-31431	21	15	of	of	ADP
ajst-31431	21	16	hybrid	hybrid	ADJ
ajst-31431	21	17	content	content	NOUN
ajst-31431	21	18	formats	format	NOUN
ajst-31431	21	19	like	like	ADP
ajst-31431	21	20	memes	meme	NOUN
ajst-31431	21	21	,	,	PUNCT
ajst-31431	21	22	which	which	PRON
ajst-31431	21	23	integrate	integrate	VERB
ajst-31431	21	24	images	image	NOUN
ajst-31431	21	25	and	and	CCONJ
ajst-31431	21	26	text	text	NOUN
ajst-31431	21	27	.	.	PUNCT
ajst-31431	22	1	leveraging	leverage	VERB
ajst-31431	22	2	their	their	PRON
ajst-31431	22	3	viral	viral	ADJ
ajst-31431	22	4	propagation	propagation	NOUN
ajst-31431	22	5	characteristics	characteristic	NOUN
ajst-31431	22	6	,	,	PUNCT
ajst-31431	22	7	these	these	DET
ajst-31431	22	8	formats	format	NOUN
ajst-31431	22	9	have	have	AUX
ajst-31431	22	10	become	become	VERB
ajst-31431	22	11	dominant	dominant	ADJ
ajst-31431	22	12	modes	mode	NOUN
ajst-31431	22	13	of	of	ADP
ajst-31431	22	14	online	online	ADJ
ajst-31431	22	15	expression	expression	NOUN
ajst-31431	22	16	.	.	PUNCT
ajst-31431	23	1	however	however	ADV
ajst-31431	23	2	,	,	PUNCT
ajst-31431	23	3	the	the	DET
ajst-31431	23	4	exponential	exponential	ADJ
ajst-31431	23	5	growth	growth	NOUN
ajst-31431	23	6	of	of	ADP
ajst-31431	23	7	content	content	NOUN
ajst-31431	23	8	within	within	ADP
ajst-31431	23	9	these	these	DET
ajst-31431	23	10	formats	format	NOUN
ajst-31431	23	11	that	that	PRON
ajst-31431	23	12	supports	support	VERB
ajst-31431	23	13	radical	radical	ADJ
ajst-31431	23	14	speech	speech	NOUN
ajst-31431	23	15	has	have	AUX
ajst-31431	23	16	created	create	VERB
ajst-31431	23	17	an	an	DET
ajst-31431	23	18	urgent	urgent	ADJ
ajst-31431	23	19	demand	demand	NOUN
ajst-31431	23	20	for	for	ADP
ajst-31431	23	21	automated	automate	VERB
ajst-31431	23	22	detection	detection	NOUN
ajst-31431	23	23	technologies	technology	NOUN
ajst-31431	23	24	.	.	PUNCT
ajst-31431	24	1	the	the	DET
ajst-31431	24	2	image	image	NOUN
ajst-31431	24	3	-	-	PUNCT
ajst-31431	24	4	text	text	NOUN
ajst-31431	24	5	hybrid	hybrid	ADJ
ajst-31431	24	6	structure	structure	NOUN
ajst-31431	24	7	of	of	ADP
ajst-31431	24	8	memes	meme	NOUN
ajst-31431	24	9	presents	present	VERB
ajst-31431	24	10	dual	dual	ADJ
ajst-31431	24	11	detection	detection	NOUN
ajst-31431	24	12	challenges	challenge	NOUN
ajst-31431	24	13	:	:	PUNCT
ajst-31431	24	14	first	first	ADV
ajst-31431	24	15	,	,	PUNCT
ajst-31431	24	16	the	the	DET
ajst-31431	24	17	open	open	ADV
ajst-31431	24	18	-	-	PUNCT
ajst-31431	24	19	ended	end	VERB
ajst-31431	24	20	nature	nature	NOUN
ajst-31431	24	21	of	of	ADP
ajst-31431	24	22	image	image	NOUN
ajst-31431	24	23	interpretation	interpretation	NOUN
ajst-31431	24	24	—	—	PUNCT
ajst-31431	24	25	where	where	SCONJ
ajst-31431	24	26	the	the	DET
ajst-31431	24	27	diverse	diverse	ADJ
ajst-31431	24	28	semantics	semantic	NOUN
ajst-31431	24	29	and	and	CCONJ
ajst-31431	24	30	cultural	cultural	ADJ
ajst-31431	24	31	specificity	specificity	NOUN
ajst-31431	24	32	of	of	ADP
ajst-31431	24	33	visual	visual	ADJ
ajst-31431	24	34	symbols	symbol	NOUN
ajst-31431	24	35	render	render	VERB
ajst-31431	24	36	the	the	DET
ajst-31431	24	37	construction	construction	NOUN
ajst-31431	24	38	of	of	ADP
ajst-31431	24	39	comprehensive	comprehensive	ADJ
ajst-31431	24	40	datasets	dataset	NOUN
ajst-31431	24	41	infeasible	infeasible	ADJ
ajst-31431	24	42	;	;	PUNCT
ajst-31431	24	43	second	second	X
ajst-31431	24	44	,	,	PUNCT
ajst-31431	24	45	the	the	DET
ajst-31431	24	46	challenge	challenge	NOUN
ajst-31431	24	47	of	of	ADP
ajst-31431	24	48	integrating	integrate	VERB
ajst-31431	24	49	cross	cross	ADJ
ajst-31431	24	50	-	-	ADJ
ajst-31431	24	51	modal	modal	ADJ
ajst-31431	24	52	information	information	NOUN
ajst-31431	24	53	for	for	ADP
ajst-31431	24	54	classification	classification	NOUN
ajst-31431	24	55	—	—	PUNCT
ajst-31431	24	56	requiring	require	VERB
ajst-31431	24	57	the	the	DET
ajst-31431	24	58	effective	effective	ADJ
ajst-31431	24	59	combination	combination	NOUN
ajst-31431	24	60	of	of	ADP
ajst-31431	24	61	imagederived	imagederived	ADJ
ajst-31431	24	62	data	datum	NOUN
ajst-31431	24	63	with	with	ADP
ajst-31431	24	64	textual	textual	ADJ
ajst-31431	24	65	information	information	NOUN
ajst-31431	24	66	to	to	PART
ajst-31431	24	67	ascertain	ascertain	VERB
ajst-31431	24	68	the	the	DET
ajst-31431	24	69	presence	presence	NOUN
ajst-31431	24	70	of	of	ADP
ajst-31431	24	71	hateful	hateful	ADJ
ajst-31431	24	72	intent	intent	NOUN
ajst-31431	24	73	.	.	PUNCT
ajst-31431	25	1	based	base	VERB
ajst-31431	25	2	on	on	ADP
ajst-31431	25	3	this	this	PRON
ajst-31431	25	4	,	,	PUNCT
ajst-31431	25	5	we	we	PRON
ajst-31431	25	6	analyze	analyze	VERB
ajst-31431	25	7	and	and	CCONJ
ajst-31431	25	8	summarize	summarize	VERB
ajst-31431	25	9	the	the	DET
ajst-31431	25	10	application	application	NOUN
ajst-31431	25	11	and	and	CCONJ
ajst-31431	25	12	challenges	challenge	NOUN
ajst-31431	25	13	of	of	ADP
ajst-31431	25	14	machine	machine	NOUN
ajst-31431	25	15	learning	learning	NOUN
ajst-31431	25	16	and	and	CCONJ
ajst-31431	25	17	deep	deep	ADJ
ajst-31431	25	18	learning	learning	NOUN
ajst-31431	25	19	in	in	ADP
ajst-31431	25	20	both	both	CCONJ
ajst-31431	25	21	unimodal	unimodal	ADJ
ajst-31431	25	22	and	and	CCONJ
ajst-31431	25	23	multimodal	multimodal	NOUN
ajst-31431	25	24	hate	hate	NOUN
ajst-31431	25	25	speech	speech	NOUN
ajst-31431	25	26	detection	detection	NOUN
ajst-31431	25	27	.	.	PUNCT
ajst-31431	26	1	our	our	PRON
ajst-31431	26	2	study	study	NOUN
ajst-31431	26	3	includes	include	VERB
ajst-31431	26	4	the	the	DET
ajst-31431	26	5	following	follow	VERB
ajst-31431	26	6	aspects	aspect	NOUN
ajst-31431	26	7	:	:	PUNCT
ajst-31431	26	8	1	1	X
ajst-31431	26	9	.	.	X
ajst-31431	26	10	we	we	PRON
ajst-31431	26	11	summarize	summarize	VERB
ajst-31431	26	12	various	various	ADJ
ajst-31431	26	13	state	state	NOUN
ajst-31431	26	14	-	-	PUNCT
ajst-31431	26	15	of	of	ADP
ajst-31431	26	16	-	-	PUNCT
ajst-31431	26	17	the	the	DET
ajst-31431	26	18	-	-	PUNCT
ajst-31431	26	19	art	art	NOUN
ajst-31431	26	20	benchmark	benchmark	NOUN
ajst-31431	26	21	hate	hate	NOUN
ajst-31431	26	22	speech	speech	NOUN
ajst-31431	26	23	datasets	dataset	NOUN
ajst-31431	26	24	employed	employ	VERB
ajst-31431	26	25	for	for	ADP
ajst-31431	26	26	this	this	DET
ajst-31431	26	27	task	task	NOUN
ajst-31431	26	28	.	.	PUNCT
ajst-31431	27	1	2	2	X
ajst-31431	27	2	.	.	X
ajst-31431	27	3	we	we	PRON
ajst-31431	27	4	provide	provide	VERB
ajst-31431	27	5	a	a	DET
ajst-31431	27	6	concise	concise	ADJ
ajst-31431	27	7	overview	overview	NOUN
ajst-31431	27	8	of	of	ADP
ajst-31431	27	9	recent	recent	ADJ
ajst-31431	27	10	relevant	relevant	ADJ
ajst-31431	27	11	work	work	NOUN
ajst-31431	27	12	pertaining	pertain	VERB
ajst-31431	27	13	to	to	PART
ajst-31431	27	14	hate	hate	VERB
ajst-31431	27	15	speech	speech	NOUN
ajst-31431	27	16	detection	detection	NOUN
ajst-31431	27	17	.	.	PUNCT
ajst-31431	28	1	3	3	X
ajst-31431	28	2	.	.	X
ajst-31431	28	3	we	we	PRON
ajst-31431	28	4	conduct	conduct	VERB
ajst-31431	28	5	a	a	DET
ajst-31431	28	6	systematic	systematic	ADJ
ajst-31431	28	7	literature	literature	NOUN
ajst-31431	28	8	review	review	NOUN
ajst-31431	28	9	across	across	ADP
ajst-31431	28	10	different	different	ADJ
ajst-31431	28	11	data	datum	NOUN
ajst-31431	28	12	modalities	modality	NOUN
ajst-31431	28	13	,	,	PUNCT
ajst-31431	28	14	specifically	specifically	ADV
ajst-31431	28	15	focusing	focus	VERB
ajst-31431	28	16	on	on	ADP
ajst-31431	28	17	text	text	NOUN
ajst-31431	28	18	-	-	PUNCT
ajst-31431	28	19	based	base	VERB
ajst-31431	28	20	hate	hate	NOUN
ajst-31431	28	21	speech	speech	NOUN
ajst-31431	28	22	detection	detection	NOUN
ajst-31431	28	23	and	and	CCONJ
ajst-31431	28	24	multimodal	multimodal	NOUN
ajst-31431	28	25	hate	hate	NOUN
ajst-31431	28	26	speech	speech	NOUN
ajst-31431	28	27	detection	detection	NOUN
ajst-31431	28	28	.	.	PUNCT
ajst-31431	29	1	2	2	X
ajst-31431	29	2	.	.	X
ajst-31431	29	3	hate	hate	NOUN
ajst-31431	29	4	speech	speech	NOUN
ajst-31431	29	5	dataset	dataset	VERB
ajst-31431	29	6	2.1	2.1	NUM
ajst-31431	29	7	.	.	PUNCT
ajst-31431	30	1	text	text	NOUN
ajst-31431	30	2	hate	hate	NOUN
ajst-31431	30	3	speech	speech	NOUN
ajst-31431	30	4	dataset	dataset	VERB
ajst-31431	30	5	while	while	SCONJ
ajst-31431	30	6	existing	exist	VERB
ajst-31431	30	7	hate	hate	NOUN
ajst-31431	30	8	speech	speech	NOUN
ajst-31431	30	9	detection	detection	NOUN
ajst-31431	30	10	datasets	dataset	NOUN
ajst-31431	30	11	offer	offer	VERB
ajst-31431	30	12	comprehensive	comprehensive	ADJ
ajst-31431	30	13	category	category	NOUN
ajst-31431	30	14	coverage	coverage	NOUN
ajst-31431	30	15	,	,	PUNCT
ajst-31431	30	16	significant	significant	ADJ
ajst-31431	30	17	variations	variation	NOUN
ajst-31431	30	18	exist	exist	VERB
ajst-31431	30	19	in	in	ADP
ajst-31431	30	20	their	their	PRON
ajst-31431	30	21	quality	quality	NOUN
ajst-31431	30	22	and	and	CCONJ
ajst-31431	30	23	reliability	reliability	NOUN
ajst-31431	30	24	.	.	PUNCT
ajst-31431	31	1	some	some	DET
ajst-31431	31	2	early	early	ADJ
ajst-31431	31	3	datasets	dataset	NOUN
ajst-31431	31	4	,	,	PUNCT
ajst-31431	31	5	such	such	ADJ
ajst-31431	31	6	as	as	ADP
ajst-31431	31	7	waseem	waseem	PROPN
ajst-31431	31	8	’s	’s	PART
ajst-31431	31	9	[	[	X
ajst-31431	31	10	2	2	NUM
ajst-31431	31	11	]	]	PUNCT
ajst-31431	31	12	,	,	PUNCT
ajst-31431	31	13	suffer	suffer	VERB
ajst-31431	31	14	from	from	ADP
ajst-31431	31	15	ambiguous	ambiguous	ADJ
ajst-31431	31	16	semantic	semantic	ADJ
ajst-31431	31	17	boundaries	boundary	NOUN
ajst-31431	31	18	,	,	PUNCT
ajst-31431	31	19	and	and	CCONJ
ajst-31431	31	20	rozental	rozental	PROPN
ajst-31431	31	21	et	et	PROPN
ajst-31431	31	22	al	al	PROPN
ajst-31431	31	23	.	.	PROPN
ajst-31431	31	24	's	's	PART
ajst-31431	31	25	hierarchical	hierarchical	ADJ
ajst-31431	31	26	dataset	dataset	NOUN
ajst-31431	32	1	[	[	X
ajst-31431	32	2	3	3	NUM
ajst-31431	32	3	]	]	PUNCT
ajst-31431	32	4	is	be	AUX
ajst-31431	32	5	limited	limit	VERB
ajst-31431	32	6	by	by	ADP
ajst-31431	32	7	insufficient	insufficient	ADJ
ajst-31431	32	8	sample	sample	NOUN
ajst-31431	32	9	representativeness	representativeness	NOUN
ajst-31431	32	10	.	.	PUNCT
ajst-31431	33	1	in	in	ADP
ajst-31431	33	2	contrast	contrast	NOUN
ajst-31431	33	3	,	,	PUNCT
ajst-31431	33	4	datasets	dataset	NOUN
ajst-31431	33	5	supported	support	VERB
ajst-31431	33	6	by	by	ADP
ajst-31431	33	7	academic	academic	ADJ
ajst-31431	33	8	conferences	conference	NOUN
ajst-31431	33	9	demonstrate	demonstrate	VERB
ajst-31431	33	10	stronger	strong	ADJ
ajst-31431	33	11	standardization	standardization	NOUN
ajst-31431	33	12	.	.	PUNCT
ajst-31431	34	1	for	for	ADP
ajst-31431	34	2	instance	instance	NOUN
ajst-31431	34	3	,	,	PUNCT
ajst-31431	34	4	the	the	DET
ajst-31431	34	5	hasoc	hasoc	ADJ
ajst-31431	34	6	benchmark	benchmark	NOUN
ajst-31431	35	1	[	[	X
ajst-31431	35	2	4	4	NUM
ajst-31431	35	3	]	]	PUNCT
ajst-31431	35	4	,	,	PUNCT
ajst-31431	35	5	employing	employ	VERB
ajst-31431	35	6	a	a	DET
ajst-31431	35	7	graded	grade	VERB
ajst-31431	35	8	task	task	NOUN
ajst-31431	35	9	structure	structure	NOUN
ajst-31431	35	10	(	(	PUNCT
ajst-31431	35	11	subtasks	subtask	VERB
ajst-31431	35	12	a	a	DET
ajst-31431	35	13	/	/	SYM
ajst-31431	35	14	b	b	NOUN
ajst-31431	35	15	)	)	PUNCT
ajst-31431	35	16	,	,	PUNCT
ajst-31431	35	17	provides	provide	VERB
ajst-31431	35	18	a	a	DET
ajst-31431	35	19	systematic	systematic	ADJ
ajst-31431	35	20	evaluation	evaluation	NOUN
ajst-31431	35	21	framework	framework	NOUN
ajst-31431	35	22	.	.	PUNCT
ajst-31431	36	1	although	although	SCONJ
ajst-31431	36	2	baseline	baseline	NOUN
ajst-31431	36	3	models	model	NOUN
ajst-31431	36	4	41	41	NUM
ajst-31431	36	5	achieve	achieve	VERB
ajst-31431	36	6	relatively	relatively	ADV
ajst-31431	36	7	low	low	ADJ
ajst-31431	36	8	f1	f1	NOUN
ajst-31431	36	9	scores	score	NOUN
ajst-31431	36	10	(	(	PUNCT
ajst-31431	36	11	e.g.	e.g.	ADV
ajst-31431	36	12	,	,	PUNCT
ajst-31431	36	13	0.52	0.52	NUM
ajst-31431	36	14	for	for	ADP
ajst-31431	36	15	binary	binary	ADJ
ajst-31431	36	16	classification	classification	NOUN
ajst-31431	36	17	)	)	PUNCT
ajst-31431	36	18	,	,	PUNCT
ajst-31431	36	19	it	it	PRON
ajst-31431	36	20	remains	remain	VERB
ajst-31431	36	21	a	a	DET
ajst-31431	36	22	crucial	crucial	ADJ
ajst-31431	36	23	research	research	NOUN
ajst-31431	36	24	platform	platform	NOUN
ajst-31431	36	25	.	.	PUNCT
ajst-31431	37	1	similarly	similarly	ADV
ajst-31431	37	2	,	,	PUNCT
ajst-31431	37	3	the	the	DET
ajst-31431	37	4	semeval	semeval	ADJ
ajst-31431	37	5	series	series	NOUN
ajst-31431	37	6	has	have	AUX
ajst-31431	37	7	advanced	advance	VERB
ajst-31431	37	8	the	the	DET
ajst-31431	37	9	field	field	NOUN
ajst-31431	37	10	,	,	PUNCT
ajst-31431	37	11	with	with	ADP
ajst-31431	37	12	task	task	NOUN
ajst-31431	37	13	5	5	NUM
ajst-31431	37	14	[	[	X
ajst-31431	37	15	5,6	5,6	NUM
ajst-31431	37	16	]	]	PUNCT
ajst-31431	37	17	focusing	focus	VERB
ajst-31431	37	18	on	on	ADP
ajst-31431	37	19	multilingual	multilingual	ADJ
ajst-31431	37	20	hate	hate	NOUN
ajst-31431	37	21	speech	speech	NOUN
ajst-31431	37	22	detection	detection	NOUN
ajst-31431	37	23	concerning	concern	VERB
ajst-31431	37	24	immigrants	immigrant	NOUN
ajst-31431	37	25	and	and	CCONJ
ajst-31431	37	26	women	woman	NOUN
ajst-31431	37	27	,	,	PUNCT
ajst-31431	37	28	and	and	CCONJ
ajst-31431	37	29	task	task	NOUN
ajst-31431	37	30	6	6	NUM
ajst-31431	37	31	[	[	X
ajst-31431	37	32	7	7	X
ajst-31431	37	33	]	]	X
ajst-31431	37	34	establishing	establish	VERB
ajst-31431	37	35	the	the	DET
ajst-31431	37	36	olid	olid	ADJ
ajst-31431	37	37	dataset	dataset	NOUN
ajst-31431	37	38	,	,	PUNCT
ajst-31431	37	39	which	which	PRON
ajst-31431	37	40	sets	set	VERB
ajst-31431	37	41	fine	fine	ADV
ajst-31431	37	42	-	-	PUNCT
ajst-31431	37	43	grained	grain	VERB
ajst-31431	37	44	classification	classification	NOUN
ajst-31431	37	45	standards	standard	NOUN
ajst-31431	37	46	and	and	CCONJ
ajst-31431	37	47	serves	serve	VERB
ajst-31431	37	48	as	as	ADP
ajst-31431	37	49	a	a	DET
ajst-31431	37	50	key	key	ADJ
ajst-31431	37	51	reference	reference	NOUN
ajst-31431	37	52	for	for	ADP
ajst-31431	37	53	hasoc	hasoc	NOUN
ajst-31431	37	54	.	.	PUNCT
ajst-31431	38	1	current	current	ADJ
ajst-31431	38	2	dataset	dataset	NOUN
ajst-31431	38	3	development	development	NOUN
ajst-31431	38	4	exhibits	exhibit	VERB
ajst-31431	38	5	a	a	DET
ajst-31431	38	6	bipolar	bipolar	ADJ
ajst-31431	38	7	trend	trend	NOUN
ajst-31431	38	8	:	:	PUNCT
ajst-31431	38	9	largescale	largescale	ADJ
ajst-31431	38	10	annotation	annotation	NOUN
ajst-31431	38	11	projects	project	NOUN
ajst-31431	38	12	enhance	enhance	VERB
ajst-31431	38	13	model	model	NOUN
ajst-31431	38	14	generalization	generalization	NOUN
ajst-31431	38	15	,	,	PUNCT
ajst-31431	38	16	while	while	SCONJ
ajst-31431	38	17	smaller	small	ADJ
ajst-31431	38	18	,	,	PUNCT
ajst-31431	38	19	specialized	specialized	ADJ
ajst-31431	38	20	datasets	dataset	NOUN
ajst-31431	38	21	offer	offer	VERB
ajst-31431	38	22	deeper	deep	ADJ
ajst-31431	38	23	insights	insight	NOUN
ajst-31431	38	24	in	in	ADP
ajst-31431	38	25	specific	specific	ADJ
ajst-31431	38	26	contexts	contexts	NOUN
ajst-31431	38	27	.	.	PUNCT
ajst-31431	39	1	however	however	ADV
ajst-31431	39	2	,	,	PUNCT
ajst-31431	39	3	existing	exist	VERB
ajst-31431	39	4	research	research	NOUN
ajst-31431	39	5	exhibits	exhibit	VERB
ajst-31431	39	6	notable	notable	ADJ
ajst-31431	39	7	limitations	limitation	NOUN
ajst-31431	39	8	:	:	PUNCT
ajst-31431	39	9	first	first	ADJ
ajst-31431	39	10	,	,	PUNCT
ajst-31431	39	11	systematic	systematic	ADJ
ajst-31431	39	12	evaluation	evaluation	NOUN
ajst-31431	39	13	of	of	ADP
ajst-31431	39	14	non	non	ADJ
ajst-31431	39	15	-	-	ADJ
ajst-31431	39	16	twitter	twitter	ADJ
ajst-31431	39	17	platform	platform	NOUN
ajst-31431	39	18	data	datum	NOUN
ajst-31431	39	19	[	[	X
ajst-31431	39	20	8	8	NUM
ajst-31431	39	21	]	]	PUNCT
ajst-31431	39	22	is	be	AUX
ajst-31431	39	23	lacking	lack	VERB
ajst-31431	39	24	;	;	PUNCT
ajst-31431	39	25	second	second	ADJ
ajst-31431	39	26	,	,	PUNCT
ajst-31431	39	27	multilingual	multilingual	ADJ
ajst-31431	39	28	extension	extension	NOUN
ajst-31431	39	29	remains	remain	VERB
ajst-31431	39	30	severely	severely	ADV
ajst-31431	39	31	underdeveloped	underdeveloped	ADJ
ajst-31431	39	32	,	,	PUNCT
ajst-31431	39	33	and	and	CCONJ
ajst-31431	39	34	non	non	ADJ
ajst-31431	39	35	-	-	ADJ
ajst-31431	39	36	english	english	ADJ
ajst-31431	39	37	datasets	dataset	NOUN
ajst-31431	39	38	[	[	X
ajst-31431	39	39	9	9	NUM
ajst-31431	39	40	]	]	PUNCT
ajst-31431	39	41	are	be	AUX
ajst-31431	39	42	often	often	ADV
ajst-31431	39	43	discarded	discard	VERB
ajst-31431	39	44	due	due	ADP
ajst-31431	39	45	to	to	ADP
ajst-31431	39	46	inconsistent	inconsistent	ADJ
ajst-31431	39	47	annotation	annotation	NOUN
ajst-31431	39	48	standards	standard	NOUN
ajst-31431	39	49	;	;	PUNCT
ajst-31431	39	50	third	third	ADV
ajst-31431	39	51	,	,	PUNCT
ajst-31431	39	52	the	the	DET
ajst-31431	39	53	trade	trade	NOUN
ajst-31431	39	54	-	-	PUNCT
ajst-31431	39	55	off	off	NOUN
ajst-31431	39	56	for	for	ADP
ajst-31431	39	57	data	datum	NOUN
ajst-31431	39	58	diversity	diversity	NOUN
ajst-31431	39	59	may	may	AUX
ajst-31431	39	60	filter	filter	VERB
ajst-31431	39	61	out	out	ADP
ajst-31431	39	62	valuable	valuable	ADJ
ajst-31431	39	63	features	feature	NOUN
ajst-31431	39	64	.	.	PUNCT
ajst-31431	40	1	these	these	DET
ajst-31431	40	2	shortcomings	shortcoming	NOUN
ajst-31431	40	3	highlight	highlight	VERB
ajst-31431	40	4	three	three	NUM
ajst-31431	40	5	critical	critical	ADJ
ajst-31431	40	6	directions	direction	NOUN
ajst-31431	40	7	for	for	ADP
ajst-31431	40	8	constructing	construct	VERB
ajst-31431	40	9	high	high	ADJ
ajst-31431	40	10	-	-	PUNCT
ajst-31431	40	11	quality	quality	NOUN
ajst-31431	40	12	datasets	dataset	NOUN
ajst-31431	40	13	:	:	PUNCT
ajst-31431	40	14	crossplatform	crossplatform	NOUN
ajst-31431	40	15	validation	validation	NOUN
ajst-31431	40	16	,	,	PUNCT
ajst-31431	40	17	culturally	culturally	ADV
ajst-31431	40	18	sensitive	sensitive	ADJ
ajst-31431	40	19	multilingual	multilingual	ADJ
ajst-31431	40	20	annotation	annotation	NOUN
ajst-31431	40	21	systems	system	NOUN
ajst-31431	40	22	,	,	PUNCT
ajst-31431	40	23	and	and	CCONJ
ajst-31431	40	24	adversarial	adversarial	ADJ
ajst-31431	40	25	sample	sample	NOUN
ajst-31431	40	26	augmentation	augmentation	NOUN
ajst-31431	40	27	mechanisms	mechanism	NOUN
ajst-31431	40	28	.	.	PUNCT
ajst-31431	41	1	2.2	2.2	NUM
ajst-31431	41	2	.	.	PUNCT
ajst-31431	42	1	hate	hate	NOUN
ajst-31431	42	2	memes	meme	NOUN
ajst-31431	42	3	dataset	dataset	VERB
ajst-31431	42	4	the	the	DET
ajst-31431	42	5	construction	construction	NOUN
ajst-31431	42	6	of	of	ADP
ajst-31431	42	7	datasets	dataset	NOUN
ajst-31431	42	8	for	for	ADP
ajst-31431	42	9	multimodal	multimodal	ADJ
ajst-31431	42	10	hateful	hateful	ADJ
ajst-31431	42	11	meme	meme	NOUN
ajst-31431	42	12	detection	detection	NOUN
ajst-31431	42	13	centers	center	NOUN
ajst-31431	42	14	on	on	ADP
ajst-31431	42	15	adversarial	adversarial	ADJ
ajst-31431	42	16	sample	sample	NOUN
ajst-31431	42	17	design	design	NOUN
ajst-31431	42	18	,	,	PUNCT
ajst-31431	42	19	cultural	cultural	ADJ
ajst-31431	42	20	context	context	NOUN
ajst-31431	42	21	sensitivity	sensitivity	NOUN
ajst-31431	42	22	,	,	PUNCT
ajst-31431	42	23	and	and	CCONJ
ajst-31431	42	24	multimodal	multimodal	ADJ
ajst-31431	42	25	fusion	fusion	NOUN
ajst-31431	42	26	challenges	challenge	NOUN
ajst-31431	42	27	.	.	PUNCT
ajst-31431	43	1	prominent	prominent	ADJ
ajst-31431	43	2	among	among	ADP
ajst-31431	43	3	these	these	PRON
ajst-31431	43	4	is	be	AUX
ajst-31431	43	5	the	the	DET
ajst-31431	43	6	hateful	hateful	ADJ
ajst-31431	43	7	memes	meme	NOUN
ajst-31431	43	8	challenge	challenge	VERB
ajst-31431	43	9	dataset	dataset	NOUN
ajst-31431	43	10	released	release	VERB
ajst-31431	43	11	by	by	ADP
ajst-31431	43	12	facebook	facebook	PROPN
ajst-31431	43	13	ai	ai	PROPN
ajst-31431	43	14	research	research	NOUN
ajst-31431	43	15	[	[	X
ajst-31431	43	16	10	10	NUM
ajst-31431	43	17	]	]	PUNCT
ajst-31431	43	18	,	,	PUNCT
ajst-31431	43	19	which	which	PRON
ajst-31431	43	20	compels	compel	VERB
ajst-31431	43	21	models	model	NOUN
ajst-31431	43	22	to	to	PART
ajst-31431	43	23	integrate	integrate	VERB
ajst-31431	43	24	text	text	NOUN
ajst-31431	43	25	-	-	PUNCT
ajst-31431	43	26	image	image	NOUN
ajst-31431	43	27	features	feature	NOUN
ajst-31431	43	28	by	by	ADP
ajst-31431	43	29	deliberately	deliberately	ADV
ajst-31431	43	30	isolating	isolate	VERB
ajst-31431	43	31	single	single	ADJ
ajst-31431	43	32	-	-	PUNCT
ajst-31431	43	33	modality	modality	NOUN
ajst-31431	43	34	information	information	NOUN
ajst-31431	43	35	,	,	PUNCT
ajst-31431	43	36	establishing	establish	VERB
ajst-31431	43	37	it	it	PRON
ajst-31431	43	38	as	as	ADP
ajst-31431	43	39	a	a	DET
ajst-31431	43	40	benchmark	benchmark	NOUN
ajst-31431	43	41	.	.	PUNCT
ajst-31431	44	1	semeval-2020	semeval-2020	NOUN
ajst-31431	44	2	task	task	NOUN
ajst-31431	44	3	8	8	NUM
ajst-31431	45	1	[	[	SYM
ajst-31431	45	2	11	11	NUM
ajst-31431	45	3	]	]	X
ajst-31431	45	4	advances	advance	NOUN
ajst-31431	45	5	model	model	NOUN
ajst-31431	45	6	understanding	understanding	NOUN
ajst-31431	45	7	of	of	ADP
ajst-31431	45	8	ironic	ironic	ADJ
ajst-31431	45	9	contexts	context	NOUN
ajst-31431	45	10	by	by	ADP
ajst-31431	45	11	distinguishing	distinguish	VERB
ajst-31431	45	12	explicit	explicit	ADJ
ajst-31431	45	13	hate	hate	NOUN
ajst-31431	45	14	,	,	PUNCT
ajst-31431	45	15	subtle	subtle	ADJ
ajst-31431	45	16	sarcasm	sarcasm	NOUN
ajst-31431	45	17	,	,	PUNCT
ajst-31431	45	18	and	and	CCONJ
ajst-31431	45	19	aggressive	aggressive	ADJ
ajst-31431	45	20	humor	humor	NOUN
ajst-31431	45	21	.	.	PUNCT
ajst-31431	46	1	the	the	DET
ajst-31431	46	2	recently	recently	ADV
ajst-31431	46	3	released	release	VERB
ajst-31431	46	4	mute	mute	NOUN
ajst-31431	46	5	-	-	PUNCT
ajst-31431	46	6	aacl22	aacl22	NOUN
ajst-31431	46	7	[	[	X
ajst-31431	46	8	12	12	NUM
ajst-31431	46	9	]	]	PUNCT
ajst-31431	46	10	further	far	ADV
ajst-31431	46	11	focuses	focus	VERB
ajst-31431	46	12	on	on	ADP
ajst-31431	46	13	identifying	identify	VERB
ajst-31431	46	14	multimodal	multimodal	ADJ
ajst-31431	46	15	insinuation	insinuation	NOUN
ajst-31431	46	16	and	and	CCONJ
ajst-31431	46	17	puns	pun	NOUN
ajst-31431	46	18	,	,	PUNCT
ajst-31431	46	19	filling	fill	VERB
ajst-31431	46	20	gaps	gap	NOUN
ajst-31431	46	21	in	in	ADP
ajst-31431	46	22	detecting	detect	VERB
ajst-31431	46	23	non	non	ADJ
ajst-31431	46	24	-	-	ADJ
ajst-31431	46	25	explicit	explicit	ADJ
ajst-31431	46	26	hate	hate	NOUN
ajst-31431	46	27	expressions	expression	NOUN
ajst-31431	46	28	through	through	ADP
ajst-31431	46	29	coupled	couple	VERB
ajst-31431	46	30	designs	design	NOUN
ajst-31431	46	31	involving	involve	VERB
ajst-31431	46	32	visual	visual	ADJ
ajst-31431	46	33	metaphors	metaphor	NOUN
ajst-31431	46	34	(	(	PUNCT
ajst-31431	46	35	e.g.	e.g.	ADV
ajst-31431	46	36	,	,	PUNCT
ajst-31431	46	37	repurposing	repurpose	VERB
ajst-31431	46	38	a	a	DET
ajst-31431	46	39	dove	dove	NOUN
ajst-31431	46	40	image	image	NOUN
ajst-31431	46	41	as	as	ADP
ajst-31431	46	42	a	a	DET
ajst-31431	46	43	gun	gun	NOUN
ajst-31431	46	44	)	)	PUNCT
ajst-31431	46	45	and	and	CCONJ
ajst-31431	46	46	cross	cross	ADJ
ajst-31431	46	47	-	-	ADJ
ajst-31431	46	48	cultural	cultural	ADJ
ajst-31431	46	49	contexts	contexts	NOUN
ajst-31431	46	50	.	.	PUNCT
ajst-31431	47	1	additionally	additionally	ADV
ajst-31431	47	2	,	,	PUNCT
ajst-31431	47	3	multioff	multioff	PROPN
ajst-31431	47	4	[	[	X
ajst-31431	47	5	13	13	NUM
ajst-31431	47	6	]	]	PUNCT
ajst-31431	47	7	extends	extend	VERB
ajst-31431	47	8	detection	detection	NOUN
ajst-31431	47	9	boundaries	boundary	NOUN
ajst-31431	47	10	by	by	ADP
ajst-31431	47	11	addressing	address	VERB
ajst-31431	47	12	cross	cross	ADJ
ajst-31431	47	13	-	-	ADJ
ajst-31431	47	14	cultural	cultural	ADJ
ajst-31431	47	15	metaphors	metaphor	NOUN
ajst-31431	47	16	(	(	PUNCT
ajst-31431	47	17	e.g.	e.g.	ADV
ajst-31431	47	18	,	,	PUNCT
ajst-31431	47	19	juxtaposing	juxtapose	VERB
ajst-31431	47	20	religious	religious	ADJ
ajst-31431	47	21	symbols	symbol	NOUN
ajst-31431	47	22	)	)	PUNCT
ajst-31431	47	23	and	and	CCONJ
ajst-31431	47	24	ocr	ocr	ADJ
ajst-31431	47	25	adversarial	adversarial	ADJ
ajst-31431	47	26	text	text	NOUN
ajst-31431	47	27	.	.	PUNCT
ajst-31431	48	1	current	current	ADJ
ajst-31431	48	2	dataset	dataset	VERB
ajst-31431	48	3	evolution	evolution	NOUN
ajst-31431	48	4	trends	trend	NOUN
ajst-31431	48	5	include	include	VERB
ajst-31431	48	6	:	:	PUNCT
ajst-31431	48	7	emphasizing	emphasize	VERB
ajst-31431	48	8	cultural	cultural	ADJ
ajst-31431	48	9	specificity	specificity	NOUN
ajst-31431	48	10	to	to	PART
ajst-31431	48	11	address	address	VERB
ajst-31431	48	12	localized	localized	ADJ
ajst-31431	48	13	hate	hate	NOUN
ajst-31431	48	14	symbols	symbol	NOUN
ajst-31431	48	15	,	,	PUNCT
ajst-31431	48	16	incorporating	incorporate	VERB
ajst-31431	48	17	multimodal	multimodal	ADJ
ajst-31431	48	18	pre	pre	ADJ
ajst-31431	48	19	-	-	ADJ
ajst-31431	48	20	trained	train	VERB
ajst-31431	48	21	models	model	NOUN
ajst-31431	48	22	(	(	PUNCT
ajst-31431	48	23	e.g.	e.g.	ADV
ajst-31431	48	24	,	,	PUNCT
ajst-31431	48	25	clip	clip	PROPN
ajst-31431	48	26	,	,	PUNCT
ajst-31431	48	27	vilt	vilt	NOUN
ajst-31431	48	28	)	)	PUNCT
ajst-31431	48	29	to	to	PART
ajst-31431	48	30	bridge	bridge	VERB
ajst-31431	48	31	the	the	DET
ajst-31431	48	32	text	text	NOUN
ajst-31431	48	33	-	-	PUNCT
ajst-31431	48	34	image	image	NOUN
ajst-31431	48	35	semantic	semantic	ADJ
ajst-31431	48	36	gap	gap	NOUN
ajst-31431	48	37	,	,	PUNCT
ajst-31431	48	38	and	and	CCONJ
ajst-31431	48	39	employing	employ	VERB
ajst-31431	48	40	adversarial	adversarial	ADJ
ajst-31431	48	41	sample	sample	NOUN
ajst-31431	48	42	generation	generation	NOUN
ajst-31431	48	43	techniques	technique	NOUN
ajst-31431	48	44	to	to	PART
ajst-31431	48	45	enhance	enhance	VERB
ajst-31431	48	46	model	model	NOUN
ajst-31431	48	47	robustness	robustness	NOUN
ajst-31431	48	48	.	.	PUNCT
ajst-31431	49	1	however	however	ADV
ajst-31431	49	2	,	,	PUNCT
ajst-31431	49	3	challenges	challenge	NOUN
ajst-31431	49	4	remain	remain	VERB
ajst-31431	49	5	,	,	PUNCT
ajst-31431	49	6	including	include	VERB
ajst-31431	49	7	metaphorical	metaphorical	ADJ
ajst-31431	49	8	ambiguity	ambiguity	NOUN
ajst-31431	49	9	,	,	PUNCT
ajst-31431	49	10	insufficient	insufficient	ADJ
ajst-31431	49	11	coverage	coverage	NOUN
ajst-31431	49	12	of	of	ADP
ajst-31431	49	13	minority	minority	NOUN
ajst-31431	49	14	languages	language	NOUN
ajst-31431	49	15	,	,	PUNCT
ajst-31431	49	16	and	and	CCONJ
ajst-31431	49	17	ethical	ethical	ADJ
ajst-31431	49	18	review	review	NOUN
ajst-31431	49	19	controversies	controversy	NOUN
ajst-31431	49	20	.	.	PUNCT
ajst-31431	50	1	table	table	NOUN
ajst-31431	50	2	1	1	NUM
ajst-31431	50	3	.	.	PUNCT
ajst-31431	50	4	data	datum	NOUN
ajst-31431	50	5	setstatistics	setstatistic	NOUN
ajst-31431	50	6	dataset	dataset	NOUN
ajst-31431	50	7	scale	scale	NOUN
ajst-31431	50	8	(	(	PUNCT
ajst-31431	50	9	number	number	NOUN
ajst-31431	50	10	of	of	ADP
ajst-31431	50	11	posts	post	NOUN
ajst-31431	50	12	)	)	PUNCT
ajst-31431	50	13	dataset	dataset	NOUN
ajst-31431	50	14	category	category	NOUN
ajst-31431	50	15	waseem	waseem	PROPN
ajst-31431	50	16	et	et	PROPN
ajst-31431	50	17	al	al	PROPN
ajst-31431	50	18	.	.	PROPN
ajst-31431	51	1	6,909	6,909	NUM
ajst-31431	51	2	posts	post	NOUN
ajst-31431	51	3	racism	racism	NOUN
ajst-31431	51	4	,	,	PUNCT
ajst-31431	51	5	gender	gender	NOUN
ajst-31431	51	6	discrimination	discrimination	NOUN
ajst-31431	51	7	,	,	PUNCT
ajst-31431	51	8	neither	neither	CCONJ
ajst-31431	51	9	,	,	PUNCT
ajst-31431	51	10	both	both	PRON
ajst-31431	51	11	olid	olid	ADJ
ajst-31431	51	12	semantic	semantic	ADJ
ajst-31431	51	13	evaluation	evaluation	NOUN
ajst-31431	51	14	task	task	NOUN
ajst-31431	51	15	6	6	NUM
ajst-31431	51	16	14,000	14,000	NUM
ajst-31431	51	17	posts	post	NOUN
ajst-31431	51	18	a	a	DET
ajst-31431	51	19	level	level	NOUN
ajst-31431	51	20	:	:	PUNCT
ajst-31431	51	21	criminality	criminality	NOUN
ajst-31431	51	22	,	,	PUNCT
ajst-31431	51	23	non	non	ADJ
ajst-31431	51	24	-	-	NOUN
ajst-31431	51	25	criminality	criminality	ADJ
ajst-31431	51	26	;	;	PUNCT
ajst-31431	51	27	b	b	X
ajst-31431	51	28	level	level	NOUN
ajst-31431	51	29	:	:	PUNCT
ajst-31431	51	30	targeted	targeted	ADJ
ajst-31431	51	31	insult	insult	NOUN
ajst-31431	51	32	,	,	PUNCT
ajst-31431	51	33	non	non	ADJ
ajst-31431	51	34	-	-	ADJ
ajst-31431	51	35	targeted	targeted	ADJ
ajst-31431	51	36	insult	insult	NOUN
ajst-31431	51	37	;	;	PUNCT
ajst-31431	51	38	c	c	NOUN
ajst-31431	51	39	level	level	NOUN
ajst-31431	51	40	:	:	PUNCT
ajst-31431	51	41	individual	individual	ADJ
ajst-31431	51	42	,	,	PUNCT
ajst-31431	51	43	group	group	NOUN
ajst-31431	51	44	,	,	PUNCT
ajst-31431	51	45	other	other	ADJ
ajst-31431	51	46	hateval	hateval	ADJ
ajst-31431	51	47	semantic	semantic	ADJ
ajst-31431	51	48	evaluation	evaluation	NOUN
ajst-31431	51	49	task	task	NOUN
ajst-31431	51	50	5	5	NUM
ajst-31431	51	51	english	english	PROPN
ajst-31431	51	52	:	:	PUNCT
ajst-31431	51	53	13,000	13,000	NUM
ajst-31431	51	54	posts	post	NOUN
ajst-31431	51	55	,	,	PUNCT
ajst-31431	51	56	spanish	spanish	ADJ
ajst-31431	51	57	:	:	PUNCT
ajst-31431	51	58	6,600	6,600	NUM
ajst-31431	51	59	posts	post	NOUN
ajst-31431	51	60	subtask	subtask	VERB
ajst-31431	51	61	a	a	DET
ajst-31431	51	62	:	:	PUNCT
ajst-31431	51	63	hate	hate	NOUN
ajst-31431	51	64	speech	speech	NOUN
ajst-31431	51	65	,	,	PUNCT
ajst-31431	51	66	non	non	ADJ
ajst-31431	51	67	-	-	ADJ
ajst-31431	51	68	hate	hate	ADJ
ajst-31431	51	69	speech	speech	NOUN
ajst-31431	51	70	;	;	PUNCT
ajst-31431	51	71	subtask	subtask	PROPN
ajst-31431	51	72	b	b	NOUN
ajst-31431	51	73	:	:	PUNCT
ajst-31431	51	74	individual	individual	ADJ
ajst-31431	51	75	target	target	NOUN
ajst-31431	51	76	,	,	PUNCT
ajst-31431	51	77	group	group	NOUN
ajst-31431	51	78	target	target	NOUN
ajst-31431	51	79	;	;	PUNCT
ajst-31431	51	80	subtask	subtask	NOUN
ajst-31431	51	81	c	c	NOUN
ajst-31431	51	82	:	:	PUNCT
ajst-31431	51	83	aggressive	aggressive	ADJ
ajst-31431	51	84	,	,	PUNCT
ajst-31431	51	85	non	non	ADJ
ajst-31431	51	86	-	-	ADJ
ajst-31431	51	87	aggressive	aggressive	ADJ
ajst-31431	51	88	hasoc	hasoc	NOUN
ajst-31431	51	89	:	:	PUNCT
ajst-31431	51	90	identifying	identify	VERB
ajst-31431	51	91	hate	hate	NOUN
ajst-31431	51	92	speech	speech	NOUN
ajst-31431	51	93	and	and	CCONJ
ajst-31431	51	94	offensiveness	offensiveness	NOUN
ajst-31431	51	95	in	in	ADP
ajst-31431	51	96	indic	indic	ADJ
ajst-31431	51	97	languages	language	NOUN
ajst-31431	51	98	hasoc	hasoc	VERB
ajst-31431	51	99	english	english	PROPN
ajst-31431	51	100	dataset	dataset	VERB
ajst-31431	51	101	2020	2020	NUM
ajst-31431	51	102	:	:	PUNCT
ajst-31431	51	103	5,335	5,335	NUM
ajst-31431	51	104	posts	post	NOUN
ajst-31431	51	105	,	,	PUNCT
ajst-31431	51	106	hasoc	hasoc	PROPN
ajst-31431	51	107	english	english	PROPN
ajst-31431	51	108	dataset	dataset	VERB
ajst-31431	51	109	2019	2019	NUM
ajst-31431	51	110	:	:	PUNCT
ajst-31431	51	111	7,005	7,005	NUM
ajst-31431	51	112	posts	post	NOUN
ajst-31431	51	113	subtask	subtask	VERB
ajst-31431	51	114	a	a	DET
ajst-31431	51	115	:	:	PUNCT
ajst-31431	51	116	hateful	hateful	ADJ
ajst-31431	51	117	but	but	CCONJ
ajst-31431	51	118	non	non	ADJ
ajst-31431	51	119	-	-	ADJ
ajst-31431	51	120	offensive	offensive	ADJ
ajst-31431	51	121	;	;	PUNCT
ajst-31431	51	122	subtask	subtask	PROPN
ajst-31431	51	123	b	b	NOUN
ajst-31431	51	124	:	:	PUNCT
ajst-31431	51	125	hate	hate	NOUN
ajst-31431	51	126	speech	speech	NOUN
ajst-31431	51	127	,	,	PUNCT
ajst-31431	51	128	offensive	offensive	ADJ
ajst-31431	51	129	speech	speech	NOUN
ajst-31431	51	130	,	,	PUNCT
ajst-31431	51	131	and	and	CCONJ
ajst-31431	51	132	profanity	profanity	PROPN
ajst-31431	51	133	elsherief	elsherief	PROPN
ajst-31431	51	134	et	et	PROPN
ajst-31431	51	135	al	al	PROPN
ajst-31431	51	136	.	.	PROPN
ajst-31431	51	137	25,278	25,278	NUM
ajst-31431	51	138	haters	hater	NOUN
ajst-31431	51	139	,	,	PUNCT
ajst-31431	51	140	22,857	22,857	NUM
ajst-31431	51	141	targets	target	NOUN
ajst-31431	51	142	,	,	PUNCT
ajst-31431	51	143	27,330	27,330	NUM
ajst-31431	51	144	posts	post	NOUN
ajst-31431	51	145	ancient	ancient	ADJ
ajst-31431	51	146	language	language	NOUN
ajst-31431	51	147	,	,	PUNCT
ajst-31431	51	148	class	class	NOUN
ajst-31431	51	149	,	,	PUNCT
ajst-31431	51	150	disability	disability	NOUN
ajst-31431	51	151	,	,	PUNCT
ajst-31431	51	152	race	race	NOUN
ajst-31431	51	153	,	,	PUNCT
ajst-31431	51	154	gender	gender	NOUN
ajst-31431	51	155	,	,	PUNCT
ajst-31431	51	156	nationality	nationality	NOUN
ajst-31431	51	157	,	,	PUNCT
ajst-31431	51	158	religion	religion	NOUN
ajst-31431	51	159	,	,	PUNCT
ajst-31431	51	160	sexual	sexual	ADJ
ajst-31431	51	161	orientation	orientation	NOUN
ajst-31431	51	162	,	,	PUNCT
ajst-31431	51	163	hate	hate	VERB
ajst-31431	51	164	motivation	motivation	NOUN
ajst-31431	51	165	,	,	PUNCT
ajst-31431	51	166	confrontational	confrontational	ADJ
ajst-31431	51	167	non	non	ADJ
ajst-31431	51	168	-	-	ADJ
ajst-31431	51	169	hate	hate	ADJ
ajst-31431	51	170	samples	sample	NOUN
ajst-31431	51	171	hateful	hateful	ADJ
ajst-31431	51	172	memes	meme	NOUN
ajst-31431	51	173	challenge	challenge	VERB
ajst-31431	51	174	dataset	dataset	VERB
ajst-31431	51	175	10,000	10,000	NUM
ajst-31431	51	176	image	image	NOUN
ajst-31431	51	177	-	-	PUNCT
ajst-31431	51	178	text	text	NOUN
ajst-31431	51	179	pairs	pair	NOUN
ajst-31431	51	180	hateful	hateful	ADJ
ajst-31431	51	181	memes	meme	NOUN
ajst-31431	51	182	,	,	PUNCT
ajst-31431	51	183	adversarial	adversarial	ADJ
ajst-31431	51	184	non	non	ADJ
ajst-31431	51	185	-	-	ADJ
ajst-31431	51	186	hateful	hateful	ADJ
ajst-31431	51	187	examples	example	NOUN
ajst-31431	51	188	semeval-2020	semeval-2020	NOUN
ajst-31431	51	189	task	task	NOUN
ajst-31431	51	190	8	8	NUM
ajst-31431	51	191	10,000	10,000	NUM
ajst-31431	51	192	labeled	label	VERB
ajst-31431	51	193	samples	sample	NOUN
ajst-31431	51	194	task	task	NOUN
ajst-31431	51	195	a	a	PRON
ajst-31431	51	196	:	:	PUNCT
ajst-31431	51	197	emotion	emotion	NOUN
ajst-31431	51	198	classification	classification	NOUN
ajst-31431	51	199	;	;	PUNCT
ajst-31431	51	200	task	task	PROPN
ajst-31431	51	201	b	b	PROPN
ajst-31431	51	202	:	:	PUNCT
ajst-31431	51	203	humor	humor	NOUN
ajst-31431	51	204	type	type	NOUN
ajst-31431	51	205	recognition	recognition	NOUN
ajst-31431	51	206	;	;	PUNCT
ajst-31431	51	207	task	task	NOUN
ajst-31431	51	208	c	c	NOUN
ajst-31431	51	209	:	:	PUNCT
ajst-31431	51	210	semantic	semantic	ADJ
ajst-31431	51	211	strength	strength	NOUN
ajst-31431	51	212	quantification	quantification	NOUN
ajst-31431	51	213	mute	mute	ADJ
ajst-31431	51	214	-	-	PUNCT
ajst-31431	51	215	aacl22	aacl22	NOUN
ajst-31431	51	216	4,158	4,158	NUM
ajst-31431	51	217	multimodal	multimodal	NOUN
ajst-31431	51	218	memes	meme	NOUN
ajst-31431	51	219	hate	hate	VERB
ajst-31431	51	220	motivation	motivation	NOUN
ajst-31431	51	221	,	,	PUNCT
ajst-31431	51	222	non	non	ADJ
ajst-31431	51	223	-	-	ADJ
ajst-31431	51	224	hate	hate	ADJ
ajst-31431	51	225	motivation	motivation	NOUN
ajst-31431	51	226	multioff	multioff	VERB
ajst-31431	51	227	1,198	1,198	NUM
ajst-31431	51	228	multimodal	multimodal	NOUN
ajst-31431	51	229	memes	meme	NOUN
ajst-31431	51	230	,	,	PUNCT
ajst-31431	51	231	hate	hate	VERB
ajst-31431	51	232	motivation	motivation	NOUN
ajst-31431	51	233	,	,	PUNCT
ajst-31431	51	234	non	non	ADJ
ajst-31431	51	235	-	-	ADJ
ajst-31431	51	236	hate	hate	ADJ
ajst-31431	51	237	motivation	motivation	NOUN
ajst-31431	51	238	3	3	X
ajst-31431	51	239	.	.	X
ajst-31431	51	240	text	text	NOUN
ajst-31431	51	241	hate	hate	NOUN
ajst-31431	51	242	speech	speech	NOUN
ajst-31431	51	243	detection	detection	NOUN
ajst-31431	51	244	3.1	3.1	NUM
ajst-31431	51	245	.	.	PUNCT
ajst-31431	52	1	traditional	traditional	ADJ
ajst-31431	52	2	machine	machine	NOUN
ajst-31431	52	3	learning	learn	VERB
ajst-31431	52	4	for	for	ADP
ajst-31431	52	5	detecting	detect	VERB
ajst-31431	52	6	hate	hate	NOUN
ajst-31431	52	7	speech	speech	NOUN
ajst-31431	52	8	the	the	DET
ajst-31431	52	9	anonymity	anonymity	NOUN
ajst-31431	52	10	of	of	ADP
ajst-31431	52	11	social	social	ADJ
ajst-31431	52	12	networks	network	NOUN
ajst-31431	52	13	fosters	foster	VERB
ajst-31431	52	14	hate	hate	NOUN
ajst-31431	52	15	speech	speech	NOUN
ajst-31431	52	16	,	,	PUNCT
ajst-31431	52	17	which	which	PRON
ajst-31431	52	18	not	not	PART
ajst-31431	52	19	only	only	ADV
ajst-31431	52	20	exacerbates	exacerbate	VERB
ajst-31431	52	21	group	group	NOUN
ajst-31431	52	22	conflicts	conflict	NOUN
ajst-31431	52	23	,	,	PUNCT
ajst-31431	52	24	threatens	threaten	VERB
ajst-31431	52	25	public	public	ADJ
ajst-31431	52	26	mental	mental	ADJ
ajst-31431	52	27	health	health	NOUN
ajst-31431	52	28	,	,	PUNCT
ajst-31431	52	29	but	but	CCONJ
ajst-31431	52	30	also	also	ADV
ajst-31431	52	31	potentially	potentially	ADV
ajst-31431	52	32	exerts	exert	VERB
ajst-31431	52	33	negative	negative	ADJ
ajst-31431	52	34	societal	societal	ADJ
ajst-31431	52	35	impacts	impact	NOUN
ajst-31431	52	36	[	[	X
ajst-31431	52	37	14	14	NUM
ajst-31431	52	38	]	]	PUNCT
ajst-31431	52	39	.	.	PUNCT
ajst-31431	53	1	with	with	ADP
ajst-31431	53	2	the	the	DET
ajst-31431	53	3	explosive	explosive	ADJ
ajst-31431	53	4	growth	growth	NOUN
ajst-31431	53	5	of	of	ADP
ajst-31431	53	6	social	social	ADJ
ajst-31431	53	7	media	medium	NOUN
ajst-31431	53	8	data	datum	NOUN
ajst-31431	53	9	,	,	PUNCT
ajst-31431	53	10	detecting	detect	VERB
ajst-31431	53	11	hate	hate	NOUN
ajst-31431	53	12	speech	speech	NOUN
ajst-31431	53	13	has	have	AUX
ajst-31431	53	14	become	become	VERB
ajst-31431	53	15	increasingly	increasingly	ADV
ajst-31431	53	16	critical	critical	ADJ
ajst-31431	53	17	.	.	PUNCT
ajst-31431	54	1	over	over	ADP
ajst-31431	54	2	the	the	DET
ajst-31431	54	3	past	past	ADJ
ajst-31431	54	4	few	few	ADJ
ajst-31431	54	5	decades	decade	NOUN
ajst-31431	54	6	,	,	PUNCT
ajst-31431	54	7	hate	hate	VERB
ajst-31431	54	8	speech	speech	NOUN
ajst-31431	54	9	and	and	CCONJ
ajst-31431	54	10	online	online	ADJ
ajst-31431	54	11	bullying	bullying	NOUN
ajst-31431	54	12	have	have	AUX
ajst-31431	54	13	been	be	AUX
ajst-31431	54	14	among	among	ADP
ajst-31431	54	15	the	the	DET
ajst-31431	54	16	most	most	ADV
ajst-31431	54	17	extensively	extensively	ADV
ajst-31431	54	18	researched	research	VERB
ajst-31431	54	19	areas	area	NOUN
ajst-31431	54	20	within	within	ADP
ajst-31431	54	21	natural	natural	ADJ
ajst-31431	54	22	language	language	NOUN
ajst-31431	54	23	processing	processing	NOUN
ajst-31431	54	24	[	[	X
ajst-31431	54	25	15	15	NUM
ajst-31431	54	26	]	]	PUNCT
ajst-31431	54	27	.	.	PUNCT
ajst-31431	55	1	machine	machine	NOUN
ajst-31431	55	2	learning	learn	VERB
ajst-31431	55	3	algorithms	algorithm	NOUN
ajst-31431	55	4	have	have	AUX
ajst-31431	55	5	made	make	VERB
ajst-31431	55	6	significant	significant	ADJ
ajst-31431	55	7	contributions	contribution	NOUN
ajst-31431	55	8	to	to	PART
ajst-31431	55	9	hate	hate	VERB
ajst-31431	55	10	speech	speech	NOUN
ajst-31431	55	11	detection	detection	NOUN
ajst-31431	55	12	and	and	CCONJ
ajst-31431	55	13	social	social	ADJ
ajst-31431	55	14	media	medium	NOUN
ajst-31431	55	15	content	content	NOUN
ajst-31431	55	16	analysis	analysis	NOUN
ajst-31431	55	17	[	[	X
ajst-31431	55	18	16	16	NUM
ajst-31431	55	19	]	]	PUNCT
ajst-31431	55	20	.	.	PUNCT
ajst-31431	56	1	in	in	ADP
ajst-31431	56	2	the	the	DET
ajst-31431	56	3	realm	realm	NOUN
ajst-31431	56	4	of	of	ADP
ajst-31431	56	5	social	social	ADJ
ajst-31431	56	6	media	medium	NOUN
ajst-31431	56	7	data	datum	NOUN
ajst-31431	56	8	analysis	analysis	NOUN
ajst-31431	56	9	,	,	PUNCT
ajst-31431	56	10	these	these	DET
ajst-31431	56	11	algorithms	algorithm	NOUN
ajst-31431	56	12	play	play	VERB
ajst-31431	56	13	a	a	DET
ajst-31431	56	14	crucial	crucial	ADJ
ajst-31431	56	15	role	role	NOUN
ajst-31431	56	16	in	in	ADP
ajst-31431	56	17	identifying	identify	VERB
ajst-31431	56	18	and	and	CCONJ
ajst-31431	56	19	classifying	classify	VERB
ajst-31431	56	20	offensive	offensive	ADJ
ajst-31431	56	21	comments	comment	NOUN
ajst-31431	56	22	[	[	X
ajst-31431	56	23	17	17	NUM
ajst-31431	56	24	]	]	PUNCT
ajst-31431	56	25	.	.	PUNCT
ajst-31431	57	1	traditional	traditional	ADJ
ajst-31431	57	2	machine	machine	NOUN
ajst-31431	57	3	learning	learn	VERB
ajst-31431	57	4	methods	method	NOUN
ajst-31431	57	5	primarily	primarily	ADV
ajst-31431	57	6	rely	rely	VERB
ajst-31431	57	7	on	on	ADP
ajst-31431	57	8	supervised	supervised	ADJ
ajst-31431	57	9	learning	learning	NOUN
ajst-31431	57	10	,	,	PUNCT
ajst-31431	57	11	dependent	dependent	ADJ
ajst-31431	57	12	on	on	ADP
ajst-31431	57	13	manually	manually	ADV
ajst-31431	57	14	annotated	annotate	VERB
ajst-31431	57	15	hate	hate	NOUN
ajst-31431	57	16	speech	speech	NOUN
ajst-31431	57	17	datasets	dataset	NOUN
ajst-31431	57	18	.	.	PUNCT
ajst-31431	58	1	they	they	PRON
ajst-31431	58	2	employ	employ	VERB
ajst-31431	58	3	classical	classical	ADJ
ajst-31431	58	4	algorithms	algorithm	NOUN
ajst-31431	58	5	such	such	ADJ
ajst-31431	58	6	as	as	ADP
ajst-31431	58	7	naive	naive	ADJ
ajst-31431	58	8	bayes	bayes	NOUN
ajst-31431	58	9	(	(	PUNCT
ajst-31431	58	10	nb	nb	NOUN
ajst-31431	58	11	)	)	PUNCT
ajst-31431	59	1	[	[	X
ajst-31431	59	2	18	18	NUM
ajst-31431	59	3	-	-	SYM
ajst-31431	59	4	20	20	NUM
ajst-31431	59	5	]	]	PUNCT
ajst-31431	59	6	,	,	PUNCT
ajst-31431	59	7	decision	decision	NOUN
ajst-31431	59	8	trees	tree	NOUN
ajst-31431	59	9	(	(	PUNCT
ajst-31431	59	10	dt	dt	PROPN
ajst-31431	59	11	)	)	PUNCT
ajst-31431	59	12	,	,	PUNCT
ajst-31431	59	13	support	support	VERB
ajst-31431	59	14	vector	vector	NOUN
ajst-31431	59	15	machines	machine	NOUN
ajst-31431	59	16	(	(	PUNCT
ajst-31431	59	17	svm	svm	PROPN
ajst-31431	59	18	)	)	PUNCT
ajst-31431	59	19	,	,	PUNCT
ajst-31431	59	20	linear	linear	ADJ
ajst-31431	59	21	regression	regression	NOUN
ajst-31431	59	22	[	[	X
ajst-31431	59	23	21	21	NUM
ajst-31431	59	24	]	]	PUNCT
ajst-31431	59	25	,	,	PUNCT
ajst-31431	59	26	and	and	CCONJ
ajst-31431	59	27	logistic	logistic	ADJ
ajst-31431	59	28	regression	regression	NOUN
ajst-31431	59	29	(	(	PUNCT
ajst-31431	59	30	lr	lr	NOUN
ajst-31431	59	31	)	)	PUNCT
ajst-31431	60	1	[	[	X
ajst-31431	60	2	22	22	NUM
ajst-31431	60	3	]	]	PUNCT
ajst-31431	60	4	,	,	PUNCT
ajst-31431	60	5	often	often	ADV
ajst-31431	60	6	combined	combine	VERB
ajst-31431	60	7	with	with	ADP
ajst-31431	60	8	statistical	statistical	ADJ
ajst-31431	60	9	feature	feature	NOUN
ajst-31431	60	10	engineering	engineering	NOUN
ajst-31431	60	11	techniques	technique	NOUN
ajst-31431	60	12	like	like	ADP
ajst-31431	60	13	term	term	NOUN
ajst-31431	60	14	frequency	frequency	NOUN
ajst-31431	60	15	-	-	PUNCT
ajst-31431	60	16	inverse	inverse	NOUN
ajst-31431	60	17	document	document	NOUN
ajst-31431	60	18	frequency	frequency	NOUN
ajst-31431	60	19	(	(	PUNCT
ajst-31431	60	20	tf	tf	PROPN
ajst-31431	60	21	-	-	PUNCT
ajst-31431	60	22	idf	idf	NOUN
ajst-31431	60	23	)	)	PUNCT
ajst-31431	60	24	and	and	CCONJ
ajst-31431	60	25	ngrams	ngram	NOUN
ajst-31431	60	26	,	,	PUNCT
ajst-31431	60	27	to	to	PART
ajst-31431	60	28	perform	perform	VERB
ajst-31431	60	29	shallow	shallow	ADJ
ajst-31431	60	30	semantic	semantic	ADJ
ajst-31431	60	31	analysis	analysis	NOUN
ajst-31431	60	32	of	of	ADP
ajst-31431	60	33	text	text	NOUN
ajst-31431	60	34	.	.	PUNCT
ajst-31431	61	1	however	however	ADV
ajst-31431	61	2	,	,	PUNCT
ajst-31431	61	3	these	these	DET
ajst-31431	61	4	approaches	approach	NOUN
ajst-31431	61	5	face	face	VERB
ajst-31431	61	6	dual	dual	ADJ
ajst-31431	61	7	challenges	challenge	NOUN
ajst-31431	61	8	:	:	PUNCT
ajst-31431	61	9	firstly	firstly	ADV
ajst-31431	61	10	,	,	PUNCT
ajst-31431	61	11	the	the	DET
ajst-31431	61	12	annotation	annotation	NOUN
ajst-31431	61	13	process	process	NOUN
ajst-31431	61	14	requires	require	VERB
ajst-31431	61	15	linguistics	linguistic	NOUN
ajst-31431	61	16	experts	expert	NOUN
ajst-31431	61	17	to	to	PART
ajst-31431	61	18	individually	individually	ADV
ajst-31431	61	19	assess	assess	VERB
ajst-31431	61	20	hate	hate	NOUN
ajst-31431	61	21	intent	intent	NOUN
ajst-31431	61	22	,	,	PUNCT
ajst-31431	61	23	proving	prove	VERB
ajst-31431	61	24	costly	costly	ADJ
ajst-31431	61	25	and	and	CCONJ
ajst-31431	61	26	susceptible	susceptible	ADJ
ajst-31431	61	27	to	to	ADP
ajst-31431	61	28	subjective	subjective	ADJ
ajst-31431	61	29	bias	bias	NOUN
ajst-31431	61	30	;	;	PUNCT
ajst-31431	61	31	secondly	secondly	ADV
ajst-31431	61	32	,	,	PUNCT
ajst-31431	61	33	the	the	DET
ajst-31431	61	34	models	model	NOUN
ajst-31431	61	35	heavily	heavily	ADV
ajst-31431	61	36	rely	rely	VERB
ajst-31431	61	37	on	on	ADP
ajst-31431	61	38	domain	domain	NOUN
ajst-31431	61	39	-	-	PUNCT
ajst-31431	61	40	specific	specific	ADJ
ajst-31431	61	41	42	42	NUM
ajst-31431	61	42	annotated	annotate	VERB
ajst-31431	61	43	data	datum	NOUN
ajst-31431	61	44	,	,	PUNCT
ajst-31431	61	45	leading	lead	VERB
ajst-31431	61	46	to	to	ADP
ajst-31431	61	47	a	a	DET
ajst-31431	61	48	sharp	sharp	ADJ
ajst-31431	61	49	decline	decline	NOUN
ajst-31431	61	50	in	in	ADP
ajst-31431	61	51	generalization	generalization	NOUN
ajst-31431	61	52	performance	performance	NOUN
ajst-31431	61	53	when	when	SCONJ
ajst-31431	61	54	encountering	encounter	VERB
ajst-31431	61	55	emerging	emerge	VERB
ajst-31431	61	56	online	online	ADJ
ajst-31431	61	57	slang	slang	PROPN
ajst-31431	61	58	(	(	PUNCT
ajst-31431	61	59	e.g.	e.g.	ADV
ajst-31431	61	60	,	,	PUNCT
ajst-31431	61	61	abbreviations	abbreviation	NOUN
ajst-31431	61	62	,	,	PUNCT
ajst-31431	61	63	puns	pun	NOUN
ajst-31431	61	64	)	)	PUNCT
ajst-31431	61	65	or	or	CCONJ
ajst-31431	61	66	cross	cross	ADJ
ajst-31431	61	67	-	-	ADJ
ajst-31431	61	68	cultural	cultural	ADJ
ajst-31431	61	69	linguistic	linguistic	ADJ
ajst-31431	61	70	variations	variation	NOUN
ajst-31431	61	71	.	.	PUNCT
ajst-31431	62	1	to	to	PART
ajst-31431	62	2	address	address	VERB
ajst-31431	62	3	this	this	DET
ajst-31431	62	4	limitation	limitation	NOUN
ajst-31431	62	5	,	,	PUNCT
ajst-31431	62	6	researchers	researcher	NOUN
ajst-31431	62	7	have	have	AUX
ajst-31431	62	8	shifted	shift	VERB
ajst-31431	62	9	towards	towards	ADP
ajst-31431	62	10	semisupervised	semisupervise	VERB
ajst-31431	62	11	and	and	CCONJ
ajst-31431	62	12	unsupervised	unsupervised	ADJ
ajst-31431	62	13	learning	learning	NOUN
ajst-31431	62	14	paradigms	paradigm	NOUN
ajst-31431	62	15	.	.	PUNCT
ajst-31431	63	1	for	for	ADP
ajst-31431	63	2	instance	instance	NOUN
ajst-31431	63	3	,	,	PUNCT
ajst-31431	63	4	gitari	gitari	PROPN
ajst-31431	63	5	et	et	PROPN
ajst-31431	63	6	al	al	PROPN
ajst-31431	63	7	.	.	PUNCT
ajst-31431	64	1	[	[	X
ajst-31431	64	2	23	23	NUM
ajst-31431	64	3	]	]	PUNCT
ajst-31431	64	4	employed	employ	VERB
ajst-31431	64	5	a	a	DET
ajst-31431	64	6	bootstrapping	bootstrappe	VERB
ajst-31431	64	7	method	method	NOUN
ajst-31431	64	8	to	to	PART
ajst-31431	64	9	construct	construct	VERB
ajst-31431	64	10	a	a	DET
ajst-31431	64	11	lexicon	lexicon	NOUN
ajst-31431	64	12	from	from	ADP
ajst-31431	64	13	hate	hate	NOUN
ajst-31431	64	14	verbs	verb	NOUN
ajst-31431	64	15	,	,	PUNCT
ajst-31431	64	16	iteratively	iteratively	ADV
ajst-31431	64	17	expanding	expand	VERB
ajst-31431	64	18	it	it	PRON
ajst-31431	64	19	,	,	PUNCT
ajst-31431	64	20	and	and	CCONJ
ajst-31431	64	21	achieved	achieve	VERB
ajst-31431	64	22	optimal	optimal	ADJ
ajst-31431	64	23	results	result	NOUN
ajst-31431	64	24	by	by	ADP
ajst-31431	64	25	incorporating	incorporate	VERB
ajst-31431	64	26	more	more	ADJ
ajst-31431	64	27	features	feature	NOUN
ajst-31431	64	28	.	.	PUNCT
ajst-31431	65	1	to	to	PART
ajst-31431	65	2	overcome	overcome	VERB
ajst-31431	65	3	the	the	DET
ajst-31431	65	4	performance	performance	NOUN
ajst-31431	65	5	limitations	limitation	NOUN
ajst-31431	65	6	of	of	ADP
ajst-31431	65	7	single	single	ADJ
ajst-31431	65	8	models	model	NOUN
ajst-31431	65	9	,	,	PUNCT
ajst-31431	65	10	ensemble	ensemble	ADJ
ajst-31431	65	11	learning	learning	NOUN
ajst-31431	65	12	methods	method	NOUN
ajst-31431	65	13	enhance	enhance	VERB
ajst-31431	65	14	detection	detection	NOUN
ajst-31431	65	15	robustness	robustness	NOUN
ajst-31431	65	16	through	through	ADP
ajst-31431	65	17	algorithmic	algorithmic	ADJ
ajst-31431	65	18	fusion	fusion	NOUN
ajst-31431	65	19	[	[	X
ajst-31431	65	20	24	24	NUM
ajst-31431	65	21	,	,	PUNCT
ajst-31431	65	22	25	25	NUM
ajst-31431	65	23	]	]	PUNCT
ajst-31431	65	24	.	.	PUNCT
ajst-31431	66	1	ensemble	ensemble	ADJ
ajst-31431	66	2	techniques	technique	NOUN
ajst-31431	66	3	are	be	AUX
ajst-31431	66	4	developed	develop	VERB
ajst-31431	66	5	to	to	PART
ajst-31431	66	6	overcome	overcome	VERB
ajst-31431	66	7	the	the	DET
ajst-31431	66	8	limitations	limitation	NOUN
ajst-31431	66	9	of	of	ADP
ajst-31431	66	10	several	several	ADJ
ajst-31431	66	11	individual	individual	ADJ
ajst-31431	66	12	machine	machine	NOUN
ajst-31431	66	13	learning	learn	VERB
ajst-31431	66	14	algorithms	algorithm	NOUN
ajst-31431	66	15	while	while	SCONJ
ajst-31431	66	16	enhancing	enhance	VERB
ajst-31431	66	17	their	their	PRON
ajst-31431	66	18	strengths	strength	NOUN
ajst-31431	66	19	.	.	PUNCT
ajst-31431	67	1	since	since	SCONJ
ajst-31431	67	2	each	each	DET
ajst-31431	67	3	model	model	NOUN
ajst-31431	67	4	possesses	possess	VERB
ajst-31431	67	5	inherent	inherent	ADJ
ajst-31431	67	6	weaknesses	weakness	NOUN
ajst-31431	67	7	and	and	CCONJ
ajst-31431	67	8	none	none	NOUN
ajst-31431	67	9	is	be	AUX
ajst-31431	67	10	perfect	perfect	ADJ
ajst-31431	67	11	,	,	PUNCT
ajst-31431	67	12	ensemble	ensemble	ADJ
ajst-31431	67	13	methods	method	NOUN
ajst-31431	67	14	attempt	attempt	VERB
ajst-31431	67	15	to	to	PART
ajst-31431	67	16	combine	combine	VERB
ajst-31431	67	17	the	the	DET
ajst-31431	67	18	advantages	advantage	NOUN
ajst-31431	67	19	of	of	ADP
ajst-31431	67	20	multiple	multiple	ADJ
ajst-31431	67	21	models	model	NOUN
ajst-31431	67	22	to	to	PART
ajst-31431	67	23	achieve	achieve	VERB
ajst-31431	67	24	better	well	ADJ
ajst-31431	67	25	performance	performance	NOUN
ajst-31431	67	26	than	than	ADP
ajst-31431	67	27	any	any	DET
ajst-31431	67	28	single	single	ADJ
ajst-31431	67	29	model	model	NOUN
ajst-31431	67	30	.	.	PUNCT
ajst-31431	68	1	statistically	statistically	ADV
ajst-31431	68	2	,	,	PUNCT
ajst-31431	68	3	combining	combine	VERB
ajst-31431	68	4	two	two	NUM
ajst-31431	68	5	or	or	CCONJ
ajst-31431	68	6	more	more	ADJ
ajst-31431	68	7	machine	machine	NOUN
ajst-31431	68	8	learning	learn	VERB
ajst-31431	68	9	algorithms	algorithm	NOUN
ajst-31431	68	10	can	can	AUX
ajst-31431	68	11	significantly	significantly	ADV
ajst-31431	68	12	reduce	reduce	VERB
ajst-31431	68	13	variance	variance	NOUN
ajst-31431	68	14	and	and	CCONJ
ajst-31431	68	15	improve	improve	VERB
ajst-31431	68	16	learning	learn	VERB
ajst-31431	68	17	capacity	capacity	NOUN
ajst-31431	68	18	.	.	PUNCT
ajst-31431	69	1	bagging	bag	VERB
ajst-31431	69	2	methodology	methodology	NOUN
ajst-31431	69	3	,	,	PUNCT
ajst-31431	69	4	random	random	ADJ
ajst-31431	69	5	forest	forest	NOUN
ajst-31431	69	6	(	(	PUNCT
ajst-31431	69	7	rf	rf	NOUN
ajst-31431	69	8	)	)	PUNCT
ajst-31431	69	9	,	,	PUNCT
ajst-31431	69	10	and	and	CCONJ
ajst-31431	69	11	boosting	boost	VERB
ajst-31431	69	12	methods	method	NOUN
ajst-31431	69	13	[	[	X
ajst-31431	69	14	26	26	NUM
ajst-31431	69	15	]	]	PUNCT
ajst-31431	69	16	are	be	AUX
ajst-31431	69	17	some	some	DET
ajst-31431	69	18	examples	example	NOUN
ajst-31431	69	19	of	of	ADP
ajst-31431	69	20	such	such	ADJ
ajst-31431	69	21	ensemble	ensemble	ADJ
ajst-31431	69	22	techniques	technique	NOUN
ajst-31431	69	23	.	.	PUNCT
ajst-31431	70	1	the	the	DET
ajst-31431	70	2	revolutionary	revolutionary	ADJ
ajst-31431	70	3	breakthrough	breakthrough	NOUN
ajst-31431	70	4	of	of	ADP
ajst-31431	70	5	word	word	NOUN
ajst-31431	70	6	embedding	embed	VERB
ajst-31431	70	7	technology	technology	NOUN
ajst-31431	70	8	has	have	AUX
ajst-31431	70	9	introduced	introduce	VERB
ajst-31431	70	10	new	new	ADJ
ajst-31431	70	11	dimensions	dimension	NOUN
ajst-31431	70	12	to	to	ADP
ajst-31431	70	13	semantic	semantic	ADJ
ajst-31431	70	14	understanding	understanding	NOUN
ajst-31431	70	15	.	.	PUNCT
ajst-31431	71	1	embeddings	embedding	NOUN
ajst-31431	71	2	learn	learn	VERB
ajst-31431	71	3	vector	vector	NOUN
ajst-31431	71	4	representations	representation	NOUN
ajst-31431	71	5	from	from	ADP
ajst-31431	71	6	discrete	discrete	ADJ
ajst-31431	71	7	inputs	input	NOUN
ajst-31431	71	8	,	,	PUNCT
ajst-31431	71	9	which	which	PRON
ajst-31431	71	10	are	be	AUX
ajst-31431	71	11	then	then	ADV
ajst-31431	71	12	utilized	utilize	VERB
ajst-31431	71	13	in	in	ADP
ajst-31431	71	14	downstream	downstream	ADJ
ajst-31431	71	15	text	text	NOUN
ajst-31431	71	16	mining	mining	NOUN
ajst-31431	71	17	tasks	task	NOUN
ajst-31431	71	18	,	,	PUNCT
ajst-31431	71	19	enabling	enable	VERB
ajst-31431	71	20	semantically	semantically	ADV
ajst-31431	71	21	related	relate	VERB
ajst-31431	71	22	phrases	phrase	NOUN
ajst-31431	71	23	to	to	PART
ajst-31431	71	24	share	share	VERB
ajst-31431	71	25	similar	similar	ADJ
ajst-31431	71	26	vector	vector	NOUN
ajst-31431	71	27	representations	representation	NOUN
ajst-31431	71	28	.	.	PUNCT
ajst-31431	72	1	over	over	ADP
ajst-31431	72	2	the	the	DET
ajst-31431	72	3	years	year	NOUN
ajst-31431	72	4	,	,	PUNCT
ajst-31431	72	5	numerous	numerous	ADJ
ajst-31431	72	6	word	word	NOUN
ajst-31431	72	7	embedding	embed	VERB
ajst-31431	72	8	algorithms	algorithm	NOUN
ajst-31431	72	9	have	have	AUX
ajst-31431	72	10	been	be	AUX
ajst-31431	72	11	developed	develop	VERB
ajst-31431	72	12	,	,	PUNCT
ajst-31431	72	13	including	include	VERB
ajst-31431	72	14	global	global	ADJ
ajst-31431	72	15	vectors	vector	NOUN
ajst-31431	72	16	for	for	ADP
ajst-31431	72	17	word	word	NOUN
ajst-31431	72	18	representation	representation	NOUN
ajst-31431	72	19	(	(	PUNCT
ajst-31431	72	20	glove	glove	NOUN
ajst-31431	72	21	)	)	PUNCT
ajst-31431	72	22	,	,	PUNCT
ajst-31431	72	23	word2vec	word2vec	X
ajst-31431	72	24	,	,	PUNCT
ajst-31431	72	25	and	and	CCONJ
ajst-31431	72	26	fasttext	fasttext	NOUN
ajst-31431	72	27	[	[	X
ajst-31431	72	28	27	27	NUM
ajst-31431	72	29	]	]	PUNCT
ajst-31431	72	30	.	.	PUNCT
ajst-31431	73	1	the	the	DET
ajst-31431	73	2	representations	representation	NOUN
ajst-31431	73	3	obtained	obtain	VERB
ajst-31431	73	4	through	through	ADP
ajst-31431	73	5	word	word	NOUN
ajst-31431	73	6	embedding	embed	VERB
ajst-31431	73	7	techniques	technique	NOUN
ajst-31431	73	8	are	be	AUX
ajst-31431	73	9	subsequently	subsequently	ADV
ajst-31431	73	10	input	input	ADJ
ajst-31431	73	11	into	into	ADP
ajst-31431	73	12	various	various	ADJ
ajst-31431	73	13	classifiers	classifier	NOUN
ajst-31431	73	14	.	.	PUNCT
ajst-31431	74	1	3.2	3.2	NUM
ajst-31431	74	2	.	.	PUNCT
ajst-31431	75	1	deep	deep	ADJ
ajst-31431	75	2	learning	learning	NOUN
ajst-31431	75	3	hate	hate	VERB
ajst-31431	75	4	speech	speech	NOUN
ajst-31431	75	5	detection	detection	NOUN
ajst-31431	75	6	deep	deep	ADJ
ajst-31431	75	7	learning	learning	NOUN
ajst-31431	75	8	utilizes	utilize	VERB
ajst-31431	75	9	multi	multi	ADJ
ajst-31431	75	10	-	-	ADJ
ajst-31431	75	11	layered	layered	ADJ
ajst-31431	75	12	neural	neural	ADJ
ajst-31431	75	13	network	network	NOUN
ajst-31431	75	14	architectures	architecture	NOUN
ajst-31431	75	15	to	to	PART
ajst-31431	75	16	automatically	automatically	ADV
ajst-31431	75	17	extract	extract	VERB
ajst-31431	75	18	deep	deep	ADJ
ajst-31431	75	19	semantic	semantic	ADJ
ajst-31431	75	20	features	feature	NOUN
ajst-31431	75	21	from	from	ADP
ajst-31431	75	22	text	text	NOUN
ajst-31431	75	23	,	,	PUNCT
ajst-31431	75	24	contrasting	contrast	VERB
ajst-31431	75	25	with	with	ADP
ajst-31431	75	26	traditional	traditional	ADJ
ajst-31431	75	27	machine	machine	NOUN
ajst-31431	75	28	learning	learning	NOUN
ajst-31431	75	29	methods	method	NOUN
ajst-31431	75	30	which	which	PRON
ajst-31431	75	31	often	often	ADV
ajst-31431	75	32	rely	rely	VERB
ajst-31431	75	33	on	on	ADP
ajst-31431	75	34	handcrafted	handcrafted	ADJ
ajst-31431	75	35	features	feature	NOUN
ajst-31431	75	36	(	(	PUNCT
ajst-31431	75	37	e.g.	e.g.	ADV
ajst-31431	75	38	,	,	PUNCT
ajst-31431	75	39	tfidf	tfidf	NOUN
ajst-31431	75	40	)	)	PUNCT
ajst-31431	75	41	or	or	CCONJ
ajst-31431	75	42	rule	rule	NOUN
ajst-31431	75	43	-	-	PUNCT
ajst-31431	75	44	based	base	VERB
ajst-31431	75	45	techniques	technique	NOUN
ajst-31431	75	46	(	(	PUNCT
ajst-31431	75	47	e.g.	e.g.	ADV
ajst-31431	75	48	,	,	PUNCT
ajst-31431	75	49	word2vec	word2vec	NOUN
ajst-31431	75	50	)	)	PUNCT
ajst-31431	75	51	.	.	PUNCT
ajst-31431	76	1	through	through	ADP
ajst-31431	76	2	end	end	NOUN
ajst-31431	76	3	-	-	PUNCT
ajst-31431	76	4	to	to	ADP
ajst-31431	76	5	-	-	PUNCT
ajst-31431	76	6	end	end	NOUN
ajst-31431	76	7	learning	learning	NOUN
ajst-31431	76	8	,	,	PUNCT
ajst-31431	76	9	deep	deep	ADJ
ajst-31431	76	10	learning	learning	NOUN
ajst-31431	76	11	models	model	NOUN
ajst-31431	76	12	autonomously	autonomously	ADV
ajst-31431	76	13	capture	capture	VERB
ajst-31431	76	14	hierarchical	hierarchical	ADJ
ajst-31431	76	15	information	information	NOUN
ajst-31431	76	16	from	from	ADP
ajst-31431	76	17	lexical	lexical	ADJ
ajst-31431	76	18	morphology	morphology	NOUN
ajst-31431	76	19	to	to	ADP
ajst-31431	76	20	contextual	contextual	ADJ
ajst-31431	76	21	relationships	relationship	NOUN
ajst-31431	76	22	,	,	PUNCT
ajst-31431	76	23	significantly	significantly	ADV
ajst-31431	76	24	enhancing	enhance	VERB
ajst-31431	76	25	the	the	DET
ajst-31431	76	26	accuracy	accuracy	NOUN
ajst-31431	76	27	and	and	CCONJ
ajst-31431	76	28	robustness	robustness	NOUN
ajst-31431	76	29	of	of	ADP
ajst-31431	76	30	hate	hate	NOUN
ajst-31431	76	31	speech	speech	NOUN
ajst-31431	76	32	detection	detection	NOUN
ajst-31431	76	33	[	[	X
ajst-31431	76	34	28,29	28,29	X
ajst-31431	76	35	]	]	X
ajst-31431	76	36	.	.	PUNCT
ajst-31431	77	1	this	this	DET
ajst-31431	77	2	capability	capability	NOUN
ajst-31431	77	3	grants	grant	VERB
ajst-31431	77	4	them	they	PRON
ajst-31431	77	5	distinct	distinct	ADJ
ajst-31431	77	6	advantages	advantage	NOUN
ajst-31431	77	7	in	in	ADP
ajst-31431	77	8	sentiment	sentiment	NOUN
ajst-31431	77	9	analysis	analysis	NOUN
ajst-31431	77	10	and	and	CCONJ
ajst-31431	77	11	hate	hate	VERB
ajst-31431	77	12	content	content	NOUN
ajst-31431	77	13	identification	identification	NOUN
ajst-31431	77	14	,	,	PUNCT
ajst-31431	77	15	leading	lead	VERB
ajst-31431	77	16	to	to	ADP
ajst-31431	77	17	widespread	widespread	ADJ
ajst-31431	77	18	application	application	NOUN
ajst-31431	77	19	in	in	ADP
ajst-31431	77	20	data	datum	NOUN
ajst-31431	77	21	mining	mining	NOUN
ajst-31431	77	22	and	and	CCONJ
ajst-31431	77	23	text	text	NOUN
ajst-31431	77	24	classification	classification	NOUN
ajst-31431	77	25	.	.	PUNCT
ajst-31431	78	1	recurrent	recurrent	ADJ
ajst-31431	78	2	neural	neural	ADJ
ajst-31431	78	3	networks	network	NOUN
ajst-31431	78	4	(	(	PUNCT
ajst-31431	78	5	rnns	rnns	PROPN
ajst-31431	78	6	)	)	PUNCT
ajst-31431	78	7	are	be	AUX
ajst-31431	78	8	specifically	specifically	ADV
ajst-31431	78	9	designed	design	VERB
ajst-31431	78	10	for	for	ADP
ajst-31431	78	11	sequential	sequential	ADJ
ajst-31431	78	12	data	datum	NOUN
ajst-31431	78	13	,	,	PUNCT
ajst-31431	78	14	overcoming	overcome	VERB
ajst-31431	78	15	the	the	DET
ajst-31431	78	16	limitation	limitation	NOUN
ajst-31431	78	17	of	of	ADP
ajst-31431	78	18	standard	standard	ADJ
ajst-31431	78	19	feedforward	feedforward	NOUN
ajst-31431	78	20	networks	network	NOUN
ajst-31431	78	21	which	which	PRON
ajst-31431	78	22	process	process	VERB
ajst-31431	78	23	independent	independent	ADJ
ajst-31431	78	24	data	datum	NOUN
ajst-31431	78	25	points	point	NOUN
ajst-31431	78	26	.	.	PUNCT
ajst-31431	79	1	rnns	rnn	NOUN
ajst-31431	79	2	retain	retain	VERB
ajst-31431	79	3	information	information	NOUN
ajst-31431	79	4	from	from	ADP
ajst-31431	79	5	previous	previous	ADJ
ajst-31431	79	6	inputs	input	NOUN
ajst-31431	79	7	,	,	PUNCT
ajst-31431	79	8	making	make	VERB
ajst-31431	79	9	them	they	PRON
ajst-31431	79	10	suitable	suitable	ADJ
ajst-31431	79	11	for	for	ADP
ajst-31431	79	12	tasks	task	NOUN
ajst-31431	79	13	like	like	ADP
ajst-31431	79	14	sentiment	sentiment	NOUN
ajst-31431	79	15	analysis	analysis	NOUN
ajst-31431	79	16	and	and	CCONJ
ajst-31431	79	17	hate	hate	NOUN
ajst-31431	79	18	speech	speech	NOUN
ajst-31431	79	19	detection	detection	NOUN
ajst-31431	79	20	,	,	PUNCT
ajst-31431	79	21	where	where	SCONJ
ajst-31431	79	22	word	word	NOUN
ajst-31431	79	23	context	context	NOUN
ajst-31431	79	24	and	and	CCONJ
ajst-31431	79	25	order	order	NOUN
ajst-31431	79	26	are	be	AUX
ajst-31431	79	27	crucial	crucial	ADJ
ajst-31431	79	28	.	.	PUNCT
ajst-31431	80	1	long	long	ADJ
ajst-31431	80	2	short	short	ADJ
ajst-31431	80	3	-	-	PUNCT
ajst-31431	80	4	term	term	NOUN
ajst-31431	80	5	memory	memory	NOUN
ajst-31431	80	6	networks	network	NOUN
ajst-31431	80	7	(	(	PUNCT
ajst-31431	80	8	lstms	lstms	ADJ
ajst-31431	80	9	)	)	PUNCT
ajst-31431	80	10	,	,	PUNCT
ajst-31431	80	11	a	a	DET
ajst-31431	80	12	type	type	NOUN
ajst-31431	80	13	of	of	ADP
ajst-31431	80	14	rnn	rnn	NOUN
ajst-31431	80	15	,	,	PUNCT
ajst-31431	80	16	effectively	effectively	ADV
ajst-31431	80	17	address	address	VERB
ajst-31431	80	18	the	the	DET
ajst-31431	80	19	vanishing	vanish	VERB
ajst-31431	80	20	gradient	gradient	NOUN
ajst-31431	80	21	problem	problem	NOUN
ajst-31431	80	22	,	,	PUNCT
ajst-31431	80	23	enabling	enable	VERB
ajst-31431	80	24	them	they	PRON
ajst-31431	80	25	to	to	PART
ajst-31431	80	26	capture	capture	VERB
ajst-31431	80	27	long	long	ADJ
ajst-31431	80	28	-	-	PUNCT
ajst-31431	80	29	range	range	NOUN
ajst-31431	80	30	dependencies	dependency	NOUN
ajst-31431	80	31	in	in	ADP
ajst-31431	80	32	sequences	sequence	NOUN
ajst-31431	80	33	.	.	PUNCT
ajst-31431	81	1	this	this	PRON
ajst-31431	81	2	makes	make	VERB
ajst-31431	81	3	lstms	lstms	ADJ
ajst-31431	81	4	particularly	particularly	ADV
ajst-31431	81	5	advantageous	advantageous	ADJ
ajst-31431	81	6	for	for	ADP
ajst-31431	81	7	longer	long	ADJ
ajst-31431	81	8	texts	text	NOUN
ajst-31431	81	9	in	in	ADP
ajst-31431	81	10	hate	hate	NOUN
ajst-31431	81	11	speech	speech	NOUN
ajst-31431	81	12	detection	detection	NOUN
ajst-31431	81	13	.	.	PUNCT
ajst-31431	82	1	they	they	PRON
ajst-31431	82	2	process	process	VERB
ajst-31431	82	3	text	text	NOUN
ajst-31431	82	4	incrementally	incrementally	ADV
ajst-31431	82	5	,	,	PUNCT
ajst-31431	82	6	updating	update	VERB
ajst-31431	82	7	a	a	DET
ajst-31431	82	8	hidden	hidden	ADJ
ajst-31431	82	9	state	state	NOUN
ajst-31431	82	10	to	to	PART
ajst-31431	82	11	capture	capture	VERB
ajst-31431	82	12	contextual	contextual	ADJ
ajst-31431	82	13	relationships	relationship	NOUN
ajst-31431	82	14	.	.	PUNCT
ajst-31431	83	1	attention	attention	NOUN
ajst-31431	83	2	mechanisms	mechanism	NOUN
ajst-31431	83	3	can	can	AUX
ajst-31431	83	4	be	be	AUX
ajst-31431	83	5	integrated	integrate	VERB
ajst-31431	83	6	to	to	PART
ajst-31431	83	7	enhance	enhance	VERB
ajst-31431	83	8	performance	performance	NOUN
ajst-31431	83	9	by	by	ADP
ajst-31431	83	10	focusing	focus	VERB
ajst-31431	83	11	on	on	ADP
ajst-31431	83	12	salient	salient	NOUN
ajst-31431	83	13	hate	hate	VERB
ajst-31431	83	14	-	-	PUNCT
ajst-31431	83	15	indicative	indicative	ADJ
ajst-31431	83	16	parts	part	NOUN
ajst-31431	83	17	of	of	ADP
ajst-31431	83	18	the	the	DET
ajst-31431	83	19	text	text	NOUN
ajst-31431	83	20	.	.	PUNCT
ajst-31431	84	1	gated	gate	VERB
ajst-31431	84	2	recurrent	recurrent	ADJ
ajst-31431	84	3	units	unit	NOUN
ajst-31431	84	4	(	(	PUNCT
ajst-31431	84	5	grus	grus	NOUN
ajst-31431	84	6	)	)	PUNCT
ajst-31431	84	7	are	be	AUX
ajst-31431	84	8	another	another	DET
ajst-31431	84	9	rnn	rnn	NOUN
ajst-31431	84	10	variant	variant	NOUN
ajst-31431	84	11	designed	design	VERB
ajst-31431	84	12	to	to	PART
ajst-31431	84	13	mitigate	mitigate	VERB
ajst-31431	84	14	the	the	DET
ajst-31431	84	15	vanishing	vanish	VERB
ajst-31431	84	16	gradient	gradient	NOUN
ajst-31431	84	17	problem	problem	NOUN
ajst-31431	84	18	and	and	CCONJ
ajst-31431	84	19	capture	capture	VERB
ajst-31431	84	20	long	long	ADJ
ajst-31431	84	21	-	-	PUNCT
ajst-31431	84	22	range	range	NOUN
ajst-31431	84	23	dependencies	dependency	NOUN
ajst-31431	84	24	,	,	PUNCT
ajst-31431	84	25	similar	similar	ADJ
ajst-31431	84	26	to	to	ADP
ajst-31431	84	27	lstms	lstms	NOUN
ajst-31431	84	28	.	.	PUNCT
ajst-31431	85	1	grus	grus	PROPN
ajst-31431	85	2	also	also	ADV
ajst-31431	85	3	update	update	VERB
ajst-31431	85	4	a	a	DET
ajst-31431	85	5	hidden	hidden	ADJ
ajst-31431	85	6	state	state	NOUN
ajst-31431	85	7	incrementally	incrementally	ADV
ajst-31431	85	8	and	and	CCONJ
ajst-31431	85	9	effectively	effectively	ADV
ajst-31431	85	10	model	model	VERB
ajst-31431	85	11	sequential	sequential	ADJ
ajst-31431	85	12	dependencies	dependency	NOUN
ajst-31431	85	13	for	for	ADP
ajst-31431	85	14	context	context	NOUN
ajst-31431	85	15	and	and	CCONJ
ajst-31431	85	16	sentiment	sentiment	NOUN
ajst-31431	85	17	understanding	understanding	NOUN
ajst-31431	85	18	in	in	ADP
ajst-31431	85	19	hate	hate	NOUN
ajst-31431	85	20	speech	speech	NOUN
ajst-31431	85	21	detection	detection	NOUN
ajst-31431	85	22	.	.	PUNCT
ajst-31431	86	1	convolutional	convolutional	ADJ
ajst-31431	86	2	neural	neural	ADJ
ajst-31431	86	3	networks	network	NOUN
ajst-31431	86	4	(	(	PUNCT
ajst-31431	86	5	cnns	cnns	PROPN
ajst-31431	86	6	)	)	PUNCT
ajst-31431	86	7	,	,	PUNCT
ajst-31431	86	8	while	while	SCONJ
ajst-31431	86	9	prominent	prominent	ADJ
ajst-31431	86	10	in	in	ADP
ajst-31431	86	11	computer	computer	NOUN
ajst-31431	86	12	vision	vision	NOUN
ajst-31431	86	13	,	,	PUNCT
ajst-31431	86	14	have	have	AUX
ajst-31431	86	15	been	be	AUX
ajst-31431	86	16	successfully	successfully	ADV
ajst-31431	86	17	applied	apply	VERB
ajst-31431	86	18	to	to	ADP
ajst-31431	86	19	nlp	nlp	ADJ
ajst-31431	86	20	tasks	task	NOUN
ajst-31431	86	21	including	include	VERB
ajst-31431	86	22	sentiment	sentiment	NOUN
ajst-31431	86	23	analysis	analysis	NOUN
ajst-31431	86	24	and	and	CCONJ
ajst-31431	86	25	hate	hate	NOUN
ajst-31431	86	26	speech	speech	NOUN
ajst-31431	86	27	detection	detection	NOUN
ajst-31431	86	28	.	.	PUNCT
ajst-31431	87	1	cnns	cnns	PROPN
ajst-31431	87	2	treat	treat	NOUN
ajst-31431	87	3	text	text	NOUN
ajst-31431	87	4	as	as	ADP
ajst-31431	87	5	a	a	DET
ajst-31431	87	6	1d	1d	NUM
ajst-31431	87	7	sequence	sequence	NOUN
ajst-31431	87	8	,	,	PUNCT
ajst-31431	87	9	applying	apply	VERB
ajst-31431	87	10	convolutional	convolutional	ADJ
ajst-31431	87	11	filters	filter	NOUN
ajst-31431	87	12	to	to	PART
ajst-31431	87	13	extract	extract	VERB
ajst-31431	87	14	local	local	ADJ
ajst-31431	87	15	patterns	pattern	NOUN
ajst-31431	87	16	and	and	CCONJ
ajst-31431	87	17	features	feature	NOUN
ajst-31431	87	18	,	,	PUNCT
ajst-31431	87	19	proving	prove	VERB
ajst-31431	87	20	effective	effective	ADJ
ajst-31431	87	21	for	for	ADP
ajst-31431	87	22	capturing	capture	VERB
ajst-31431	87	23	such	such	ADJ
ajst-31431	87	24	structures	structure	NOUN
ajst-31431	87	25	in	in	ADP
ajst-31431	87	26	text	text	NOUN
ajst-31431	87	27	data	datum	NOUN
ajst-31431	87	28	.	.	PUNCT
ajst-31431	88	1	3.3	3.3	NUM
ajst-31431	88	2	.	.	PUNCT
ajst-31431	88	3	transformer	transformer	NOUN
ajst-31431	88	4	-	-	PUNCT
ajst-31431	88	5	based	base	VERB
ajst-31431	88	6	models	model	NOUN
ajst-31431	88	7	.	.	PUNCT
ajst-31431	89	1	since	since	SCONJ
ajst-31431	89	2	their	their	PRON
ajst-31431	89	3	introduction	introduction	NOUN
ajst-31431	89	4	by	by	ADP
ajst-31431	89	5	vaswani	vaswani	PROPN
ajst-31431	89	6	et	et	PROPN
ajst-31431	89	7	al	al	PROPN
ajst-31431	89	8	.	.	PUNCT
ajst-31431	90	1	[	[	X
ajst-31431	90	2	30	30	NUM
ajst-31431	90	3	]	]	PUNCT
ajst-31431	90	4	in	in	ADP
ajst-31431	90	5	"	"	PUNCT
ajst-31431	90	6	attention	attention	NOUN
ajst-31431	90	7	is	be	AUX
ajst-31431	90	8	all	all	PRON
ajst-31431	90	9	you	you	PRON
ajst-31431	90	10	need	need	VERB
ajst-31431	90	11	,	,	PUNCT
ajst-31431	90	12	"	"	PUNCT
ajst-31431	90	13	transformer	transformer	NOUN
ajst-31431	90	14	-	-	PUNCT
ajst-31431	90	15	based	base	VERB
ajst-31431	90	16	models	model	NOUN
ajst-31431	90	17	have	have	AUX
ajst-31431	90	18	revolutionized	revolutionize	VERB
ajst-31431	90	19	natural	natural	ADJ
ajst-31431	90	20	language	language	NOUN
ajst-31431	90	21	processing	processing	NOUN
ajst-31431	90	22	(	(	PUNCT
ajst-31431	90	23	nlp	nlp	NOUN
ajst-31431	90	24	)	)	PUNCT
ajst-31431	90	25	.	.	PUNCT
ajst-31431	91	1	these	these	DET
ajst-31431	91	2	models	model	NOUN
ajst-31431	91	3	have	have	AUX
ajst-31431	91	4	become	become	VERB
ajst-31431	91	5	state	state	NOUN
ajst-31431	91	6	-	-	PUNCT
ajst-31431	91	7	of	of	ADP
ajst-31431	91	8	-	-	PUNCT
ajst-31431	91	9	the	the	DET
ajst-31431	91	10	-	-	PUNCT
ajst-31431	91	11	art	art	NOUN
ajst-31431	91	12	for	for	ADP
ajst-31431	91	13	various	various	ADJ
ajst-31431	91	14	nlp	nlp	NOUN
ajst-31431	91	15	tasks	task	NOUN
ajst-31431	91	16	,	,	PUNCT
ajst-31431	91	17	primarily	primarily	ADV
ajst-31431	91	18	due	due	ADP
ajst-31431	91	19	to	to	ADP
ajst-31431	91	20	their	their	PRON
ajst-31431	91	21	ability	ability	NOUN
ajst-31431	91	22	to	to	PART
ajst-31431	91	23	effectively	effectively	ADV
ajst-31431	91	24	handle	handle	VERB
ajst-31431	91	25	long	long	ADJ
ajst-31431	91	26	-	-	PUNCT
ajst-31431	91	27	range	range	NOUN
ajst-31431	91	28	dependencies	dependency	NOUN
ajst-31431	91	29	and	and	CCONJ
ajst-31431	91	30	process	process	NOUN
ajst-31431	91	31	text	text	NOUN
ajst-31431	91	32	in	in	ADP
ajst-31431	91	33	parallel	parallel	NOUN
ajst-31431	91	34	,	,	PUNCT
ajst-31431	91	35	offering	offer	VERB
ajst-31431	91	36	superior	superior	ADJ
ajst-31431	91	37	efficiency	efficiency	NOUN
ajst-31431	91	38	and	and	CCONJ
ajst-31431	91	39	scalability	scalability	NOUN
ajst-31431	91	40	compared	compare	VERB
ajst-31431	91	41	to	to	ADP
ajst-31431	91	42	traditional	traditional	ADJ
ajst-31431	91	43	rnns	rnn	NOUN
ajst-31431	91	44	and	and	CCONJ
ajst-31431	91	45	cnns	cnn	NOUN
ajst-31431	91	46	.	.	PUNCT
ajst-31431	92	1	the	the	DET
ajst-31431	92	2	core	core	NOUN
ajst-31431	92	3	innovation	innovation	NOUN
ajst-31431	92	4	lies	lie	VERB
ajst-31431	92	5	in	in	ADP
ajst-31431	92	6	the	the	DET
ajst-31431	92	7	self	self	NOUN
ajst-31431	92	8	-	-	PUNCT
ajst-31431	92	9	attention	attention	NOUN
ajst-31431	92	10	mechanism	mechanism	NOUN
ajst-31431	92	11	,	,	PUNCT
ajst-31431	92	12	which	which	PRON
ajst-31431	92	13	allows	allow	VERB
ajst-31431	92	14	the	the	DET
ajst-31431	92	15	model	model	NOUN
ajst-31431	92	16	to	to	PART
ajst-31431	92	17	dynamically	dynamically	ADV
ajst-31431	92	18	weigh	weigh	VERB
ajst-31431	92	19	the	the	DET
ajst-31431	92	20	relevance	relevance	NOUN
ajst-31431	92	21	of	of	ADP
ajst-31431	92	22	all	all	DET
ajst-31431	92	23	words	word	NOUN
ajst-31431	92	24	in	in	ADP
ajst-31431	92	25	the	the	DET
ajst-31431	92	26	input	input	NOUN
ajst-31431	92	27	sequence	sequence	NOUN
ajst-31431	92	28	when	when	SCONJ
ajst-31431	92	29	making	make	VERB
ajst-31431	92	30	predictions	prediction	NOUN
ajst-31431	92	31	,	,	PUNCT
ajst-31431	92	32	thereby	thereby	ADV
ajst-31431	92	33	capturing	capture	VERB
ajst-31431	92	34	comprehensive	comprehensive	ADJ
ajst-31431	92	35	contextual	contextual	ADJ
ajst-31431	92	36	information	information	NOUN
ajst-31431	92	37	.	.	PUNCT
ajst-31431	93	1	prominent	prominent	ADJ
ajst-31431	93	2	transformer	transformer	NOUN
ajst-31431	93	3	architectures	architecture	NOUN
ajst-31431	93	4	,	,	PUNCT
ajst-31431	93	5	including	include	VERB
ajst-31431	93	6	bidirectional	bidirectional	ADJ
ajst-31431	93	7	encoder	encoder	NOUN
ajst-31431	93	8	representations	representation	VERB
ajst-31431	93	9	from	from	ADP
ajst-31431	93	10	transformers	transformer	NOUN
ajst-31431	93	11	(	(	PUNCT
ajst-31431	93	12	bert	bert	PROPN
ajst-31431	93	13	)	)	PUNCT
ajst-31431	94	1	[	[	X
ajst-31431	94	2	31	31	NUM
ajst-31431	94	3	]	]	PUNCT
ajst-31431	94	4	,	,	PUNCT
ajst-31431	94	5	generative	generative	ADJ
ajst-31431	94	6	pre	pre	ADJ
ajst-31431	94	7	-	-	ADJ
ajst-31431	94	8	trained	train	VERB
ajst-31431	94	9	transformer	transformer	NOUN
ajst-31431	94	10	(	(	PUNCT
ajst-31431	94	11	gpt	gpt	NOUN
ajst-31431	94	12	)	)	PUNCT
ajst-31431	94	13	,	,	PUNCT
ajst-31431	94	14	and	and	CCONJ
ajst-31431	94	15	a	a	DET
ajst-31431	94	16	robustly	robustly	ADV
ajst-31431	94	17	optimized	optimize	VERB
ajst-31431	94	18	bert	bert	NOUN
ajst-31431	94	19	pretraining	pretraine	VERB
ajst-31431	94	20	approach	approach	NOUN
ajst-31431	94	21	(	(	PUNCT
ajst-31431	94	22	roberta	roberta	PROPN
ajst-31431	94	23	)	)	PUNCT
ajst-31431	95	1	[	[	X
ajst-31431	95	2	32	32	NUM
ajst-31431	95	3	]	]	PUNCT
ajst-31431	95	4	,	,	PUNCT
ajst-31431	95	5	have	have	AUX
ajst-31431	95	6	demonstrated	demonstrate	VERB
ajst-31431	95	7	exceptional	exceptional	ADJ
ajst-31431	95	8	potential	potential	NOUN
ajst-31431	95	9	across	across	ADP
ajst-31431	95	10	nlp	nlp	NOUN
ajst-31431	95	11	applications	application	NOUN
ajst-31431	95	12	such	such	ADJ
ajst-31431	95	13	as	as	ADP
ajst-31431	95	14	sentiment	sentiment	NOUN
ajst-31431	95	15	analysis	analysis	NOUN
ajst-31431	95	16	and	and	CCONJ
ajst-31431	95	17	hate	hate	NOUN
ajst-31431	95	18	speech	speech	NOUN
ajst-31431	95	19	detection	detection	NOUN
ajst-31431	95	20	.	.	PUNCT
ajst-31431	96	1	typically	typically	ADV
ajst-31431	96	2	pre	pre	VERB
ajst-31431	96	3	-	-	VERB
ajst-31431	96	4	trained	train	VERB
ajst-31431	96	5	on	on	ADP
ajst-31431	96	6	massive	massive	ADJ
ajst-31431	96	7	text	text	NOUN
ajst-31431	96	8	corpora	corpora	NOUN
ajst-31431	96	9	and	and	CCONJ
ajst-31431	96	10	then	then	ADV
ajst-31431	96	11	fine	fine	ADV
ajst-31431	96	12	-	-	PUNCT
ajst-31431	96	13	tuned	tune	VERB
ajst-31431	96	14	for	for	ADP
ajst-31431	96	15	specific	specific	ADJ
ajst-31431	96	16	downstream	downstream	ADJ
ajst-31431	96	17	tasks	task	NOUN
ajst-31431	96	18	,	,	PUNCT
ajst-31431	96	19	these	these	DET
ajst-31431	96	20	models	model	NOUN
ajst-31431	96	21	achieve	achieve	VERB
ajst-31431	96	22	high	high	ADJ
ajst-31431	96	23	performance	performance	NOUN
ajst-31431	96	24	even	even	ADV
ajst-31431	96	25	with	with	ADP
ajst-31431	96	26	limited	limited	ADJ
ajst-31431	96	27	task	task	NOUN
ajst-31431	96	28	-	-	PUNCT
ajst-31431	96	29	specific	specific	ADJ
ajst-31431	96	30	data	datum	NOUN
ajst-31431	96	31	.	.	PUNCT
ajst-31431	97	1	their	their	PRON
ajst-31431	97	2	effectiveness	effectiveness	NOUN
ajst-31431	97	3	in	in	ADP
ajst-31431	97	4	hate	hate	NOUN
ajst-31431	97	5	speech	speech	NOUN
ajst-31431	97	6	detection	detection	NOUN
ajst-31431	97	7	stems	stem	VERB
ajst-31431	97	8	from	from	ADP
ajst-31431	97	9	deep	deep	ADJ
ajst-31431	97	10	contextual	contextual	ADJ
ajst-31431	97	11	understanding	understanding	NOUN
ajst-31431	97	12	,	,	PUNCT
ajst-31431	97	13	bidirectional	bidirectional	ADJ
ajst-31431	97	14	context	context	NOUN
ajst-31431	97	15	modeling	modeling	NOUN
ajst-31431	97	16	,	,	PUNCT
ajst-31431	97	17	large	large	ADJ
ajst-31431	97	18	-	-	PUNCT
ajst-31431	97	19	scale	scale	NOUN
ajst-31431	97	20	pre	pre	NOUN
ajst-31431	97	21	-	-	NOUN
ajst-31431	97	22	training	training	ADJ
ajst-31431	97	23	,	,	PUNCT
ajst-31431	97	24	transfer	transfer	NOUN
ajst-31431	97	25	learning	learning	NOUN
ajst-31431	97	26	,	,	PUNCT
ajst-31431	97	27	and	and	CCONJ
ajst-31431	97	28	the	the	DET
ajst-31431	97	29	powerful	powerful	ADJ
ajst-31431	97	30	attention	attention	NOUN
ajst-31431	97	31	mechanism	mechanism	NOUN
ajst-31431	97	32	.	.	PUNCT
ajst-31431	98	1	3.4	3.4	NUM
ajst-31431	98	2	.	.	PUNCT
ajst-31431	98	3	hybrid	hybrid	ADJ
ajst-31431	98	4	models	model	NOUN
ajst-31431	98	5	.	.	PUNCT
ajst-31431	99	1	beyond	beyond	ADP
ajst-31431	99	2	the	the	DET
ajst-31431	99	3	aforementioned	aforementioned	ADJ
ajst-31431	99	4	models	model	NOUN
ajst-31431	99	5	,	,	PUNCT
ajst-31431	99	6	hybrid	hybrid	ADJ
ajst-31431	99	7	approaches	approach	NOUN
ajst-31431	99	8	and	and	CCONJ
ajst-31431	99	9	graph	graph	NOUN
ajst-31431	99	10	-	-	PUNCT
ajst-31431	99	11	based	base	VERB
ajst-31431	99	12	methods	method	NOUN
ajst-31431	99	13	have	have	AUX
ajst-31431	99	14	also	also	ADV
ajst-31431	99	15	emerged	emerge	VERB
ajst-31431	99	16	for	for	ADP
ajst-31431	99	17	hate	hate	NOUN
ajst-31431	99	18	speech	speech	NOUN
ajst-31431	99	19	detection	detection	NOUN
ajst-31431	99	20	.	.	PUNCT
ajst-31431	100	1	hybrid	hybrid	ADJ
ajst-31431	100	2	models	model	NOUN
ajst-31431	100	3	combine	combine	VERB
ajst-31431	100	4	architectures	architecture	NOUN
ajst-31431	100	5	to	to	ADP
ajst-31431	100	6	leverage	leverage	NOUN
ajst-31431	100	7	complementary	complementary	ADJ
ajst-31431	100	8	strengths	strength	NOUN
ajst-31431	100	9	;	;	PUNCT
ajst-31431	100	10	for	for	ADP
ajst-31431	100	11	instance	instance	NOUN
ajst-31431	100	12	,	,	PUNCT
ajst-31431	100	13	mathew	mathew	PROPN
ajst-31431	100	14	et	et	PROPN
ajst-31431	100	15	al	al	PROPN
ajst-31431	100	16	.	.	PUNCT
ajst-31431	101	1	[	[	X
ajst-31431	101	2	33	33	NUM
ajst-31431	101	3	]	]	X
ajst-31431	101	4	integrated	integrate	VERB
ajst-31431	101	5	cnns	cnn	NOUN
ajst-31431	101	6	with	with	ADP
ajst-31431	101	7	grus	grus	NOUN
ajst-31431	101	8	,	,	PUNCT
ajst-31431	101	9	while	while	SCONJ
ajst-31431	101	10	khan	khan	PROPN
ajst-31431	101	11	et	et	PROPN
ajst-31431	101	12	al	al	PROPN
ajst-31431	101	13	.	.	PUNCT
ajst-31431	102	1	[	[	X
ajst-31431	102	2	56	56	NUM
ajst-31431	102	3	]	]	PUNCT
ajst-31431	102	4	employed	employ	VERB
ajst-31431	102	5	a	a	DET
ajst-31431	102	6	cnn	cnn	PROPN
ajst-31431	102	7	-	-	PUNCT
ajst-31431	102	8	lstm	lstm	ADJ
ajst-31431	102	9	framework	framework	NOUN
ajst-31431	102	10	for	for	ADP
ajst-31431	102	11	sentiment	sentiment	NOUN
ajst-31431	102	12	analysis	analysis	NOUN
ajst-31431	102	13	in	in	ADP
ajst-31431	102	14	roman	roman	PROPN
ajst-31431	102	15	urdu	urdu	PROPN
ajst-31431	102	16	and	and	CCONJ
ajst-31431	102	17	english	english	ADJ
ajst-31431	102	18	dialects	dialect	NOUN
ajst-31431	102	19	,	,	PUNCT
ajst-31431	102	20	enabling	enable	VERB
ajst-31431	102	21	the	the	DET
ajst-31431	102	22	capture	capture	NOUN
ajst-31431	102	23	of	of	ADP
ajst-31431	102	24	both	both	CCONJ
ajst-31431	102	25	local	local	ADJ
ajst-31431	102	26	patterns	pattern	NOUN
ajst-31431	102	27	and	and	CCONJ
ajst-31431	102	28	long	long	ADJ
ajst-31431	102	29	-	-	PUNCT
ajst-31431	102	30	term	term	NOUN
ajst-31431	102	31	dependencies	dependency	NOUN
ajst-31431	102	32	in	in	ADP
ajst-31431	102	33	text	text	NOUN
ajst-31431	102	34	.	.	PUNCT
ajst-31431	103	1	recently	recently	ADV
ajst-31431	103	2	,	,	PUNCT
ajst-31431	103	3	graph	graph	NOUN
ajst-31431	103	4	neural	neural	ADJ
ajst-31431	103	5	networks	network	NOUN
ajst-31431	103	6	(	(	PUNCT
ajst-31431	103	7	gnns	gnns	NOUN
ajst-31431	103	8	)	)	PUNCT
ajst-31431	103	9	have	have	AUX
ajst-31431	103	10	gained	gain	VERB
ajst-31431	103	11	prominence	prominence	NOUN
ajst-31431	103	12	in	in	ADP
ajst-31431	103	13	nlp	nlp	NOUN
ajst-31431	103	14	by	by	ADP
ajst-31431	103	15	representing	represent	VERB
ajst-31431	103	16	text	text	NOUN
ajst-31431	103	17	as	as	ADP
ajst-31431	103	18	graph	graph	NOUN
ajst-31431	103	19	structures	structure	NOUN
ajst-31431	103	20	(	(	PUNCT
ajst-31431	103	21	where	where	SCONJ
ajst-31431	103	22	nodes	node	NOUN
ajst-31431	103	23	denote	denote	VERB
ajst-31431	103	24	words	word	NOUN
ajst-31431	103	25	/	/	SYM
ajst-31431	103	26	entities	entity	NOUN
ajst-31431	103	27	and	and	CCONJ
ajst-31431	103	28	edges	edge	NOUN
ajst-31431	103	29	represent	represent	VERB
ajst-31431	103	30	relationships	relationship	NOUN
ajst-31431	103	31	,	,	PUNCT
ajst-31431	103	32	constructed	construct	VERB
ajst-31431	103	33	via	via	ADP
ajst-31431	103	34	techniques	technique	NOUN
ajst-31431	103	35	like	like	ADP
ajst-31431	103	36	dependency	dependency	NOUN
ajst-31431	103	37	parsing	parsing	NOUN
ajst-31431	103	38	or	or	CCONJ
ajst-31431	103	39	co	co	NOUN
ajst-31431	103	40	-	-	NOUN
ajst-31431	103	41	occurrence	occurrence	NOUN
ajst-31431	103	42	)	)	PUNCT
ajst-31431	103	43	to	to	PART
ajst-31431	103	44	exploit	exploit	VERB
ajst-31431	103	45	inherent	inherent	ADJ
ajst-31431	103	46	relational	relational	ADJ
ajst-31431	103	47	information	information	NOUN
ajst-31431	103	48	,	,	PUNCT
ajst-31431	103	49	achieving	achieve	VERB
ajst-31431	103	50	state	state	NOUN
ajst-31431	103	51	-	-	PUNCT
ajst-31431	103	52	ofthe	ofthe	NOUN
ajst-31431	103	53	-	-	PUNCT
ajst-31431	103	54	art	art	NOUN
ajst-31431	103	55	results	result	NOUN
ajst-31431	103	56	.	.	PUNCT
ajst-31431	104	1	sarracén	sarracén	VERB
ajst-31431	104	2	et	et	PROPN
ajst-31431	104	3	al	al	PROPN
ajst-31431	104	4	.	.	PUNCT
ajst-31431	105	1	[	[	X
ajst-31431	105	2	34	34	NUM
ajst-31431	105	3	]	]	PUNCT
ajst-31431	105	4	proposed	propose	VERB
ajst-31431	105	5	an	an	DET
ajst-31431	105	6	unsupervised	unsupervised	ADJ
ajst-31431	105	7	method	method	NOUN
ajst-31431	105	8	using	use	VERB
ajst-31431	105	9	graph	graph	NOUN
ajst-31431	105	10	autoencoders	autoencoder	NOUN
ajst-31431	105	11	(	(	PUNCT
ajst-31431	105	12	gaes	gaes	PROPN
ajst-31431	105	13	)	)	PUNCT
ajst-31431	105	14	,	,	PUNCT
ajst-31431	105	15	representing	represent	VERB
ajst-31431	105	16	texts	text	NOUN
ajst-31431	105	17	as	as	ADP
ajst-31431	105	18	nodes	node	NOUN
ajst-31431	105	19	encoded	encode	VERB
ajst-31431	105	20	via	via	ADP
ajst-31431	105	21	transformer	transformer	NOUN
ajst-31431	105	22	and	and	CCONJ
ajst-31431	105	23	convolutional	convolutional	ADJ
ajst-31431	105	24	layers	layer	NOUN
ajst-31431	105	25	within	within	ADP
ajst-31431	105	26	a	a	DET
ajst-31431	105	27	low	low	ADJ
ajst-31431	105	28	-	-	PUNCT
ajst-31431	105	29	dimensional	dimensional	ADJ
ajst-31431	105	30	space	space	NOUN
ajst-31431	105	31	,	,	PUNCT
ajst-31431	105	32	with	with	ADP
ajst-31431	105	33	reconstruction	reconstruction	NOUN
ajst-31431	105	34	guiding	guide	VERB
ajst-31431	105	35	hate	hate	NOUN
ajst-31431	105	36	speech	speech	NOUN
ajst-31431	105	37	detection	detection	NOUN
ajst-31431	105	38	across	across	ADP
ajst-31431	105	39	domains	domain	NOUN
ajst-31431	105	40	and	and	CCONJ
ajst-31431	105	41	languages	language	NOUN
ajst-31431	105	42	.	.	PUNCT
ajst-31431	106	1	convolutional	convolutional	ADJ
ajst-31431	106	2	graph	graph	NOUN
ajst-31431	106	3	neural	neural	ADJ
ajst-31431	106	4	networks	network	NOUN
ajst-31431	106	5	(	(	PUNCT
ajst-31431	106	6	cgnns	cgnns	PROPN
ajst-31431	106	7	)	)	PUNCT
ajst-31431	106	8	specifically	specifically	ADV
ajst-31431	106	9	show	show	VERB
ajst-31431	106	10	promise	promise	NOUN
ajst-31431	106	11	in	in	ADP
ajst-31431	106	12	capturing	capture	VERB
ajst-31431	106	13	structural	structural	ADJ
ajst-31431	106	14	and	and	CCONJ
ajst-31431	106	15	linguistic	linguistic	ADJ
ajst-31431	106	16	patterns	pattern	NOUN
ajst-31431	106	17	,	,	PUNCT
ajst-31431	106	18	demonstrated	demonstrate	VERB
ajst-31431	106	19	effectively	effectively	ADV
ajst-31431	106	20	by	by	ADP
ajst-31431	106	21	sarracén	sarracén	PROPN
ajst-31431	106	22	et	et	PROPN
ajst-31431	106	23	al	al	PROPN
ajst-31431	106	24	.	.	PROPN
ajst-31431	107	1	through	through	ADP
ajst-31431	107	2	graph	graph	NOUN
ajst-31431	107	3	construction	construction	NOUN
ajst-31431	107	4	and	and	CCONJ
ajst-31431	107	5	embedding	embed	VERB
ajst-31431	107	6	learning	learning	NOUN
ajst-31431	107	7	.	.	PUNCT
ajst-31431	108	1	furthermore	furthermore	ADV
ajst-31431	108	2	,	,	PUNCT
ajst-31431	108	3	their	their	PRON
ajst-31431	108	4	work	work	NOUN
ajst-31431	108	5	introduced	introduce	VERB
ajst-31431	108	6	an	an	DET
ajst-31431	108	7	unsupervised	unsupervised	ADJ
ajst-31431	108	8	hybrid	hybrid	ADJ
ajst-31431	108	9	approach	approach	NOUN
ajst-31431	108	10	combining	combine	VERB
ajst-31431	108	11	bert	bert	PROPN
ajst-31431	108	12	's	's	PART
ajst-31431	108	13	multi	multi	ADJ
ajst-31431	108	14	-	-	ADJ
ajst-31431	108	15	headed	headed	ADJ
ajst-31431	108	16	self	self	NOUN
ajst-31431	108	17	-	-	PUNCT
ajst-31431	108	18	attention	attention	NOUN
ajst-31431	108	19	for	for	ADP
ajst-31431	108	20	contextual	contextual	ADJ
ajst-31431	108	21	word	word	NOUN
ajst-31431	108	22	relations	relation	NOUN
ajst-31431	108	23	with	with	ADP
ajst-31431	108	24	graph	graph	NOUN
ajst-31431	108	25	-	-	PUNCT
ajst-31431	108	26	based	base	VERB
ajst-31431	108	27	word	word	NOUN
ajst-31431	108	28	inference	inference	NOUN
ajst-31431	108	29	,	,	PUNCT
ajst-31431	108	30	significantly	significantly	ADV
ajst-31431	108	31	43	43	NUM
ajst-31431	108	32	enhancing	enhance	VERB
ajst-31431	108	33	detection	detection	NOUN
ajst-31431	108	34	efficiency	efficiency	NOUN
ajst-31431	108	35	.	.	PUNCT
ajst-31431	109	1	(	(	PUNCT
ajst-31431	109	2	figure	figure	NOUN
ajst-31431	109	3	1	1	NUM
ajst-31431	109	4	illustrates	illustrate	VERB
ajst-31431	109	5	the	the	DET
ajst-31431	109	6	general	general	ADJ
ajst-31431	109	7	hate	hate	NOUN
ajst-31431	109	8	speech	speech	NOUN
ajst-31431	109	9	detection	detection	NOUN
ajst-31431	109	10	pipeline	pipeline	NOUN
ajst-31431	109	11	.	.	PUNCT
ajst-31431	109	12	)	)	PUNCT
ajst-31431	110	1	figure	figure	VERB
ajst-31431	110	2	1	1	NUM
ajst-31431	110	3	.	.	PUNCT
ajst-31431	111	1	text	text	NOUN
ajst-31431	111	2	hate	hate	NOUN
ajst-31431	111	3	speech	speech	NOUN
ajst-31431	111	4	detection	detection	NOUN
ajst-31431	111	5	flow	flow	NOUN
ajst-31431	111	6	chart	chart	NOUN
ajst-31431	111	7	4	4	NUM
ajst-31431	111	8	.	.	PUNCT
ajst-31431	111	9	multimodal	multimodal	NOUN
ajst-31431	111	10	hate	hate	NOUN
ajst-31431	111	11	speech	speech	NOUN
ajst-31431	111	12	detection	detection	NOUN
ajst-31431	111	13	multimodal	multimodal	NOUN
ajst-31431	111	14	analysis	analysis	NOUN
ajst-31431	111	15	,	,	PUNCT
ajst-31431	111	16	integrating	integrate	VERB
ajst-31431	111	17	diverse	diverse	ADJ
ajst-31431	111	18	data	datum	NOUN
ajst-31431	111	19	modalities	modality	NOUN
ajst-31431	111	20	(	(	PUNCT
ajst-31431	111	21	e.g.	e.g.	ADV
ajst-31431	111	22	,	,	PUNCT
ajst-31431	111	23	audio	audio	ADJ
ajst-31431	111	24	,	,	PUNCT
ajst-31431	111	25	visual	visual	ADJ
ajst-31431	111	26	,	,	PUNCT
ajst-31431	111	27	textual	textual	ADJ
ajst-31431	111	28	)	)	PUNCT
ajst-31431	111	29	,	,	PUNCT
ajst-31431	111	30	has	have	AUX
ajst-31431	111	31	gained	gain	VERB
ajst-31431	111	32	significant	significant	ADJ
ajst-31431	111	33	recent	recent	ADJ
ajst-31431	111	34	attention	attention	NOUN
ajst-31431	111	35	.	.	PUNCT
ajst-31431	112	1	hateful	hateful	ADJ
ajst-31431	112	2	meme	meme	NOUN
ajst-31431	112	3	classification	classification	NOUN
ajst-31431	112	4	exemplifies	exemplify	VERB
ajst-31431	112	5	a	a	DET
ajst-31431	112	6	key	key	ADJ
ajst-31431	112	7	challenge	challenge	NOUN
ajst-31431	112	8	in	in	ADP
ajst-31431	112	9	this	this	DET
ajst-31431	112	10	domain	domain	NOUN
ajst-31431	112	11	[	[	X
ajst-31431	112	12	35	35	NUM
ajst-31431	112	13	]	]	PUNCT
ajst-31431	112	14	.	.	PUNCT
ajst-31431	113	1	memes	meme	NOUN
ajst-31431	113	2	,	,	PUNCT
ajst-31431	113	3	as	as	ADP
ajst-31431	113	4	pervasive	pervasive	ADJ
ajst-31431	113	5	cultural	cultural	ADJ
ajst-31431	113	6	vehicles	vehicle	NOUN
ajst-31431	113	7	on	on	ADP
ajst-31431	113	8	the	the	DET
ajst-31431	113	9	internet	internet	NOUN
ajst-31431	113	10	(	(	PUNCT
ajst-31431	113	11	images	image	NOUN
ajst-31431	113	12	,	,	PUNCT
ajst-31431	113	13	videos	video	NOUN
ajst-31431	113	14	,	,	PUNCT
ajst-31431	113	15	posts	post	NOUN
ajst-31431	113	16	)	)	PUNCT
ajst-31431	113	17	,	,	PUNCT
ajst-31431	113	18	pose	pose	VERB
ajst-31431	113	19	a	a	DET
ajst-31431	113	20	significant	significant	ADJ
ajst-31431	113	21	societal	societal	ADJ
ajst-31431	113	22	threat	threat	NOUN
ajst-31431	113	23	due	due	ADP
ajst-31431	113	24	to	to	ADP
ajst-31431	113	25	their	their	PRON
ajst-31431	113	26	inherent	inherent	ADJ
ajst-31431	113	27	complexity	complexity	NOUN
ajst-31431	113	28	:	:	PUNCT
ajst-31431	113	29	meaning	meaning	NOUN
ajst-31431	113	30	often	often	ADV
ajst-31431	113	31	emerges	emerge	VERB
ajst-31431	113	32	not	not	PART
ajst-31431	113	33	from	from	ADP
ajst-31431	113	34	isolated	isolated	ADJ
ajst-31431	113	35	text	text	NOUN
ajst-31431	113	36	or	or	CCONJ
ajst-31431	113	37	visual	visual	ADJ
ajst-31431	113	38	components	component	NOUN
ajst-31431	113	39	,	,	PUNCT
ajst-31431	113	40	which	which	PRON
ajst-31431	113	41	may	may	AUX
ajst-31431	113	42	appear	appear	VERB
ajst-31431	113	43	benign	benign	ADJ
ajst-31431	113	44	individually	individually	ADV
ajst-31431	113	45	,	,	PUNCT
ajst-31431	113	46	but	but	CCONJ
ajst-31431	113	47	from	from	ADP
ajst-31431	113	48	their	their	PRON
ajst-31431	113	49	contextual	contextual	ADJ
ajst-31431	113	50	interplay	interplay	NOUN
ajst-31431	113	51	.	.	PUNCT
ajst-31431	114	1	this	this	DET
ajst-31431	114	2	multimodal	multimodal	NOUN
ajst-31431	114	3	fusion	fusion	NOUN
ajst-31431	114	4	frequently	frequently	ADV
ajst-31431	114	5	employs	employ	VERB
ajst-31431	114	6	irony	irony	NOUN
ajst-31431	114	7	,	,	PUNCT
ajst-31431	114	8	humor	humor	NOUN
ajst-31431	114	9	,	,	PUNCT
ajst-31431	114	10	or	or	CCONJ
ajst-31431	114	11	contextual	contextual	ADJ
ajst-31431	114	12	association	association	NOUN
ajst-31431	114	13	to	to	PART
ajst-31431	114	14	generate	generate	VERB
ajst-31431	114	15	novel	novel	NOUN
ajst-31431	114	16	—	—	PUNCT
ajst-31431	114	17	and	and	CCONJ
ajst-31431	114	18	often	often	ADV
ajst-31431	114	19	offensive	offensive	ADJ
ajst-31431	114	20	—	—	PUNCT
ajst-31431	114	21	implicit	implicit	ADJ
ajst-31431	114	22	meanings	meaning	NOUN
ajst-31431	114	23	,	,	PUNCT
ajst-31431	114	24	facilitating	facilitate	VERB
ajst-31431	114	25	the	the	DET
ajst-31431	114	26	creation	creation	NOUN
ajst-31431	114	27	and	and	CCONJ
ajst-31431	114	28	spread	spread	VERB
ajst-31431	114	29	of	of	ADP
ajst-31431	114	30	hate	hate	NOUN
ajst-31431	114	31	speech	speech	NOUN
ajst-31431	114	32	.	.	PUNCT
ajst-31431	115	1	consequently	consequently	ADV
ajst-31431	115	2	,	,	PUNCT
ajst-31431	115	3	while	while	SCONJ
ajst-31431	115	4	detection	detection	NOUN
ajst-31431	115	5	of	of	ADP
ajst-31431	115	6	hateful	hateful	ADJ
ajst-31431	115	7	,	,	PUNCT
ajst-31431	115	8	offensive	offensive	ADJ
ajst-31431	115	9	,	,	PUNCT
ajst-31431	115	10	or	or	CCONJ
ajst-31431	115	11	aggressive	aggressive	ADJ
ajst-31431	115	12	content	content	NOUN
ajst-31431	115	13	within	within	ADP
ajst-31431	115	14	unimodal	unimodal	ADJ
ajst-31431	115	15	data	datum	NOUN
ajst-31431	115	16	(	(	PUNCT
ajst-31431	115	17	text	text	NOUN
ajst-31431	115	18	-	-	PUNCT
ajst-31431	115	19	only	only	ADV
ajst-31431	115	20	or	or	CCONJ
ajst-31431	115	21	image	image	NOUN
ajst-31431	115	22	-	-	PUNCT
ajst-31431	115	23	only	only	ADV
ajst-31431	115	24	)	)	PUNCT
ajst-31431	115	25	is	be	AUX
ajst-31431	115	26	relatively	relatively	ADV
ajst-31431	115	27	mature	mature	ADJ
ajst-31431	115	28	,	,	PUNCT
ajst-31431	115	29	accurately	accurately	ADV
ajst-31431	115	30	identifying	identify	VERB
ajst-31431	115	31	hateful	hateful	ADJ
ajst-31431	115	32	memes	meme	NOUN
ajst-31431	115	33	necessitates	necessitate	VERB
ajst-31431	115	34	the	the	DET
ajst-31431	115	35	joint	joint	ADJ
ajst-31431	115	36	analysis	analysis	NOUN
ajst-31431	115	37	and	and	CCONJ
ajst-31431	115	38	comprehension	comprehension	NOUN
ajst-31431	115	39	of	of	ADP
ajst-31431	115	40	both	both	CCONJ
ajst-31431	115	41	textual	textual	ADJ
ajst-31431	115	42	and	and	CCONJ
ajst-31431	115	43	visual	visual	ADJ
ajst-31431	115	44	modalities	modality	NOUN
ajst-31431	115	45	.	.	PUNCT
ajst-31431	116	1	research	research	NOUN
ajst-31431	116	2	focused	focus	VERB
ajst-31431	116	3	on	on	ADP
ajst-31431	116	4	effectively	effectively	ADV
ajst-31431	116	5	combining	combine	VERB
ajst-31431	116	6	these	these	DET
ajst-31431	116	7	modalities	modality	NOUN
ajst-31431	116	8	for	for	ADP
ajst-31431	116	9	harmful	harmful	ADJ
ajst-31431	116	10	content	content	NOUN
ajst-31431	116	11	identification	identification	NOUN
ajst-31431	116	12	remains	remain	VERB
ajst-31431	116	13	in	in	ADP
ajst-31431	116	14	its	its	PRON
ajst-31431	116	15	nascent	nascent	ADJ
ajst-31431	116	16	stages	stage	NOUN
ajst-31431	116	17	.	.	PUNCT
ajst-31431	117	1	references	reference	NOUN
ajst-31431	117	2	are	be	AUX
ajst-31431	117	3	cited	cite	VERB
ajst-31431	117	4	in	in	ADP
ajst-31431	117	5	the	the	DET
ajst-31431	117	6	text	text	NOUN
ajst-31431	117	7	just	just	ADV
ajst-31431	117	8	by	by	ADP
ajst-31431	117	9	square	square	ADJ
ajst-31431	117	10	brackets	bracket	NOUN
ajst-31431	117	11	.	.	PUNCT
ajst-31431	118	1	(	(	PUNCT
ajst-31431	118	2	if	if	SCONJ
ajst-31431	118	3	square	square	ADJ
ajst-31431	118	4	brackets	bracket	NOUN
ajst-31431	118	5	are	be	AUX
ajst-31431	118	6	not	not	PART
ajst-31431	118	7	available	available	ADJ
ajst-31431	118	8	,	,	PUNCT
ajst-31431	118	9	slashes	slash	VERB
ajst-31431	118	10	may	may	AUX
ajst-31431	118	11	be	be	AUX
ajst-31431	118	12	used	use	VERB
ajst-31431	118	13	instead	instead	ADV
ajst-31431	118	14	,	,	PUNCT
ajst-31431	118	15	e.g.	e.g.	ADV
ajst-31431	118	16	)	)	PUNCT
ajst-31431	118	17	two	two	NUM
ajst-31431	118	18	or	or	CCONJ
ajst-31431	118	19	more	more	ADJ
ajst-31431	118	20	references	reference	NOUN
ajst-31431	118	21	at	at	ADP
ajst-31431	118	22	a	a	DET
ajst-31431	118	23	time	time	NOUN
ajst-31431	118	24	may	may	AUX
ajst-31431	118	25	be	be	AUX
ajst-31431	118	26	put	put	VERB
ajst-31431	118	27	in	in	ADP
ajst-31431	118	28	one	one	NUM
ajst-31431	118	29	set	set	NOUN
ajst-31431	118	30	of	of	ADP
ajst-31431	118	31	brackets	bracket	NOUN
ajst-31431	118	32	.	.	PUNCT
ajst-31431	119	1	the	the	DET
ajst-31431	119	2	references	reference	NOUN
ajst-31431	119	3	are	be	AUX
ajst-31431	119	4	to	to	PART
ajst-31431	119	5	be	be	AUX
ajst-31431	119	6	numbered	number	VERB
ajst-31431	119	7	in	in	ADP
ajst-31431	119	8	the	the	DET
ajst-31431	119	9	order	order	NOUN
ajst-31431	119	10	in	in	ADP
ajst-31431	119	11	which	which	PRON
ajst-31431	119	12	they	they	PRON
ajst-31431	119	13	are	be	AUX
ajst-31431	119	14	cited	cite	VERB
ajst-31431	119	15	in	in	ADP
ajst-31431	119	16	the	the	DET
ajst-31431	119	17	text	text	NOUN
ajst-31431	119	18	and	and	CCONJ
ajst-31431	119	19	are	be	AUX
ajst-31431	119	20	to	to	PART
ajst-31431	119	21	be	be	AUX
ajst-31431	119	22	listed	list	VERB
ajst-31431	119	23	at	at	ADP
ajst-31431	119	24	the	the	DET
ajst-31431	119	25	end	end	NOUN
ajst-31431	119	26	of	of	ADP
ajst-31431	119	27	the	the	DET
ajst-31431	119	28	contribution	contribution	NOUN
ajst-31431	119	29	under	under	ADP
ajst-31431	119	30	a	a	DET
ajst-31431	119	31	heading	heading	ADJ
ajst-31431	119	32	references	reference	NOUN
ajst-31431	119	33	,	,	PUNCT
ajst-31431	119	34	see	see	VERB
ajst-31431	119	35	our	our	PRON
ajst-31431	119	36	example	example	NOUN
ajst-31431	119	37	below	below	ADV
ajst-31431	119	38	.	.	PUNCT
ajst-31431	120	1	4.1	4.1	NUM
ajst-31431	120	2	.	.	PUNCT
ajst-31431	120	3	.	.	PUNCT
ajst-31431	121	1	general	general	ADJ
ajst-31431	121	2	structure	structure	NOUN
ajst-31431	121	3	for	for	ADP
ajst-31431	121	4	meme	meme	NOUN
ajst-31431	121	5	classification	classification	NOUN
ajst-31431	121	6	meme	meme	NOUN
ajst-31431	121	7	classification	classification	NOUN
ajst-31431	121	8	constitutes	constitute	VERB
ajst-31431	121	9	a	a	DET
ajst-31431	121	10	complex	complex	ADJ
ajst-31431	121	11	vision	vision	NOUN
ajst-31431	121	12	-	-	PUNCT
ajst-31431	121	13	language	language	NOUN
ajst-31431	121	14	multimodal	multimodal	NOUN
ajst-31431	121	15	task	task	NOUN
ajst-31431	121	16	.	.	PUNCT
ajst-31431	122	1	it	it	PRON
ajst-31431	122	2	differs	differ	VERB
ajst-31431	122	3	significantly	significantly	ADV
ajst-31431	122	4	from	from	ADP
ajst-31431	122	5	other	other	ADJ
ajst-31431	122	6	visionlanguage	visionlanguage	NOUN
ajst-31431	122	7	problems	problem	NOUN
ajst-31431	122	8	,	,	PUNCT
ajst-31431	122	9	such	such	ADJ
ajst-31431	122	10	as	as	ADP
ajst-31431	122	11	image	image	NOUN
ajst-31431	122	12	captioning	captioning	NOUN
ajst-31431	122	13	,	,	PUNCT
ajst-31431	122	14	where	where	SCONJ
ajst-31431	122	15	the	the	DET
ajst-31431	122	16	goal	goal	NOUN
ajst-31431	122	17	is	be	AUX
ajst-31431	122	18	to	to	PART
ajst-31431	122	19	generate	generate	VERB
ajst-31431	122	20	descriptive	descriptive	ADJ
ajst-31431	122	21	text	text	NOUN
ajst-31431	122	22	for	for	ADP
ajst-31431	122	23	an	an	DET
ajst-31431	122	24	image	image	NOUN
ajst-31431	122	25	.	.	PUNCT
ajst-31431	123	1	in	in	ADP
ajst-31431	123	2	contrast	contrast	NOUN
ajst-31431	123	3	,	,	PUNCT
ajst-31431	123	4	meme	meme	NOUN
ajst-31431	123	5	classification	classification	NOUN
ajst-31431	123	6	requires	require	VERB
ajst-31431	123	7	a	a	DET
ajst-31431	123	8	decision	decision	NOUN
ajst-31431	123	9	based	base	VERB
ajst-31431	123	10	on	on	ADP
ajst-31431	123	11	text	text	NOUN
ajst-31431	123	12	that	that	PRON
ajst-31431	123	13	is	be	AUX
ajst-31431	123	14	semantically	semantically	ADV
ajst-31431	123	15	related	relate	VERB
ajst-31431	123	16	to	to	ADP
ajst-31431	123	17	the	the	DET
ajst-31431	123	18	accompanying	accompanying	ADJ
ajst-31431	123	19	visual	visual	ADJ
ajst-31431	123	20	content	content	NOUN
ajst-31431	123	21	.	.	PUNCT
ajst-31431	124	1	consequently	consequently	ADV
ajst-31431	124	2	,	,	PUNCT
ajst-31431	124	3	cross	cross	ADJ
ajst-31431	124	4	-	-	ADJ
ajst-31431	124	5	modal	modal	ADJ
ajst-31431	124	6	approaches	approach	NOUN
ajst-31431	124	7	integrating	integrate	VERB
ajst-31431	124	8	both	both	DET
ajst-31431	124	9	visual	visual	ADJ
ajst-31431	124	10	and	and	CCONJ
ajst-31431	124	11	textual	textual	ADJ
ajst-31431	124	12	information	information	NOUN
ajst-31431	124	13	are	be	AUX
ajst-31431	124	14	likely	likely	ADJ
ajst-31431	124	15	to	to	PART
ajst-31431	124	16	yield	yield	VERB
ajst-31431	124	17	superior	superior	ADJ
ajst-31431	124	18	performance	performance	NOUN
ajst-31431	124	19	for	for	ADP
ajst-31431	124	20	this	this	DET
ajst-31431	124	21	task	task	NOUN
ajst-31431	124	22	.	.	PUNCT
ajst-31431	125	1	while	while	SCONJ
ajst-31431	125	2	traditional	traditional	ADJ
ajst-31431	125	3	vision	vision	NOUN
ajst-31431	125	4	-	-	PUNCT
ajst-31431	125	5	language	language	NOUN
ajst-31431	125	6	methods	method	NOUN
ajst-31431	125	7	rely	rely	VERB
ajst-31431	125	8	on	on	ADP
ajst-31431	125	9	simplistic	simplistic	ADJ
ajst-31431	125	10	early	early	ADJ
ajst-31431	125	11	or	or	CCONJ
ajst-31431	125	12	late	late	ADJ
ajst-31431	125	13	fusion	fusion	NOUN
ajst-31431	125	14	of	of	ADP
ajst-31431	125	15	separately	separately	ADV
ajst-31431	125	16	learned	learn	VERB
ajst-31431	125	17	unimodal	unimodal	ADJ
ajst-31431	125	18	features	feature	NOUN
ajst-31431	125	19	,	,	PUNCT
ajst-31431	125	20	models	model	NOUN
ajst-31431	125	21	benefiting	benefit	VERB
ajst-31431	125	22	from	from	ADP
ajst-31431	125	23	multimodal	multimodal	ADJ
ajst-31431	125	24	pre	pre	ADJ
ajst-31431	125	25	-	-	ADJ
ajst-31431	125	26	training	training	ADJ
ajst-31431	125	27	demonstrate	demonstrate	NOUN
ajst-31431	125	28	enhanced	enhance	VERB
ajst-31431	125	29	potential	potential	NOUN
ajst-31431	125	30	for	for	ADP
ajst-31431	125	31	meme	meme	NOUN
ajst-31431	125	32	classification	classification	NOUN
ajst-31431	125	33	.	.	PUNCT
ajst-31431	126	1	based	base	VERB
ajst-31431	126	2	on	on	ADP
ajst-31431	126	3	a	a	DET
ajst-31431	126	4	comprehensive	comprehensive	ADJ
ajst-31431	126	5	literature	literature	NOUN
ajst-31431	126	6	review	review	NOUN
ajst-31431	126	7	,	,	PUNCT
ajst-31431	126	8	we	we	PRON
ajst-31431	126	9	propose	propose	VERB
ajst-31431	126	10	a	a	DET
ajst-31431	126	11	generalized	generalized	ADJ
ajst-31431	126	12	multimodal	multimodal	NOUN
ajst-31431	126	13	architecture	architecture	NOUN
ajst-31431	126	14	for	for	ADP
ajst-31431	126	15	meme	meme	NOUN
ajst-31431	126	16	classification	classification	NOUN
ajst-31431	126	17	,	,	PUNCT
ajst-31431	126	18	illustrated	illustrate	VERB
ajst-31431	126	19	in	in	ADP
ajst-31431	126	20	figure	figure	NOUN
ajst-31431	126	21	2	2	NUM
ajst-31431	126	22	.	.	PUNCT
ajst-31431	127	1	this	this	DET
ajst-31431	127	2	architecture	architecture	NOUN
ajst-31431	127	3	comprises	comprise	VERB
ajst-31431	127	4	two	two	NUM
ajst-31431	127	5	primary	primary	ADJ
ajst-31431	127	6	processing	processing	NOUN
ajst-31431	127	7	flows	flow	NOUN
ajst-31431	127	8	:	:	PUNCT
ajst-31431	127	9	a	a	DET
ajst-31431	127	10	language	language	NOUN
ajst-31431	127	11	processing	processing	NOUN
ajst-31431	127	12	flow	flow	NOUN
ajst-31431	127	13	(	(	PUNCT
ajst-31431	127	14	lpf	lpf	PROPN
ajst-31431	127	15	)	)	PUNCT
ajst-31431	127	16	and	and	CCONJ
ajst-31431	127	17	a	a	DET
ajst-31431	127	18	visual	visual	ADJ
ajst-31431	127	19	processing	processing	NOUN
ajst-31431	127	20	flow	flow	NOUN
ajst-31431	127	21	(	(	PUNCT
ajst-31431	127	22	vpf	vpf	NOUN
ajst-31431	127	23	)	)	PUNCT
ajst-31431	127	24	.	.	PUNCT
ajst-31431	128	1	a	a	DET
ajst-31431	128	2	central	central	ADJ
ajst-31431	128	3	fusion	fusion	NOUN
ajst-31431	128	4	and	and	CCONJ
ajst-31431	128	5	pre	pre	NOUN
ajst-31431	128	6	-	-	NOUN
ajst-31431	128	7	training	training	ADJ
ajst-31431	128	8	(	(	PUNCT
ajst-31431	128	9	fpt	fpt	NOUN
ajst-31431	128	10	)	)	PUNCT
ajst-31431	128	11	stage	stage	NOUN
ajst-31431	128	12	defines	define	VERB
ajst-31431	128	13	the	the	DET
ajst-31431	128	14	strategies	strategy	NOUN
ajst-31431	128	15	for	for	ADP
ajst-31431	128	16	merging	merge	VERB
ajst-31431	128	17	these	these	DET
ajst-31431	128	18	flows	flow	NOUN
ajst-31431	128	19	and	and	CCONJ
ajst-31431	128	20	incorporating	incorporate	VERB
ajst-31431	128	21	pretrained	pretraine	VERB
ajst-31431	128	22	knowledge	knowledge	NOUN
ajst-31431	128	23	both	both	CCONJ
ajst-31431	128	24	nlp	nlp	NOUN
ajst-31431	128	25	and	and	CCONJ
ajst-31431	128	26	computer	computer	NOUN
ajst-31431	128	27	vision	vision	NOUN
ajst-31431	128	28	possess	possess	VERB
ajst-31431	128	29	rich	rich	ADJ
ajst-31431	128	30	histories	history	NOUN
ajst-31431	128	31	in	in	ADP
ajst-31431	128	32	machine	machine	NOUN
ajst-31431	128	33	learning	learning	NOUN
ajst-31431	128	34	,	,	PUNCT
ajst-31431	128	35	which	which	PRON
ajst-31431	128	36	we	we	PRON
ajst-31431	128	37	categorize	categorize	VERB
ajst-31431	128	38	into	into	ADP
ajst-31431	128	39	first	first	ADJ
ajst-31431	128	40	-	-	PUNCT
ajst-31431	128	41	generation	generation	NOUN
ajst-31431	128	42	(	(	PUNCT
ajst-31431	128	43	1	1	NUM
ajst-31431	128	44	g	g	NOUN
ajst-31431	128	45	)	)	PUNCT
ajst-31431	128	46	and	and	CCONJ
ajst-31431	128	47	second	second	ADJ
ajst-31431	128	48	-	-	PUNCT
ajst-31431	128	49	generation	generation	NOUN
ajst-31431	128	50	(	(	PUNCT
ajst-31431	128	51	2	2	NUM
ajst-31431	128	52	g	g	NOUN
ajst-31431	128	53	)	)	PUNCT
ajst-31431	128	54	approaches	approach	NOUN
ajst-31431	128	55	.	.	PUNCT
ajst-31431	129	1	the	the	DET
ajst-31431	129	2	advent	advent	NOUN
ajst-31431	129	3	of	of	ADP
ajst-31431	129	4	deep	deep	ADJ
ajst-31431	129	5	learning	learning	NOUN
ajst-31431	129	6	,	,	PUNCT
ajst-31431	129	7	catalyzed	catalyze	VERB
ajst-31431	129	8	by	by	ADP
ajst-31431	129	9	breakthroughs	breakthrough	NOUN
ajst-31431	129	10	like	like	ADP
ajst-31431	129	11	alexnet	alexnet	NOUN
ajst-31431	129	12	[	[	X
ajst-31431	129	13	36	36	NUM
ajst-31431	129	14	]	]	PUNCT
ajst-31431	129	15	for	for	ADP
ajst-31431	129	16	vision	vision	NOUN
ajst-31431	129	17	(	(	PUNCT
ajst-31431	129	18	particularly	particularly	ADV
ajst-31431	129	19	cnns	cnns	ADJ
ajst-31431	129	20	)	)	PUNCT
ajst-31431	129	21	and	and	CCONJ
ajst-31431	129	22	bert	bert	PROPN
ajst-31431	129	23	for	for	ADP
ajst-31431	129	24	nlp	nlp	NOUN
ajst-31431	129	25	,	,	PUNCT
ajst-31431	129	26	marks	mark	VERB
ajst-31431	129	27	the	the	DET
ajst-31431	129	28	transition	transition	NOUN
ajst-31431	129	29	to	to	ADP
ajst-31431	129	30	these	these	DET
ajst-31431	129	31	second	second	ADJ
ajst-31431	129	32	-	-	PUNCT
ajst-31431	129	33	generation	generation	NOUN
ajst-31431	129	34	methods	method	NOUN
ajst-31431	129	35	.	.	PUNCT
ajst-31431	130	1	accordingly	accordingly	ADV
ajst-31431	130	2	,	,	PUNCT
ajst-31431	130	3	we	we	PRON
ajst-31431	130	4	delineate	delineate	VERB
ajst-31431	130	5	both	both	CCONJ
ajst-31431	130	6	the	the	DET
ajst-31431	130	7	lpf	lpf	PROPN
ajst-31431	130	8	and	and	CCONJ
ajst-31431	130	9	vpf	vpf	VERB
ajst-31431	130	10	into	into	ADP
ajst-31431	130	11	their	their	PRON
ajst-31431	130	12	respective	respective	ADJ
ajst-31431	130	13	1	1	NUM
ajst-31431	130	14	g	g	NOUN
ajst-31431	130	15	and	and	CCONJ
ajst-31431	130	16	2	2	NUM
ajst-31431	130	17	g	g	NOUN
ajst-31431	130	18	implementations	implementation	NOUN
ajst-31431	130	19	.	.	PUNCT
ajst-31431	131	1	each	each	DET
ajst-31431	131	2	generation	generation	NOUN
ajst-31431	131	3	will	will	AUX
ajst-31431	131	4	be	be	AUX
ajst-31431	131	5	elaborated	elaborate	VERB
ajst-31431	131	6	upon	upon	SCONJ
ajst-31431	131	7	in	in	ADP
ajst-31431	131	8	subsequent	subsequent	ADJ
ajst-31431	131	9	sections	section	NOUN
ajst-31431	131	10	.	.	PUNCT
ajst-31431	132	1	figure	figure	NOUN
ajst-31431	132	2	2	2	NUM
ajst-31431	132	3	.	.	PUNCT
ajst-31431	132	4	meme	meme	NOUN
ajst-31431	132	5	classification	classification	NOUN
ajst-31431	132	6	flowchart	flowchart	NOUN
ajst-31431	132	7	44	44	NUM
ajst-31431	132	8	4.1.1	4.1.1	NUM
ajst-31431	132	9	.	.	PUNCT
ajst-31431	133	1	language	language	NOUN
ajst-31431	133	2	processing	processing	NOUN
ajst-31431	133	3	flow	flow	NOUN
ajst-31431	133	4	prior	prior	ADV
ajst-31431	133	5	to	to	ADP
ajst-31431	133	6	the	the	DET
ajst-31431	133	7	rise	rise	NOUN
ajst-31431	133	8	of	of	ADP
ajst-31431	133	9	deep	deep	ADJ
ajst-31431	133	10	learning	learning	NOUN
ajst-31431	133	11	(	(	PUNCT
ajst-31431	133	12	dl	dl	NOUN
ajst-31431	133	13	)	)	PUNCT
ajst-31431	133	14	,	,	PUNCT
ajst-31431	133	15	natural	natural	ADJ
ajst-31431	133	16	language	language	NOUN
ajst-31431	133	17	processing	processing	NOUN
ajst-31431	133	18	(	(	PUNCT
ajst-31431	133	19	nlp	nlp	NOUN
ajst-31431	133	20	)	)	PUNCT
ajst-31431	133	21	pipelines	pipeline	NOUN
ajst-31431	133	22	predominantly	predominantly	ADV
ajst-31431	133	23	relied	rely	VERB
ajst-31431	133	24	on	on	ADP
ajst-31431	133	25	machine	machine	NOUN
ajst-31431	133	26	learning	learning	NOUN
ajst-31431	133	27	(	(	PUNCT
ajst-31431	133	28	ml	ml	NOUN
ajst-31431	133	29	)	)	PUNCT
ajst-31431	133	30	techniques	technique	NOUN
ajst-31431	133	31	involving	involve	VERB
ajst-31431	133	32	sequential	sequential	ADJ
ajst-31431	133	33	stages	stage	NOUN
ajst-31431	133	34	:	:	PUNCT
ajst-31431	133	35	preprocessing	preprocesse	VERB
ajst-31431	133	36	(	(	PUNCT
ajst-31431	133	37	stopword	stopword	PROPN
ajst-31431	133	38	removal	removal	NOUN
ajst-31431	133	39	,	,	PUNCT
ajst-31431	133	40	tokenization	tokenization	NOUN
ajst-31431	133	41	,	,	PUNCT
ajst-31431	133	42	lemmatization	lemmatization	NOUN
ajst-31431	133	43	)	)	PUNCT
ajst-31431	133	44	,	,	PUNCT
ajst-31431	133	45	feature	feature	NOUN
ajst-31431	133	46	engineering	engineering	NOUN
ajst-31431	133	47	(	(	PUNCT
ajst-31431	133	48	handcrafted	handcrafted	ADJ
ajst-31431	133	49	/	/	SYM
ajst-31431	133	50	statistical	statistical	ADJ
ajst-31431	133	51	features	feature	NOUN
ajst-31431	133	52	like	like	ADP
ajst-31431	133	53	bag	bag	NOUN
ajst-31431	133	54	-	-	PUNCT
ajst-31431	133	55	of	of	ADP
ajst-31431	133	56	-	-	PUNCT
ajst-31431	133	57	words	word	NOUN
ajst-31431	133	58	,	,	PUNCT
ajst-31431	133	59	n	n	CCONJ
ajst-31431	133	60	-	-	PUNCT
ajst-31431	133	61	grams	gram	NOUN
ajst-31431	133	62	,	,	PUNCT
ajst-31431	133	63	tf	tf	PROPN
ajst-31431	133	64	-	-	PUNCT
ajst-31431	133	65	idf	idf	PROPN
ajst-31431	133	66	,	,	PUNCT
ajst-31431	133	67	or	or	CCONJ
ajst-31431	133	68	static	static	ADJ
ajst-31431	133	69	word	word	NOUN
ajst-31431	133	70	embeddings	embedding	NOUN
ajst-31431	133	71	from	from	ADP
ajst-31431	133	72	word2vec	word2vec	ADJ
ajst-31431	133	73	,	,	PUNCT
ajst-31431	133	74	glove	glove	NOUN
ajst-31431	133	75	,	,	PUNCT
ajst-31431	133	76	fasttext	fasttext	PROPN
ajst-31431	133	77	)	)	PUNCT
ajst-31431	133	78	,	,	PUNCT
ajst-31431	133	79	dimensionality	dimensionality	NOUN
ajst-31431	133	80	reduction	reduction	NOUN
ajst-31431	133	81	(	(	PUNCT
ajst-31431	133	82	e.g.	e.g.	ADV
ajst-31431	133	83	,	,	PUNCT
ajst-31431	133	84	pca	pca	PROPN
ajst-31431	133	85	,	,	PUNCT
ajst-31431	133	86	lda	lda	PROPN
ajst-31431	133	87	)	)	PUNCT
ajst-31431	133	88	,	,	PUNCT
ajst-31431	133	89	and	and	CCONJ
ajst-31431	133	90	classification	classification	NOUN
ajst-31431	133	91	(	(	PUNCT
ajst-31431	133	92	e.g.	e.g.	ADV
ajst-31431	133	93	,	,	PUNCT
ajst-31431	133	94	svm	svm	ADJ
ajst-31431	133	95	,	,	PUNCT
ajst-31431	133	96	naive	naive	ADJ
ajst-31431	133	97	bayes	bayes	NOUN
ajst-31431	133	98	)	)	PUNCT
ajst-31431	133	99	.	.	PUNCT
ajst-31431	134	1	while	while	SCONJ
ajst-31431	134	2	static	static	ADJ
ajst-31431	134	3	embeddings	embedding	NOUN
ajst-31431	134	4	captured	capture	VERB
ajst-31431	134	5	general	general	ADJ
ajst-31431	134	6	semantics	semantic	NOUN
ajst-31431	134	7	better	well	ADV
ajst-31431	134	8	than	than	ADP
ajst-31431	134	9	bow	bow	VERB
ajst-31431	134	10	/	/	SYM
ajst-31431	134	11	n	n	CCONJ
ajst-31431	134	12	-	-	PUNCT
ajst-31431	134	13	grams	gram	NOUN
ajst-31431	134	14	,	,	PUNCT
ajst-31431	134	15	they	they	PRON
ajst-31431	134	16	lacked	lack	VERB
ajst-31431	134	17	contextual	contextual	ADJ
ajst-31431	134	18	awareness	awareness	NOUN
ajst-31431	134	19	,	,	PUNCT
ajst-31431	134	20	failing	fail	VERB
ajst-31431	134	21	to	to	PART
ajst-31431	134	22	model	model	NOUN
ajst-31431	134	23	polysemy	polysemy	PROPN
ajst-31431	134	24	,	,	PUNCT
ajst-31431	134	25	syntax	syntax	NOUN
ajst-31431	134	26	,	,	PUNCT
ajst-31431	134	27	semantics	semantic	NOUN
ajst-31431	134	28	,	,	PUNCT
ajst-31431	134	29	and	and	CCONJ
ajst-31431	134	30	coreference	coreference	NOUN
ajst-31431	134	31	resolution	resolution	NOUN
ajst-31431	134	32	effectively	effectively	ADV
ajst-31431	134	33	.	.	PUNCT
ajst-31431	135	1	the	the	DET
ajst-31431	135	2	advent	advent	NOUN
ajst-31431	135	3	of	of	ADP
ajst-31431	135	4	dl	dl	PROPN
ajst-31431	135	5	revolutionized	revolutionize	VERB
ajst-31431	135	6	nlp	nlp	NOUN
ajst-31431	135	7	through	through	ADP
ajst-31431	135	8	its	its	PRON
ajst-31431	135	9	capacity	capacity	NOUN
ajst-31431	135	10	for	for	ADP
ajst-31431	135	11	modeling	model	VERB
ajst-31431	135	12	complex	complex	ADJ
ajst-31431	135	13	nonlinear	nonlinear	ADJ
ajst-31431	135	14	relationships	relationship	NOUN
ajst-31431	135	15	.	.	PUNCT
ajst-31431	136	1	this	this	DET
ajst-31431	136	2	"	"	PUNCT
ajst-31431	136	3	secondgeneration	secondgeneration	NOUN
ajst-31431	136	4	"	"	PUNCT
ajst-31431	136	5	shift	shift	NOUN
ajst-31431	136	6	introduced	introduce	VERB
ajst-31431	136	7	neural	neural	ADJ
ajst-31431	136	8	architectures	architecture	NOUN
ajst-31431	136	9	like	like	ADP
ajst-31431	136	10	rnns	rnn	NOUN
ajst-31431	136	11	,	,	PUNCT
ajst-31431	136	12	cnns	cnn	NOUN
ajst-31431	136	13	,	,	PUNCT
ajst-31431	136	14	and	and	CCONJ
ajst-31431	136	15	transformative	transformative	ADJ
ajst-31431	136	16	transformer	transformer	NOUN
ajst-31431	136	17	-	-	PUNCT
ajst-31431	136	18	based	base	VERB
ajst-31431	136	19	models	model	NOUN
ajst-31431	136	20	(	(	PUNCT
ajst-31431	136	21	e.g.	e.g.	ADV
ajst-31431	136	22	,	,	PUNCT
ajst-31431	136	23	bert	bert	PROPN
ajst-31431	136	24	,	,	PUNCT
ajst-31431	136	25	gpt-2	gpt-2	NOUN
ajst-31431	136	26	,	,	PUNCT
ajst-31431	136	27	roberta	roberta	PROPN
ajst-31431	136	28	,	,	PUNCT
ajst-31431	136	29	albert	albert	PROPN
ajst-31431	137	1	[	[	X
ajst-31431	137	2	37	37	NUM
ajst-31431	137	3	]	]	NUM
ajst-31431	137	4	)	)	PUNCT
ajst-31431	137	5	.	.	PUNCT
ajst-31431	138	1	bert	bert	PROPN
ajst-31431	138	2	's	's	PART
ajst-31431	138	3	breakthrough	breakthrough	NOUN
ajst-31431	138	4	,	,	PUNCT
ajst-31431	138	5	building	build	VERB
ajst-31431	138	6	upon	upon	SCONJ
ajst-31431	138	7	semi	semi	ADJ
ajst-31431	138	8	-	-	ADJ
ajst-31431	138	9	supervised	supervised	ADJ
ajst-31431	138	10	learning	learning	NOUN
ajst-31431	138	11	,	,	PUNCT
ajst-31431	138	12	contextual	contextual	ADJ
ajst-31431	138	13	representations	representation	NOUN
ajst-31431	138	14	(	(	PUNCT
ajst-31431	138	15	e.g.	e.g.	ADV
ajst-31431	138	16	,	,	PUNCT
ajst-31431	138	17	elmo	elmo	PROPN
ajst-31431	138	18	)	)	PUNCT
ajst-31431	138	19	,	,	PUNCT
ajst-31431	138	20	transfer	transfer	NOUN
ajst-31431	138	21	learning	learning	NOUN
ajst-31431	138	22	(	(	PUNCT
ajst-31431	138	23	ulmfit	ulmfit	VERB
ajst-31431	138	24	[	[	X
ajst-31431	138	25	38	38	NUM
ajst-31431	138	26	]	]	PUNCT
ajst-31431	138	27	)	)	PUNCT
ajst-31431	138	28	,	,	PUNCT
ajst-31431	138	29	and	and	CCONJ
ajst-31431	138	30	the	the	DET
ajst-31431	138	31	transformer	transformer	NOUN
ajst-31431	138	32	architecture	architecture	NOUN
ajst-31431	138	33	,	,	PUNCT
ajst-31431	138	34	marked	mark	VERB
ajst-31431	138	35	a	a	DET
ajst-31431	138	36	new	new	ADJ
ajst-31431	138	37	era	era	NOUN
ajst-31431	138	38	.	.	PUNCT
ajst-31431	139	1	transformers	transformer	NOUN
ajst-31431	139	2	utilize	utilize	VERB
ajst-31431	139	3	self	self	NOUN
ajst-31431	139	4	-	-	PUNCT
ajst-31431	139	5	attention	attention	NOUN
ajst-31431	139	6	to	to	PART
ajst-31431	139	7	capture	capture	VERB
ajst-31431	139	8	intrasequence	intrasequence	NOUN
ajst-31431	139	9	relationships	relationship	NOUN
ajst-31431	139	10	and	and	CCONJ
ajst-31431	139	11	enable	enable	ADJ
ajst-31431	139	12	parallelization	parallelization	NOUN
ajst-31431	139	13	,	,	PUNCT
ajst-31431	139	14	drastically	drastically	ADV
ajst-31431	139	15	reducing	reduce	VERB
ajst-31431	139	16	training	training	NOUN
ajst-31431	139	17	time	time	NOUN
ajst-31431	139	18	.	.	PUNCT
ajst-31431	140	1	the	the	DET
ajst-31431	140	2	core	core	ADJ
ajst-31431	140	3	architecture	architecture	NOUN
ajst-31431	140	4	stacks	stack	NOUN
ajst-31431	140	5	encoder	encoder	NOUN
ajst-31431	140	6	/	/	SYM
ajst-31431	140	7	decoder	decoder	NOUN
ajst-31431	140	8	layers	layer	NOUN
ajst-31431	140	9	,	,	PUNCT
ajst-31431	140	10	each	each	PRON
ajst-31431	140	11	containing	contain	VERB
ajst-31431	140	12	multi	multi	ADJ
ajst-31431	140	13	-	-	ADJ
ajst-31431	140	14	head	head	ADJ
ajst-31431	140	15	selfattention	selfattention	NOUN
ajst-31431	140	16	(	(	PUNCT
ajst-31431	140	17	capturing	capture	VERB
ajst-31431	140	18	diverse	diverse	ADJ
ajst-31431	140	19	representational	representational	ADJ
ajst-31431	140	20	subspaces	subspace	NOUN
ajst-31431	140	21	)	)	PUNCT
ajst-31431	140	22	and	and	CCONJ
ajst-31431	140	23	position	position	NOUN
ajst-31431	140	24	-	-	PUNCT
ajst-31431	140	25	wise	wise	ADJ
ajst-31431	140	26	feed	feed	NOUN
ajst-31431	140	27	-	-	PUNCT
ajst-31431	140	28	forward	forward	NOUN
ajst-31431	140	29	networks	network	NOUN
ajst-31431	140	30	.	.	PUNCT
ajst-31431	141	1	positional	positional	ADJ
ajst-31431	141	2	encodings	encoding	NOUN
ajst-31431	141	3	inject	inject	NOUN
ajst-31431	141	4	order	order	NOUN
ajst-31431	141	5	information	information	NOUN
ajst-31431	141	6	,	,	PUNCT
ajst-31431	141	7	though	though	SCONJ
ajst-31431	141	8	they	they	PRON
ajst-31431	141	9	offer	offer	VERB
ajst-31431	141	10	weaker	weak	ADJ
ajst-31431	141	11	sequential	sequential	ADJ
ajst-31431	141	12	modeling	modeling	NOUN
ajst-31431	141	13	than	than	ADP
ajst-31431	141	14	rnns	rnns	NOUN
ajst-31431	141	15	/	/	SYM
ajst-31431	141	16	lstms	lstms	NOUN
ajst-31431	141	17	for	for	ADP
ajst-31431	141	18	precise	precise	ADJ
ajst-31431	141	19	positional	positional	ADJ
ajst-31431	141	20	dependencies	dependency	NOUN
ajst-31431	141	21	.	.	PUNCT
ajst-31431	142	1	crucially	crucially	ADV
ajst-31431	142	2	,	,	PUNCT
ajst-31431	142	3	models	model	NOUN
ajst-31431	142	4	like	like	ADP
ajst-31431	142	5	bert	bert	PROPN
ajst-31431	142	6	simplify	simplify	NOUN
ajst-31431	142	7	application	application	NOUN
ajst-31431	142	8	by	by	ADP
ajst-31431	142	9	learning	learn	VERB
ajst-31431	142	10	rich	rich	ADJ
ajst-31431	142	11	linguistic	linguistic	ADJ
ajst-31431	142	12	knowledge	knowledge	NOUN
ajst-31431	142	13	via	via	ADP
ajst-31431	142	14	largescale	largescale	NOUN
ajst-31431	142	15	pretraining	pretraine	VERB
ajst-31431	142	16	and	and	CCONJ
ajst-31431	142	17	adapting	adapt	VERB
ajst-31431	142	18	efficiently	efficiently	ADV
ajst-31431	142	19	to	to	ADP
ajst-31431	142	20	downstream	downstream	ADJ
ajst-31431	142	21	tasks	task	NOUN
ajst-31431	142	22	through	through	ADP
ajst-31431	142	23	fine	fine	ADV
ajst-31431	142	24	-	-	PUNCT
ajst-31431	142	25	tuning	tuning	NOUN
ajst-31431	142	26	,	,	PUNCT
ajst-31431	142	27	eliminating	eliminate	VERB
ajst-31431	142	28	extensive	extensive	ADJ
ajst-31431	142	29	feature	feature	NOUN
ajst-31431	142	30	engineering	engineering	NOUN
ajst-31431	142	31	.	.	PUNCT
ajst-31431	143	1	4.1.2	4.1.2	X
ajst-31431	143	2	.	.	PUNCT
ajst-31431	143	3	visual	visual	ADJ
ajst-31431	143	4	processing	processing	NOUN
ajst-31431	143	5	flow	flow	NOUN
ajst-31431	143	6	the	the	DET
ajst-31431	143	7	success	success	NOUN
ajst-31431	143	8	of	of	ADP
ajst-31431	143	9	alexnet	alexnet	NOUN
ajst-31431	143	10	[	[	X
ajst-31431	143	11	36	36	NUM
ajst-31431	143	12	]	]	PUNCT
ajst-31431	143	13	shifted	shift	VERB
ajst-31431	143	14	research	research	NOUN
ajst-31431	143	15	focus	focus	NOUN
ajst-31431	143	16	from	from	ADP
ajst-31431	143	17	traditional	traditional	ADJ
ajst-31431	143	18	computer	computer	NOUN
ajst-31431	143	19	vision	vision	NOUN
ajst-31431	143	20	methods	method	NOUN
ajst-31431	143	21	—	—	PUNCT
ajst-31431	143	22	reliant	reliant	ADJ
ajst-31431	143	23	on	on	ADP
ajst-31431	143	24	handcrafted	handcraft	VERB
ajst-31431	143	25	features	feature	NOUN
ajst-31431	143	26	like	like	ADP
ajst-31431	143	27	lbp	lbp	NOUN
ajst-31431	143	28	,	,	PUNCT
ajst-31431	143	29	sift	sift	PROPN
ajst-31431	143	30	,	,	PUNCT
ajst-31431	143	31	hog	hog	NOUN
ajst-31431	143	32	,	,	PUNCT
ajst-31431	143	33	surf	surf	NOUN
ajst-31431	143	34	,	,	PUNCT
ajst-31431	143	35	and	and	CCONJ
ajst-31431	143	36	brief	brief	ADJ
ajst-31431	143	37	—	—	PUNCT
ajst-31431	143	38	towards	towards	ADP
ajst-31431	143	39	deep	deep	ADJ
ajst-31431	143	40	learning	learning	NOUN
ajst-31431	143	41	.	.	PUNCT
ajst-31431	144	1	traditional	traditional	ADJ
ajst-31431	144	2	approaches	approach	NOUN
ajst-31431	144	3	faced	face	VERB
ajst-31431	144	4	significant	significant	ADJ
ajst-31431	144	5	limitations	limitation	NOUN
ajst-31431	144	6	:	:	PUNCT
ajst-31431	144	7	feature	feature	NOUN
ajst-31431	144	8	engineering	engineering	NOUN
ajst-31431	144	9	required	require	VERB
ajst-31431	144	10	task	task	NOUN
ajst-31431	144	11	-	-	PUNCT
ajst-31431	144	12	specific	specific	ADJ
ajst-31431	144	13	redesign	redesign	NOUN
ajst-31431	144	14	,	,	PUNCT
ajst-31431	144	15	classifiers	classifier	NOUN
ajst-31431	144	16	lacked	lack	VERB
ajst-31431	144	17	transferability	transferability	NOUN
ajst-31431	144	18	,	,	PUNCT
ajst-31431	144	19	and	and	CCONJ
ajst-31431	144	20	the	the	DET
ajst-31431	144	21	overall	overall	ADJ
ajst-31431	144	22	pipeline	pipeline	NOUN
ajst-31431	144	23	(	(	PUNCT
ajst-31431	144	24	preprocessing	preprocessing	NOUN
ajst-31431	144	25	,	,	PUNCT
ajst-31431	144	26	feature	feature	NOUN
ajst-31431	144	27	extraction	extraction	NOUN
ajst-31431	144	28	,	,	PUNCT
ajst-31431	144	29	feature	feature	NOUN
ajst-31431	144	30	selection	selection	NOUN
ajst-31431	144	31	,	,	PUNCT
ajst-31431	144	32	classification	classification	NOUN
ajst-31431	144	33	)	)	PUNCT
ajst-31431	144	34	necessitated	necessitate	VERB
ajst-31431	144	35	inefficient	inefficient	ADJ
ajst-31431	144	36	optimization	optimization	NOUN
ajst-31431	144	37	of	of	ADP
ajst-31431	144	38	each	each	DET
ajst-31431	144	39	sub	sub	NOUN
ajst-31431	144	40	-	-	NOUN
ajst-31431	144	41	step	step	NOUN
ajst-31431	144	42	.	.	PUNCT
ajst-31431	145	1	convolutional	convolutional	ADJ
ajst-31431	145	2	neural	neural	ADJ
ajst-31431	145	3	networks	network	NOUN
ajst-31431	145	4	(	(	PUNCT
ajst-31431	145	5	cnns	cnns	PROPN
ajst-31431	145	6	)	)	PUNCT
ajst-31431	145	7	overcame	overcome	VERB
ajst-31431	145	8	these	these	DET
ajst-31431	145	9	issues	issue	NOUN
ajst-31431	145	10	by	by	ADP
ajst-31431	145	11	automatically	automatically	ADV
ajst-31431	145	12	learning	learn	VERB
ajst-31431	145	13	features	feature	NOUN
ajst-31431	145	14	from	from	ADP
ajst-31431	145	15	data	datum	NOUN
ajst-31431	145	16	and	and	CCONJ
ajst-31431	145	17	enabling	enable	VERB
ajst-31431	145	18	efficient	efficient	ADJ
ajst-31431	145	19	fine	fine	NOUN
ajst-31431	145	20	-	-	PUNCT
ajst-31431	145	21	tuning	tuning	NOUN
ajst-31431	145	22	for	for	ADP
ajst-31431	145	23	related	related	ADJ
ajst-31431	145	24	tasks	task	NOUN
ajst-31431	145	25	via	via	ADP
ajst-31431	145	26	transfer	transfer	NOUN
ajst-31431	145	27	learning	learning	NOUN
ajst-31431	145	28	.	.	PUNCT
ajst-31431	146	1	alexnet	alexnet	PROPN
ajst-31431	146	2	's	's	PART
ajst-31431	146	3	breakthrough	breakthrough	ADJ
ajst-31431	146	4	performance	performance	NOUN
ajst-31431	146	5	on	on	ADP
ajst-31431	146	6	imagenet	imagenet	NOUN
ajst-31431	146	7	catalyzed	catalyze	VERB
ajst-31431	146	8	a	a	DET
ajst-31431	146	9	paradigm	paradigm	NOUN
ajst-31431	146	10	shift	shift	NOUN
ajst-31431	146	11	,	,	PUNCT
ajst-31431	146	12	marking	mark	VERB
ajst-31431	146	13	the	the	DET
ajst-31431	146	14	advent	advent	NOUN
ajst-31431	146	15	of	of	ADP
ajst-31431	146	16	the	the	DET
ajst-31431	146	17	secondgeneration	secondgeneration	NOUN
ajst-31431	146	18	(	(	PUNCT
ajst-31431	146	19	2	2	NUM
ajst-31431	146	20	g	g	NOUN
ajst-31431	146	21	)	)	PUNCT
ajst-31431	146	22	era	era	NOUN
ajst-31431	146	23	for	for	ADP
ajst-31431	146	24	the	the	DET
ajst-31431	146	25	visual	visual	ADJ
ajst-31431	146	26	processing	processing	NOUN
ajst-31431	146	27	flow	flow	NOUN
ajst-31431	146	28	(	(	PUNCT
ajst-31431	146	29	vpf	vpf	NOUN
ajst-31431	146	30	)	)	PUNCT
ajst-31431	146	31	.	.	PUNCT
ajst-31431	147	1	deep	deep	ADJ
ajst-31431	147	2	learning	learning	NOUN
ajst-31431	147	3	models	model	NOUN
ajst-31431	147	4	,	,	PUNCT
ajst-31431	147	5	particularly	particularly	ADV
ajst-31431	147	6	cnns	cnns	ADJ
ajst-31431	147	7	,	,	PUNCT
ajst-31431	147	8	facilitate	facilitate	VERB
ajst-31431	147	9	the	the	DET
ajst-31431	147	10	integration	integration	NOUN
ajst-31431	147	11	of	of	ADP
ajst-31431	147	12	diverse	diverse	ADJ
ajst-31431	147	13	,	,	PUNCT
ajst-31431	147	14	large	large	ADJ
ajst-31431	147	15	-	-	PUNCT
ajst-31431	147	16	scale	scale	NOUN
ajst-31431	147	17	image	image	NOUN
ajst-31431	147	18	and	and	CCONJ
ajst-31431	147	19	video	video	NOUN
ajst-31431	147	20	datasets	dataset	NOUN
ajst-31431	147	21	,	,	PUNCT
ajst-31431	147	22	learning	learn	VERB
ajst-31431	147	23	robust	robust	ADJ
ajst-31431	147	24	features	feature	NOUN
ajst-31431	147	25	that	that	PRON
ajst-31431	147	26	generalize	generalize	VERB
ajst-31431	147	27	effectively	effectively	ADV
ajst-31431	147	28	.	.	PUNCT
ajst-31431	148	1	a	a	DET
ajst-31431	148	2	cnn	cnn	PROPN
ajst-31431	148	3	typically	typically	ADV
ajst-31431	148	4	combines	combine	VERB
ajst-31431	148	5	convolutional	convolutional	ADJ
ajst-31431	148	6	layers	layer	NOUN
ajst-31431	148	7	with	with	ADP
ajst-31431	148	8	activation	activation	NOUN
ajst-31431	148	9	functions	function	NOUN
ajst-31431	148	10	(	(	PUNCT
ajst-31431	148	11	e.g.	e.g.	ADV
ajst-31431	148	12	,	,	PUNCT
ajst-31431	148	13	relu	relu	NOUN
ajst-31431	148	14	variants	variant	NOUN
ajst-31431	148	15	)	)	PUNCT
ajst-31431	148	16	,	,	PUNCT
ajst-31431	148	17	downsampling	downsample	VERB
ajst-31431	148	18	layers	layer	NOUN
ajst-31431	148	19	(	(	PUNCT
ajst-31431	148	20	e.g.	e.g.	ADV
ajst-31431	148	21	,	,	PUNCT
ajst-31431	148	22	max	max	PROPN
ajst-31431	148	23	pooling	pooling	NOUN
ajst-31431	148	24	)	)	PUNCT
ajst-31431	148	25	,	,	PUNCT
ajst-31431	148	26	fully	fully	ADV
ajst-31431	148	27	convolutional	convolutional	ADJ
ajst-31431	148	28	layers	layer	NOUN
ajst-31431	148	29	,	,	PUNCT
ajst-31431	148	30	dense	dense	ADJ
ajst-31431	148	31	layers	layer	NOUN
ajst-31431	148	32	,	,	PUNCT
ajst-31431	148	33	and	and	CCONJ
ajst-31431	148	34	a	a	DET
ajst-31431	148	35	final	final	ADJ
ajst-31431	148	36	softmax	softmax	NOUN
ajst-31431	148	37	layer	layer	NOUN
ajst-31431	148	38	.	.	PUNCT
ajst-31431	149	1	post	post	ADJ
ajst-31431	149	2	-	-	ADJ
ajst-31431	149	3	alexnet	alexnet	ADJ
ajst-31431	149	4	,	,	PUNCT
ajst-31431	149	5	research	research	NOUN
ajst-31431	149	6	focused	focus	VERB
ajst-31431	149	7	on	on	ADP
ajst-31431	149	8	enhancing	enhance	VERB
ajst-31431	149	9	cnn	cnn	PROPN
ajst-31431	149	10	architectures	architecture	NOUN
ajst-31431	149	11	.	.	PUNCT
ajst-31431	150	1	vgg	vgg	VERB
ajst-31431	151	1	[	[	X
ajst-31431	151	2	39	39	NUM
ajst-31431	151	3	]	]	PUNCT
ajst-31431	151	4	and	and	CCONJ
ajst-31431	151	5	googlenet	googlenet	NOUN
ajst-31431	151	6	's	's	PART
ajst-31431	151	7	inception	inception	NOUN
ajst-31431	151	8	modules	module	NOUN
ajst-31431	151	9	demonstrated	demonstrate	VERB
ajst-31431	151	10	benefits	benefit	NOUN
ajst-31431	151	11	from	from	ADP
ajst-31431	151	12	increasing	increase	VERB
ajst-31431	151	13	depth	depth	NOUN
ajst-31431	151	14	and	and	CCONJ
ajst-31431	151	15	width	width	NOUN
ajst-31431	151	16	.	.	PUNCT
ajst-31431	152	1	resnets	resnet	NOUN
ajst-31431	153	1	[	[	X
ajst-31431	153	2	40	40	NUM
ajst-31431	153	3	]	]	PUNCT
ajst-31431	153	4	introduced	introduce	VERB
ajst-31431	153	5	residual	residual	ADJ
ajst-31431	153	6	learning	learning	NOUN
ajst-31431	153	7	blocks	block	NOUN
ajst-31431	153	8	with	with	ADP
ajst-31431	153	9	identity	identity	NOUN
ajst-31431	153	10	shortcut	shortcut	NOUN
ajst-31431	153	11	connections	connection	NOUN
ajst-31431	153	12	,	,	PUNCT
ajst-31431	153	13	enabling	enable	VERB
ajst-31431	153	14	training	training	NOUN
ajst-31431	153	15	of	of	ADP
ajst-31431	153	16	networks	network	NOUN
ajst-31431	153	17	hundreds	hundred	NOUN
ajst-31431	153	18	or	or	CCONJ
ajst-31431	153	19	thousands	thousand	NOUN
ajst-31431	153	20	of	of	ADP
ajst-31431	153	21	layers	layer	NOUN
ajst-31431	153	22	deep	deep	ADJ
ajst-31431	153	23	.	.	PUNCT
ajst-31431	154	1	densenet	densenet	NOUN
ajst-31431	155	1	[	[	X
ajst-31431	155	2	41	41	NUM
ajst-31431	155	3	]	]	PUNCT
ajst-31431	155	4	further	far	ADV
ajst-31431	155	5	enhanced	enhance	VERB
ajst-31431	155	6	representational	representational	ADJ
ajst-31431	155	7	power	power	NOUN
ajst-31431	155	8	through	through	ADP
ajst-31431	155	9	dense	dense	ADJ
ajst-31431	155	10	interlayer	interlayer	NOUN
ajst-31431	155	11	connectivity	connectivity	NOUN
ajst-31431	155	12	.	.	PUNCT
ajst-31431	156	1	significant	significant	ADJ
ajst-31431	156	2	progress	progress	NOUN
ajst-31431	156	3	was	be	AUX
ajst-31431	156	4	also	also	ADV
ajst-31431	156	5	made	make	VERB
ajst-31431	156	6	in	in	ADP
ajst-31431	156	7	object	object	NOUN
ajst-31431	156	8	detection	detection	NOUN
ajst-31431	156	9	with	with	ADP
ajst-31431	156	10	numerous	numerous	ADJ
ajst-31431	156	11	cnn	cnn	PROPN
ajst-31431	156	12	-	-	PUNCT
ajst-31431	156	13	based	base	VERB
ajst-31431	156	14	methods	method	NOUN
ajst-31431	156	15	(	(	PUNCT
ajst-31431	156	16	e.g.	e.g.	ADV
ajst-31431	156	17	,	,	PUNCT
ajst-31431	156	18	rcnn	rcnn	PROPN
ajst-31431	156	19	,	,	PUNCT
ajst-31431	156	20	faster	fast	ADJ
ajst-31431	156	21	r	r	NOUN
ajst-31431	156	22	-	-	PUNCT
ajst-31431	156	23	cnn	cnn	PROPN
ajst-31431	156	24	,	,	PUNCT
ajst-31431	156	25	yolo	yolo	PROPN
ajst-31431	156	26	)	)	PUNCT
ajst-31431	156	27	.	.	PUNCT
ajst-31431	157	1	4.1.3	4.1.3	X
ajst-31431	157	2	.	.	X
ajst-31431	157	3	fusion	fusion	NOUN
ajst-31431	157	4	and	and	CCONJ
ajst-31431	157	5	pre	pre	NOUN
ajst-31431	157	6	-	-	NOUN
ajst-31431	157	7	training	training	NOUN
ajst-31431	157	8	:	:	PUNCT
ajst-31431	157	9	towards	towards	ADP
ajst-31431	157	10	multimodality	multimodality	NOUN
ajst-31431	157	11	.	.	PUNCT
ajst-31431	158	1	following	follow	VERB
ajst-31431	158	2	,	,	PUNCT
ajst-31431	158	3	visual	visual	ADJ
ajst-31431	158	4	-	-	PUNCT
ajst-31431	158	5	language	language	NOUN
ajst-31431	158	6	(	(	PUNCT
ajst-31431	158	7	vl	vl	NOUN
ajst-31431	158	8	)	)	PUNCT
ajst-31431	158	9	multimodal	multimodal	NOUN
ajst-31431	158	10	fusion	fusion	NOUN
ajst-31431	158	11	strategies	strategy	NOUN
ajst-31431	158	12	are	be	AUX
ajst-31431	158	13	categorized	categorize	VERB
ajst-31431	158	14	into	into	ADP
ajst-31431	158	15	three	three	NUM
ajst-31431	158	16	types	type	NOUN
ajst-31431	158	17	:	:	PUNCT
ajst-31431	158	18	early	early	ADJ
ajst-31431	158	19	fusion	fusion	NOUN
ajst-31431	158	20	(	(	PUNCT
ajst-31431	158	21	combining	combine	VERB
ajst-31431	158	22	modalities	modality	NOUN
ajst-31431	158	23	at	at	ADP
ajst-31431	158	24	the	the	DET
ajst-31431	158	25	feature	feature	NOUN
ajst-31431	158	26	level	level	NOUN
ajst-31431	158	27	immediately	immediately	ADV
ajst-31431	158	28	after	after	ADP
ajst-31431	158	29	extraction	extraction	NOUN
ajst-31431	158	30	)	)	PUNCT
ajst-31431	158	31	,	,	PUNCT
ajst-31431	158	32	late	late	ADJ
ajst-31431	158	33	fusion	fusion	NOUN
ajst-31431	158	34	(	(	PUNCT
ajst-31431	158	35	integrating	integrate	VERB
ajst-31431	158	36	decisions	decision	NOUN
ajst-31431	158	37	made	make	VERB
ajst-31431	158	38	independently	independently	ADV
ajst-31431	158	39	by	by	ADP
ajst-31431	158	40	each	each	DET
ajst-31431	158	41	modality	modality	NOUN
ajst-31431	158	42	)	)	PUNCT
ajst-31431	158	43	,	,	PUNCT
ajst-31431	158	44	and	and	CCONJ
ajst-31431	158	45	hybrid	hybrid	ADJ
ajst-31431	158	46	fusion	fusion	NOUN
ajst-31431	158	47	(	(	PUNCT
ajst-31431	158	48	merging	merge	VERB
ajst-31431	158	49	outputs	output	NOUN
ajst-31431	158	50	from	from	ADP
ajst-31431	158	51	both	both	DET
ajst-31431	158	52	unimodal	unimodal	ADJ
ajst-31431	158	53	predictors	predictor	NOUN
ajst-31431	158	54	and	and	CCONJ
ajst-31431	158	55	an	an	DET
ajst-31431	158	56	early	early	ADJ
ajst-31431	158	57	fusion	fusion	NOUN
ajst-31431	158	58	component	component	NOUN
ajst-31431	158	59	)	)	PUNCT
ajst-31431	158	60	.	.	PUNCT
ajst-31431	159	1	pre	pre	ADJ
ajst-31431	159	2	-	-	ADJ
ajst-31431	159	3	training	training	ADJ
ajst-31431	159	4	approaches	approach	NOUN
ajst-31431	159	5	are	be	AUX
ajst-31431	159	6	similarly	similarly	ADV
ajst-31431	159	7	distinguished	distinguish	VERB
ajst-31431	159	8	:	:	PUNCT
ajst-31431	159	9	models	model	NOUN
ajst-31431	159	10	combining	combine	VERB
ajst-31431	159	11	unimodally	unimodally	ADV
ajst-31431	159	12	pre	pre	ADJ
ajst-31431	159	13	-	-	ADJ
ajst-31431	159	14	trained	train	VERB
ajst-31431	159	15	language	language	NOUN
ajst-31431	159	16	and	and	CCONJ
ajst-31431	159	17	vision	vision	NOUN
ajst-31431	159	18	components	component	NOUN
ajst-31431	159	19	via	via	ADP
ajst-31431	159	20	fusion	fusion	NOUN
ajst-31431	159	21	are	be	AUX
ajst-31431	159	22	termed	term	VERB
ajst-31431	159	23	unimodally	unimodally	ADV
ajst-31431	159	24	pre	pre	ADJ
ajst-31431	159	25	-	-	ADJ
ajst-31431	159	26	trained	train	VERB
ajst-31431	159	27	multimodal	multimodal	NOUN
ajst-31431	159	28	models	model	NOUN
ajst-31431	159	29	.	.	PUNCT
ajst-31431	160	1	conversely	conversely	ADV
ajst-31431	160	2	,	,	PUNCT
ajst-31431	160	3	models	model	NOUN
ajst-31431	160	4	pre	pre	ADJ
ajst-31431	160	5	-	-	VERB
ajst-31431	160	6	trained	train	VERB
ajst-31431	160	7	jointly	jointly	ADV
ajst-31431	160	8	on	on	ADP
ajst-31431	160	9	multimodal	multimodal	ADJ
ajst-31431	160	10	data	datum	NOUN
ajst-31431	160	11	are	be	AUX
ajst-31431	160	12	referred	refer	VERB
ajst-31431	160	13	to	to	ADP
ajst-31431	160	14	as	as	ADP
ajst-31431	160	15	multimodally	multimodally	ADV
ajst-31431	160	16	pre	pre	ADJ
ajst-31431	160	17	-	-	ADJ
ajst-31431	160	18	trained	train	VERB
ajst-31431	160	19	models	model	NOUN
ajst-31431	160	20	.	.	PUNCT
ajst-31431	161	1	4.2	4.2	NUM
ajst-31431	161	2	.	.	PUNCT
ajst-31431	162	1	the	the	DET
ajst-31431	162	2	latest	late	ADJ
ajst-31431	162	3	advances	advance	NOUN
ajst-31431	162	4	in	in	ADP
ajst-31431	162	5	meme	meme	NOUN
ajst-31431	162	6	classification	classification	NOUN
ajst-31431	162	7	given	give	VERB
ajst-31431	162	8	the	the	DET
ajst-31431	162	9	limited	limited	ADJ
ajst-31431	162	10	research	research	NOUN
ajst-31431	162	11	specifically	specifically	ADV
ajst-31431	162	12	on	on	ADP
ajst-31431	162	13	multimodal	multimodal	NOUN
ajst-31431	162	14	meme	meme	NOUN
ajst-31431	162	15	classification	classification	NOUN
ajst-31431	162	16	,	,	PUNCT
ajst-31431	162	17	we	we	PRON
ajst-31431	162	18	draw	draw	VERB
ajst-31431	162	19	inspiration	inspiration	NOUN
ajst-31431	162	20	from	from	ADP
ajst-31431	162	21	state	state	NOUN
ajst-31431	162	22	-	-	PUNCT
ajst-31431	162	23	of	of	ADP
ajst-31431	162	24	-	-	PUNCT
ajst-31431	162	25	the	the	DET
ajst-31431	162	26	-	-	PUNCT
ajst-31431	162	27	art	art	NOUN
ajst-31431	162	28	approaches	approach	NOUN
ajst-31431	162	29	in	in	ADP
ajst-31431	162	30	related	related	ADJ
ajst-31431	162	31	vision	vision	NOUN
ajst-31431	162	32	-	-	PUNCT
ajst-31431	162	33	language	language	NOUN
ajst-31431	162	34	(	(	PUNCT
ajst-31431	162	35	vl	vl	NOUN
ajst-31431	162	36	)	)	PUNCT
ajst-31431	162	37	tasks	task	NOUN
ajst-31431	162	38	.	.	PUNCT
ajst-31431	163	1	a	a	DET
ajst-31431	163	2	foundational	foundational	ADJ
ajst-31431	163	3	objective	objective	NOUN
ajst-31431	163	4	in	in	ADP
ajst-31431	163	5	vl	vl	PROPN
ajst-31431	163	6	exploration	exploration	NOUN
ajst-31431	163	7	is	be	AUX
ajst-31431	163	8	understanding	understand	VERB
ajst-31431	163	9	alignment	alignment	NOUN
ajst-31431	163	10	relationships	relationship	NOUN
ajst-31431	163	11	between	between	ADP
ajst-31431	163	12	multimodal	multimodal	ADJ
ajst-31431	163	13	feature	feature	NOUN
ajst-31431	163	14	spaces	space	NOUN
ajst-31431	163	15	.	.	PUNCT
ajst-31431	164	1	architectures	architecture	NOUN
ajst-31431	164	2	combining	combine	VERB
ajst-31431	164	3	cnns	cnn	NOUN
ajst-31431	164	4	and	and	CCONJ
ajst-31431	164	5	rnns	rnn	NOUN
ajst-31431	164	6	are	be	AUX
ajst-31431	164	7	commonly	commonly	ADV
ajst-31431	164	8	trained	train	VERB
ajst-31431	164	9	on	on	ADP
ajst-31431	164	10	aligned	aligned	ADJ
ajst-31431	164	11	multimodal	multimodal	NOUN
ajst-31431	164	12	data	datum	NOUN
ajst-31431	164	13	to	to	PART
ajst-31431	164	14	learn	learn	VERB
ajst-31431	164	15	a	a	DET
ajst-31431	164	16	joint	joint	ADJ
ajst-31431	164	17	embedding	embed	VERB
ajst-31431	164	18	space	space	NOUN
ajst-31431	164	19	,	,	PUNCT
ajst-31431	164	20	a	a	DET
ajst-31431	164	21	standard	standard	ADJ
ajst-31431	164	22	approach	approach	NOUN
ajst-31431	164	23	for	for	ADP
ajst-31431	164	24	tasks	task	NOUN
ajst-31431	164	25	like	like	ADP
ajst-31431	164	26	image	image	NOUN
ajst-31431	164	27	captioning	captioning	NOUN
ajst-31431	164	28	[	[	X
ajst-31431	164	29	75	75	NUM
ajst-31431	164	30	]	]	PUNCT
ajst-31431	164	31	.	.	PUNCT
ajst-31431	165	1	in	in	ADP
ajst-31431	165	2	contrast	contrast	NOUN
ajst-31431	165	3	,	,	PUNCT
ajst-31431	165	4	visual	visual	ADJ
ajst-31431	165	5	question	question	NOUN
ajst-31431	165	6	answering	answering	NOUN
ajst-31431	165	7	(	(	PUNCT
ajst-31431	165	8	vqa	vqa	ADJ
ajst-31431	165	9	)	)	PUNCT
ajst-31431	165	10	focuses	focus	VERB
ajst-31431	165	11	on	on	ADP
ajst-31431	165	12	fusing	fuse	VERB
ajst-31431	165	13	vl	vl	NOUN
ajst-31431	165	14	modalities	modality	NOUN
ajst-31431	165	15	to	to	PART
ajst-31431	165	16	determine	determine	VERB
ajst-31431	165	17	the	the	DET
ajst-31431	165	18	correct	correct	ADJ
ajst-31431	165	19	answer	answer	NOUN
ajst-31431	165	20	,	,	PUNCT
ajst-31431	165	21	necessitating	necessitate	VERB
ajst-31431	165	22	precise	precise	ADJ
ajst-31431	165	23	correlation	correlation	NOUN
ajst-31431	165	24	modeling	modeling	NOUN
ajst-31431	165	25	between	between	ADP
ajst-31431	165	26	image	image	NOUN
ajst-31431	165	27	and	and	CCONJ
ajst-31431	165	28	question	question	NOUN
ajst-31431	165	29	representations	representation	NOUN
ajst-31431	165	30	rather	rather	ADV
ajst-31431	165	31	than	than	ADP
ajst-31431	165	32	explicit	explicit	ADJ
ajst-31431	165	33	alignment	alignment	NOUN
ajst-31431	165	34	learning	learning	NOUN
ajst-31431	165	35	.	.	PUNCT
ajst-31431	166	1	hateful	hateful	ADJ
ajst-31431	166	2	meme	meme	NOUN
ajst-31431	166	3	detection	detection	NOUN
ajst-31431	166	4	similarly	similarly	ADV
ajst-31431	166	5	requires	require	VERB
ajst-31431	166	6	precise	precise	ADJ
ajst-31431	166	7	correlation	correlation	NOUN
ajst-31431	166	8	modeling	modeling	NOUN
ajst-31431	166	9	between	between	ADP
ajst-31431	166	10	image	image	NOUN
ajst-31431	166	11	and	and	CCONJ
ajst-31431	166	12	text	text	NOUN
ajst-31431	166	13	to	to	PART
ajst-31431	166	14	uncover	uncover	VERB
ajst-31431	166	15	latent	latent	ADJ
ajst-31431	166	16	inter	inter	ADJ
ajst-31431	166	17	-	-	ADJ
ajst-31431	166	18	modal	modal	ADJ
ajst-31431	166	19	relationships	relationship	NOUN
ajst-31431	166	20	for	for	ADP
ajst-31431	166	21	accurate	accurate	ADJ
ajst-31431	166	22	classification	classification	NOUN
ajst-31431	166	23	.	.	PUNCT
ajst-31431	167	1	consequently	consequently	ADV
ajst-31431	167	2	,	,	PUNCT
ajst-31431	167	3	we	we	PRON
ajst-31431	167	4	leverage	leverage	VERB
ajst-31431	167	5	insights	insight	NOUN
ajst-31431	167	6	from	from	ADP
ajst-31431	167	7	vqa	vqa	PROPN
ajst-31431	167	8	literature	literature	NOUN
ajst-31431	167	9	to	to	PART
ajst-31431	167	10	inform	inform	VERB
ajst-31431	167	11	the	the	DET
ajst-31431	167	12	development	development	NOUN
ajst-31431	167	13	of	of	ADP
ajst-31431	167	14	state	state	NOUN
ajst-31431	167	15	-	-	PUNCT
ajst-31431	167	16	of	of	ADP
ajst-31431	167	17	-	-	PUNCT
ajst-31431	167	18	the	the	DET
ajst-31431	167	19	-	-	PUNCT
ajst-31431	167	20	art	art	NOUN
ajst-31431	167	21	meme	meme	NOUN
ajst-31431	167	22	classification	classification	NOUN
ajst-31431	167	23	models	model	NOUN
ajst-31431	167	24	.	.	PUNCT
ajst-31431	168	1	early	early	ADJ
ajst-31431	168	2	vqa	vqa	PROPN
ajst-31431	168	3	research	research	NOUN
ajst-31431	168	4	employed	employ	VERB
ajst-31431	168	5	simple	simple	ADJ
ajst-31431	168	6	feature	feature	NOUN
ajst-31431	168	7	concatenation	concatenation	NOUN
ajst-31431	168	8	(	(	PUNCT
ajst-31431	168	9	early	early	ADJ
ajst-31431	168	10	fusion	fusion	NOUN
ajst-31431	168	11	)	)	PUNCT
ajst-31431	168	12	,	,	PUNCT
ajst-31431	168	13	while	while	SCONJ
ajst-31431	168	14	later	later	ADV
ajst-31431	168	15	methods	method	NOUN
ajst-31431	168	16	utilized	utilize	VERB
ajst-31431	169	1	bilinear	bilinear	PROPN
ajst-31431	169	2	pooling	pool	VERB
ajst-31431	169	3	for	for	ADP
ajst-31431	169	4	multimodal	multimodal	ADJ
ajst-31431	169	5	feature	feature	NOUN
ajst-31431	169	6	learning	learning	NOUN
ajst-31431	169	7	.	.	PUNCT
ajst-31431	170	1	however	however	ADV
ajst-31431	170	2	,	,	PUNCT
ajst-31431	170	3	these	these	DET
ajst-31431	170	4	approaches	approach	NOUN
ajst-31431	170	5	faced	face	VERB
ajst-31431	170	6	significant	significant	ADJ
ajst-31431	170	7	limitations	limitation	NOUN
ajst-31431	170	8	:	:	PUNCT
ajst-31431	170	9	fusing	fuse	VERB
ajst-31431	170	10	features	feature	NOUN
ajst-31431	170	11	late	late	ADV
ajst-31431	170	12	in	in	ADP
ajst-31431	170	13	the	the	DET
ajst-31431	170	14	model	model	NOUN
ajst-31431	170	15	hindered	hinder	VERB
ajst-31431	170	16	effective	effective	ADJ
ajst-31431	170	17	vl	vl	ADJ
ajst-31431	170	18	alignment	alignment	NOUN
ajst-31431	170	19	extraction	extraction	NOUN
ajst-31431	170	20	,	,	PUNCT
ajst-31431	170	21	and	and	CCONJ
ajst-31431	170	22	representing	represent	VERB
ajst-31431	170	23	cnn	cnn	PROPN
ajst-31431	170	24	outputs	output	NOUN
ajst-31431	170	25	as	as	SCONJ
ajst-31431	170	26	1d	1d	NUM
ajst-31431	170	27	vectors	vector	NOUN
ajst-31431	170	28	led	lead	VERB
ajst-31431	170	29	to	to	ADP
ajst-31431	170	30	substantial	substantial	ADJ
ajst-31431	170	31	degradation	degradation	NOUN
ajst-31431	170	32	of	of	ADP
ajst-31431	170	33	spatial	spatial	ADJ
ajst-31431	170	34	information	information	NOUN
ajst-31431	170	35	[	[	X
ajst-31431	170	36	42	42	NUM
ajst-31431	170	37	]	]	PUNCT
ajst-31431	170	38	.	.	PUNCT
ajst-31431	171	1	recent	recent	ADJ
ajst-31431	171	2	efforts	effort	NOUN
ajst-31431	171	3	have	have	AUX
ajst-31431	171	4	shifted	shift	VERB
ajst-31431	171	5	towards	towards	ADP
ajst-31431	171	6	cross	cross	ADJ
ajst-31431	171	7	-	-	ADJ
ajst-31431	171	8	modal	modal	ADJ
ajst-31431	171	9	learning	learning	NOUN
ajst-31431	171	10	via	via	ADP
ajst-31431	171	11	multimodal	multimodal	ADJ
ajst-31431	171	12	pretrained	pretraine	VERB
ajst-31431	171	13	models	model	NOUN
ajst-31431	171	14	like	like	ADP
ajst-31431	171	15	visualbert	visualbert	NOUN
ajst-31431	171	16	and	and	CCONJ
ajst-31431	171	17	uniter	uniter	NOUN
ajst-31431	171	18	[	[	X
ajst-31431	171	19	43	43	NUM
ajst-31431	171	20	]	]	PUNCT
ajst-31431	171	21	,	,	PUNCT
ajst-31431	171	22	which	which	PRON
ajst-31431	171	23	have	have	AUX
ajst-31431	171	24	surpassed	surpass	VERB
ajst-31431	171	25	previous	previous	ADJ
ajst-31431	171	26	methods	method	NOUN
ajst-31431	171	27	across	across	ADP
ajst-31431	171	28	multiple	multiple	ADJ
ajst-31431	171	29	vl	vl	PROPN
ajst-31431	171	30	benchmarks	benchmark	NOUN
ajst-31431	171	31	.	.	PUNCT
ajst-31431	172	1	5	5	X
ajst-31431	172	2	.	.	X
ajst-31431	172	3	conclusion	conclusion	NOUN
ajst-31431	172	4	this	this	DET
ajst-31431	172	5	paper	paper	NOUN
ajst-31431	172	6	presents	present	VERB
ajst-31431	172	7	a	a	DET
ajst-31431	172	8	comprehensive	comprehensive	ADJ
ajst-31431	172	9	review	review	NOUN
ajst-31431	172	10	of	of	ADP
ajst-31431	172	11	recent	recent	ADJ
ajst-31431	172	12	advancements	advancement	NOUN
ajst-31431	172	13	in	in	ADP
ajst-31431	172	14	hate	hate	NOUN
ajst-31431	172	15	speech	speech	NOUN
ajst-31431	172	16	detection	detection	NOUN
ajst-31431	172	17	on	on	ADP
ajst-31431	172	18	social	social	ADJ
ajst-31431	172	19	media	medium	NOUN
ajst-31431	172	20	.	.	PUNCT
ajst-31431	173	1	although	although	SCONJ
ajst-31431	173	2	hate	hate	NOUN
ajst-31431	173	3	speech	speech	NOUN
ajst-31431	173	4	represents	represent	VERB
ajst-31431	173	5	a	a	DET
ajst-31431	173	6	relatively	relatively	ADV
ajst-31431	173	7	nascent	nascent	ADJ
ajst-31431	173	8	research	research	NOUN
ajst-31431	173	9	domain	domain	NOUN
ajst-31431	173	10	within	within	ADP
ajst-31431	173	11	computer	computer	NOUN
ajst-31431	173	12	science	science	NOUN
ajst-31431	173	13	,	,	PUNCT
ajst-31431	173	14	it	it	PRON
ajst-31431	173	15	has	have	AUX
ajst-31431	173	16	long	long	ADV
ajst-31431	173	17	been	be	AUX
ajst-31431	173	18	a	a	DET
ajst-31431	173	19	significant	significant	ADJ
ajst-31431	173	20	concern	concern	NOUN
ajst-31431	173	21	in	in	ADP
ajst-31431	173	22	the	the	DET
ajst-31431	173	23	humanities	humanity	NOUN
ajst-31431	173	24	and	and	CCONJ
ajst-31431	173	25	arts	art	NOUN
ajst-31431	173	26	.	.	PUNCT
ajst-31431	174	1	consequently	consequently	ADV
ajst-31431	174	2	,	,	PUNCT
ajst-31431	174	3	this	this	DET
ajst-31431	174	4	review	review	NOUN
ajst-31431	174	5	aims	aim	VERB
ajst-31431	174	6	to	to	PART
ajst-31431	174	7	provide	provide	VERB
ajst-31431	174	8	researchers	researcher	NOUN
ajst-31431	174	9	with	with	ADP
ajst-31431	174	10	a	a	DET
ajst-31431	174	11	thorough	thorough	ADJ
ajst-31431	174	12	understanding	understanding	NOUN
ajst-31431	174	13	of	of	ADP
ajst-31431	174	14	current	current	ADJ
ajst-31431	174	15	developments	development	NOUN
ajst-31431	174	16	in	in	ADP
ajst-31431	174	17	the	the	DET
ajst-31431	174	18	field	field	NOUN
ajst-31431	174	19	.	.	PUNCT
ajst-31431	175	1	the	the	DET
ajst-31431	175	2	analysis	analysis	NOUN
ajst-31431	175	3	encompasses	encompass	VERB
ajst-31431	175	4	a	a	DET
ajst-31431	175	5	broad	broad	ADJ
ajst-31431	175	6	spectrum	spectrum	NOUN
ajst-31431	175	7	of	of	ADP
ajst-31431	175	8	methods	method	NOUN
ajst-31431	175	9	,	,	PUNCT
ajst-31431	175	10	including	include	VERB
ajst-31431	175	11	traditional	traditional	ADJ
ajst-31431	175	12	machine	machine	NOUN
ajst-31431	175	13	learning	learning	NOUN
ajst-31431	175	14	,	,	PUNCT
ajst-31431	175	15	ensemble	ensemble	ADJ
ajst-31431	175	16	techniques	technique	NOUN
ajst-31431	175	17	,	,	PUNCT
ajst-31431	175	18	and	and	CCONJ
ajst-31431	175	19	deep	deep	ADJ
ajst-31431	175	20	learning	learning	NOUN
ajst-31431	175	21	approaches	approach	NOUN
ajst-31431	175	22	specifically	specifically	ADV
ajst-31431	175	23	applied	apply	VERB
ajst-31431	175	24	to	to	PART
ajst-31431	175	25	hate	hate	VERB
ajst-31431	175	26	speech	speech	NOUN
ajst-31431	175	27	detection	detection	NOUN
ajst-31431	175	28	in	in	ADP
ajst-31431	175	29	social	social	ADJ
ajst-31431	175	30	media	medium	NOUN
ajst-31431	175	31	content	content	NOUN
ajst-31431	175	32	.	.	PUNCT
ajst-31431	176	1	by	by	ADP
ajst-31431	176	2	synthesizing	synthesize	VERB
ajst-31431	176	3	insights	insight	NOUN
ajst-31431	176	4	from	from	ADP
ajst-31431	176	5	published	publish	VERB
ajst-31431	176	6	literature	literature	NOUN
ajst-31431	176	7	,	,	PUNCT
ajst-31431	176	8	45	45	NUM
ajst-31431	176	9	this	this	DET
ajst-31431	176	10	work	work	NOUN
ajst-31431	176	11	serves	serve	VERB
ajst-31431	176	12	as	as	ADP
ajst-31431	176	13	a	a	DET
ajst-31431	176	14	valuable	valuable	ADJ
ajst-31431	176	15	reference	reference	NOUN
ajst-31431	176	16	for	for	ADP
ajst-31431	176	17	researchers	researcher	NOUN
ajst-31431	176	18	.	.	PUNCT
ajst-31431	177	1	it	it	PRON
ajst-31431	177	2	underscores	underscore	VERB
ajst-31431	177	3	the	the	DET
ajst-31431	177	4	critical	critical	ADJ
ajst-31431	177	5	importance	importance	NOUN
ajst-31431	177	6	of	of	ADP
ajst-31431	177	7	ongoing	ongoing	ADJ
ajst-31431	177	8	research	research	NOUN
ajst-31431	177	9	in	in	ADP
ajst-31431	177	10	this	this	DET
ajst-31431	177	11	area	area	NOUN
ajst-31431	177	12	,	,	PUNCT
ajst-31431	177	13	enabling	enable	VERB
ajst-31431	177	14	the	the	DET
ajst-31431	177	15	academic	academic	ADJ
ajst-31431	177	16	community	community	NOUN
ajst-31431	177	17	to	to	PART
ajst-31431	177	18	focus	focus	VERB
ajst-31431	177	19	on	on	ADP
ajst-31431	177	20	advancing	advance	VERB
ajst-31431	177	21	methodologies	methodology	NOUN
ajst-31431	177	22	for	for	ADP
ajst-31431	177	23	hate	hate	NOUN
ajst-31431	177	24	speech	speech	NOUN
ajst-31431	177	25	detection	detection	NOUN
ajst-31431	177	26	.	.	PUNCT
ajst-31431	178	1	acknowledgements	acknowledgement	NOUN
ajst-31431	178	2	the	the	DET
ajst-31431	178	3	authors	author	NOUN
ajst-31431	178	4	gratefully	gratefully	ADV
ajst-31431	178	5	acknowledge	acknowledge	VERB
ajst-31431	178	6	the	the	DET
ajst-31431	178	7	financial	financial	ADJ
ajst-31431	178	8	support	support	NOUN
ajst-31431	178	9	from	from	ADP
ajst-31431	178	10	the	the	DET
ajst-31431	178	11	double	double	ADJ
ajst-31431	178	12	first	first	ADJ
ajst-31431	178	13	-	-	PUNCT
ajst-31431	178	14	class	class	NOUN
ajst-31431	178	15	special	special	ADJ
ajst-31431	178	16	task	task	NOUN
ajst-31431	178	17	for	for	ADP
ajst-31431	178	18	security	security	NOUN
ajst-31431	178	19	and	and	CCONJ
ajst-31431	178	20	prevention	prevention	NOUN
ajst-31431	178	21	engineering	engineering	NOUN
ajst-31431	178	22	at	at	ADP
ajst-31431	178	23	the	the	DET
ajst-31431	178	24	chinese	chinese	ADJ
ajst-31431	178	25	people	people	NOUN
ajst-31431	178	26	’s	’s	PART
ajst-31431	178	27	public	public	ADJ
ajst-31431	178	28	security	security	NOUN
ajst-31431	178	29	university	university	NOUN
ajst-31431	178	30	.	.	PUNCT
ajst-31431	179	1	(	(	PUNCT
ajst-31431	179	2	2023syl08	2023syl08	NUM
ajst-31431	179	3	)	)	PUNCT
ajst-31431	179	4	references	reference	NOUN
ajst-31431	179	5	[	[	X
ajst-31431	179	6	1	1	NUM
ajst-31431	179	7	]	]	X
ajst-31431	179	8	blaya	blaya	PROPN
ajst-31431	179	9	c.	c.	PROPN
ajst-31431	179	10	cyberhate	cyberhate	PROPN
ajst-31431	179	11	:	:	PUNCT
ajst-31431	179	12	a	a	DET
ajst-31431	179	13	review	review	NOUN
ajst-31431	179	14	and	and	CCONJ
ajst-31431	179	15	content	content	VERB
ajst-31431	179	16	analysis	analysis	NOUN
ajst-31431	179	17	of	of	ADP
ajst-31431	179	18	intervention	intervention	NOUN
ajst-31431	179	19	strategies[j	strategies[j	PROPN
ajst-31431	179	20	]	]	PUNCT
ajst-31431	179	21	.	.	PUNCT
ajst-31431	180	1	aggression	aggression	NOUN
ajst-31431	180	2	and	and	CCONJ
ajst-31431	180	3	violent	violent	ADJ
ajst-31431	180	4	behavior	behavior	NOUN
ajst-31431	180	5	,	,	PUNCT
ajst-31431	180	6	2019	2019	NUM
ajst-31431	180	7	,	,	PUNCT
ajst-31431	180	8	45	45	NUM
ajst-31431	180	9	:	:	SYM
ajst-31431	180	10	163	163	NUM
ajst-31431	180	11	-	-	SYM
ajst-31431	180	12	172	172	NUM
ajst-31431	180	13	.	.	PUNCT
ajst-31431	181	1	[	[	X
ajst-31431	181	2	2	2	NUM
ajst-31431	181	3	]	]	X
ajst-31431	181	4	talat	talat	PROPN
ajst-31431	181	5	z	z	PROPN
ajst-31431	181	6	,	,	PUNCT
ajst-31431	181	7	hovy	hovy	PROPN
ajst-31431	181	8	d.	d.	PROPN
ajst-31431	181	9	hateful	hateful	ADJ
ajst-31431	181	10	symbols	symbol	NOUN
ajst-31431	181	11	or	or	CCONJ
ajst-31431	181	12	hateful	hateful	ADJ
ajst-31431	181	13	people	people	NOUN
ajst-31431	181	14	?	?	PUNCT
ajst-31431	182	1	predictive	predictive	ADJ
ajst-31431	182	2	features	feature	NOUN
ajst-31431	182	3	for	for	ADP
ajst-31431	182	4	hate	hate	NOUN
ajst-31431	182	5	speech	speech	NOUN
ajst-31431	182	6	detection	detection	NOUN
ajst-31431	182	7	on	on	ADP
ajst-31431	182	8	twitter[c]//proceedings	twitter[c]//proceeding	NOUN
ajst-31431	182	9	of	of	ADP
ajst-31431	182	10	the	the	DET
ajst-31431	182	11	naacl	naacl	PROPN
ajst-31431	182	12	student	student	NOUN
ajst-31431	182	13	research	research	NOUN
ajst-31431	182	14	workshop	workshop	NOUN
ajst-31431	182	15	.	.	PUNCT
ajst-31431	183	1	2016	2016	NUM
ajst-31431	183	2	:	:	PUNCT
ajst-31431	183	3	88	88	NUM
ajst-31431	183	4	-	-	SYM
ajst-31431	183	5	93	93	NUM
ajst-31431	183	6	.	.	PUNCT
ajst-31431	184	1	[	[	X
ajst-31431	184	2	3	3	X
ajst-31431	184	3	]	]	X
ajst-31431	184	4	rozental	rozental	NOUN
ajst-31431	184	5	a	a	PRON
ajst-31431	184	6	,	,	PUNCT
ajst-31431	184	7	biton	biton	PROPN
ajst-31431	184	8	d.	d.	PROPN
ajst-31431	184	9	amobee	amobee	PROPN
ajst-31431	184	10	at	at	ADP
ajst-31431	184	11	semeval-2019	semeval-2019	ADJ
ajst-31431	184	12	tasks	task	NOUN
ajst-31431	184	13	5	5	NUM
ajst-31431	184	14	and	and	CCONJ
ajst-31431	184	15	6	6	NUM
ajst-31431	184	16	:	:	PUNCT
ajst-31431	184	17	multiple	multiple	ADJ
ajst-31431	184	18	choice	choice	NOUN
ajst-31431	184	19	cnn	cnn	PROPN
ajst-31431	184	20	over	over	ADP
ajst-31431	184	21	contextual	contextual	ADJ
ajst-31431	184	22	embedding[j	embedding[j	PROPN
ajst-31431	184	23	]	]	PUNCT
ajst-31431	184	24	.	.	PUNCT
ajst-31431	185	1	arxiv	arxiv	PROPN
ajst-31431	185	2	preprint	preprint	PROPN
ajst-31431	185	3	arxiv:1904.08292	arxiv:1904.08292	NOUN
ajst-31431	185	4	,	,	PUNCT
ajst-31431	185	5	2019	2019	NUM
ajst-31431	185	6	.	.	PUNCT
ajst-31431	186	1	[	[	X
ajst-31431	186	2	4	4	X
ajst-31431	186	3	]	]	X
ajst-31431	186	4	mandl	mandl	PROPN
ajst-31431	186	5	t	t	PROPN
ajst-31431	186	6	,	,	PUNCT
ajst-31431	186	7	modha	modha	VERB
ajst-31431	186	8	s	s	PART
ajst-31431	186	9	,	,	PUNCT
ajst-31431	186	10	majumder	majumder	NOUN
ajst-31431	186	11	p	p	NOUN
ajst-31431	186	12	,	,	PUNCT
ajst-31431	186	13	et	et	PROPN
ajst-31431	186	14	al	al	PROPN
ajst-31431	186	15	.	.	PROPN
ajst-31431	187	1	overview	overview	NOUN
ajst-31431	187	2	of	of	ADP
ajst-31431	187	3	the	the	DET
ajst-31431	187	4	hasoc	hasoc	ADJ
ajst-31431	187	5	track	track	NOUN
ajst-31431	187	6	at	at	ADP
ajst-31431	187	7	fire	fire	NOUN
ajst-31431	187	8	2019	2019	NUM
ajst-31431	187	9	:	:	PUNCT
ajst-31431	187	10	hate	hate	VERB
ajst-31431	187	11	speech	speech	NOUN
ajst-31431	187	12	and	and	CCONJ
ajst-31431	187	13	offensive	offensive	ADJ
ajst-31431	187	14	content	content	NOUN
ajst-31431	187	15	identification	identification	NOUN
ajst-31431	187	16	in	in	ADP
ajst-31431	187	17	indo	indo	NOUN
ajst-31431	187	18	-	-	PUNCT
ajst-31431	187	19	european	european	ADJ
ajst-31431	187	20	languages[c]//proceedings	languages[c]//proceedings	PROPN
ajst-31431	187	21	of	of	ADP
ajst-31431	187	22	the	the	DET
ajst-31431	187	23	11th	11th	ADJ
ajst-31431	187	24	annual	annual	ADJ
ajst-31431	187	25	meeting	meeting	NOUN
ajst-31431	187	26	of	of	ADP
ajst-31431	187	27	the	the	DET
ajst-31431	187	28	forum	forum	NOUN
ajst-31431	187	29	for	for	ADP
ajst-31431	187	30	information	information	NOUN
ajst-31431	187	31	retrieval	retrieval	NOUN
ajst-31431	187	32	evaluation	evaluation	NOUN
ajst-31431	187	33	.	.	PUNCT
ajst-31431	188	1	2019	2019	NUM
ajst-31431	188	2	:	:	PUNCT
ajst-31431	188	3	14	14	NUM
ajst-31431	188	4	-	-	SYM
ajst-31431	188	5	17	17	NUM
ajst-31431	188	6	.	.	PUNCT
ajst-31431	189	1	[	[	X
ajst-31431	189	2	5	5	X
ajst-31431	189	3	]	]	PUNCT
ajst-31431	189	4	wang	wang	PROPN
ajst-31431	189	5	b	b	PROPN
ajst-31431	189	6	,	,	PUNCT
ajst-31431	189	7	ding	ding	PROPN
ajst-31431	189	8	h.	h.	PROPN
ajst-31431	189	9	ynu	ynu	PROPN
ajst-31431	189	10	nlp	nlp	PROPN
ajst-31431	189	11	at	at	ADP
ajst-31431	189	12	semeval-2019	semeval-2019	ADJ
ajst-31431	189	13	task	task	NOUN
ajst-31431	189	14	5	5	NUM
ajst-31431	189	15	:	:	PUNCT
ajst-31431	189	16	attention	attention	NOUN
ajst-31431	189	17	and	and	CCONJ
ajst-31431	189	18	capsule	capsule	NOUN
ajst-31431	189	19	ensemble	ensemble	ADJ
ajst-31431	189	20	for	for	ADP
ajst-31431	189	21	identifying	identify	VERB
ajst-31431	189	22	hate	hate	NOUN
ajst-31431	189	23	speech[c]//proceedings	speech[c]//proceeding	NOUN
ajst-31431	189	24	of	of	ADP
ajst-31431	189	25	the	the	DET
ajst-31431	189	26	13th	13th	ADJ
ajst-31431	189	27	international	international	ADJ
ajst-31431	189	28	workshop	workshop	NOUN
ajst-31431	189	29	on	on	ADP
ajst-31431	189	30	semantic	semantic	ADJ
ajst-31431	189	31	evaluation	evaluation	NOUN
ajst-31431	189	32	.	.	PUNCT
ajst-31431	190	1	2019	2019	NUM
ajst-31431	190	2	:	:	PUNCT
ajst-31431	190	3	529	529	NUM
ajst-31431	190	4	-	-	SYM
ajst-31431	190	5	534	534	NUM
ajst-31431	190	6	.	.	PUNCT
ajst-31431	191	1	[	[	X
ajst-31431	191	2	6	6	NUM
ajst-31431	191	3	]	]	X
ajst-31431	191	4	yang	yang	PROPN
ajst-31431	191	5	x	x	PROPN
ajst-31431	191	6	,	,	PUNCT
ajst-31431	191	7	obadinma	obadinma	X
ajst-31431	191	8	s	s	PROPN
ajst-31431	191	9	,	,	PUNCT
ajst-31431	191	10	zhao	zhao	PROPN
ajst-31431	191	11	h	h	PROPN
ajst-31431	191	12	,	,	PUNCT
ajst-31431	191	13	et	et	PROPN
ajst-31431	191	14	al	al	PROPN
ajst-31431	191	15	.	.	PUNCT
ajst-31431	191	16	semeval-2020	semeval-2020	PROPN
ajst-31431	191	17	task	task	NOUN
ajst-31431	191	18	5	5	NUM
ajst-31431	191	19	:	:	PUNCT
ajst-31431	191	20	counterfactual	counterfactual	ADJ
ajst-31431	191	21	recognition[j	recognition[j	NOUN
ajst-31431	191	22	]	]	PUNCT
ajst-31431	191	23	.	.	PUNCT
ajst-31431	192	1	arxiv	arxiv	PROPN
ajst-31431	192	2	preprint	preprint	VERB
ajst-31431	192	3	arxiv:2008.00563	arxiv:2008.00563	ADV
ajst-31431	192	4	,	,	PUNCT
ajst-31431	192	5	2020	2020	NUM
ajst-31431	192	6	.	.	PUNCT
ajst-31431	193	1	[	[	X
ajst-31431	193	2	7	7	NUM
ajst-31431	193	3	]	]	X
ajst-31431	193	4	zampieri	zampieri	PROPN
ajst-31431	193	5	m	m	PROPN
ajst-31431	193	6	,	,	PUNCT
ajst-31431	193	7	malmasi	malmasi	NOUN
ajst-31431	193	8	s	s	NOUN
ajst-31431	193	9	,	,	PUNCT
ajst-31431	193	10	nakov	nakov	NOUN
ajst-31431	193	11	p	p	NOUN
ajst-31431	193	12	,	,	PUNCT
ajst-31431	193	13	et	et	PROPN
ajst-31431	193	14	al	al	PROPN
ajst-31431	193	15	.	.	PUNCT
ajst-31431	194	1	semeval-2019	semeval-2019	ADJ
ajst-31431	194	2	task	task	NOUN
ajst-31431	194	3	6	6	NUM
ajst-31431	194	4	:	:	PUNCT
ajst-31431	194	5	identifying	identify	VERB
ajst-31431	194	6	and	and	CCONJ
ajst-31431	194	7	categorizing	categorize	VERB
ajst-31431	194	8	offensive	offensive	ADJ
ajst-31431	194	9	language	language	NOUN
ajst-31431	194	10	in	in	ADP
ajst-31431	194	11	social	social	ADJ
ajst-31431	194	12	media	medium	NOUN
ajst-31431	194	13	(	(	PUNCT
ajst-31431	194	14	offenseval)[j	offenseval)[j	PROPN
ajst-31431	194	15	]	]	PUNCT
ajst-31431	194	16	.	.	PUNCT
ajst-31431	195	1	arxiv	arxiv	PROPN
ajst-31431	195	2	preprint	preprint	VERB
ajst-31431	195	3	arxiv:1903.08983	arxiv:1903.08983	ADJ
ajst-31431	195	4	,	,	PUNCT
ajst-31431	195	5	2019	2019	NUM
ajst-31431	195	6	.	.	PUNCT
ajst-31431	196	1	[	[	X
ajst-31431	196	2	8	8	NUM
ajst-31431	196	3	]	]	X
ajst-31431	196	4	guest	guest	NOUN
ajst-31431	196	5	e	e	NOUN
ajst-31431	196	6	,	,	PUNCT
ajst-31431	196	7	vidgen	vidgen	PROPN
ajst-31431	196	8	b	b	PROPN
ajst-31431	196	9	,	,	PUNCT
ajst-31431	196	10	mittos	mitto	NOUN
ajst-31431	196	11	a	a	PRON
ajst-31431	196	12	,	,	PUNCT
ajst-31431	196	13	et	et	PROPN
ajst-31431	196	14	al	al	PROPN
ajst-31431	196	15	.	.	PUNCT
ajst-31431	197	1	an	an	DET
ajst-31431	197	2	expert	expert	NOUN
ajst-31431	197	3	annotated	annotate	VERB
ajst-31431	197	4	dataset	dataset	NOUN
ajst-31431	197	5	for	for	ADP
ajst-31431	197	6	the	the	DET
ajst-31431	197	7	detection	detection	NOUN
ajst-31431	197	8	of	of	ADP
ajst-31431	197	9	online	online	ADJ
ajst-31431	197	10	misogyny[c]//proceedings	misogyny[c]//proceeding	NOUN
ajst-31431	197	11	of	of	ADP
ajst-31431	197	12	the	the	DET
ajst-31431	197	13	16th	16th	ADJ
ajst-31431	197	14	conference	conference	NOUN
ajst-31431	197	15	of	of	ADP
ajst-31431	197	16	the	the	DET
ajst-31431	197	17	european	european	ADJ
ajst-31431	197	18	chapter	chapter	NOUN
ajst-31431	197	19	of	of	ADP
ajst-31431	197	20	the	the	DET
ajst-31431	197	21	association	association	NOUN
ajst-31431	197	22	for	for	ADP
ajst-31431	197	23	computational	computational	ADJ
ajst-31431	197	24	linguistics	linguistic	NOUN
ajst-31431	197	25	:	:	PUNCT
ajst-31431	197	26	main	main	ADJ
ajst-31431	197	27	volume	volume	NOUN
ajst-31431	197	28	.	.	PUNCT
ajst-31431	198	1	2021	2021	NUM
ajst-31431	198	2	:	:	PUNCT
ajst-31431	198	3	1336	1336	NUM
ajst-31431	198	4	-	-	SYM
ajst-31431	198	5	1350	1350	NUM
ajst-31431	198	6	.	.	PUNCT
ajst-31431	199	1	2021	2021	NUM
ajst-31431	200	1	[	[	X
ajst-31431	200	2	9	9	NUM
ajst-31431	200	3	]	]	SYM
ajst-31431	200	4	mulki	mulki	NOUN
ajst-31431	200	5	h	h	NOUN
ajst-31431	200	6	,	,	PUNCT
ajst-31431	200	7	haddad	haddad	PROPN
ajst-31431	200	8	h	h	PROPN
ajst-31431	200	9	,	,	PUNCT
ajst-31431	200	10	ali	ali	PROPN
ajst-31431	200	11	c	c	PROPN
ajst-31431	200	12	b	b	PROPN
ajst-31431	200	13	,	,	PUNCT
ajst-31431	200	14	et	et	PROPN
ajst-31431	200	15	al	al	PROPN
ajst-31431	200	16	.	.	PUNCT
ajst-31431	201	1	l	l	NOUN
ajst-31431	201	2	-	-	PUNCT
ajst-31431	201	3	hsab	hsab	NOUN
ajst-31431	201	4	:	:	PUNCT
ajst-31431	201	5	a	a	DET
ajst-31431	201	6	levantine	levantine	PROPN
ajst-31431	201	7	twitter	twitter	NOUN
ajst-31431	201	8	dataset	dataset	VERB
ajst-31431	201	9	for	for	ADP
ajst-31431	201	10	hate	hate	NOUN
ajst-31431	201	11	speech	speech	NOUN
ajst-31431	201	12	and	and	CCONJ
ajst-31431	201	13	abusive	abusive	ADJ
ajst-31431	201	14	language[c]//proceedings	language[c]//proceeding	NOUN
ajst-31431	201	15	of	of	ADP
ajst-31431	201	16	the	the	DET
ajst-31431	201	17	third	third	ADJ
ajst-31431	201	18	workshop	workshop	NOUN
ajst-31431	201	19	on	on	ADP
ajst-31431	201	20	abusive	abusive	ADJ
ajst-31431	201	21	language	language	NOUN
ajst-31431	201	22	online	online	NOUN
ajst-31431	201	23	.	.	PUNCT
ajst-31431	202	1	2019	2019	NUM
ajst-31431	202	2	:	:	PUNCT
ajst-31431	202	3	111118	111118	NUM
ajst-31431	202	4	.	.	PUNCT
ajst-31431	203	1	[	[	X
ajst-31431	203	2	10	10	NUM
ajst-31431	203	3	]	]	X
ajst-31431	203	4	kiela	kiela	NOUN
ajst-31431	203	5	d	d	PROPN
ajst-31431	203	6	,	,	PUNCT
ajst-31431	203	7	firooz	firooz	ADJ
ajst-31431	203	8	h	h	NOUN
ajst-31431	203	9	,	,	PUNCT
ajst-31431	203	10	mohan	mohan	PROPN
ajst-31431	203	11	a	a	PROPN
ajst-31431	203	12	,	,	PUNCT
ajst-31431	203	13	et	et	PROPN
ajst-31431	203	14	al	al	PROPN
ajst-31431	203	15	.	.	PUNCT
ajst-31431	204	1	the	the	DET
ajst-31431	204	2	hateful	hateful	ADJ
ajst-31431	204	3	memes	meme	NOUN
ajst-31431	204	4	challenge	challenge	VERB
ajst-31431	204	5	:	:	PUNCT
ajst-31431	204	6	detecting	detect	VERB
ajst-31431	204	7	hate	hate	NOUN
ajst-31431	204	8	speech	speech	NOUN
ajst-31431	204	9	in	in	ADP
ajst-31431	204	10	multimodal	multimodal	NOUN
ajst-31431	204	11	memes[j	memes[j	NOUN
ajst-31431	204	12	]	]	PUNCT
ajst-31431	204	13	.	.	PUNCT
ajst-31431	205	1	advances	advance	NOUN
ajst-31431	205	2	in	in	ADP
ajst-31431	205	3	neural	neural	ADJ
ajst-31431	205	4	information	information	NOUN
ajst-31431	205	5	processing	processing	NOUN
ajst-31431	205	6	systems	system	NOUN
ajst-31431	205	7	,	,	PUNCT
ajst-31431	205	8	2020	2020	NUM
ajst-31431	205	9	,	,	PUNCT
ajst-31431	205	10	33	33	NUM
ajst-31431	205	11	:	:	SYM
ajst-31431	205	12	2611	2611	NUM
ajst-31431	205	13	-	-	SYM
ajst-31431	205	14	2624	2624	NUM
ajst-31431	205	15	.	.	PUNCT
ajst-31431	206	1	[	[	X
ajst-31431	206	2	11	11	NUM
ajst-31431	206	3	]	]	PUNCT
ajst-31431	206	4	pàmies	pàmie	NOUN
ajst-31431	206	5	m	m	PROPN
ajst-31431	206	6	,	,	PUNCT
ajst-31431	206	7	öhman	öhman	NOUN
ajst-31431	206	8	e	e	NOUN
ajst-31431	206	9	,	,	PUNCT
ajst-31431	206	10	kajava	kajava	PROPN
ajst-31431	206	11	k	k	PROPN
ajst-31431	206	12	,	,	PUNCT
ajst-31431	206	13	et	et	PROPN
ajst-31431	206	14	al	al	PROPN
ajst-31431	206	15	.	.	PUNCT
ajst-31431	207	1	lt@	lt@	PROPN
ajst-31431	207	2	helsinki	helsinki	PROPN
ajst-31431	207	3	at	at	ADP
ajst-31431	207	4	semeval-2020	semeval-2020	NOUN
ajst-31431	207	5	task	task	NOUN
ajst-31431	207	6	12	12	NUM
ajst-31431	207	7	:	:	PUNCT
ajst-31431	207	8	multilingual	multilingual	ADJ
ajst-31431	207	9	or	or	CCONJ
ajst-31431	207	10	language	language	NOUN
ajst-31431	207	11	-	-	PUNCT
ajst-31431	207	12	specific	specific	NOUN
ajst-31431	207	13	bert?[j	bert?[j	NOUN
ajst-31431	207	14	]	]	PUNCT
ajst-31431	207	15	.	.	PUNCT
ajst-31431	208	1	arxiv	arxiv	PROPN
ajst-31431	208	2	preprint	preprint	PROPN
ajst-31431	208	3	arxiv:2008.00805	arxiv:2008.00805	PROPN
ajst-31431	208	4	,	,	PUNCT
ajst-31431	208	5	2020	2020	NUM
ajst-31431	208	6	.	.	PUNCT
ajst-31431	209	1	[	[	X
ajst-31431	209	2	12	12	NUM
ajst-31431	209	3	]	]	X
ajst-31431	209	4	hossain	hossain	PROPN
ajst-31431	209	5	e	e	PROPN
ajst-31431	209	6	,	,	PUNCT
ajst-31431	209	7	sharif	sharif	PROPN
ajst-31431	209	8	o	o	NOUN
ajst-31431	209	9	,	,	PUNCT
ajst-31431	209	10	hoque	hoque	NOUN
ajst-31431	209	11	m	m	VERB
ajst-31431	209	12	m	m	PROPN
ajst-31431	209	13	,	,	PUNCT
ajst-31431	209	14	et	et	PROPN
ajst-31431	209	15	al	al	PROPN
ajst-31431	209	16	.	.	PUNCT
ajst-31431	210	1	deciphering	decipher	VERB
ajst-31431	210	2	hate	hate	NOUN
ajst-31431	210	3	:	:	PUNCT
ajst-31431	210	4	identifying	identify	VERB
ajst-31431	210	5	hateful	hateful	ADJ
ajst-31431	210	6	memes	meme	NOUN
ajst-31431	210	7	and	and	CCONJ
ajst-31431	210	8	their	their	PRON
ajst-31431	210	9	targets[j	targets[j	NOUN
ajst-31431	210	10	]	]	PUNCT
ajst-31431	210	11	.	.	PUNCT
ajst-31431	211	1	arxiv	arxiv	PROPN
ajst-31431	211	2	preprint	preprint	PROPN
ajst-31431	211	3	arxiv:2403.10829	arxiv:2403.10829	NOUN
ajst-31431	211	4	,	,	PUNCT
ajst-31431	211	5	2024	2024	NUM
ajst-31431	211	6	.	.	PUNCT
ajst-31431	212	1	[	[	X
ajst-31431	212	2	13	13	NUM
ajst-31431	212	3	]	]	PUNCT
ajst-31431	212	4	suryawanshi	suryawanshi	NOUN
ajst-31431	212	5	s	s	PROPN
ajst-31431	212	6	,	,	PUNCT
ajst-31431	212	7	chakravarthi	chakravarthi	NOUN
ajst-31431	212	8	b	b	NOUN
ajst-31431	212	9	r	r	NOUN
ajst-31431	212	10	,	,	PUNCT
ajst-31431	212	11	arcan	arcan	PROPN
ajst-31431	212	12	m	m	PROPN
ajst-31431	212	13	,	,	PUNCT
ajst-31431	212	14	et	et	PROPN
ajst-31431	212	15	al	al	PROPN
ajst-31431	212	16	.	.	PUNCT
ajst-31431	212	17	multimodal	multimodal	ADJ
ajst-31431	212	18	meme	meme	NOUN
ajst-31431	212	19	dataset	dataset	NOUN
ajst-31431	212	20	(	(	PUNCT
ajst-31431	212	21	multioff	multioff	PROPN
ajst-31431	212	22	)	)	PUNCT
ajst-31431	212	23	for	for	ADP
ajst-31431	212	24	identifying	identify	VERB
ajst-31431	212	25	offensive	offensive	ADJ
ajst-31431	212	26	content	content	NOUN
ajst-31431	212	27	in	in	ADP
ajst-31431	212	28	image	image	NOUN
ajst-31431	212	29	and	and	CCONJ
ajst-31431	212	30	text[c]//proceedings	text[c]//proceeding	NOUN
ajst-31431	212	31	of	of	ADP
ajst-31431	212	32	the	the	DET
ajst-31431	212	33	second	second	ADJ
ajst-31431	212	34	workshop	workshop	NOUN
ajst-31431	212	35	on	on	ADP
ajst-31431	212	36	trolling	trolling	NOUN
ajst-31431	212	37	,	,	PUNCT
ajst-31431	212	38	aggression	aggression	NOUN
ajst-31431	212	39	and	and	CCONJ
ajst-31431	212	40	cyberbullying	cyberbullying	NOUN
ajst-31431	212	41	.	.	PUNCT
ajst-31431	213	1	2020	2020	NUM
ajst-31431	213	2	:	:	PUNCT
ajst-31431	214	1	32	32	NUM
ajst-31431	214	2	-	-	SYM
ajst-31431	214	3	41	41	NUM
ajst-31431	214	4	.	.	PUNCT
ajst-31431	215	1	[	[	X
ajst-31431	215	2	14	14	NUM
ajst-31431	215	3	]	]	X
ajst-31431	215	4	wu	wu	PROPN
ajst-31431	215	5	j	j	PROPN
ajst-31431	215	6	,	,	PUNCT
ajst-31431	215	7	hong	hong	PROPN
ajst-31431	215	8	q	q	PROPN
ajst-31431	215	9	,	,	PUNCT
ajst-31431	215	10	cao	cao	PROPN
ajst-31431	215	11	m	m	PROPN
ajst-31431	215	12	,	,	PUNCT
ajst-31431	215	13	et	et	PROPN
ajst-31431	215	14	al	al	PROPN
ajst-31431	215	15	.	.	PUNCT
ajst-31431	216	1	a	a	DET
ajst-31431	216	2	group	group	NOUN
ajst-31431	216	3	consensus	consensus	NOUN
ajst-31431	216	4	-	-	PUNCT
ajst-31431	216	5	based	base	VERB
ajst-31431	216	6	travel	travel	NOUN
ajst-31431	216	7	destination	destination	NOUN
ajst-31431	216	8	evaluation	evaluation	NOUN
ajst-31431	216	9	method	method	NOUN
ajst-31431	216	10	with	with	ADP
ajst-31431	216	11	online	online	PROPN
ajst-31431	216	12	reviews[j	reviews[j	PROPN
ajst-31431	216	13	]	]	PUNCT
ajst-31431	216	14	.	.	PUNCT
ajst-31431	217	1	applied	apply	VERB
ajst-31431	217	2	intelligence	intelligence	NOUN
ajst-31431	217	3	,	,	PUNCT
ajst-31431	217	4	2022	2022	NUM
ajst-31431	217	5	,	,	PUNCT
ajst-31431	217	6	52(2	52(2	NUM
ajst-31431	217	7	):	):	PUNCT
ajst-31431	217	8	1306	1306	NUM
ajst-31431	217	9	-	-	SYM
ajst-31431	217	10	1324	1324	NUM
ajst-31431	217	11	.	.	PUNCT
ajst-31431	218	1	[	[	X
ajst-31431	218	2	15	15	NUM
ajst-31431	218	3	]	]	X
ajst-31431	218	4	rodriguez	rodriguez	NOUN
ajst-31431	218	5	a	a	X
ajst-31431	218	6	,	,	PUNCT
ajst-31431	218	7	argueta	argueta	PROPN
ajst-31431	218	8	c	c	NOUN
ajst-31431	218	9	,	,	PUNCT
ajst-31431	218	10	chen	chen	PROPN
ajst-31431	218	11	y	y	PROPN
ajst-31431	218	12	l.	l.	PROPN
ajst-31431	218	13	automatic	automatic	ADJ
ajst-31431	218	14	detection	detection	NOUN
ajst-31431	218	15	of	of	ADP
ajst-31431	218	16	hate	hate	NOUN
ajst-31431	218	17	speech	speech	NOUN
ajst-31431	218	18	on	on	ADP
ajst-31431	218	19	facebook	facebook	NOUN
ajst-31431	218	20	using	use	VERB
ajst-31431	218	21	sentiment	sentiment	NOUN
ajst-31431	218	22	and	and	CCONJ
ajst-31431	218	23	emotion	emotion	NOUN
ajst-31431	218	24	analysis[c]//2019	analysis[c]//2019	ADP
ajst-31431	218	25	international	international	ADJ
ajst-31431	218	26	conference	conference	NOUN
ajst-31431	218	27	on	on	ADP
ajst-31431	218	28	artificial	artificial	ADJ
ajst-31431	218	29	intelligence	intelligence	NOUN
ajst-31431	218	30	in	in	ADP
ajst-31431	218	31	information	information	NOUN
ajst-31431	218	32	and	and	CCONJ
ajst-31431	218	33	communication	communication	NOUN
ajst-31431	218	34	(	(	PUNCT
ajst-31431	218	35	icaiic	icaiic	ADJ
ajst-31431	218	36	)	)	PUNCT
ajst-31431	218	37	.	.	PUNCT
ajst-31431	219	1	ieee	ieee	PROPN
ajst-31431	219	2	,	,	PUNCT
ajst-31431	219	3	2019	2019	NUM
ajst-31431	219	4	:	:	PUNCT
ajst-31431	219	5	169	169	NUM
ajst-31431	219	6	-	-	SYM
ajst-31431	219	7	174	174	NUM
ajst-31431	219	8	.	.	PUNCT
ajst-31431	220	1	[	[	X
ajst-31431	220	2	16	16	NUM
ajst-31431	220	3	]	]	X
ajst-31431	220	4	[	[	X
ajst-31431	220	5	19	19	NUM
ajst-31431	220	6	]	]	X
ajst-31431	220	7	al	al	PROPN
ajst-31431	220	8	-	-	PUNCT
ajst-31431	220	9	garadi	garadi	PROPN
ajst-31431	220	10	m	m	PROPN
ajst-31431	220	11	a	a	NOUN
ajst-31431	220	12	,	,	PUNCT
ajst-31431	220	13	hussain	hussain	PROPN
ajst-31431	220	14	m	m	PROPN
ajst-31431	220	15	r	r	PROPN
ajst-31431	220	16	,	,	PUNCT
ajst-31431	220	17	khan	khan	PROPN
ajst-31431	220	18	n	n	CCONJ
ajst-31431	220	19	,	,	PUNCT
ajst-31431	220	20	et	et	PROPN
ajst-31431	220	21	al	al	PROPN
ajst-31431	220	22	.	.	PROPN
ajst-31431	220	23	predicting	predict	VERB
ajst-31431	220	24	cyberbullying	cyberbullye	VERB
ajst-31431	220	25	on	on	ADP
ajst-31431	220	26	social	social	ADJ
ajst-31431	220	27	media	medium	NOUN
ajst-31431	220	28	in	in	ADP
ajst-31431	220	29	the	the	DET
ajst-31431	220	30	big	big	ADJ
ajst-31431	220	31	data	data	NOUN
ajst-31431	220	32	era	era	NOUN
ajst-31431	220	33	using	use	VERB
ajst-31431	220	34	machine	machine	NOUN
ajst-31431	220	35	learning	learn	VERB
ajst-31431	220	36	algorithms	algorithm	NOUN
ajst-31431	220	37	:	:	PUNCT
ajst-31431	220	38	review	review	NOUN
ajst-31431	220	39	of	of	ADP
ajst-31431	220	40	literature	literature	NOUN
ajst-31431	220	41	and	and	CCONJ
ajst-31431	220	42	open	open	ADJ
ajst-31431	220	43	challenges[j	challenges[j	PROPN
ajst-31431	220	44	]	]	PUNCT
ajst-31431	220	45	.	.	PUNCT
ajst-31431	221	1	ieee	ieee	NOUN
ajst-31431	221	2	access	access	NOUN
ajst-31431	221	3	,	,	PUNCT
ajst-31431	221	4	2019	2019	NUM
ajst-31431	221	5	,	,	PUNCT
ajst-31431	221	6	7	7	NUM
ajst-31431	221	7	:	:	SYM
ajst-31431	221	8	70701	70701	NUM
ajst-31431	221	9	-	-	SYM
ajst-31431	221	10	70718	70718	NUM
ajst-31431	221	11	.	.	PUNCT
ajst-31431	222	1	[	[	X
ajst-31431	222	2	17	17	NUM
ajst-31431	222	3	]	]	PUNCT
ajst-31431	222	4	weir	weir	NOUN
ajst-31431	222	5	g	g	PROPN
ajst-31431	222	6	,	,	PUNCT
ajst-31431	222	7	owoeye	owoeye	NOUN
ajst-31431	222	8	k	k	PROPN
ajst-31431	222	9	,	,	PUNCT
ajst-31431	222	10	oberacker	oberacker	NOUN
ajst-31431	222	11	a	a	PROPN
ajst-31431	222	12	,	,	PUNCT
ajst-31431	222	13	et	et	PROPN
ajst-31431	222	14	al	al	PROPN
ajst-31431	222	15	.	.	PROPN
ajst-31431	222	16	cloud	cloud	NOUN
ajst-31431	222	17	-	-	PUNCT
ajst-31431	222	18	based	base	VERB
ajst-31431	222	19	textual	textual	ADJ
ajst-31431	222	20	analysis	analysis	NOUN
ajst-31431	222	21	as	as	ADP
ajst-31431	222	22	a	a	DET
ajst-31431	222	23	basis	basis	NOUN
ajst-31431	222	24	for	for	ADP
ajst-31431	222	25	document	document	NOUN
ajst-31431	222	26	classification[c]//2018	classification[c]//2018	VERB
ajst-31431	222	27	international	international	ADJ
ajst-31431	222	28	conference	conference	NOUN
ajst-31431	222	29	on	on	ADP
ajst-31431	222	30	high	high	ADJ
ajst-31431	222	31	performance	performance	NOUN
ajst-31431	222	32	computing	computing	NOUN
ajst-31431	222	33	&	&	CCONJ
ajst-31431	222	34	simulation	simulation	PROPN
ajst-31431	222	35	(	(	PUNCT
ajst-31431	222	36	hpcs	hpcs	PROPN
ajst-31431	222	37	)	)	PUNCT
ajst-31431	222	38	.	.	PUNCT
ajst-31431	223	1	ieee	ieee	NOUN
ajst-31431	223	2	,	,	PUNCT
ajst-31431	223	3	2018	2018	NUM
ajst-31431	223	4	:	:	PUNCT
ajst-31431	223	5	672	672	NUM
ajst-31431	223	6	-	-	SYM
ajst-31431	223	7	676	676	NUM
ajst-31431	223	8	.	.	PUNCT
ajst-31431	224	1	[	[	X
ajst-31431	224	2	18	18	NUM
ajst-31431	224	3	]	]	PUNCT
ajst-31431	224	4	nurce	nurce	NOUN
ajst-31431	224	5	e	e	NOUN
ajst-31431	224	6	,	,	PUNCT
ajst-31431	224	7	keci	keci	PROPN
ajst-31431	224	8	j	j	PROPN
ajst-31431	224	9	,	,	PUNCT
ajst-31431	224	10	derczynski	derczynski	PROPN
ajst-31431	224	11	l.	l.	PROPN
ajst-31431	224	12	detecting	detect	VERB
ajst-31431	224	13	abusive	abusive	ADJ
ajst-31431	224	14	albanian[j	albanian[j	NOUN
ajst-31431	224	15	]	]	PUNCT
ajst-31431	224	16	.	.	PUNCT
ajst-31431	225	1	arxiv	arxiv	PROPN
ajst-31431	225	2	preprint	preprint	NOUN
ajst-31431	225	3	arxiv:2107.13592	arxiv:2107.13592	NOUN
ajst-31431	225	4	,	,	PUNCT
ajst-31431	225	5	2021	2021	NUM
ajst-31431	225	6	.	.	PUNCT
ajst-31431	226	1	[	[	X
ajst-31431	226	2	19	19	NUM
ajst-31431	226	3	]	]	PUNCT
ajst-31431	226	4	albadi	albadi	NOUN
ajst-31431	226	5	,	,	PUNCT
ajst-31431	226	6	n.	n.	NOUN
ajst-31431	226	7	,m.kurdi	,m.kurdi	PUNCT
ajst-31431	226	8	,	,	PUNCT
ajst-31431	226	9	ands.mishra.aretheyourbrothers?analysis	ands.mishra.aretheyourbrothers?analysis	AUX
ajst-31431	226	10	anddetection	anddetection	NOUN
ajst-31431	226	11	ofreligioushatespeechinthearabictwittersphere.in2018ieee	ofreligioushatespeechinthearabictwittersphere.in2018ieee	NOUN
ajst-31431	226	12	/	/	SYM
ajst-31431	226	13	a	a	DET
ajst-31431	226	14	cm	cm	NOUN
ajst-31431	226	15	internationalconferenceonadvancesinsocialnetworksanalysi	internationalconferenceonadvancesinsocialnetworksanalysi	NOUN
ajst-31431	226	16	sandmining	sandmine	VERB
ajst-31431	226	17	(	(	PUNCT
ajst-31431	226	18	asonam).2018.ieee	asonam).2018.ieee	NOUN
ajst-31431	226	19	.	.	PUNCT
ajst-31431	227	1	[	[	X
ajst-31431	227	2	20	20	NUM
ajst-31431	227	3	]	]	X
ajst-31431	227	4	davidson	davidson	PROPN
ajst-31431	227	5	t	t	PROPN
ajst-31431	227	6	,	,	PUNCT
ajst-31431	227	7	warmsley	warmsley	NOUN
ajst-31431	227	8	d	d	PROPN
ajst-31431	227	9	,	,	PUNCT
ajst-31431	227	10	macy	macy	PROPN
ajst-31431	227	11	m	m	PROPN
ajst-31431	227	12	,	,	PUNCT
ajst-31431	227	13	et	et	PROPN
ajst-31431	227	14	al	al	PROPN
ajst-31431	227	15	.	.	PROPN
ajst-31431	228	1	automated	automate	VERB
ajst-31431	228	2	hate	hate	NOUN
ajst-31431	228	3	speech	speech	NOUN
ajst-31431	228	4	detection	detection	NOUN
ajst-31431	228	5	and	and	CCONJ
ajst-31431	228	6	the	the	DET
ajst-31431	228	7	problem	problem	NOUN
ajst-31431	228	8	of	of	ADP
ajst-31431	228	9	offensive	offensive	ADJ
ajst-31431	228	10	language[c]//proceedings	language[c]//proceeding	NOUN
ajst-31431	228	11	of	of	ADP
ajst-31431	228	12	the	the	DET
ajst-31431	228	13	international	international	ADJ
ajst-31431	228	14	aaai	aaai	PROPN
ajst-31431	228	15	conference	conference	PROPN
ajst-31431	228	16	on	on	ADP
ajst-31431	228	17	web	web	NOUN
ajst-31431	228	18	and	and	CCONJ
ajst-31431	228	19	social	social	ADJ
ajst-31431	228	20	media	medium	NOUN
ajst-31431	228	21	.	.	PUNCT
ajst-31431	229	1	2017	2017	NUM
ajst-31431	229	2	,	,	PUNCT
ajst-31431	229	3	11(1	11(1	NUM
ajst-31431	229	4	):	):	PUNCT
ajst-31431	229	5	512	512	NUM
ajst-31431	229	6	-	-	SYM
ajst-31431	229	7	515	515	NUM
ajst-31431	229	8	.	.	PUNCT
ajst-31431	230	1	[	[	X
ajst-31431	230	2	21	21	NUM
ajst-31431	230	3	]	]	X
ajst-31431	230	4	mollas	mollas	PROPN
ajst-31431	230	5	i	i	PROPN
ajst-31431	230	6	,	,	PUNCT
ajst-31431	230	7	chrysopoulou	chrysopoulou	PROPN
ajst-31431	230	8	z	z	PROPN
ajst-31431	230	9	,	,	PUNCT
ajst-31431	230	10	karlos	karlos	PROPN
ajst-31431	230	11	s	s	PROPN
ajst-31431	230	12	,	,	PUNCT
ajst-31431	230	13	et	et	PROPN
ajst-31431	230	14	al	al	PROPN
ajst-31431	230	15	.	.	PUNCT
ajst-31431	231	1	ethos	ethos	PROPN
ajst-31431	231	2	:	:	PUNCT
ajst-31431	231	3	a	a	DET
ajst-31431	231	4	multilabel	multilabel	NOUN
ajst-31431	231	5	hate	hate	VERB
ajst-31431	231	6	speech	speech	NOUN
ajst-31431	231	7	detection	detection	NOUN
ajst-31431	231	8	dataset[j	dataset[j	NOUN
ajst-31431	231	9	]	]	PUNCT
ajst-31431	231	10	.	.	PUNCT
ajst-31431	232	1	complex	complex	ADJ
ajst-31431	232	2	&	&	CCONJ
ajst-31431	232	3	intelligent	intelligent	ADJ
ajst-31431	232	4	systems	system	NOUN
ajst-31431	232	5	,	,	PUNCT
ajst-31431	232	6	2022	2022	NUM
ajst-31431	232	7	,	,	PUNCT
ajst-31431	232	8	8(6	8(6	NUM
ajst-31431	232	9	):	):	PUNCT
ajst-31431	232	10	4663	4663	NUM
ajst-31431	232	11	-	-	SYM
ajst-31431	232	12	4678	4678	NUM
ajst-31431	232	13	.	.	PUNCT
ajst-31431	233	1	[	[	X
ajst-31431	233	2	22	22	NUM
ajst-31431	233	3	]	]	PUNCT
ajst-31431	233	4	wiegand	wiegand	NOUN
ajst-31431	233	5	m	m	PROPN
ajst-31431	233	6	,	,	PUNCT
ajst-31431	233	7	siegel	siegel	PROPN
ajst-31431	233	8	m	m	PROPN
ajst-31431	233	9	,	,	PUNCT
ajst-31431	233	10	ruppenhofer	ruppenhofer	VERB
ajst-31431	233	11	j.	j.	PROPN
ajst-31431	233	12	overview	overview	NOUN
ajst-31431	233	13	of	of	ADP
ajst-31431	233	14	the	the	DET
ajst-31431	233	15	germeval	germeval	ADJ
ajst-31431	233	16	2018	2018	NUM
ajst-31431	233	17	shared	share	VERB
ajst-31431	233	18	task	task	NOUN
ajst-31431	233	19	on	on	ADP
ajst-31431	233	20	the	the	DET
ajst-31431	233	21	identification	identification	NOUN
ajst-31431	233	22	of	of	ADP
ajst-31431	233	23	offensive	offensive	ADJ
ajst-31431	233	24	language[j	language[j	PROPN
ajst-31431	233	25	]	]	PUNCT
ajst-31431	233	26	.	.	PUNCT
ajst-31431	234	1	2018	2018	NUM
ajst-31431	234	2	.	.	PUNCT
ajst-31431	235	1	2018	2018	NUM
ajst-31431	236	1	[	[	X
ajst-31431	236	2	23	23	NUM
ajst-31431	236	3	]	]	X
ajst-31431	236	4	gitari	gitari	PROPN
ajst-31431	236	5	n	n	PROPN
ajst-31431	236	6	d	d	NOUN
ajst-31431	236	7	,	,	PUNCT
ajst-31431	236	8	zu	zu	NOUN
ajst-31431	236	9	*	*	PUNCT
ajst-31431	236	10	*	*	PUNCT
ajst-31431	236	11	z	z	X
ajst-31431	236	12	,	,	PUNCT
ajst-31431	236	13	damien	damien	PROPN
ajst-31431	236	14	h	h	PROPN
ajst-31431	236	15	,	,	PUNCT
ajst-31431	236	16	et	et	PROPN
ajst-31431	236	17	al	al	PROPN
ajst-31431	236	18	.	.	PUNCT
ajst-31431	237	1	a	a	DET
ajst-31431	237	2	lexicon	lexicon	NOUN
ajst-31431	237	3	-	-	PUNCT
ajst-31431	237	4	based	base	VERB
ajst-31431	237	5	approach	approach	NOUN
ajst-31431	237	6	for	for	ADP
ajst-31431	237	7	hate	hate	NOUN
ajst-31431	237	8	speech	speech	NOUN
ajst-31431	237	9	detection[j	detection[j	PROPN
ajst-31431	237	10	]	]	PUNCT
ajst-31431	237	11	.	.	PUNCT
ajst-31431	238	1	international	international	ADJ
ajst-31431	238	2	journal	journal	PROPN
ajst-31431	238	3	of	of	ADP
ajst-31431	238	4	multimedia	multimedia	NOUN
ajst-31431	238	5	and	and	CCONJ
ajst-31431	238	6	ubiquitous	ubiquitous	ADJ
ajst-31431	238	7	engineering	engineering	NOUN
ajst-31431	238	8	,	,	PUNCT
ajst-31431	238	9	2015	2015	NUM
ajst-31431	238	10	,	,	PUNCT
ajst-31431	238	11	10(4	10(4	NUM
ajst-31431	238	12	):	):	PUNCT
ajst-31431	238	13	215	215	NUM
ajst-31431	238	14	-	-	SYM
ajst-31431	238	15	230	230	NUM
ajst-31431	238	16	.	.	PUNCT
ajst-31431	239	1	[	[	X
ajst-31431	239	2	24	24	NUM
ajst-31431	239	3	]	]	X
ajst-31431	239	4	liao	liao	PROPN
ajst-31431	239	5	w	w	PROPN
ajst-31431	239	6	,	,	PUNCT
ajst-31431	239	7	zeng	zeng	PROPN
ajst-31431	239	8	b	b	PROPN
ajst-31431	239	9	,	,	PUNCT
ajst-31431	239	10	yin	yin	PROPN
ajst-31431	239	11	x	x	X
ajst-31431	239	12	,	,	PUNCT
ajst-31431	239	13	et	et	PROPN
ajst-31431	239	14	al	al	PROPN
ajst-31431	239	15	.	.	PUNCT
ajst-31431	240	1	an	an	DET
ajst-31431	240	2	improved	improve	VERB
ajst-31431	240	3	aspect	aspect	NOUN
ajst-31431	240	4	-	-	PUNCT
ajst-31431	240	5	category	category	NOUN
ajst-31431	240	6	sentiment	sentiment	NOUN
ajst-31431	240	7	analysis	analysis	NOUN
ajst-31431	240	8	model	model	NOUN
ajst-31431	240	9	for	for	ADP
ajst-31431	240	10	text	text	NOUN
ajst-31431	240	11	sentiment	sentiment	NOUN
ajst-31431	240	12	analysis	analysis	NOUN
ajst-31431	240	13	based	base	VERB
ajst-31431	240	14	on	on	ADP
ajst-31431	240	15	roberta[j	roberta[j	NOUN
ajst-31431	240	16	]	]	PUNCT
ajst-31431	240	17	.	.	PUNCT
ajst-31431	241	1	applied	apply	VERB
ajst-31431	241	2	intelligence	intelligence	NOUN
ajst-31431	241	3	,	,	PUNCT
ajst-31431	241	4	2021	2021	NUM
ajst-31431	241	5	,	,	PUNCT
ajst-31431	241	6	51	51	NUM
ajst-31431	241	7	:	:	SYM
ajst-31431	241	8	3522	3522	NUM
ajst-31431	241	9	-	-	SYM
ajst-31431	241	10	3533	3533	NUM
ajst-31431	241	11	.	.	PUNCT
ajst-31431	242	1	appl	appl	PROPN
ajst-31431	242	2	intell	intell	PROPN
ajst-31431	242	3	,	,	PUNCT
ajst-31431	242	4	2021	2021	NUM
ajst-31431	242	5	[	[	X
ajst-31431	242	6	25	25	NUM
ajst-31431	242	7	]	]	PUNCT
ajst-31431	242	8	fersini	fersini	NOUN
ajst-31431	242	9	e	e	NOUN
ajst-31431	242	10	,	,	PUNCT
ajst-31431	242	11	rosso	rosso	NOUN
ajst-31431	242	12	p	p	NOUN
ajst-31431	242	13	,	,	PUNCT
ajst-31431	242	14	anzovino	anzovino	NOUN
ajst-31431	242	15	m.	m.	NOUN
ajst-31431	242	16	overview	overview	NOUN
ajst-31431	242	17	of	of	ADP
ajst-31431	242	18	the	the	DET
ajst-31431	242	19	task	task	NOUN
ajst-31431	242	20	on	on	ADP
ajst-31431	242	21	automatic	automatic	ADJ
ajst-31431	242	22	misogyny	misogyny	NOUN
ajst-31431	242	23	identification	identification	NOUN
ajst-31431	242	24	at	at	ADP
ajst-31431	242	25	ibereval	ibereval	NOUN
ajst-31431	242	26	2018[j	2018[j	NOUN
ajst-31431	242	27	]	]	PUNCT
ajst-31431	242	28	.	.	PUNCT
ajst-31431	243	1	ibereval@	ibereval@	ADV
ajst-31431	243	2	sepln	sepln	PROPN
ajst-31431	243	3	,	,	PUNCT
ajst-31431	243	4	2018	2018	NUM
ajst-31431	243	5	,	,	PUNCT
ajst-31431	243	6	2150	2150	NUM
ajst-31431	243	7	:	:	PUNCT
ajst-31431	244	1	214	214	NUM
ajst-31431	244	2	-	-	SYM
ajst-31431	244	3	228	228	NUM
ajst-31431	244	4	.	.	PUNCT
ajst-31431	245	1	[	[	X
ajst-31431	245	2	26	26	NUM
ajst-31431	245	3	]	]	X
ajst-31431	245	4	salminen	salminen	PROPN
ajst-31431	245	5	j	j	PROPN
ajst-31431	245	6	,	,	PUNCT
ajst-31431	245	7	almerekhi	almerekhi	PROPN
ajst-31431	245	8	h	h	PROPN
ajst-31431	245	9	,	,	PUNCT
ajst-31431	245	10	milenković	milenković	PROPN
ajst-31431	245	11	m	m	PROPN
ajst-31431	245	12	,	,	PUNCT
ajst-31431	245	13	et	et	PROPN
ajst-31431	245	14	al	al	PROPN
ajst-31431	245	15	.	.	PROPN
ajst-31431	245	16	anatomy	anatomy	NOUN
ajst-31431	245	17	of	of	ADP
ajst-31431	245	18	online	online	ADJ
ajst-31431	245	19	hate	hate	NOUN
ajst-31431	245	20	:	:	PUNCT
ajst-31431	245	21	develo	develo	VERB
ajst-31431	245	22	*	*	PROPN
ajst-31431	245	23	*	*	PUNCT
ajst-31431	245	24	a	a	DET
ajst-31431	245	25	taxonomy	taxonomy	NOUN
ajst-31431	245	26	and	and	CCONJ
ajst-31431	245	27	machine	machine	NOUN
ajst-31431	245	28	learning	learning	NOUN
ajst-31431	245	29	models	model	NOUN
ajst-31431	245	30	for	for	ADP
ajst-31431	245	31	identifying	identify	VERB
ajst-31431	245	32	and	and	CCONJ
ajst-31431	245	33	classifying	classify	VERB
ajst-31431	245	34	hate	hate	NOUN
ajst-31431	245	35	in	in	ADP
ajst-31431	245	36	online	online	ADJ
ajst-31431	245	37	news	news	NOUN
ajst-31431	245	38	media[c]//proceedings	media[c]//proceeding	NOUN
ajst-31431	245	39	of	of	ADP
ajst-31431	245	40	the	the	DET
ajst-31431	245	41	international	international	ADJ
ajst-31431	245	42	aaai	aaai	PROPN
ajst-31431	245	43	conference	conference	PROPN
ajst-31431	245	44	on	on	ADP
ajst-31431	245	45	web	web	NOUN
ajst-31431	245	46	and	and	CCONJ
ajst-31431	245	47	social	social	ADJ
ajst-31431	245	48	media	medium	NOUN
ajst-31431	245	49	.	.	PUNCT
ajst-31431	246	1	2018	2018	NUM
ajst-31431	246	2	,	,	PUNCT
ajst-31431	246	3	12(1	12(1	NUM
ajst-31431	246	4	)	)	PUNCT
ajst-31431	246	5	.	.	PUNCT
ajst-31431	247	1	2018	2018	NUM
ajst-31431	248	1	[	[	X
ajst-31431	248	2	27	27	NUM
ajst-31431	248	3	]	]	X
ajst-31431	248	4	salminen	salminen	PROPN
ajst-31431	248	5	j	j	PROPN
ajst-31431	248	6	,	,	PUNCT
ajst-31431	248	7	almerekhi	almerekhi	PROPN
ajst-31431	248	8	h	h	PROPN
ajst-31431	248	9	,	,	PUNCT
ajst-31431	248	10	milenković	milenković	PROPN
ajst-31431	248	11	m	m	PROPN
ajst-31431	248	12	,	,	PUNCT
ajst-31431	248	13	et	et	PROPN
ajst-31431	248	14	al	al	PROPN
ajst-31431	248	15	.	.	PROPN
ajst-31431	248	16	anatomy	anatomy	NOUN
ajst-31431	248	17	of	of	ADP
ajst-31431	248	18	online	online	ADJ
ajst-31431	248	19	hate	hate	NOUN
ajst-31431	248	20	:	:	PUNCT
ajst-31431	248	21	develo	develo	VERB
ajst-31431	248	22	*	*	PROPN
ajst-31431	248	23	*	*	PUNCT
ajst-31431	248	24	a	a	DET
ajst-31431	248	25	taxonomy	taxonomy	NOUN
ajst-31431	248	26	and	and	CCONJ
ajst-31431	248	27	machine	machine	NOUN
ajst-31431	248	28	learning	learning	NOUN
ajst-31431	248	29	models	model	NOUN
ajst-31431	248	30	for	for	ADP
ajst-31431	248	31	identifying	identify	VERB
ajst-31431	248	32	and	and	CCONJ
ajst-31431	248	33	classifying	classify	VERB
ajst-31431	248	34	hate	hate	NOUN
ajst-31431	248	35	in	in	ADP
ajst-31431	248	36	online	online	ADJ
ajst-31431	248	37	news	news	NOUN
ajst-31431	248	38	media[c]//proceedings	media[c]//proceeding	NOUN
ajst-31431	248	39	of	of	ADP
ajst-31431	248	40	the	the	DET
ajst-31431	248	41	international	international	ADJ
ajst-31431	248	42	aaai	aaai	PROPN
ajst-31431	248	43	conference	conference	PROPN
ajst-31431	248	44	on	on	ADP
ajst-31431	248	45	web	web	NOUN
ajst-31431	248	46	and	and	CCONJ
ajst-31431	248	47	social	social	ADJ
ajst-31431	248	48	media	medium	NOUN
ajst-31431	248	49	.	.	PUNCT
ajst-31431	249	1	2018	2018	NUM
ajst-31431	249	2	,	,	PUNCT
ajst-31431	249	3	12(1	12(1	NUM
ajst-31431	249	4	)	)	PUNCT
ajst-31431	250	1	[	[	X
ajst-31431	250	2	28	28	NUM
ajst-31431	250	3	]	]	AUX
ajst-31431	250	4	pitsilis	pitsilis	PROPN
ajst-31431	250	5	g	g	PROPN
ajst-31431	250	6	k	k	PROPN
ajst-31431	250	7	,	,	PUNCT
ajst-31431	250	8	ramampiaro	ramampiaro	NOUN
ajst-31431	250	9	h	h	NOUN
ajst-31431	250	10	,	,	PUNCT
ajst-31431	250	11	langseth	langseth	PROPN
ajst-31431	250	12	h.	h.	PROPN
ajst-31431	250	13	detecting	detect	VERB
ajst-31431	250	14	offensive	offensive	ADJ
ajst-31431	250	15	language	language	NOUN
ajst-31431	250	16	in	in	ADP
ajst-31431	250	17	tweets	tweet	NOUN
ajst-31431	250	18	using	use	VERB
ajst-31431	250	19	deep	deep	ADJ
ajst-31431	250	20	learning[j	learning[j	NOUN
ajst-31431	250	21	]	]	PUNCT
ajst-31431	250	22	.	.	PUNCT
ajst-31431	251	1	arxiv	arxiv	PROPN
ajst-31431	251	2	preprint	preprint	NOUN
ajst-31431	251	3	arxiv:1801.04433	arxiv:1801.04433	NOUN
ajst-31431	251	4	,	,	PUNCT
ajst-31431	251	5	2018	2018	NUM
ajst-31431	251	6	.	.	PUNCT
ajst-31431	252	1	[	[	X
ajst-31431	252	2	29	29	NUM
ajst-31431	252	3	]	]	X
ajst-31431	252	4	fu	fu	NOUN
ajst-31431	252	5	e	e	NOUN
ajst-31431	252	6	,	,	PUNCT
ajst-31431	252	7	*	*	PUNCT
ajst-31431	252	8	*	*	PUNCT
ajst-31431	252	9	ang	ang	PROPN
ajst-31431	252	10	j	j	PROPN
ajst-31431	252	11	,	,	PUNCT
ajst-31431	252	12	*	*	PUNCT
ajst-31431	252	13	*	*	PROPN
ajst-31431	252	14	ong	ong	PROPN
ajst-31431	252	15	c.	c.	PROPN
ajst-31431	252	16	deep	deep	PROPN
ajst-31431	252	17	learning	learn	VERB
ajst-31431	252	18	techniques	technique	NOUN
ajst-31431	252	19	for	for	ADP
ajst-31431	252	20	sentiment	sentiment	NOUN
ajst-31431	252	21	analysis[j	analysis[j	PROPN
ajst-31431	252	22	]	]	PUNCT
ajst-31431	252	23	.	.	PUNCT
ajst-31431	253	1	highlights	highlight	NOUN
ajst-31431	253	2	in	in	ADP
ajst-31431	253	3	science	science	NOUN
ajst-31431	253	4	,	,	PUNCT
ajst-31431	253	5	engineering	engineering	NOUN
ajst-31431	253	6	and	and	CCONJ
ajst-31431	253	7	technology	technology	NOUN
ajst-31431	253	8	,	,	PUNCT
ajst-31431	253	9	2022	2022	NUM
ajst-31431	253	10	,	,	PUNCT
ajst-31431	253	11	16	16	NUM
ajst-31431	253	12	:	:	SYM
ajst-31431	253	13	1	1	NUM
ajst-31431	253	14	-	-	SYM
ajst-31431	253	15	7	7	NUM
ajst-31431	253	16	.	.	PUNCT
ajst-31431	254	1	[	[	X
ajst-31431	254	2	30	30	NUM
ajst-31431	254	3	]	]	X
ajst-31431	254	4	vaswani	vaswani	NOUN
ajst-31431	254	5	a	a	PRON
ajst-31431	254	6	,	,	PUNCT
ajst-31431	254	7	shazeer	shazeer	NOUN
ajst-31431	254	8	n	n	SYM
ajst-31431	254	9	,	,	PUNCT
ajst-31431	254	10	parmar	parmar	PROPN
ajst-31431	254	11	n	n	CCONJ
ajst-31431	254	12	,	,	PUNCT
ajst-31431	254	13	et	et	PROPN
ajst-31431	254	14	al	al	PROPN
ajst-31431	254	15	.	.	PUNCT
ajst-31431	254	16	attention	attention	NOUN
ajst-31431	254	17	is	be	AUX
ajst-31431	254	18	all	all	PRON
ajst-31431	254	19	you	you	PRON
ajst-31431	254	20	need[j	need[j	VERB
ajst-31431	254	21	]	]	PUNCT
ajst-31431	254	22	.	.	PUNCT
ajst-31431	255	1	advances	advance	NOUN
ajst-31431	255	2	in	in	ADP
ajst-31431	255	3	neural	neural	ADJ
ajst-31431	255	4	information	information	NOUN
ajst-31431	255	5	processing	processing	NOUN
ajst-31431	255	6	systems	system	NOUN
ajst-31431	255	7	,	,	PUNCT
ajst-31431	255	8	2017	2017	NUM
ajst-31431	255	9	,	,	PUNCT
ajst-31431	255	10	30	30	NUM
ajst-31431	255	11	.	.	PUNCT
ajst-31431	256	1	[	[	X
ajst-31431	256	2	31	31	NUM
ajst-31431	256	3	]	]	X
ajst-31431	256	4	kennedy	kennedy	PROPN
ajst-31431	256	5	c	c	PROPN
ajst-31431	256	6	j	j	PROPN
ajst-31431	256	7	,	,	PUNCT
ajst-31431	256	8	bacon	bacon	NOUN
ajst-31431	256	9	g	g	NOUN
ajst-31431	256	10	,	,	PUNCT
ajst-31431	256	11	sahn	sahn	PROPN
ajst-31431	256	12	a	a	PRON
ajst-31431	256	13	,	,	PUNCT
ajst-31431	256	14	et	et	PROPN
ajst-31431	256	15	al	al	PROPN
ajst-31431	256	16	.	.	PUNCT
ajst-31431	257	1	constructing	construct	VERB
ajst-31431	257	2	interval	interval	NOUN
ajst-31431	257	3	variables	variable	NOUN
ajst-31431	257	4	via	via	ADP
ajst-31431	257	5	faceted	faceted	PROPN
ajst-31431	257	6	rasch	rasch	PROPN
ajst-31431	257	7	measurement	measurement	PROPN
ajst-31431	257	8	and	and	CCONJ
ajst-31431	257	9	multitask	multitask	ADJ
ajst-31431	257	10	deep	deep	ADJ
ajst-31431	257	11	learning	learning	NOUN
ajst-31431	257	12	:	:	PUNCT
ajst-31431	257	13	a	a	DET
ajst-31431	257	14	hate	hate	NOUN
ajst-31431	257	15	speech	speech	NOUN
ajst-31431	257	16	application[j	application[j	PROPN
ajst-31431	257	17	]	]	PUNCT
ajst-31431	257	18	.	.	PUNCT
ajst-31431	258	1	arxiv	arxiv	PROPN
ajst-31431	258	2	preprint	preprint	NOUN
ajst-31431	258	3	arxiv:2009.10277	arxiv:2009.10277	NOUN
ajst-31431	258	4	,	,	PUNCT
ajst-31431	258	5	2020	2020	NUM
ajst-31431	258	6	.	.	PUNCT
ajst-31431	259	1	46	46	NUM
ajst-31431	260	1	[	[	X
ajst-31431	260	2	32	32	NUM
ajst-31431	260	3	]	]	X
ajst-31431	260	4	pavlopoulos	pavlopoulos	PROPN
ajst-31431	260	5	j	j	PROPN
ajst-31431	260	6	,	,	PUNCT
ajst-31431	260	7	sorensen	sorensen	PROPN
ajst-31431	260	8	j	j	PROPN
ajst-31431	260	9	,	,	PUNCT
ajst-31431	260	10	laugier	laugier	PROPN
ajst-31431	260	11	l	l	NOUN
ajst-31431	260	12	,	,	PUNCT
ajst-31431	260	13	et	et	PROPN
ajst-31431	260	14	al	al	PROPN
ajst-31431	260	15	.	.	PUNCT
ajst-31431	261	1	semeval-2021	semeval-2021	NOUN
ajst-31431	261	2	task	task	NOUN
ajst-31431	261	3	5	5	NUM
ajst-31431	261	4	:	:	SYM
ajst-31431	261	5	toxic	toxic	ADJ
ajst-31431	261	6	spans	span	NOUN
ajst-31431	261	7	detection[c]//proceedings	detection[c]//proceeding	NOUN
ajst-31431	261	8	of	of	ADP
ajst-31431	261	9	the	the	DET
ajst-31431	261	10	15th	15th	ADJ
ajst-31431	261	11	international	international	ADJ
ajst-31431	261	12	workshop	workshop	NOUN
ajst-31431	261	13	on	on	ADP
ajst-31431	261	14	semantic	semantic	ADJ
ajst-31431	261	15	evaluation	evaluation	NOUN
ajst-31431	261	16	(	(	PUNCT
ajst-31431	261	17	semeval2021	semeval2021	PROPN
ajst-31431	261	18	)	)	PUNCT
ajst-31431	261	19	.	.	PUNCT
ajst-31431	262	1	2021	2021	NUM
ajst-31431	262	2	:	:	PUNCT
ajst-31431	262	3	59	59	NUM
ajst-31431	262	4	-	-	SYM
ajst-31431	262	5	69	69	NUM
ajst-31431	262	6	.	.	PUNCT
ajst-31431	263	1	[	[	X
ajst-31431	263	2	33	33	NUM
ajst-31431	263	3	]	]	PUNCT
ajst-31431	263	4	mathew	mathew	PROPN
ajst-31431	263	5	b	b	PROPN
ajst-31431	263	6	,	,	PUNCT
ajst-31431	263	7	saha	saha	PROPN
ajst-31431	263	8	p	p	NOUN
ajst-31431	263	9	,	,	PUNCT
ajst-31431	263	10	yimam	yimam	PROPN
ajst-31431	263	11	s	s	PART
ajst-31431	263	12	m	m	PROPN
ajst-31431	263	13	,	,	PUNCT
ajst-31431	264	1	et	et	PROPN
ajst-31431	264	2	al	al	PROPN
ajst-31431	264	3	.	.	PROPN
ajst-31431	264	4	hatexplain	hatexplain	PROPN
ajst-31431	264	5	:	:	PUNCT
ajst-31431	264	6	a	a	DET
ajst-31431	264	7	benchmark	benchmark	NOUN
ajst-31431	264	8	dataset	dataset	NOUN
ajst-31431	264	9	for	for	ADP
ajst-31431	264	10	explainable	explainable	ADJ
ajst-31431	264	11	hate	hate	NOUN
ajst-31431	264	12	speech	speech	NOUN
ajst-31431	264	13	detection[c]//proceedings	detection[c]//proceeding	NOUN
ajst-31431	264	14	of	of	ADP
ajst-31431	264	15	the	the	DET
ajst-31431	264	16	aaai	aaai	PROPN
ajst-31431	264	17	conference	conference	NOUN
ajst-31431	264	18	on	on	ADP
ajst-31431	264	19	artificial	artificial	ADJ
ajst-31431	264	20	intelligence	intelligence	NOUN
ajst-31431	264	21	.	.	PUNCT
ajst-31431	265	1	2021	2021	NUM
ajst-31431	265	2	,	,	PUNCT
ajst-31431	265	3	35(17	35(17	NUM
ajst-31431	265	4	):	):	PUNCT
ajst-31431	265	5	14867	14867	NUM
ajst-31431	265	6	-	-	SYM
ajst-31431	265	7	14875	14875	NUM
ajst-31431	265	8	.	.	PUNCT
ajst-31431	266	1	2021	2021	NUM
ajst-31431	266	2	[	[	X
ajst-31431	266	3	34	34	NUM
ajst-31431	266	4	]	]	X
ajst-31431	266	5	de	de	X
ajst-31431	266	6	la	la	X
ajst-31431	266	7	peña	peña	PROPN
ajst-31431	266	8	sarracén	sarracén	PROPN
ajst-31431	266	9	g	g	PROPN
ajst-31431	266	10	l	l	PROPN
ajst-31431	266	11	,	,	PUNCT
ajst-31431	266	12	rosso	rosso	PROPN
ajst-31431	266	13	p.	p.	NOUN
ajst-31431	266	14	unsupervised	unsupervised	ADJ
ajst-31431	266	15	embeddings	embedding	NOUN
ajst-31431	266	16	with	with	ADP
ajst-31431	266	17	graph	graph	NOUN
ajst-31431	266	18	auto	auto	NOUN
ajst-31431	266	19	-	-	PUNCT
ajst-31431	266	20	encoders	encoder	NOUN
ajst-31431	266	21	for	for	ADP
ajst-31431	266	22	multi	multi	ADJ
ajst-31431	266	23	-	-	ADJ
ajst-31431	266	24	domain	domain	ADJ
ajst-31431	266	25	and	and	CCONJ
ajst-31431	266	26	multilingual	multilingual	ADJ
ajst-31431	266	27	hate	hate	NOUN
ajst-31431	266	28	speech	speech	NOUN
ajst-31431	266	29	detection[c]//proceedings	detection[c]//proceeding	NOUN
ajst-31431	266	30	of	of	ADP
ajst-31431	266	31	the	the	DET
ajst-31431	266	32	thirteenth	thirteenth	ADJ
ajst-31431	266	33	language	language	NOUN
ajst-31431	266	34	resources	resource	NOUN
ajst-31431	266	35	and	and	CCONJ
ajst-31431	266	36	evaluation	evaluation	NOUN
ajst-31431	266	37	conference	conference	NOUN
ajst-31431	266	38	.	.	PUNCT
ajst-31431	267	1	2022	2022	NUM
ajst-31431	267	2	:	:	PUNCT
ajst-31431	267	3	21962204	21962204	NUM
ajst-31431	267	4	.	.	PUNCT
ajst-31431	268	1	2196～2204	2196～2204	NUM
ajst-31431	269	1	[	[	X
ajst-31431	269	2	35	35	NUM
ajst-31431	269	3	]	]	PUNCT
ajst-31431	269	4	gandhi	gandhi	PROPN
ajst-31431	269	5	a	a	DET
ajst-31431	269	6	,	,	PUNCT
ajst-31431	269	7	adhvaryu	adhvaryu	NOUN
ajst-31431	269	8	k	k	PROPN
ajst-31431	269	9	,	,	PUNCT
ajst-31431	269	10	khanduja	khanduja	PROPN
ajst-31431	269	11	v.	v.	ADP
ajst-31431	269	12	multimodal	multimodal	ADJ
ajst-31431	269	13	sentiment	sentiment	NOUN
ajst-31431	269	14	analysis	analysis	NOUN
ajst-31431	269	15	:	:	PUNCT
ajst-31431	269	16	review	review	NOUN
ajst-31431	269	17	,	,	PUNCT
ajst-31431	269	18	application	application	NOUN
ajst-31431	269	19	domains	domain	NOUN
ajst-31431	269	20	and	and	CCONJ
ajst-31431	269	21	future	future	ADJ
ajst-31431	269	22	directions[c]//2021	directions[c]//2021	PROPN
ajst-31431	269	23	ieee	ieee	NOUN
ajst-31431	269	24	pune	pune	PROPN
ajst-31431	269	25	section	section	PROPN
ajst-31431	269	26	international	international	ADJ
ajst-31431	269	27	conference	conference	PROPN
ajst-31431	269	28	(	(	PUNCT
ajst-31431	269	29	punecon	punecon	NOUN
ajst-31431	269	30	)	)	PUNCT
ajst-31431	269	31	.	.	PUNCT
ajst-31431	270	1	ieee	ieee	NOUN
ajst-31431	270	2	,	,	PUNCT
ajst-31431	270	3	2021	2021	NUM
ajst-31431	270	4	:	:	PUNCT
ajst-31431	270	5	1	1	NUM
ajst-31431	270	6	-	-	SYM
ajst-31431	270	7	5	5	NUM
ajst-31431	270	8	.	.	PUNCT
ajst-31431	271	1	[	[	X
ajst-31431	271	2	36	36	NUM
ajst-31431	271	3	]	]	PUNCT
ajst-31431	271	4	krizhevsky	krizhevsky	PROPN
ajst-31431	271	5	a	a	PROPN
ajst-31431	271	6	,	,	PUNCT
ajst-31431	271	7	sutskever	sutskever	VERB
ajst-31431	271	8	i	i	PRON
ajst-31431	271	9	,	,	PUNCT
ajst-31431	271	10	hinton	hinton	PROPN
ajst-31431	271	11	g	g	PROPN
ajst-31431	271	12	e.	e.	PROPN
ajst-31431	271	13	imagenet	imagenet	PROPN
ajst-31431	271	14	classification	classification	NOUN
ajst-31431	271	15	with	with	ADP
ajst-31431	271	16	deep	deep	ADJ
ajst-31431	271	17	convolutional	convolutional	ADJ
ajst-31431	271	18	neural	neural	ADJ
ajst-31431	271	19	networks[j	networks[j	NOUN
ajst-31431	271	20	]	]	X
ajst-31431	271	21	.	.	PUNCT
ajst-31431	272	1	advances	advance	NOUN
ajst-31431	272	2	in	in	ADP
ajst-31431	272	3	neural	neural	ADJ
ajst-31431	272	4	information	information	NOUN
ajst-31431	272	5	processing	processing	NOUN
ajst-31431	272	6	systems	system	NOUN
ajst-31431	272	7	,	,	PUNCT
ajst-31431	272	8	2012	2012	NUM
ajst-31431	272	9	,	,	PUNCT
ajst-31431	272	10	25	25	NUM
ajst-31431	272	11	.	.	PUNCT
ajst-31431	273	1	[	[	X
ajst-31431	273	2	37	37	NUM
ajst-31431	273	3	]	]	X
ajst-31431	273	4	lan	lan	PROPN
ajst-31431	273	5	z	z	PROPN
ajst-31431	273	6	,	,	PUNCT
ajst-31431	273	7	chen	chen	PROPN
ajst-31431	273	8	m	m	PROPN
ajst-31431	273	9	,	,	PUNCT
ajst-31431	273	10	goodman	goodman	PROPN
ajst-31431	273	11	s	s	PROPN
ajst-31431	273	12	,	,	PUNCT
ajst-31431	273	13	et	et	PROPN
ajst-31431	273	14	al	al	PROPN
ajst-31431	273	15	.	.	PROPN
ajst-31431	273	16	albert	albert	PROPN
ajst-31431	273	17	:	:	PUNCT
ajst-31431	273	18	a	a	DET
ajst-31431	273	19	lite	lite	ADJ
ajst-31431	273	20	bert	bert	NOUN
ajst-31431	273	21	for	for	ADP
ajst-31431	273	22	selfsupervised	selfsupervised	ADJ
ajst-31431	273	23	learning	learning	NOUN
ajst-31431	273	24	of	of	ADP
ajst-31431	273	25	language	language	NOUN
ajst-31431	273	26	representations[j	representations[j	PROPN
ajst-31431	273	27	]	]	PUNCT
ajst-31431	273	28	.	.	PUNCT
ajst-31431	274	1	arxiv	arxiv	PROPN
ajst-31431	274	2	preprint	preprint	PROPN
ajst-31431	274	3	arxiv:1909.11942	arxiv:1909.11942	NOUN
ajst-31431	274	4	,	,	PUNCT
ajst-31431	274	5	2019	2019	NUM
ajst-31431	274	6	.	.	PUNCT
ajst-31431	275	1	[	[	X
ajst-31431	275	2	38	38	NUM
ajst-31431	275	3	]	]	PUNCT
ajst-31431	275	4	howard	howard	PROPN
ajst-31431	275	5	j	j	PROPN
ajst-31431	275	6	,	,	PUNCT
ajst-31431	275	7	ruder	rude	ADJ
ajst-31431	275	8	s.	s.	PROPN
ajst-31431	275	9	universal	universal	PROPN
ajst-31431	275	10	language	language	PROPN
ajst-31431	275	11	model	model	NOUN
ajst-31431	275	12	fine	fine	ADV
ajst-31431	275	13	-	-	PUNCT
ajst-31431	275	14	tuning	tuning	NOUN
ajst-31431	275	15	for	for	ADP
ajst-31431	275	16	text	text	NOUN
ajst-31431	275	17	classification[j	classification[j	NOUN
ajst-31431	275	18	]	]	X
ajst-31431	275	19	.	.	PUNCT
ajst-31431	276	1	arxiv	arxiv	PROPN
ajst-31431	276	2	preprint	preprint	NOUN
ajst-31431	276	3	arxiv:1801.06146	arxiv:1801.06146	NOUN
ajst-31431	276	4	,	,	PUNCT
ajst-31431	276	5	2018	2018	NUM
ajst-31431	276	6	.	.	PUNCT
ajst-31431	277	1	[	[	X
ajst-31431	277	2	39	39	NUM
ajst-31431	277	3	]	]	PUNCT
ajst-31431	277	4	simonyan	simonyan	PROPN
ajst-31431	277	5	k	k	NOUN
ajst-31431	277	6	,	,	PUNCT
ajst-31431	277	7	zisserman	zisserman	NOUN
ajst-31431	277	8	a.	a.	NOUN
ajst-31431	277	9	very	very	ADV
ajst-31431	277	10	deep	deep	ADJ
ajst-31431	277	11	convolutional	convolutional	ADJ
ajst-31431	277	12	networks	network	NOUN
ajst-31431	277	13	for	for	ADP
ajst-31431	277	14	large	large	ADJ
ajst-31431	277	15	-	-	PUNCT
ajst-31431	277	16	scale	scale	NOUN
ajst-31431	277	17	image	image	NOUN
ajst-31431	277	18	recognition[j	recognition[j	NOUN
ajst-31431	277	19	]	]	PUNCT
ajst-31431	277	20	.	.	PUNCT
ajst-31431	278	1	arxiv	arxiv	PROPN
ajst-31431	278	2	preprint	preprint	PROPN
ajst-31431	278	3	arxiv:1409.1556	arxiv:1409.1556	NOUN
ajst-31431	278	4	,	,	PUNCT
ajst-31431	278	5	2014	2014	NUM
ajst-31431	278	6	.	.	PUNCT
ajst-31431	279	1	[	[	X
ajst-31431	279	2	40	40	NUM
ajst-31431	279	3	]	]	PUNCT
ajst-31431	279	4	he	he	PRON
ajst-31431	279	5	k	k	PROPN
ajst-31431	279	6	,	,	PUNCT
ajst-31431	279	7	zhang	zhang	PROPN
ajst-31431	279	8	x	x	PROPN
ajst-31431	279	9	,	,	PUNCT
ajst-31431	279	10	ren	ren	PROPN
ajst-31431	279	11	s	s	PROPN
ajst-31431	279	12	,	,	PUNCT
ajst-31431	279	13	et	et	PROPN
ajst-31431	279	14	al	al	PROPN
ajst-31431	279	15	.	.	PUNCT
ajst-31431	280	1	deep	deep	ADJ
ajst-31431	280	2	residual	residual	ADJ
ajst-31431	280	3	learning	learning	NOUN
ajst-31431	280	4	for	for	ADP
ajst-31431	280	5	image	image	NOUN
ajst-31431	280	6	recognition[c]//proceedings	recognition[c]//proceeding	NOUN
ajst-31431	280	7	of	of	ADP
ajst-31431	280	8	the	the	DET
ajst-31431	280	9	ieee	ieee	NOUN
ajst-31431	280	10	conference	conference	NOUN
ajst-31431	280	11	on	on	ADP
ajst-31431	280	12	computer	computer	NOUN
ajst-31431	280	13	vision	vision	NOUN
ajst-31431	280	14	and	and	CCONJ
ajst-31431	280	15	pattern	pattern	NOUN
ajst-31431	280	16	recognition	recognition	NOUN
ajst-31431	280	17	.	.	PUNCT
ajst-31431	281	1	2016	2016	NUM
ajst-31431	281	2	:	:	PUNCT
ajst-31431	282	1	770	770	NUM
ajst-31431	282	2	-	-	SYM
ajst-31431	282	3	778	778	NUM
ajst-31431	282	4	.	.	PUNCT
ajst-31431	283	1	[	[	X
ajst-31431	283	2	41	41	NUM
ajst-31431	283	3	]	]	X
ajst-31431	283	4	huang	huang	PROPN
ajst-31431	283	5	g	g	PROPN
ajst-31431	283	6	,	,	PUNCT
ajst-31431	283	7	liu	liu	PROPN
ajst-31431	283	8	z	z	PROPN
ajst-31431	283	9	,	,	PUNCT
ajst-31431	283	10	van	van	PROPN
ajst-31431	283	11	der	der	NOUN
ajst-31431	283	12	maaten	maaten	VERB
ajst-31431	283	13	l	l	NOUN
ajst-31431	283	14	,	,	PUNCT
ajst-31431	283	15	et	et	PROPN
ajst-31431	283	16	al	al	PROPN
ajst-31431	283	17	.	.	PROPN
ajst-31431	284	1	densely	densely	ADV
ajst-31431	284	2	connected	connect	VERB
ajst-31431	284	3	convolutional	convolutional	ADJ
ajst-31431	284	4	networks[c]//proceedings	networks[c]//proceeding	NOUN
ajst-31431	284	5	of	of	ADP
ajst-31431	284	6	the	the	DET
ajst-31431	284	7	ieee	ieee	NOUN
ajst-31431	284	8	conference	conference	NOUN
ajst-31431	284	9	on	on	ADP
ajst-31431	284	10	computer	computer	NOUN
ajst-31431	284	11	vision	vision	NOUN
ajst-31431	284	12	and	and	CCONJ
ajst-31431	284	13	pattern	pattern	NOUN
ajst-31431	284	14	recognition	recognition	NOUN
ajst-31431	284	15	.	.	PUNCT
ajst-31431	285	1	2017	2017	NUM
ajst-31431	285	2	:	:	SYM
ajst-31431	285	3	4700	4700	NUM
ajst-31431	285	4	-	-	SYM
ajst-31431	285	5	4708	4708	NUM
ajst-31431	285	6	.	.	PUNCT
ajst-31431	286	1	4700～4708	4700～4708	NUM
ajst-31431	286	2	[	[	X
ajst-31431	286	3	42	42	NUM
ajst-31431	286	4	]	]	PUNCT
ajst-31431	286	5	gomez	gomez	PROPN
ajst-31431	286	6	r	r	PROPN
ajst-31431	286	7	,	,	PUNCT
ajst-31431	286	8	gibert	gibert	PROPN
ajst-31431	286	9	j	j	PROPN
ajst-31431	286	10	,	,	PUNCT
ajst-31431	286	11	gomez	gomez	PROPN
ajst-31431	286	12	l	l	PROPN
ajst-31431	286	13	,	,	PUNCT
ajst-31431	286	14	et	et	PROPN
ajst-31431	286	15	al	al	PROPN
ajst-31431	286	16	.	.	PUNCT
ajst-31431	286	17	exploring	explore	VERB
ajst-31431	286	18	hate	hate	NOUN
ajst-31431	286	19	speech	speech	NOUN
ajst-31431	286	20	detection	detection	NOUN
ajst-31431	286	21	in	in	ADP
ajst-31431	286	22	multimodal	multimodal	ADJ
ajst-31431	286	23	publications[c]//proceedings	publications[c]//proceeding	NOUN
ajst-31431	286	24	of	of	ADP
ajst-31431	286	25	the	the	DET
ajst-31431	286	26	ieee	ieee	NOUN
ajst-31431	286	27	/	/	SYM
ajst-31431	286	28	cvf	cvf	NOUN
ajst-31431	286	29	winter	winter	NOUN
ajst-31431	286	30	conference	conference	NOUN
ajst-31431	286	31	on	on	ADP
ajst-31431	286	32	applications	application	NOUN
ajst-31431	286	33	of	of	ADP
ajst-31431	286	34	computer	computer	NOUN
ajst-31431	286	35	vision	vision	NOUN
ajst-31431	286	36	.	.	PUNCT
ajst-31431	287	1	2020	2020	NUM
ajst-31431	287	2	:	:	PUNCT
ajst-31431	287	3	1470	1470	NUM
ajst-31431	287	4	-	-	SYM
ajst-31431	287	5	1478	1478	NUM
ajst-31431	287	6	.	.	PUNCT
ajst-31431	288	1	[	[	X
ajst-31431	288	2	43	43	NUM
ajst-31431	288	3	]	]	X
ajst-31431	288	4	chen	chen	PROPN
ajst-31431	288	5	y	y	PROPN
ajst-31431	288	6	c	c	PROPN
ajst-31431	288	7	,	,	PUNCT
ajst-31431	288	8	li	li	PROPN
ajst-31431	288	9	l	l	PROPN
ajst-31431	288	10	,	,	PUNCT
ajst-31431	288	11	yu	yu	PROPN
ajst-31431	288	12	l	l	PROPN
ajst-31431	288	13	,	,	PUNCT
ajst-31431	288	14	et	et	PROPN
ajst-31431	288	15	al	al	PROPN
ajst-31431	288	16	.	.	PUNCT
ajst-31431	288	17	uniter	uniter	NOUN
ajst-31431	288	18	:	:	PUNCT
ajst-31431	288	19	universal	universal	ADJ
ajst-31431	288	20	image	image	NOUN
ajst-31431	288	21	-	-	PUNCT
ajst-31431	288	22	text	text	NOUN
ajst-31431	288	23	representation	representation	NOUN
ajst-31431	288	24	learning[c]//european	learning[c]//european	PROPN
ajst-31431	288	25	conference	conference	NOUN
ajst-31431	288	26	on	on	ADP
ajst-31431	288	27	computer	computer	NOUN
ajst-31431	288	28	vision	vision	NOUN
ajst-31431	288	29	.	.	PUNCT
ajst-31431	289	1	cham	cham	PROPN
ajst-31431	289	2	:	:	PUNCT
ajst-31431	289	3	springer	springer	NOUN
ajst-31431	289	4	international	international	ADJ
ajst-31431	289	5	publishing	publishing	NOUN
ajst-31431	289	6	,	,	PUNCT
ajst-31431	289	7	2020	2020	NUM
ajst-31431	289	8	:	:	PUNCT
ajst-31431	289	9	104120	104120	NUM
ajst-31431	289	10	.	.	PUNCT