id	sid	tid	token	lemma	pos
easat-10905	1	1	edelweiss	edelweiss	PROPN
easat-10905	1	2	applied	apply	VERB
easat-10905	1	3	science	science	NOUN
easat-10905	1	4	and	and	CCONJ
easat-10905	1	5	technology	technology	NOUN
easat-10905	1	6	issn	issn	PROPN
easat-10905	1	7	:	:	PUNCT
easat-10905	1	8	2576	2576	NUM
easat-10905	1	9	-	-	SYM
easat-10905	1	10	8484	8484	NUM
easat-10905	1	11	vol	vol	NOUN
easat-10905	1	12	.	.	PROPN
easat-10905	2	1	9	9	NUM
easat-10905	2	2	,	,	PUNCT
easat-10905	2	3	no	no	INTJ
easat-10905	2	4	.	.	NOUN
easat-10905	2	5	11	11	NUM
easat-10905	2	6	,	,	PUNCT
easat-10905	2	7	448	448	NUM
easat-10905	2	8	-	-	SYM
easat-10905	2	9	466	466	NUM
easat-10905	2	10	2025	2025	NUM
easat-10905	2	11	publisher	publisher	NOUN
easat-10905	2	12	:	:	PUNCT
easat-10905	2	13	learning	learn	VERB
easat-10905	2	14	gate	gate	NOUN
easat-10905	2	15	doi	doi	PROPN
easat-10905	2	16	:	:	PUNCT
easat-10905	2	17	10.55214/2576	10.55214/2576	NUM
easat-10905	2	18	-	-	PUNCT
easat-10905	2	19	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	3	1	©	©	PROPN
easat-10905	3	2	2025	2025	NUM
easat-10905	3	3	by	by	ADP
easat-10905	3	4	the	the	DET
easat-10905	3	5	authors	author	NOUN
easat-10905	3	6	;	;	PUNCT
easat-10905	3	7	licensee	licensee	PROPN
easat-10905	3	8	learning	learning	NOUN
easat-10905	3	9	gate	gate	NOUN
easat-10905	3	10	©	©	PROPN
easat-10905	3	11	2025	2025	NUM
easat-10905	3	12	by	by	ADP
easat-10905	3	13	the	the	DET
easat-10905	3	14	authors	author	NOUN
easat-10905	3	15	;	;	PUNCT
easat-10905	3	16	licensee	licensee	PROPN
easat-10905	3	17	learning	learning	NOUN
easat-10905	3	18	gate	gate	NOUN
easat-10905	3	19	history	history	NOUN
easat-10905	3	20	:	:	PUNCT
easat-10905	3	21	received	receive	VERB
easat-10905	3	22	:	:	PUNCT
easat-10905	3	23	2	2	NUM
easat-10905	3	24	september	september	PROPN
easat-10905	3	25	2025	2025	NUM
easat-10905	3	26	;	;	PUNCT
easat-10905	3	27	revised	revise	VERB
easat-10905	3	28	:	:	PUNCT
easat-10905	3	29	10	10	NUM
easat-10905	3	30	october	october	NOUN
easat-10905	3	31	2025	2025	NUM
easat-10905	3	32	;	;	PUNCT
easat-10905	3	33	accepted	accept	VERB
easat-10905	3	34	:	:	PUNCT
easat-10905	3	35	14	14	NUM
easat-10905	3	36	october	october	NOUN
easat-10905	3	37	2025	2025	NUM
easat-10905	3	38	;	;	PUNCT
easat-10905	3	39	published	publish	VERB
easat-10905	3	40	:	:	PUNCT
easat-10905	3	41	7	7	NUM
easat-10905	3	42	november	november	PROPN
easat-10905	3	43	2025	2025	NUM
easat-10905	3	44	*	*	PUNCT
easat-10905	3	45	correspondence	correspondence	NOUN
easat-10905	3	46	:	:	PUNCT
easat-10905	4	1	mallinishanth72@gmail.com	mallinishanth72@gmail.com	X
easat-10905	4	2	learning	learn	VERB
easat-10905	4	3	without	without	ADP
easat-10905	4	4	forgetting	forget	VERB
easat-10905	4	5	in	in	ADP
easat-10905	4	6	nlp	nlp	NOUN
easat-10905	4	7	:	:	PUNCT
easat-10905	4	8	deep	deep	ADJ
easat-10905	4	9	learning	learning	NOUN
easat-10905	4	10	based	base	VERB
easat-10905	4	11	transformer	transformer	NOUN
easat-10905	4	12	models	model	NOUN
easat-10905	4	13	for	for	ADP
easat-10905	4	14	lifelong	lifelong	ADJ
easat-10905	4	15	task	task	NOUN
easat-10905	4	16	incremental	incremental	ADJ
easat-10905	4	17	learning	learning	PROPN
easat-10905	4	18	ambi	ambi	PROPN
easat-10905	4	19	rachel	rachel	PROPN
easat-10905	4	20	alex1	alex1	PROPN
easat-10905	4	21	,	,	PUNCT
easat-10905	4	22	malliga	malliga	ADJ
easat-10905	4	23	subramanian2	subramanian2	NOUN
easat-10905	4	24	*	*	NOUN
easat-10905	4	25	,	,	PUNCT
easat-10905	4	26	jayanth	jayanth	PROPN
easat-10905	4	27	j	j	PROPN
easat-10905	4	28	r2	r2	PROPN
easat-10905	4	29	,	,	PUNCT
easat-10905	4	30	keerthi	keerthi	PROPN
easat-10905	4	31	bala	bala	VERB
easat-10905	4	32	a	a	DET
easat-10905	4	33	t2	t2	NOUN
easat-10905	4	34	,	,	PUNCT
easat-10905	4	35	taniya	taniya	NOUN
easat-10905	4	36	mukherjee3	mukherjee3	NOUN
easat-10905	4	37	1college	1college	NUM
easat-10905	4	38	of	of	ADP
easat-10905	4	39	engineering	engineering	PROPN
easat-10905	4	40	,	,	PUNCT
easat-10905	4	41	gulf	gulf	PROPN
easat-10905	4	42	university	university	NOUN
easat-10905	4	43	,	,	PUNCT
easat-10905	4	44	sanad	sanad	NOUN
easat-10905	4	45	26489	26489	NUM
easat-10905	4	46	,	,	PUNCT
easat-10905	4	47	kingdom	kingdom	NOUN
easat-10905	4	48	of	of	ADP
easat-10905	4	49	bahrain	bahrain	NOUN
easat-10905	4	50	.	.	PUNCT
easat-10905	5	1	2department	2department	NUM
easat-10905	5	2	of	of	ADP
easat-10905	5	3	computer	computer	NOUN
easat-10905	5	4	science	science	NOUN
easat-10905	5	5	and	and	CCONJ
easat-10905	5	6	engineering	engineering	NOUN
easat-10905	5	7	;	;	PUNCT
easat-10905	5	8	kongu	kongu	PROPN
easat-10905	5	9	engineering	engineering	PROPN
easat-10905	5	10	college	college	PROPN
easat-10905	5	11	,	,	PUNCT
easat-10905	5	12	india	india	PROPN
easat-10905	5	13	;	;	PUNCT
easat-10905	5	14	mallinishanth72@gmail.com	mallinishanth72@gmail.com	X
easat-10905	5	15	(	(	PUNCT
easat-10905	5	16	m.s	m.s	PROPN
easat-10905	5	17	.	.	PROPN
easat-10905	5	18	)	)	PUNCT
easat-10905	5	19	.	.	PUNCT
easat-10905	6	1	3general	3general	NUM
easat-10905	6	2	studies	study	NOUN
easat-10905	6	3	unit	unit	NOUN
easat-10905	6	4	,	,	PUNCT
easat-10905	6	5	gulf	gulf	PROPN
easat-10905	6	6	university	university	PROPN
easat-10905	6	7	,	,	PUNCT
easat-10905	6	8	sanad	sanad	NOUN
easat-10905	6	9	26489	26489	NUM
easat-10905	6	10	,	,	PUNCT
easat-10905	6	11	kingdom	kingdom	NOUN
easat-10905	6	12	of	of	ADP
easat-10905	6	13	bahrain	bahrain	NOUN
easat-10905	6	14	.	.	PUNCT
easat-10905	7	1	abstract	abstract	ADJ
easat-10905	7	2	:	:	PUNCT
easat-10905	7	3	this	this	DET
easat-10905	7	4	study	study	NOUN
easat-10905	7	5	investigates	investigate	VERB
easat-10905	7	6	task	task	NOUN
easat-10905	7	7	incremental	incremental	ADJ
easat-10905	7	8	learning	learning	NOUN
easat-10905	7	9	for	for	ADP
easat-10905	7	10	developing	develop	VERB
easat-10905	7	11	transformer	transformer	NOUN
easat-10905	7	12	-	-	PUNCT
easat-10905	7	13	based	base	VERB
easat-10905	7	14	natural	natural	ADJ
easat-10905	7	15	language	language	NOUN
easat-10905	7	16	processing	processing	NOUN
easat-10905	7	17	(	(	PUNCT
easat-10905	7	18	nlp	nlp	NOUN
easat-10905	7	19	)	)	PUNCT
easat-10905	7	20	models	model	NOUN
easat-10905	7	21	capable	capable	ADJ
easat-10905	7	22	of	of	ADP
easat-10905	7	23	sequentially	sequentially	ADV
easat-10905	7	24	learning	learn	VERB
easat-10905	7	25	new	new	ADJ
easat-10905	7	26	tasks	task	NOUN
easat-10905	7	27	without	without	ADP
easat-10905	7	28	forgetting	forget	VERB
easat-10905	7	29	previously	previously	ADV
easat-10905	7	30	acquired	acquire	VERB
easat-10905	7	31	knowledge	knowledge	NOUN
easat-10905	7	32	.	.	PUNCT
easat-10905	8	1	it	it	PRON
easat-10905	8	2	focuses	focus	VERB
easat-10905	8	3	on	on	ADP
easat-10905	8	4	enabling	enable	VERB
easat-10905	8	5	models	model	NOUN
easat-10905	8	6	to	to	PART
easat-10905	8	7	learn	learn	VERB
easat-10905	8	8	from	from	ADP
easat-10905	8	9	abusive	abusive	ADJ
easat-10905	8	10	comment	comment	NOUN
easat-10905	8	11	detection	detection	NOUN
easat-10905	8	12	to	to	PART
easat-10905	8	13	offensive	offensive	ADJ
easat-10905	8	14	comment	comment	NOUN
easat-10905	8	15	detection	detection	NOUN
easat-10905	8	16	while	while	SCONJ
easat-10905	8	17	maintaining	maintain	VERB
easat-10905	8	18	previous	previous	ADJ
easat-10905	8	19	knowledge	knowledge	NOUN
easat-10905	8	20	.	.	PUNCT
easat-10905	9	1	pre	pre	VERB
easat-10905	9	2	-	-	ADJ
easat-10905	9	3	trained	train	VERB
easat-10905	9	4	transformer	transformer	NOUN
easat-10905	9	5	models	model	NOUN
easat-10905	9	6	such	such	ADJ
easat-10905	9	7	as	as	ADP
easat-10905	9	8	ert	ert	PROPN
easat-10905	9	9	,	,	PUNCT
easat-10905	9	10	albert	albert	PROPN
easat-10905	9	11	,	,	PUNCT
easat-10905	9	12	roberta	roberta	PROPN
easat-10905	9	13	,	,	PUNCT
easat-10905	9	14	and	and	CCONJ
easat-10905	9	15	distilbert	distilbert	PROPN
easat-10905	9	16	were	be	AUX
easat-10905	9	17	integrated	integrate	VERB
easat-10905	9	18	with	with	ADP
easat-10905	9	19	continual	continual	ADJ
easat-10905	9	20	learning	learning	NOUN
easat-10905	9	21	strategies	strategy	NOUN
easat-10905	9	22	,	,	PUNCT
easat-10905	9	23	including	include	VERB
easat-10905	9	24	rehearsal	rehearsal	NOUN
easat-10905	9	25	,	,	PUNCT
easat-10905	9	26	pseudo	pseudo	NOUN
easat-10905	9	27	-	-	NOUN
easat-10905	9	28	rehearsal	rehearsal	NOUN
easat-10905	9	29	,	,	PUNCT
easat-10905	9	30	and	and	CCONJ
easat-10905	9	31	regularization	regularization	NOUN
easat-10905	9	32	-	-	PUNCT
easat-10905	9	33	based	base	VERB
easat-10905	9	34	methods	method	NOUN
easat-10905	9	35	.	.	PUNCT
easat-10905	10	1	their	their	PRON
easat-10905	10	2	effectiveness	effectiveness	NOUN
easat-10905	10	3	was	be	AUX
easat-10905	10	4	evaluated	evaluate	VERB
easat-10905	10	5	on	on	ADP
easat-10905	10	6	sequential	sequential	ADJ
easat-10905	10	7	abusive	abusive	ADJ
easat-10905	10	8	and	and	CCONJ
easat-10905	10	9	offensive	offensive	ADJ
easat-10905	10	10	comment	comment	NOUN
easat-10905	10	11	detection	detection	NOUN
easat-10905	10	12	tasks	task	NOUN
easat-10905	10	13	using	use	VERB
easat-10905	10	14	measures	measure	NOUN
easat-10905	10	15	of	of	ADP
easat-10905	10	16	classification	classification	NOUN
easat-10905	10	17	accuracy	accuracy	NOUN
easat-10905	10	18	and	and	CCONJ
easat-10905	10	19	knowledge	knowledge	NOUN
easat-10905	10	20	retention	retention	NOUN
easat-10905	10	21	across	across	ADP
easat-10905	10	22	multiple	multiple	ADJ
easat-10905	10	23	datasets	dataset	NOUN
easat-10905	10	24	.	.	PUNCT
easat-10905	11	1	among	among	ADP
easat-10905	11	2	the	the	DET
easat-10905	11	3	evaluated	evaluate	VERB
easat-10905	11	4	models	model	NOUN
easat-10905	11	5	,	,	PUNCT
easat-10905	11	6	bert	bert	PROPN
easat-10905	11	7	integrated	integrate	VERB
easat-10905	11	8	with	with	ADP
easat-10905	11	9	the	the	DET
easat-10905	11	10	learning	learning	NOUN
easat-10905	11	11	without	without	ADP
easat-10905	11	12	forgetting	forget	VERB
easat-10905	11	13	(	(	PUNCT
easat-10905	11	14	lwf	lwf	PROPN
easat-10905	11	15	)	)	PUNCT
easat-10905	11	16	approach	approach	NOUN
easat-10905	11	17	achieved	achieve	VERB
easat-10905	11	18	the	the	DET
easat-10905	11	19	best	good	ADJ
easat-10905	11	20	trade	trade	NOUN
easat-10905	11	21	-	-	PUNCT
easat-10905	11	22	off	off	NOUN
easat-10905	11	23	between	between	ADP
easat-10905	11	24	stability	stability	NOUN
easat-10905	11	25	and	and	CCONJ
easat-10905	11	26	plasticity	plasticity	NOUN
easat-10905	11	27	,	,	PUNCT
easat-10905	11	28	attaining	attain	VERB
easat-10905	11	29	accuracies	accuracy	NOUN
easat-10905	11	30	of	of	ADP
easat-10905	11	31	91.06	91.06	NUM
easat-10905	11	32	%	%	NOUN
easat-10905	11	33	for	for	ADP
easat-10905	11	34	abusive	abusive	ADJ
easat-10905	11	35	and	and	CCONJ
easat-10905	11	36	98.8	98.8	NUM
easat-10905	11	37	%	%	NOUN
easat-10905	11	38	for	for	ADP
easat-10905	11	39	offensive	offensive	ADJ
easat-10905	11	40	comment	comment	NOUN
easat-10905	11	41	detection	detection	NOUN
easat-10905	11	42	.	.	PUNCT
easat-10905	12	1	this	this	PRON
easat-10905	12	2	demonstrates	demonstrate	VERB
easat-10905	12	3	the	the	DET
easat-10905	12	4	ability	ability	NOUN
easat-10905	12	5	of	of	ADP
easat-10905	12	6	continual	continual	ADJ
easat-10905	12	7	learning	learning	NOUN
easat-10905	12	8	to	to	PART
easat-10905	12	9	avoid	avoid	VERB
easat-10905	12	10	catastrophic	catastrophic	ADJ
easat-10905	12	11	forgetting	forgetting	NOUN
easat-10905	12	12	in	in	ADP
easat-10905	12	13	sequential	sequential	ADJ
easat-10905	12	14	nlp	nlp	NOUN
easat-10905	12	15	tasks	task	NOUN
easat-10905	12	16	.	.	PUNCT
easat-10905	13	1	task	task	NOUN
easat-10905	13	2	incremental	incremental	ADJ
easat-10905	13	3	learning	learning	NOUN
easat-10905	13	4	enables	enable	VERB
easat-10905	13	5	transformer	transformer	NOUN
easat-10905	13	6	models	model	NOUN
easat-10905	13	7	to	to	PART
easat-10905	13	8	adapt	adapt	VERB
easat-10905	13	9	to	to	ADP
easat-10905	13	10	new	new	ADJ
easat-10905	13	11	linguistic	linguistic	ADJ
easat-10905	13	12	challenges	challenge	NOUN
easat-10905	13	13	while	while	SCONJ
easat-10905	13	14	retaining	retain	VERB
easat-10905	13	15	prior	prior	ADJ
easat-10905	13	16	knowledge	knowledge	NOUN
easat-10905	13	17	,	,	PUNCT
easat-10905	13	18	supporting	support	VERB
easat-10905	13	19	lifelong	lifelong	ADJ
easat-10905	13	20	learning	learning	NOUN
easat-10905	13	21	.	.	PUNCT
easat-10905	14	1	the	the	DET
easat-10905	14	2	proposed	propose	VERB
easat-10905	14	3	framework	framework	NOUN
easat-10905	14	4	provides	provide	VERB
easat-10905	14	5	valuable	valuable	ADJ
easat-10905	14	6	insights	insight	NOUN
easat-10905	14	7	for	for	ADP
easat-10905	14	8	building	build	VERB
easat-10905	14	9	adaptable	adaptable	ADJ
easat-10905	14	10	nlp	nlp	NOUN
easat-10905	14	11	systems	system	NOUN
easat-10905	14	12	applicable	applicable	ADJ
easat-10905	14	13	to	to	AUX
easat-10905	14	14	content	content	NOUN
easat-10905	14	15	moderation	moderation	NOUN
easat-10905	14	16	,	,	PUNCT
easat-10905	14	17	toxic	toxic	ADJ
easat-10905	14	18	speech	speech	NOUN
easat-10905	14	19	detection	detection	NOUN
easat-10905	14	20	,	,	PUNCT
easat-10905	14	21	and	and	CCONJ
easat-10905	14	22	other	other	ADJ
easat-10905	14	23	evolving	evolve	VERB
easat-10905	14	24	language	language	NOUN
easat-10905	14	25	-	-	PUNCT
easat-10905	14	26	based	base	VERB
easat-10905	14	27	applications	application	NOUN
easat-10905	14	28	.	.	PUNCT
easat-10905	15	1	keywords	keyword	NOUN
easat-10905	15	2	:	:	PUNCT
easat-10905	15	3	abusive	abusive	ADJ
easat-10905	15	4	task	task	NOUN
easat-10905	15	5	,	,	PUNCT
easat-10905	15	6	learning	learn	VERB
easat-10905	15	7	without	without	ADP
easat-10905	15	8	forgetting	forget	VERB
easat-10905	15	9	,	,	PUNCT
easat-10905	15	10	offensive	offensive	ADJ
easat-10905	15	11	dataset	dataset	NOUN
easat-10905	15	12	,	,	PUNCT
easat-10905	15	13	task	task	NOUN
easat-10905	15	14	incremental	incremental	ADJ
easat-10905	15	15	learning	learning	NOUN
easat-10905	15	16	,	,	PUNCT
easat-10905	15	17	transformer	transformer	NOUN
easat-10905	15	18	models	model	NOUN
easat-10905	15	19	.	.	PUNCT
easat-10905	16	1	1	1	X
easat-10905	16	2	.	.	X
easat-10905	16	3	introduction	introduction	NOUN
easat-10905	16	4	social	social	ADJ
easat-10905	16	5	media	medium	NOUN
easat-10905	16	6	platforms	platform	NOUN
easat-10905	16	7	such	such	ADJ
easat-10905	16	8	as	as	ADP
easat-10905	16	9	youtube	youtube	NOUN
easat-10905	16	10	,	,	PUNCT
easat-10905	16	11	twitter	twitter	NOUN
easat-10905	16	12	,	,	PUNCT
easat-10905	16	13	facebook	facebook	NOUN
easat-10905	16	14	,	,	PUNCT
easat-10905	16	15	etc	etc	X
easat-10905	16	16	.	.	X
easat-10905	16	17	,	,	PUNCT
easat-10905	16	18	have	have	AUX
easat-10905	16	19	become	become	VERB
easat-10905	16	20	an	an	DET
easat-10905	16	21	integral	integral	ADJ
easat-10905	16	22	part	part	NOUN
easat-10905	16	23	of	of	ADP
easat-10905	16	24	modern	modern	ADJ
easat-10905	16	25	communication	communication	NOUN
easat-10905	16	26	,	,	PUNCT
easat-10905	16	27	enabling	enable	VERB
easat-10905	16	28	the	the	DET
easat-10905	16	29	exchange	exchange	NOUN
easat-10905	16	30	of	of	ADP
easat-10905	16	31	ideas	idea	NOUN
easat-10905	16	32	and	and	CCONJ
easat-10905	16	33	information	information	NOUN
easat-10905	16	34	[	[	X
easat-10905	16	35	1	1	NUM
easat-10905	16	36	]	]	PUNCT
easat-10905	16	37	.	.	PUNCT
easat-10905	17	1	the	the	DET
easat-10905	17	2	proliferation	proliferation	NOUN
easat-10905	17	3	of	of	ADP
easat-10905	17	4	these	these	DET
easat-10905	17	5	platforms	platform	NOUN
easat-10905	17	6	has	have	AUX
easat-10905	17	7	transformed	transform	VERB
easat-10905	17	8	the	the	DET
easat-10905	17	9	way	way	NOUN
easat-10905	17	10	individuals	individual	NOUN
easat-10905	17	11	communicate	communicate	VERB
easat-10905	17	12	.	.	PUNCT
easat-10905	18	1	while	while	SCONJ
easat-10905	18	2	these	these	DET
easat-10905	18	3	platforms	platform	NOUN
easat-10905	18	4	provide	provide	VERB
easat-10905	18	5	numerous	numerous	ADJ
easat-10905	18	6	benefits	benefit	NOUN
easat-10905	18	7	,	,	PUNCT
easat-10905	18	8	they	they	PRON
easat-10905	18	9	also	also	ADV
easat-10905	18	10	become	become	VERB
easat-10905	18	11	channels	channel	NOUN
easat-10905	18	12	for	for	ADP
easat-10905	18	13	spreading	spread	VERB
easat-10905	18	14	abusive	abusive	ADJ
easat-10905	18	15	comments	comment	NOUN
easat-10905	18	16	.	.	PUNCT
easat-10905	19	1	abusive	abusive	ADJ
easat-10905	19	2	language	language	NOUN
easat-10905	19	3	triggers	trigger	VERB
easat-10905	19	4	cyberbullying	cyberbullye	VERB
easat-10905	19	5	,	,	PUNCT
easat-10905	19	6	which	which	PRON
easat-10905	19	7	targets	target	VERB
easat-10905	19	8	individuals	individual	NOUN
easat-10905	19	9	such	such	ADJ
easat-10905	19	10	as	as	ADP
easat-10905	19	11	celebrities	celebrity	NOUN
easat-10905	19	12	,	,	PUNCT
easat-10905	19	13	politicians	politician	NOUN
easat-10905	19	14	,	,	PUNCT
easat-10905	19	15	etc	etc	X
easat-10905	19	16	.	.	X
easat-10905	19	17	,	,	PUNCT
easat-10905	19	18	and	and	CCONJ
easat-10905	19	19	groups	group	NOUN
easat-10905	19	20	such	such	ADJ
easat-10905	19	21	as	as	ADP
easat-10905	19	22	specific	specific	ADJ
easat-10905	19	23	countries	country	NOUN
easat-10905	19	24	,	,	PUNCT
easat-10905	19	25	genders	gender	NOUN
easat-10905	19	26	,	,	PUNCT
easat-10905	19	27	religions	religion	NOUN
easat-10905	19	28	,	,	PUNCT
easat-10905	19	29	etc	etc	X
easat-10905	20	1	[	[	X
easat-10905	20	2	2	2	NUM
easat-10905	20	3	]	]	PUNCT
easat-10905	20	4	.	.	PUNCT
easat-10905	21	1	such	such	ADJ
easat-10905	21	2	comments	comment	NOUN
easat-10905	21	3	can	can	AUX
easat-10905	21	4	have	have	VERB
easat-10905	21	5	detrimental	detrimental	ADJ
easat-10905	21	6	effects	effect	NOUN
easat-10905	21	7	,	,	PUNCT
easat-10905	21	8	including	include	VERB
easat-10905	21	9	psychological	psychological	ADJ
easat-10905	21	10	impacts	impact	NOUN
easat-10905	21	11	on	on	ADP
easat-10905	21	12	individuals	individual	NOUN
easat-10905	21	13	and	and	CCONJ
easat-10905	21	14	disrupting	disrupt	VERB
easat-10905	21	15	community	community	NOUN
easat-10905	21	16	harmony	harmony	NOUN
easat-10905	22	1	[	[	X
easat-10905	22	2	3	3	NUM
easat-10905	22	3	]	]	PUNCT
easat-10905	22	4	.	.	PUNCT
easat-10905	23	1	consequently	consequently	ADV
easat-10905	23	2	,	,	PUNCT
easat-10905	23	3	detecting	detect	VERB
easat-10905	23	4	and	and	CCONJ
easat-10905	23	5	mitigating	mitigate	VERB
easat-10905	23	6	abusive	abusive	ADJ
easat-10905	23	7	comments	comment	NOUN
easat-10905	23	8	has	have	AUX
easat-10905	23	9	become	become	VERB
easat-10905	23	10	a	a	DET
easat-10905	23	11	significant	significant	ADJ
easat-10905	23	12	area	area	NOUN
easat-10905	23	13	of	of	ADP
easat-10905	23	14	research	research	NOUN
easat-10905	23	15	within	within	ADP
easat-10905	23	16	natural	natural	ADJ
easat-10905	23	17	language	language	NOUN
easat-10905	23	18	processing	processing	NOUN
easat-10905	23	19	(	(	PUNCT
easat-10905	23	20	nlp	nlp	NOUN
easat-10905	23	21	)	)	PUNCT
easat-10905	23	22	and	and	CCONJ
easat-10905	23	23	artificial	artificial	ADJ
easat-10905	23	24	intelligence	intelligence	NOUN
easat-10905	23	25	(	(	PUNCT
easat-10905	23	26	ai	ai	NOUN
easat-10905	23	27	)	)	PUNCT
easat-10905	23	28	.	.	PUNCT
easat-10905	24	1	manually	manually	ADV
easat-10905	24	2	monitoring	monitor	VERB
easat-10905	24	3	and	and	CCONJ
easat-10905	24	4	detecting	detect	VERB
easat-10905	24	5	such	such	ADJ
easat-10905	24	6	social	social	ADJ
easat-10905	24	7	media	medium	NOUN
easat-10905	24	8	content	content	NOUN
easat-10905	24	9	is	be	AUX
easat-10905	24	10	impossible	impossible	ADJ
easat-10905	24	11	due	due	ADP
easat-10905	24	12	to	to	ADP
easat-10905	24	13	the	the	DET
easat-10905	24	14	large	large	ADJ
easat-10905	24	15	amount	amount	NOUN
easat-10905	24	16	of	of	ADP
easat-10905	24	17	data	datum	NOUN
easat-10905	24	18	generated	generate	VERB
easat-10905	24	19	[	[	PUNCT
easat-10905	24	20	4	4	NUM
easat-10905	24	21	]	]	PUNCT
easat-10905	24	22	.	.	PUNCT
easat-10905	25	1	traditional	traditional	ADJ
easat-10905	25	2	methods	method	NOUN
easat-10905	25	3	for	for	ADP
easat-10905	25	4	detecting	detect	VERB
easat-10905	25	5	abusive	abusive	ADJ
easat-10905	25	6	language	language	NOUN
easat-10905	25	7	in	in	ADP
easat-10905	25	8	text	text	NOUN
easat-10905	25	9	have	have	AUX
easat-10905	25	10	relied	rely	VERB
easat-10905	25	11	on	on	ADP
easat-10905	25	12	rule	rule	NOUN
easat-10905	25	13	-	-	PUNCT
easat-10905	25	14	based	base	VERB
easat-10905	25	15	systems	system	NOUN
easat-10905	25	16	,	,	PUNCT
easat-10905	25	17	classical	classical	ADJ
easat-10905	25	18	machine	machine	NOUN
easat-10905	25	19	learning	learning	NOUN
easat-10905	25	20	,	,	PUNCT
easat-10905	25	21	and	and	CCONJ
easat-10905	25	22	deep	deep	ADJ
easat-10905	25	23	learning	learning	NOUN
easat-10905	25	24	algorithms	algorithm	NOUN
easat-10905	25	25	[	[	X
easat-10905	25	26	4	4	NUM
easat-10905	25	27	,	,	PUNCT
easat-10905	25	28	5	5	NUM
easat-10905	25	29	]	]	PUNCT
easat-10905	25	30	.	.	PUNCT
easat-10905	26	1	common	common	ADJ
easat-10905	26	2	machine	machine	NOUN
easat-10905	26	3	learning	learn	VERB
easat-10905	26	4	techniques	technique	NOUN
easat-10905	26	5	for	for	ADP
easat-10905	26	6	detecting	detect	VERB
easat-10905	26	7	offensive	offensive	ADJ
easat-10905	26	8	language	language	NOUN
easat-10905	26	9	include	include	VERB
easat-10905	26	10	lexicon	lexicon	NOUN
easat-10905	26	11	-	-	PUNCT
easat-10905	26	12	based	base	VERB
easat-10905	26	13	methods	method	NOUN
easat-10905	26	14	[	[	X
easat-10905	26	15	6	6	NUM
easat-10905	26	16	]	]	PUNCT
easat-10905	26	17	,	,	PUNCT
easat-10905	26	18	n	n	CCONJ
easat-10905	26	19	-	-	PUNCT
easat-10905	26	20	grams	gram	NOUN
easat-10905	26	21	[	[	X
easat-10905	26	22	7	7	NUM
easat-10905	26	23	-	-	SYM
easat-10905	26	24	9	9	NUM
easat-10905	26	25	]	]	PUNCT
easat-10905	26	26	and	and	CCONJ
easat-10905	26	27	ensemble	ensemble	ADJ
easat-10905	26	28	-	-	PUNCT
easat-10905	26	29	based	base	VERB
easat-10905	26	30	learning	learning	NOUN
easat-10905	26	31	[	[	X
easat-10905	26	32	10	10	NUM
easat-10905	26	33	,	,	PUNCT
easat-10905	26	34	11	11	NUM
easat-10905	26	35	]	]	PUNCT
easat-10905	26	36	.	.	PUNCT
easat-10905	27	1	some	some	DET
easat-10905	27	2	studies	study	NOUN
easat-10905	27	3	have	have	AUX
easat-10905	27	4	applied	apply	VERB
easat-10905	27	5	and	and	CCONJ
easat-10905	27	6	compared	compare	VERB
easat-10905	27	7	deep	deep	ADJ
easat-10905	27	8	learning	learning	NOUN
easat-10905	27	9	models	model	NOUN
easat-10905	27	10	,	,	PUNCT
easat-10905	27	11	such	such	ADJ
easat-10905	27	12	as	as	ADP
easat-10905	27	13	long	long	ADJ
easat-10905	27	14	short	short	ADJ
easat-10905	27	15	-	-	PUNCT
easat-10905	27	16	term	term	NOUN
easat-10905	27	17	memory	memory	NOUN
easat-10905	27	18	(	(	PUNCT
easat-10905	27	19	lstm	lstm	NOUN
easat-10905	27	20	)	)	PUNCT
easat-10905	28	1	[	[	X
easat-10905	28	2	4	4	NUM
easat-10905	28	3	,	,	PUNCT
easat-10905	28	4	12	12	NUM
easat-10905	28	5	]	]	PUNCT
easat-10905	28	6	,	,	PUNCT
easat-10905	28	7	bidirectional	bidirectional	ADJ
easat-10905	28	8	long	long	ADJ
easat-10905	28	9	short	short	ADJ
easat-10905	28	10	-	-	PUNCT
easat-10905	28	11	term	term	NOUN
easat-10905	28	12	memory	memory	NOUN
easat-10905	28	13	(	(	PUNCT
easat-10905	28	14	bilstm	bilstm	NOUN
easat-10905	28	15	)	)	PUNCT
easat-10905	29	1	[	[	X
easat-10905	29	2	13	13	NUM
easat-10905	29	3	,	,	PUNCT
easat-10905	29	4	14	14	NUM
easat-10905	29	5	]	]	PUNCT
easat-10905	29	6	,	,	PUNCT
easat-10905	29	7	recurrent	recurrent	ADJ
easat-10905	29	8	neural	neural	ADJ
easat-10905	29	9	network	network	NOUN
easat-10905	29	10	(	(	PUNCT
easat-10905	29	11	rnn	rnn	PROPN
easat-10905	29	12	)	)	PUNCT
easat-10905	29	13	and	and	CCONJ
easat-10905	29	14	convolutional	convolutional	ADJ
easat-10905	29	15	neural	neural	ADJ
easat-10905	29	16	network	network	NOUN
easat-10905	29	17	(	(	PUNCT
easat-10905	29	18	cnn	cnn	PROPN
easat-10905	29	19	)	)	PUNCT
easat-10905	30	1	[	[	X
easat-10905	30	2	12	12	NUM
easat-10905	30	3	]	]	PUNCT
easat-10905	30	4	to	to	PART
easat-10905	30	5	detect	detect	VERB
easat-10905	30	6	abusive	abusive	ADJ
easat-10905	30	7	language	language	NOUN
easat-10905	30	8	.	.	PUNCT
easat-10905	31	1	recent	recent	ADJ
easat-10905	31	2	advancements	advancement	NOUN
easat-10905	31	3	in	in	ADP
easat-10905	31	4	deep	deep	ADJ
easat-10905	31	5	learning	learning	NOUN
easat-10905	31	6	,	,	PUNCT
easat-10905	31	7	particularly	particularly	ADV
easat-10905	31	8	the	the	DET
easat-10905	31	9	development	development	NOUN
easat-10905	31	10	of	of	ADP
easat-10905	31	11	transformer	transformer	NOUN
easat-10905	31	12	-	-	PUNCT
easat-10905	31	13	based	base	VERB
easat-10905	31	14	models	model	NOUN
easat-10905	31	15	such	such	ADJ
easat-10905	31	16	as	as	ADP
easat-10905	31	17	bert	bert	PROPN
easat-10905	31	18	,	,	PUNCT
easat-10905	31	19	gpt	gpt	NOUN
easat-10905	31	20	,	,	PUNCT
easat-10905	31	21	and	and	CCONJ
easat-10905	31	22	their	their	PRON
easat-10905	31	23	449	449	NUM
easat-10905	31	24	edelweiss	edelweiss	PROPN
easat-10905	31	25	applied	apply	VERB
easat-10905	31	26	science	science	NOUN
easat-10905	31	27	and	and	CCONJ
easat-10905	31	28	technology	technology	NOUN
easat-10905	31	29	issn	issn	PROPN
easat-10905	31	30	:	:	PUNCT
easat-10905	31	31	2576	2576	NUM
easat-10905	31	32	-	-	SYM
easat-10905	31	33	8484	8484	NUM
easat-10905	31	34	vol	vol	NOUN
easat-10905	31	35	.	.	PROPN
easat-10905	32	1	9	9	NUM
easat-10905	32	2	,	,	PUNCT
easat-10905	32	3	no	no	INTJ
easat-10905	32	4	.	.	NOUN
easat-10905	32	5	11	11	NUM
easat-10905	32	6	:	:	PUNCT
easat-10905	32	7	448	448	NUM
easat-10905	32	8	-	-	SYM
easat-10905	32	9	466	466	NUM
easat-10905	32	10	,	,	PUNCT
easat-10905	32	11	2025	2025	NUM
easat-10905	32	12	doi	doi	NOUN
easat-10905	32	13	:	:	PUNCT
easat-10905	32	14	10.55214/2576	10.55214/2576	NUM
easat-10905	32	15	-	-	PUNCT
easat-10905	32	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	33	1	©	©	PROPN
easat-10905	33	2	2025	2025	NUM
easat-10905	33	3	by	by	ADP
easat-10905	33	4	the	the	DET
easat-10905	33	5	authors	author	NOUN
easat-10905	33	6	;	;	PUNCT
easat-10905	33	7	licensee	licensee	PROPN
easat-10905	33	8	learning	learn	VERB
easat-10905	33	9	gate	gate	NOUN
easat-10905	33	10	variants	variant	NOUN
easat-10905	33	11	,	,	PUNCT
easat-10905	33	12	have	have	AUX
easat-10905	33	13	significantly	significantly	ADV
easat-10905	33	14	enhanced	enhance	VERB
easat-10905	33	15	the	the	DET
easat-10905	33	16	capabilities	capability	NOUN
easat-10905	33	17	of	of	ADP
easat-10905	33	18	nlp	nlp	NOUN
easat-10905	33	19	systems	system	NOUN
easat-10905	33	20	.	.	PUNCT
easat-10905	34	1	these	these	DET
easat-10905	34	2	models	model	NOUN
easat-10905	34	3	capture	capture	VERB
easat-10905	34	4	contextual	contextual	ADJ
easat-10905	34	5	information	information	NOUN
easat-10905	34	6	and	and	CCONJ
easat-10905	34	7	have	have	AUX
easat-10905	34	8	demonstrated	demonstrate	VERB
easat-10905	34	9	remarkable	remarkable	ADJ
easat-10905	34	10	success	success	NOUN
easat-10905	34	11	in	in	ADP
easat-10905	34	12	various	various	ADJ
easat-10905	34	13	language	language	NOUN
easat-10905	34	14	understanding	understanding	NOUN
easat-10905	34	15	tasks	task	NOUN
easat-10905	34	16	,	,	PUNCT
easat-10905	34	17	including	include	VERB
easat-10905	34	18	sentiment	sentiment	NOUN
easat-10905	34	19	analysis	analysis	NOUN
easat-10905	34	20	,	,	PUNCT
easat-10905	34	21	text	text	NOUN
easat-10905	34	22	classification	classification	NOUN
easat-10905	34	23	,	,	PUNCT
easat-10905	34	24	and	and	CCONJ
easat-10905	34	25	machine	machine	NOUN
easat-10905	34	26	translation	translation	NOUN
easat-10905	34	27	.	.	PUNCT
easat-10905	35	1	in	in	ADP
easat-10905	35	2	this	this	DET
easat-10905	35	3	research	research	NOUN
easat-10905	35	4	,	,	PUNCT
easat-10905	35	5	we	we	PRON
easat-10905	35	6	explore	explore	VERB
easat-10905	35	7	the	the	DET
easat-10905	35	8	application	application	NOUN
easat-10905	35	9	of	of	ADP
easat-10905	35	10	transformer	transformer	NOUN
easat-10905	35	11	-	-	PUNCT
easat-10905	35	12	based	base	VERB
easat-10905	35	13	models	model	NOUN
easat-10905	35	14	for	for	ADP
easat-10905	35	15	detecting	detect	VERB
easat-10905	35	16	abusive	abusive	ADJ
easat-10905	35	17	comments	comment	NOUN
easat-10905	35	18	in	in	ADP
easat-10905	35	19	social	social	ADJ
easat-10905	35	20	media	medium	NOUN
easat-10905	35	21	texts	text	NOUN
easat-10905	35	22	.	.	PUNCT
easat-10905	36	1	specifically	specifically	ADV
easat-10905	36	2	,	,	PUNCT
easat-10905	36	3	we	we	PRON
easat-10905	36	4	leverage	leverage	VERB
easat-10905	36	5	the	the	DET
easat-10905	36	6	contextual	contextual	ADJ
easat-10905	36	7	understanding	understanding	NOUN
easat-10905	36	8	and	and	CCONJ
easat-10905	36	9	language	language	NOUN
easat-10905	36	10	representation	representation	NOUN
easat-10905	36	11	capabilities	capability	NOUN
easat-10905	36	12	of	of	ADP
easat-10905	36	13	models	model	NOUN
easat-10905	36	14	such	such	ADJ
easat-10905	36	15	as	as	ADP
easat-10905	36	16	bert	bert	PROPN
easat-10905	36	17	,	,	PUNCT
easat-10905	36	18	roberta	roberta	PROPN
easat-10905	36	19	,	,	PUNCT
easat-10905	36	20	albert	albert	PROPN
easat-10905	36	21	,	,	PUNCT
easat-10905	36	22	and	and	CCONJ
easat-10905	36	23	distilbert	distilbert	VERB
easat-10905	36	24	to	to	PART
easat-10905	36	25	identify	identify	VERB
easat-10905	36	26	and	and	CCONJ
easat-10905	36	27	classify	classify	VERB
easat-10905	36	28	abusive	abusive	ADJ
easat-10905	36	29	content	content	NOUN
easat-10905	36	30	accurately	accurately	ADV
easat-10905	36	31	.	.	PUNCT
easat-10905	37	1	by	by	ADP
easat-10905	37	2	fine	fine	ADV
easat-10905	37	3	-	-	PUNCT
easat-10905	37	4	tuning	tune	VERB
easat-10905	37	5	pre	pre	ADJ
easat-10905	37	6	-	-	ADJ
easat-10905	37	7	trained	train	VERB
easat-10905	37	8	transformer	transformer	NOUN
easat-10905	37	9	models	model	NOUN
easat-10905	37	10	on	on	ADP
easat-10905	37	11	annotated	annotated	ADJ
easat-10905	37	12	datasets	dataset	NOUN
easat-10905	37	13	of	of	ADP
easat-10905	37	14	social	social	ADJ
easat-10905	37	15	media	medium	NOUN
easat-10905	37	16	comments	comment	NOUN
easat-10905	37	17	,	,	PUNCT
easat-10905	37	18	we	we	PRON
easat-10905	37	19	aim	aim	VERB
easat-10905	37	20	to	to	PART
easat-10905	37	21	achieve	achieve	VERB
easat-10905	37	22	high	high	ADJ
easat-10905	37	23	accuracy	accuracy	NOUN
easat-10905	37	24	in	in	ADP
easat-10905	37	25	abusive	abusive	ADJ
easat-10905	37	26	comment	comment	NOUN
easat-10905	37	27	detection	detection	NOUN
easat-10905	37	28	.	.	PUNCT
easat-10905	38	1	however	however	ADV
easat-10905	38	2	,	,	PUNCT
easat-10905	38	3	these	these	DET
easat-10905	38	4	models	model	NOUN
easat-10905	38	5	typically	typically	ADV
easat-10905	38	6	involve	involve	VERB
easat-10905	38	7	training	training	NOUN
easat-10905	38	8	on	on	ADP
easat-10905	38	9	a	a	DET
easat-10905	38	10	fixed	fix	VERB
easat-10905	38	11	task	task	NOUN
easat-10905	38	12	.	.	PUNCT
easat-10905	39	1	the	the	DET
easat-10905	39	2	task	task	NOUN
easat-10905	39	3	is	be	AUX
easat-10905	39	4	available	available	ADJ
easat-10905	39	5	at	at	ADP
easat-10905	39	6	once	once	ADV
easat-10905	39	7	,	,	PUNCT
easat-10905	39	8	and	and	CCONJ
easat-10905	39	9	the	the	DET
easat-10905	39	10	model	model	NOUN
easat-10905	39	11	is	be	AUX
easat-10905	39	12	trained	train	VERB
easat-10905	39	13	in	in	ADP
easat-10905	39	14	a	a	DET
easat-10905	39	15	single	single	ADJ
easat-10905	39	16	batch	batch	NOUN
easat-10905	39	17	or	or	CCONJ
easat-10905	39	18	over	over	ADP
easat-10905	39	19	several	several	ADJ
easat-10905	39	20	epochs	epoch	NOUN
easat-10905	39	21	using	use	VERB
easat-10905	39	22	this	this	DET
easat-10905	39	23	task	task	NOUN
easat-10905	39	24	.	.	PUNCT
easat-10905	40	1	after	after	ADP
easat-10905	40	2	deployment	deployment	NOUN
easat-10905	40	3	,	,	PUNCT
easat-10905	40	4	the	the	DET
easat-10905	40	5	model	model	NOUN
easat-10905	40	6	does	do	AUX
easat-10905	40	7	not	not	PART
easat-10905	40	8	continue	continue	VERB
easat-10905	40	9	to	to	PART
easat-10905	40	10	learn	learn	VERB
easat-10905	40	11	.	.	PUNCT
easat-10905	41	1	any	any	DET
easat-10905	41	2	new	new	ADJ
easat-10905	41	3	task	task	NOUN
easat-10905	41	4	would	would	AUX
easat-10905	41	5	require	require	VERB
easat-10905	41	6	retraining	retrain	VERB
easat-10905	41	7	the	the	DET
easat-10905	41	8	models	model	NOUN
easat-10905	41	9	from	from	ADP
easat-10905	41	10	scratch	scratch	NOUN
easat-10905	41	11	or	or	CCONJ
easat-10905	41	12	updating	update	VERB
easat-10905	41	13	them	they	PRON
easat-10905	41	14	periodically	periodically	ADV
easat-10905	41	15	with	with	ADP
easat-10905	41	16	new	new	ADJ
easat-10905	41	17	batches	batch	NOUN
easat-10905	41	18	of	of	ADP
easat-10905	41	19	data	datum	NOUN
easat-10905	41	20	.	.	PUNCT
easat-10905	42	1	in	in	ADP
easat-10905	42	2	contrast	contrast	NOUN
easat-10905	42	3	,	,	PUNCT
easat-10905	42	4	continual	continual	ADJ
easat-10905	42	5	learning	learning	NOUN
easat-10905	42	6	involves	involve	VERB
easat-10905	42	7	training	training	NOUN
easat-10905	42	8	models	model	NOUN
easat-10905	42	9	incrementally	incrementally	ADV
easat-10905	42	10	and	and	CCONJ
easat-10905	42	11	learning	learn	VERB
easat-10905	42	12	from	from	ADP
easat-10905	42	13	a	a	DET
easat-10905	42	14	stream	stream	NOUN
easat-10905	42	15	of	of	ADP
easat-10905	42	16	tasks	task	NOUN
easat-10905	42	17	that	that	PRON
easat-10905	42	18	arrive	arrive	VERB
easat-10905	42	19	over	over	ADP
easat-10905	42	20	time	time	NOUN
easat-10905	42	21	,	,	PUNCT
easat-10905	42	22	updating	update	VERB
easat-10905	42	23	their	their	PRON
easat-10905	42	24	knowledge	knowledge	NOUN
easat-10905	42	25	continually	continually	ADV
easat-10905	42	26	.	.	PUNCT
easat-10905	43	1	continual	continual	ADJ
easat-10905	43	2	learning	learning	NOUN
easat-10905	43	3	has	have	AUX
easat-10905	43	4	been	be	AUX
easat-10905	43	5	extensively	extensively	ADV
easat-10905	43	6	researched	research	VERB
easat-10905	43	7	in	in	ADP
easat-10905	43	8	machine	machine	NOUN
easat-10905	43	9	learning	learning	NOUN
easat-10905	43	10	,	,	PUNCT
easat-10905	43	11	computer	computer	NOUN
easat-10905	43	12	vision	vision	NOUN
easat-10905	43	13	,	,	PUNCT
easat-10905	43	14	and	and	CCONJ
easat-10905	43	15	nlp	nlp	NOUN
easat-10905	43	16	.	.	PUNCT
easat-10905	44	1	it	it	PRON
easat-10905	44	2	involves	involve	VERB
easat-10905	44	3	incremental	incremental	ADJ
easat-10905	44	4	learning	learning	NOUN
easat-10905	44	5	of	of	ADP
easat-10905	44	6	tasks	task	NOUN
easat-10905	44	7	,	,	PUNCT
easat-10905	44	8	assuming	assume	VERB
easat-10905	44	9	that	that	SCONJ
easat-10905	44	10	once	once	SCONJ
easat-10905	44	11	a	a	DET
easat-10905	44	12	task	task	NOUN
easat-10905	44	13	‘	'	PUNCT
easat-10905	44	14	𝑡	𝑡	X
easat-10905	44	15	’	'	PUNCT
easat-10905	44	16	is	be	AUX
easat-10905	44	17	learned	learn	VERB
easat-10905	44	18	,	,	PUNCT
easat-10905	44	19	its	its	PRON
easat-10905	44	20	training	training	NOUN
easat-10905	44	21	data	datum	NOUN
easat-10905	44	22	,	,	PUNCT
easat-10905	44	23	or	or	CCONJ
easat-10905	44	24	most	most	ADJ
easat-10905	44	25	of	of	ADP
easat-10905	44	26	it	it	PRON
easat-10905	44	27	,	,	PUNCT
easat-10905	44	28	is	be	AUX
easat-10905	44	29	no	no	ADV
easat-10905	44	30	longer	long	ADV
easat-10905	44	31	available	available	ADJ
easat-10905	44	32	.	.	PUNCT
easat-10905	45	1	in	in	ADP
easat-10905	45	2	nlp	nlp	NOUN
easat-10905	45	3	,	,	PUNCT
easat-10905	45	4	tasks	task	NOUN
easat-10905	45	5	can	can	AUX
easat-10905	45	6	range	range	VERB
easat-10905	45	7	from	from	ADP
easat-10905	45	8	end	end	NOUN
easat-10905	45	9	tasks	task	NOUN
easat-10905	45	10	like	like	ADP
easat-10905	45	11	text	text	NOUN
easat-10905	45	12	classification	classification	NOUN
easat-10905	45	13	and	and	CCONJ
easat-10905	45	14	summarization	summarization	NOUN
easat-10905	45	15	to	to	ADP
easat-10905	45	16	domain	domain	NOUN
easat-10905	45	17	-	-	PUNCT
easat-10905	45	18	specific	specific	ADJ
easat-10905	45	19	corpora	corpora	NOUN
easat-10905	45	20	used	use	VERB
easat-10905	45	21	for	for	ADP
easat-10905	45	22	further	further	ADJ
easat-10905	45	23	pre	pre	ADJ
easat-10905	45	24	-	-	NOUN
easat-10905	45	25	training	training	NOUN
easat-10905	45	26	of	of	ADP
easat-10905	45	27	language	language	NOUN
easat-10905	45	28	models	model	NOUN
easat-10905	45	29	.	.	PUNCT
easat-10905	46	1	the	the	DET
easat-10905	46	2	nature	nature	NOUN
easat-10905	46	3	of	of	ADP
easat-10905	46	4	tasks	task	NOUN
easat-10905	46	5	in	in	ADP
easat-10905	46	6	nlp	nlp	NOUN
easat-10905	46	7	often	often	ADV
easat-10905	46	8	differs	differ	VERB
easat-10905	46	9	from	from	ADP
easat-10905	46	10	those	those	PRON
easat-10905	46	11	in	in	ADP
easat-10905	46	12	computer	computer	NOUN
easat-10905	46	13	vision	vision	NOUN
easat-10905	46	14	due	due	ADP
easat-10905	46	15	to	to	ADP
easat-10905	46	16	their	their	PRON
easat-10905	46	17	distinct	distinct	ADJ
easat-10905	46	18	applications	application	NOUN
easat-10905	46	19	.	.	PUNCT
easat-10905	47	1	additionally	additionally	ADV
easat-10905	47	2	,	,	PUNCT
easat-10905	47	3	social	social	ADJ
easat-10905	47	4	media	medium	NOUN
easat-10905	47	5	language	language	NOUN
easat-10905	47	6	is	be	AUX
easat-10905	47	7	highly	highly	ADV
easat-10905	47	8	dynamic	dynamic	ADJ
easat-10905	47	9	,	,	PUNCT
easat-10905	47	10	with	with	ADP
easat-10905	47	11	new	new	ADJ
easat-10905	47	12	usage	usage	NOUN
easat-10905	47	13	patterns	pattern	NOUN
easat-10905	47	14	constantly	constantly	ADV
easat-10905	47	15	emerging	emerge	VERB
easat-10905	47	16	.	.	PUNCT
easat-10905	48	1	in	in	ADP
easat-10905	48	2	this	this	DET
easat-10905	48	3	work	work	NOUN
easat-10905	48	4	,	,	PUNCT
easat-10905	48	5	we	we	PRON
easat-10905	48	6	incorporate	incorporate	VERB
easat-10905	48	7	continual	continual	ADJ
easat-10905	48	8	learning	learn	VERB
easat-10905	48	9	techniques	technique	NOUN
easat-10905	48	10	into	into	ADP
easat-10905	48	11	our	our	PRON
easat-10905	48	12	model	model	NOUN
easat-10905	48	13	development	development	NOUN
easat-10905	48	14	process	process	NOUN
easat-10905	48	15	to	to	PART
easat-10905	48	16	address	address	VERB
easat-10905	48	17	this	this	DET
easat-10905	48	18	challenge	challenge	NOUN
easat-10905	48	19	.	.	PUNCT
easat-10905	49	1	continual	continual	ADJ
easat-10905	49	2	learning	learning	NOUN
easat-10905	49	3	enables	enable	VERB
easat-10905	49	4	models	model	NOUN
easat-10905	49	5	to	to	PART
easat-10905	49	6	learn	learn	VERB
easat-10905	49	7	from	from	ADP
easat-10905	49	8	new	new	ADJ
easat-10905	49	9	data	datum	NOUN
easat-10905	49	10	incrementally	incrementally	ADV
easat-10905	49	11	without	without	ADP
easat-10905	49	12	forgetting	forget	VERB
easat-10905	49	13	previously	previously	ADV
easat-10905	49	14	acquired	acquire	VERB
easat-10905	49	15	knowledge	knowledge	NOUN
easat-10905	49	16	.	.	PUNCT
easat-10905	50	1	initially	initially	ADV
easat-10905	50	2	,	,	PUNCT
easat-10905	50	3	we	we	PRON
easat-10905	50	4	train	train	VERB
easat-10905	50	5	the	the	DET
easat-10905	50	6	models	model	NOUN
easat-10905	50	7	using	use	VERB
easat-10905	50	8	abusive	abusive	ADJ
easat-10905	50	9	comments	comment	NOUN
easat-10905	50	10	and	and	CCONJ
easat-10905	50	11	then	then	ADV
easat-10905	50	12	apply	apply	VERB
easat-10905	50	13	continual	continual	ADJ
easat-10905	50	14	learning	learning	NOUN
easat-10905	50	15	to	to	PART
easat-10905	50	16	enhance	enhance	VERB
easat-10905	50	17	the	the	DET
easat-10905	50	18	ability	ability	NOUN
easat-10905	50	19	of	of	ADP
easat-10905	50	20	the	the	DET
easat-10905	50	21	models	model	NOUN
easat-10905	50	22	in	in	ADP
easat-10905	50	23	offensive	offensive	ADJ
easat-10905	50	24	text	text	NOUN
easat-10905	50	25	classification	classification	NOUN
easat-10905	50	26	.	.	PUNCT
easat-10905	51	1	we	we	PRON
easat-10905	51	2	integrate	integrate	VERB
easat-10905	51	3	several	several	ADJ
easat-10905	51	4	continual	continual	ADJ
easat-10905	51	5	learning	learning	NOUN
easat-10905	51	6	approaches	approach	NOUN
easat-10905	51	7	into	into	ADP
easat-10905	51	8	our	our	PRON
easat-10905	51	9	models	model	NOUN
easat-10905	51	10	and	and	CCONJ
easat-10905	51	11	evaluate	evaluate	VERB
easat-10905	51	12	their	their	PRON
easat-10905	51	13	performance	performance	NOUN
easat-10905	51	14	on	on	ADP
easat-10905	51	15	both	both	CCONJ
easat-10905	51	16	abusive	abusive	ADJ
easat-10905	51	17	and	and	CCONJ
easat-10905	51	18	offensive	offensive	ADJ
easat-10905	51	19	tasks	task	NOUN
easat-10905	51	20	.	.	PUNCT
easat-10905	52	1	the	the	DET
easat-10905	52	2	major	major	ADJ
easat-10905	52	3	contributions	contribution	NOUN
easat-10905	52	4	of	of	ADP
easat-10905	52	5	the	the	DET
easat-10905	52	6	proposed	propose	VERB
easat-10905	52	7	work	work	NOUN
easat-10905	52	8	include	include	VERB
easat-10905	52	9	:	:	PUNCT
easat-10905	52	10	i.	i.	PROPN
easat-10905	52	11	model	model	PROPN
easat-10905	52	12	development	development	PROPN
easat-10905	52	13	:	:	PUNCT
easat-10905	52	14	we	we	PRON
easat-10905	52	15	develop	develop	VERB
easat-10905	52	16	and	and	CCONJ
easat-10905	52	17	fine	fine	ADJ
easat-10905	52	18	-	-	PUNCT
easat-10905	52	19	tune	tune	NOUN
easat-10905	52	20	transformer	transformer	NOUN
easat-10905	52	21	-	-	PUNCT
easat-10905	52	22	based	base	VERB
easat-10905	52	23	models	model	NOUN
easat-10905	52	24	specifically	specifically	ADV
easat-10905	52	25	for	for	ADP
easat-10905	52	26	the	the	DET
easat-10905	52	27	task	task	NOUN
easat-10905	52	28	of	of	ADP
easat-10905	52	29	abusive	abusive	ADJ
easat-10905	52	30	comment	comment	NOUN
easat-10905	52	31	detection	detection	NOUN
easat-10905	52	32	,	,	PUNCT
easat-10905	52	33	demonstrating	demonstrate	VERB
easat-10905	52	34	their	their	PRON
easat-10905	52	35	effectiveness	effectiveness	NOUN
easat-10905	52	36	compared	compare	VERB
easat-10905	52	37	to	to	ADP
easat-10905	52	38	traditional	traditional	ADJ
easat-10905	52	39	methods	method	NOUN
easat-10905	52	40	.	.	PUNCT
easat-10905	53	1	ii	ii	X
easat-10905	53	2	.	.	PUNCT
easat-10905	54	1	we	we	PRON
easat-10905	54	2	incorporate	incorporate	VERB
easat-10905	54	3	continual	continual	ADJ
easat-10905	54	4	learning	learn	VERB
easat-10905	54	5	strategies	strategy	NOUN
easat-10905	54	6	to	to	PART
easat-10905	54	7	ensure	ensure	VERB
easat-10905	54	8	that	that	SCONJ
easat-10905	54	9	our	our	PRON
easat-10905	54	10	models	model	NOUN
easat-10905	54	11	remain	remain	VERB
easat-10905	54	12	effective	effective	ADJ
easat-10905	54	13	over	over	ADP
easat-10905	54	14	time	time	NOUN
easat-10905	54	15	,	,	PUNCT
easat-10905	54	16	adapting	adapt	VERB
easat-10905	54	17	to	to	ADP
easat-10905	54	18	new	new	ADJ
easat-10905	54	19	types	type	NOUN
easat-10905	54	20	of	of	ADP
easat-10905	54	21	offensive	offensive	ADJ
easat-10905	54	22	language	language	NOUN
easat-10905	54	23	as	as	SCONJ
easat-10905	54	24	they	they	PRON
easat-10905	54	25	appear	appear	VERB
easat-10905	54	26	.	.	PUNCT
easat-10905	55	1	iii	iii	X
easat-10905	55	2	.	.	PUNCT
easat-10905	55	3	evaluation	evaluation	NOUN
easat-10905	55	4	and	and	CCONJ
easat-10905	55	5	validation	validation	NOUN
easat-10905	55	6	:	:	PUNCT
easat-10905	55	7	we	we	PRON
easat-10905	55	8	rigorously	rigorously	ADV
easat-10905	55	9	evaluate	evaluate	VERB
easat-10905	55	10	the	the	DET
easat-10905	55	11	developed	develop	VERB
easat-10905	55	12	models	model	NOUN
easat-10905	55	13	using	use	VERB
easat-10905	55	14	abusive	abusive	ADJ
easat-10905	55	15	and	and	CCONJ
easat-10905	55	16	offensive	offensive	ADJ
easat-10905	55	17	tasks	task	NOUN
easat-10905	55	18	,	,	PUNCT
easat-10905	55	19	ensuring	ensure	VERB
easat-10905	55	20	their	their	PRON
easat-10905	55	21	robustness	robustness	NOUN
easat-10905	55	22	and	and	CCONJ
easat-10905	55	23	generalizability	generalizability	NOUN
easat-10905	55	24	across	across	ADP
easat-10905	55	25	different	different	ADJ
easat-10905	55	26	tasks	task	NOUN
easat-10905	55	27	.	.	PUNCT
easat-10905	56	1	the	the	DET
easat-10905	56	2	novelty	novelty	NOUN
easat-10905	56	3	of	of	ADP
easat-10905	56	4	the	the	DET
easat-10905	56	5	proposed	propose	VERB
easat-10905	56	6	work	work	NOUN
easat-10905	56	7	lies	lie	VERB
easat-10905	56	8	in	in	ADP
easat-10905	56	9	its	its	PRON
easat-10905	56	10	application	application	NOUN
easat-10905	56	11	of	of	ADP
easat-10905	56	12	continual	continual	ADJ
easat-10905	56	13	learning	learning	NOUN
easat-10905	56	14	specifically	specifically	ADV
easat-10905	56	15	to	to	ADP
easat-10905	56	16	nlp	nlp	NOUN
easat-10905	56	17	tasks	task	NOUN
easat-10905	56	18	,	,	PUNCT
easat-10905	56	19	a	a	DET
easat-10905	56	20	focus	focus	NOUN
easat-10905	56	21	that	that	PRON
easat-10905	56	22	has	have	AUX
easat-10905	56	23	not	not	PART
easat-10905	56	24	been	be	AUX
easat-10905	56	25	addressed	address	VERB
easat-10905	56	26	by	by	ADP
easat-10905	56	27	existing	exist	VERB
easat-10905	56	28	studies	study	NOUN
easat-10905	56	29	,	,	PUNCT
easat-10905	56	30	which	which	PRON
easat-10905	56	31	predominantly	predominantly	ADV
easat-10905	56	32	apply	apply	VERB
easat-10905	56	33	continual	continual	ADJ
easat-10905	56	34	learning	learning	NOUN
easat-10905	56	35	to	to	PART
easat-10905	56	36	computer	computer	VERB
easat-10905	56	37	vision	vision	NOUN
easat-10905	56	38	tasks	task	NOUN
easat-10905	56	39	[	[	X
easat-10905	56	40	15	15	NUM
easat-10905	56	41	]	]	PUNCT
easat-10905	56	42	.	.	PUNCT
easat-10905	57	1	from	from	ADP
easat-10905	57	2	our	our	PRON
easat-10905	57	3	literature	literature	NOUN
easat-10905	57	4	survey	survey	NOUN
easat-10905	57	5	,	,	PUNCT
easat-10905	57	6	we	we	PRON
easat-10905	57	7	identified	identify	VERB
easat-10905	57	8	that	that	SCONJ
easat-10905	57	9	only	only	ADV
easat-10905	57	10	a	a	DET
easat-10905	57	11	few	few	ADJ
easat-10905	57	12	studies	study	NOUN
easat-10905	57	13	review	review	VERB
easat-10905	57	14	continual	continual	ADJ
easat-10905	57	15	learning	learning	NOUN
easat-10905	57	16	approaches	approach	NOUN
easat-10905	57	17	for	for	ADP
easat-10905	57	18	nlp	nlp	ADJ
easat-10905	57	19	tasks	task	NOUN
easat-10905	57	20	such	such	ADJ
easat-10905	57	21	as	as	ADP
easat-10905	57	22	topic	topic	NOUN
easat-10905	57	23	classification	classification	NOUN
easat-10905	57	24	,	,	PUNCT
easat-10905	57	25	intent	intent	ADJ
easat-10905	57	26	classification	classification	NOUN
easat-10905	57	27	,	,	PUNCT
easat-10905	57	28	and	and	CCONJ
easat-10905	57	29	question	question	VERB
easat-10905	57	30	answering	answer	VERB
easat-10905	57	31	[	[	X
easat-10905	57	32	16	16	NUM
easat-10905	57	33	]	]	PUNCT
easat-10905	57	34	.	.	PUNCT
easat-10905	58	1	this	this	DET
easat-10905	58	2	highlights	highlight	VERB
easat-10905	58	3	the	the	DET
easat-10905	58	4	unique	unique	ADJ
easat-10905	58	5	contribution	contribution	NOUN
easat-10905	58	6	of	of	ADP
easat-10905	58	7	our	our	PRON
easat-10905	58	8	work	work	NOUN
easat-10905	58	9	in	in	ADP
easat-10905	58	10	advancing	advance	VERB
easat-10905	58	11	the	the	DET
easat-10905	58	12	use	use	NOUN
easat-10905	58	13	of	of	ADP
easat-10905	58	14	continual	continual	ADJ
easat-10905	58	15	learning	learning	NOUN
easat-10905	58	16	within	within	ADP
easat-10905	58	17	the	the	DET
easat-10905	58	18	nlp	nlp	NOUN
easat-10905	58	19	domain	domain	NOUN
easat-10905	58	20	.	.	PUNCT
easat-10905	59	1	to	to	PART
easat-10905	59	2	be	be	AUX
easat-10905	59	3	more	more	ADV
easat-10905	59	4	specific	specific	ADJ
easat-10905	59	5	,	,	PUNCT
easat-10905	59	6	we	we	PRON
easat-10905	59	7	could	could	AUX
easat-10905	59	8	not	not	PART
easat-10905	59	9	find	find	VERB
easat-10905	59	10	any	any	DET
easat-10905	59	11	work	work	NOUN
easat-10905	59	12	focusing	focus	VERB
easat-10905	59	13	on	on	ADP
easat-10905	59	14	abusive	abusive	ADJ
easat-10905	59	15	and	and	CCONJ
easat-10905	59	16	offensive	offensive	ADJ
easat-10905	59	17	texts	text	NOUN
easat-10905	59	18	.	.	PUNCT
easat-10905	60	1	understanding	understand	VERB
easat-10905	60	2	the	the	DET
easat-10905	60	3	relationship	relationship	NOUN
easat-10905	60	4	between	between	ADP
easat-10905	60	5	these	these	DET
easat-10905	60	6	texts	text	NOUN
easat-10905	60	7	helps	help	VERB
easat-10905	60	8	in	in	ADP
easat-10905	60	9	developing	develop	VERB
easat-10905	60	10	better	well	ADJ
easat-10905	60	11	detection	detection	NOUN
easat-10905	60	12	and	and	CCONJ
easat-10905	60	13	mitigation	mitigation	NOUN
easat-10905	60	14	strategies	strategy	NOUN
easat-10905	60	15	,	,	PUNCT
easat-10905	60	16	ensuring	ensure	VERB
easat-10905	60	17	that	that	SCONJ
easat-10905	60	18	social	social	ADJ
easat-10905	60	19	media	medium	NOUN
easat-10905	60	20	platforms	platform	NOUN
easat-10905	60	21	can	can	AUX
easat-10905	60	22	effectively	effectively	ADV
easat-10905	60	23	address	address	VERB
easat-10905	60	24	harmful	harmful	ADJ
easat-10905	60	25	content	content	NOUN
easat-10905	60	26	while	while	SCONJ
easat-10905	60	27	fostering	foster	VERB
easat-10905	60	28	respectful	respectful	ADJ
easat-10905	60	29	communication	communication	NOUN
easat-10905	60	30	.	.	PUNCT
easat-10905	61	1	the	the	DET
easat-10905	61	2	rest	rest	NOUN
easat-10905	61	3	of	of	ADP
easat-10905	61	4	the	the	DET
easat-10905	61	5	article	article	NOUN
easat-10905	61	6	is	be	AUX
easat-10905	61	7	structured	structure	VERB
easat-10905	61	8	as	as	SCONJ
easat-10905	61	9	follows	follow	VERB
easat-10905	61	10	:	:	PUNCT
easat-10905	61	11	section	section	NOUN
easat-10905	61	12	2	2	NUM
easat-10905	61	13	reviews	review	NOUN
easat-10905	61	14	related	relate	VERB
easat-10905	61	15	work	work	NOUN
easat-10905	61	16	on	on	ADP
easat-10905	61	17	continual	continual	ADJ
easat-10905	61	18	learning	learning	NOUN
easat-10905	61	19	.	.	PUNCT
easat-10905	62	1	section	section	NOUN
easat-10905	62	2	3	3	NUM
easat-10905	62	3	introduces	introduce	NOUN
easat-10905	62	4	the	the	DET
easat-10905	62	5	task	task	NOUN
easat-10905	62	6	sets	set	NOUN
easat-10905	62	7	used	use	VERB
easat-10905	62	8	in	in	ADP
easat-10905	62	9	our	our	PRON
easat-10905	62	10	study	study	NOUN
easat-10905	62	11	and	and	CCONJ
easat-10905	62	12	describes	describe	VERB
easat-10905	62	13	the	the	DET
easat-10905	62	14	continual	continual	ADJ
easat-10905	62	15	learning	learning	NOUN
easat-10905	62	16	methods	method	NOUN
easat-10905	62	17	implemented	implement	VERB
easat-10905	62	18	.	.	PUNCT
easat-10905	63	1	section	section	NOUN
easat-10905	63	2	4	4	NUM
easat-10905	63	3	details	detail	NOUN
easat-10905	63	4	the	the	DET
easat-10905	63	5	experimental	experimental	ADJ
easat-10905	63	6	setup	setup	NOUN
easat-10905	63	7	and	and	CCONJ
easat-10905	63	8	presents	present	VERB
easat-10905	63	9	the	the	DET
easat-10905	63	10	results	result	NOUN
easat-10905	63	11	.	.	PUNCT
easat-10905	64	1	section	section	NOUN
easat-10905	64	2	5	5	NUM
easat-10905	64	3	discusses	discuss	VERB
easat-10905	64	4	the	the	DET
easat-10905	64	5	findings	finding	NOUN
easat-10905	64	6	from	from	ADP
easat-10905	64	7	our	our	PRON
easat-10905	64	8	experiments	experiment	NOUN
easat-10905	64	9	.	.	PUNCT
easat-10905	65	1	lastly	lastly	ADV
easat-10905	65	2	,	,	PUNCT
easat-10905	65	3	section	section	NOUN
easat-10905	65	4	6	6	NUM
easat-10905	65	5	concludes	conclude	VERB
easat-10905	65	6	the	the	DET
easat-10905	65	7	article	article	NOUN
easat-10905	65	8	and	and	CCONJ
easat-10905	65	9	suggests	suggest	VERB
easat-10905	65	10	potential	potential	ADJ
easat-10905	65	11	future	future	ADJ
easat-10905	65	12	research	research	NOUN
easat-10905	65	13	directions	direction	NOUN
easat-10905	65	14	.	.	PUNCT
easat-10905	66	1	450	450	NUM
easat-10905	66	2	edelweiss	edelweiss	PROPN
easat-10905	66	3	applied	apply	VERB
easat-10905	66	4	science	science	NOUN
easat-10905	66	5	and	and	CCONJ
easat-10905	66	6	technology	technology	NOUN
easat-10905	66	7	issn	issn	PROPN
easat-10905	66	8	:	:	PUNCT
easat-10905	66	9	2576	2576	NUM
easat-10905	66	10	-	-	SYM
easat-10905	66	11	8484	8484	NUM
easat-10905	66	12	vol	vol	NOUN
easat-10905	66	13	.	.	PROPN
easat-10905	67	1	9	9	NUM
easat-10905	67	2	,	,	PUNCT
easat-10905	67	3	no	no	INTJ
easat-10905	67	4	.	.	NOUN
easat-10905	67	5	11	11	NUM
easat-10905	67	6	:	:	PUNCT
easat-10905	67	7	448	448	NUM
easat-10905	67	8	-	-	SYM
easat-10905	67	9	466	466	NUM
easat-10905	67	10	,	,	PUNCT
easat-10905	67	11	2025	2025	NUM
easat-10905	67	12	doi	doi	NOUN
easat-10905	67	13	:	:	PUNCT
easat-10905	67	14	10.55214/2576	10.55214/2576	NUM
easat-10905	67	15	-	-	PUNCT
easat-10905	67	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	68	1	©	©	PROPN
easat-10905	68	2	2025	2025	NUM
easat-10905	68	3	by	by	ADP
easat-10905	68	4	the	the	DET
easat-10905	68	5	authors	author	NOUN
easat-10905	68	6	;	;	PUNCT
easat-10905	68	7	licensee	licensee	PROPN
easat-10905	68	8	learning	learning	NOUN
easat-10905	68	9	gate	gate	NOUN
easat-10905	68	10	2	2	NUM
easat-10905	68	11	.	.	PUNCT
easat-10905	69	1	literature	literature	NOUN
easat-10905	69	2	survey	survey	VERB
easat-10905	69	3	a	a	DET
easat-10905	69	4	significant	significant	ADJ
easat-10905	69	5	challenge	challenge	NOUN
easat-10905	69	6	in	in	ADP
easat-10905	69	7	incremental	incremental	ADJ
easat-10905	69	8	learning	learning	NOUN
easat-10905	69	9	is	be	AUX
easat-10905	69	10	catastrophic	catastrophic	ADJ
easat-10905	69	11	forgetting	forgetting	NOUN
easat-10905	69	12	(	(	PUNCT
easat-10905	69	13	cf	cf	NOUN
easat-10905	69	14	)	)	PUNCT
easat-10905	69	15	,	,	PUNCT
easat-10905	69	16	where	where	SCONJ
easat-10905	69	17	a	a	DET
easat-10905	69	18	model	model	NOUN
easat-10905	69	19	loses	lose	VERB
easat-10905	69	20	previously	previously	ADV
easat-10905	69	21	acquired	acquire	VERB
easat-10905	69	22	knowledge	knowledge	NOUN
easat-10905	69	23	when	when	SCONJ
easat-10905	69	24	learning	learn	VERB
easat-10905	69	25	new	new	ADJ
easat-10905	69	26	tasks	task	NOUN
easat-10905	69	27	.	.	PUNCT
easat-10905	70	1	continual	continual	ADJ
easat-10905	70	2	learning	learning	NOUN
easat-10905	70	3	strategies	strategy	NOUN
easat-10905	70	4	aim	aim	VERB
easat-10905	70	5	to	to	PART
easat-10905	70	6	mitigate	mitigate	VERB
easat-10905	70	7	this	this	DET
easat-10905	70	8	issue	issue	NOUN
easat-10905	70	9	,	,	PUNCT
easat-10905	70	10	ensuring	ensure	VERB
easat-10905	70	11	that	that	SCONJ
easat-10905	70	12	the	the	DET
easat-10905	70	13	model	model	NOUN
easat-10905	70	14	retains	retain	VERB
easat-10905	70	15	essential	essential	ADJ
easat-10905	70	16	information	information	NOUN
easat-10905	70	17	from	from	ADP
easat-10905	70	18	previous	previous	ADJ
easat-10905	70	19	tasks	task	NOUN
easat-10905	70	20	.	.	PUNCT
easat-10905	71	1	in	in	ADP
easat-10905	71	2	this	this	DET
easat-10905	71	3	section	section	NOUN
easat-10905	71	4	,	,	PUNCT
easat-10905	71	5	we	we	PRON
easat-10905	71	6	review	review	VERB
easat-10905	71	7	existing	exist	VERB
easat-10905	71	8	research	research	NOUN
easat-10905	71	9	on	on	ADP
easat-10905	71	10	utilizing	utilize	VERB
easat-10905	71	11	continual	continual	ADJ
easat-10905	71	12	learning	learning	NOUN
easat-10905	71	13	for	for	ADP
easat-10905	71	14	task	task	NOUN
easat-10905	71	15	-	-	PUNCT
easat-10905	71	16	incremental	incremental	ADJ
easat-10905	71	17	learning	learning	NOUN
easat-10905	71	18	.	.	PUNCT
easat-10905	72	1	we	we	PRON
easat-10905	72	2	begin	begin	VERB
easat-10905	72	3	by	by	ADP
easat-10905	72	4	summarizing	summarize	VERB
easat-10905	72	5	works	work	NOUN
easat-10905	72	6	that	that	PRON
easat-10905	72	7	reviewed	review	VERB
easat-10905	72	8	existing	exist	VERB
easat-10905	72	9	research	research	NOUN
easat-10905	72	10	on	on	ADP
easat-10905	72	11	continual	continual	ADJ
easat-10905	72	12	learning	learning	NOUN
easat-10905	72	13	.	.	PUNCT
easat-10905	73	1	the	the	DET
easat-10905	73	2	authors	author	NOUN
easat-10905	73	3	of	of	ADP
easat-10905	73	4	parisi	parisi	PROPN
easat-10905	73	5	et	et	PROPN
easat-10905	73	6	al	al	PROPN
easat-10905	73	7	.	.	PUNCT
easat-10905	74	1	[	[	X
easat-10905	74	2	17	17	NUM
easat-10905	74	3	]	]	PUNCT
easat-10905	74	4	and	and	CCONJ
easat-10905	74	5	wang	wang	PROPN
easat-10905	74	6	et	et	PROPN
easat-10905	74	7	al	al	PROPN
easat-10905	74	8	.	.	PUNCT
easat-10905	75	1	[	[	X
easat-10905	75	2	18	18	NUM
easat-10905	75	3	]	]	PUNCT
easat-10905	75	4	highlighted	highlight	VERB
easat-10905	75	5	the	the	DET
easat-10905	75	6	main	main	ADJ
easat-10905	75	7	challenges	challenge	NOUN
easat-10905	75	8	associated	associate	VERB
easat-10905	75	9	with	with	ADP
easat-10905	75	10	lifelong	lifelong	ADJ
easat-10905	75	11	learning	learning	NOUN
easat-10905	75	12	in	in	ADP
easat-10905	75	13	artificial	artificial	ADJ
easat-10905	75	14	systems	system	NOUN
easat-10905	75	15	and	and	CCONJ
easat-10905	75	16	compared	compare	VERB
easat-10905	75	17	various	various	ADJ
easat-10905	75	18	neural	neural	ADJ
easat-10905	75	19	network	network	NOUN
easat-10905	75	20	approaches	approach	NOUN
easat-10905	75	21	that	that	DET
easat-10905	75	22	address	address	NOUN
easat-10905	75	23	cf	cf	NOUN
easat-10905	75	24	to	to	ADP
easat-10905	75	25	varying	vary	VERB
easat-10905	75	26	degrees	degree	NOUN
easat-10905	75	27	.	.	PUNCT
easat-10905	76	1	gepperth	gepperth	NOUN
easat-10905	76	2	and	and	CCONJ
easat-10905	76	3	hammer	hammer	NOUN
easat-10905	76	4	[	[	X
easat-10905	76	5	19	19	NUM
easat-10905	76	6	]	]	PUNCT
easat-10905	76	7	formalized	formalize	VERB
easat-10905	76	8	the	the	DET
easat-10905	76	9	concept	concept	NOUN
easat-10905	76	10	of	of	ADP
easat-10905	76	11	incremental	incremental	ADJ
easat-10905	76	12	learning	learning	NOUN
easat-10905	76	13	,	,	PUNCT
easat-10905	76	14	discussed	discuss	VERB
easat-10905	76	15	the	the	DET
easat-10905	76	16	associated	associated	ADJ
easat-10905	76	17	challenges	challenge	NOUN
easat-10905	76	18	,	,	PUNCT
easat-10905	76	19	and	and	CCONJ
easat-10905	76	20	provided	provide	VERB
easat-10905	76	21	an	an	DET
easat-10905	76	22	overview	overview	NOUN
easat-10905	76	23	of	of	ADP
easat-10905	76	24	standard	standard	ADJ
easat-10905	76	25	approaches	approach	NOUN
easat-10905	76	26	,	,	PUNCT
easat-10905	76	27	their	their	PRON
easat-10905	76	28	theoretical	theoretical	ADJ
easat-10905	76	29	foundations	foundation	NOUN
easat-10905	76	30	,	,	PUNCT
easat-10905	76	31	and	and	CCONJ
easat-10905	76	32	recent	recent	ADJ
easat-10905	76	33	applications	application	NOUN
easat-10905	76	34	.	.	PUNCT
easat-10905	77	1	de	de	X
easat-10905	77	2	lange	lange	PROPN
easat-10905	77	3	et	et	PROPN
easat-10905	77	4	al	al	PROPN
easat-10905	77	5	.	.	PUNCT
easat-10905	78	1	[	[	X
easat-10905	78	2	20	20	NUM
easat-10905	78	3	]	]	PUNCT
easat-10905	78	4	focused	focus	VERB
easat-10905	78	5	on	on	ADP
easat-10905	78	6	task	task	NOUN
easat-10905	78	7	-	-	PUNCT
easat-10905	78	8	incremental	incremental	ADJ
easat-10905	78	9	classification	classification	NOUN
easat-10905	78	10	,	,	PUNCT
easat-10905	78	11	where	where	SCONJ
easat-10905	78	12	tasks	task	NOUN
easat-10905	78	13	are	be	AUX
easat-10905	78	14	presented	present	VERB
easat-10905	78	15	in	in	ADP
easat-10905	78	16	a	a	DET
easat-10905	78	17	batch	batch	NOUN
easat-10905	78	18	-	-	PUNCT
easat-10905	78	19	like	like	ADJ
easat-10905	78	20	manner	manner	NOUN
easat-10905	78	21	.	.	PUNCT
easat-10905	79	1	their	their	PRON
easat-10905	79	2	work	work	NOUN
easat-10905	79	3	provided	provide	VERB
easat-10905	79	4	a	a	DET
easat-10905	79	5	taxonomy	taxonomy	NOUN
easat-10905	79	6	and	and	CCONJ
easat-10905	79	7	comprehensive	comprehensive	ADJ
easat-10905	79	8	overview	overview	NOUN
easat-10905	79	9	of	of	ADP
easat-10905	79	10	state	state	NOUN
easat-10905	79	11	-	-	PUNCT
easat-10905	79	12	of	of	ADP
easat-10905	79	13	-	-	PUNCT
easat-10905	79	14	the	the	DET
easat-10905	79	15	-	-	PUNCT
easat-10905	79	16	art	art	NOUN
easat-10905	79	17	approaches	approach	NOUN
easat-10905	79	18	for	for	ADP
easat-10905	79	19	incremental	incremental	ADJ
easat-10905	79	20	learning	learning	NOUN
easat-10905	79	21	,	,	PUNCT
easat-10905	79	22	a	a	DET
easat-10905	79	23	novel	novel	ADJ
easat-10905	79	24	framework	framework	NOUN
easat-10905	79	25	to	to	PART
easat-10905	79	26	continuously	continuously	ADV
easat-10905	79	27	determine	determine	VERB
easat-10905	79	28	the	the	DET
easat-10905	79	29	stability	stability	NOUN
easat-10905	79	30	-	-	PUNCT
easat-10905	79	31	plasticity	plasticity	NOUN
easat-10905	79	32	trade	trade	NOUN
easat-10905	79	33	-	-	PUNCT
easat-10905	79	34	off	off	NOUN
easat-10905	79	35	of	of	ADP
easat-10905	79	36	the	the	DET
easat-10905	79	37	continual	continual	ADJ
easat-10905	79	38	learner	learner	NOUN
easat-10905	79	39	,	,	PUNCT
easat-10905	79	40	and	and	CCONJ
easat-10905	79	41	an	an	DET
easat-10905	79	42	extensive	extensive	ADJ
easat-10905	79	43	experimental	experimental	ADJ
easat-10905	79	44	comparison	comparison	NOUN
easat-10905	79	45	of	of	ADP
easat-10905	79	46	10	10	NUM
easat-10905	79	47	state	state	NOUN
easat-10905	79	48	-	-	PUNCT
easat-10905	79	49	of	of	ADP
easat-10905	79	50	-	-	PUNCT
easat-10905	79	51	the	the	DET
easat-10905	79	52	-	-	PUNCT
easat-10905	79	53	art	art	NOUN
easat-10905	79	54	continual	continual	ADJ
easat-10905	79	55	learning	learning	NOUN
easat-10905	79	56	methods	method	NOUN
easat-10905	79	57	and	and	CCONJ
easat-10905	79	58	4	4	NUM
easat-10905	79	59	baselines	baseline	NOUN
easat-10905	79	60	.	.	PUNCT
easat-10905	80	1	the	the	DET
easat-10905	80	2	tiny	tiny	ADJ
easat-10905	80	3	imagenet	imagenet	NOUN
easat-10905	80	4	and	and	CCONJ
easat-10905	80	5	a	a	DET
easat-10905	80	6	large	large	ADJ
easat-10905	80	7	-	-	PUNCT
easat-10905	80	8	scale	scale	NOUN
easat-10905	80	9	unbalanced	unbalanced	ADJ
easat-10905	80	10	inaturalist	inaturalist	ADJ
easat-10905	80	11	datasets	dataset	NOUN
easat-10905	80	12	were	be	AUX
easat-10905	80	13	used	use	VERB
easat-10905	80	14	to	to	PART
easat-10905	80	15	evaluate	evaluate	VERB
easat-10905	80	16	the	the	DET
easat-10905	80	17	reviewed	review	VERB
easat-10905	80	18	approaches	approach	NOUN
easat-10905	80	19	.	.	PUNCT
easat-10905	81	1	ke	ke	NOUN
easat-10905	81	2	and	and	CCONJ
easat-10905	81	3	liu	liu	PROPN
easat-10905	82	1	[	[	X
easat-10905	82	2	16	16	NUM
easat-10905	82	3	]	]	PUNCT
easat-10905	82	4	conducted	conduct	VERB
easat-10905	82	5	a	a	DET
easat-10905	82	6	comprehensive	comprehensive	ADJ
easat-10905	82	7	review	review	NOUN
easat-10905	82	8	and	and	CCONJ
easat-10905	82	9	analysis	analysis	NOUN
easat-10905	82	10	of	of	ADP
easat-10905	82	11	recent	recent	ADJ
easat-10905	82	12	advancements	advancement	NOUN
easat-10905	82	13	in	in	ADP
easat-10905	82	14	continual	continual	ADJ
easat-10905	82	15	learning	learning	NOUN
easat-10905	82	16	within	within	ADP
easat-10905	82	17	the	the	DET
easat-10905	82	18	field	field	NOUN
easat-10905	82	19	of	of	ADP
easat-10905	82	20	nlp	nlp	NOUN
easat-10905	82	21	.	.	PUNCT
easat-10905	83	1	their	their	PRON
easat-10905	83	2	work	work	NOUN
easat-10905	83	3	covered	cover	VERB
easat-10905	83	4	all	all	DET
easat-10905	83	5	settings	setting	NOUN
easat-10905	83	6	of	of	ADP
easat-10905	83	7	continual	continual	ADJ
easat-10905	83	8	learning	learning	NOUN
easat-10905	83	9	,	,	PUNCT
easat-10905	83	10	presenting	present	VERB
easat-10905	83	11	a	a	DET
easat-10905	83	12	taxonomy	taxonomy	NOUN
easat-10905	83	13	of	of	ADP
easat-10905	83	14	existing	exist	VERB
easat-10905	83	15	techniques	technique	NOUN
easat-10905	83	16	,	,	PUNCT
easat-10905	83	17	methods	method	NOUN
easat-10905	83	18	to	to	PART
easat-10905	83	19	prevent	prevent	VERB
easat-10905	83	20	cf	cf	NOUN
easat-10905	83	21	,	,	PUNCT
easat-10905	83	22	and	and	CCONJ
easat-10905	83	23	strategies	strategy	NOUN
easat-10905	83	24	for	for	ADP
easat-10905	83	25	knowledge	knowledge	NOUN
easat-10905	83	26	transfer	transfer	NOUN
easat-10905	83	27	,	,	PUNCT
easat-10905	83	28	which	which	PRON
easat-10905	83	29	are	be	AUX
easat-10905	83	30	significant	significant	ADJ
easat-10905	83	31	for	for	ADP
easat-10905	83	32	nlp	nlp	NOUN
easat-10905	83	33	tasks	task	NOUN
easat-10905	83	34	.	.	PUNCT
easat-10905	84	1	similarly	similarly	ADV
easat-10905	84	2	,	,	PUNCT
easat-10905	84	3	biesialska	biesialska	VERB
easat-10905	84	4	et	et	PROPN
easat-10905	84	5	al	al	PROPN
easat-10905	84	6	.	.	PUNCT
easat-10905	85	1	[	[	X
easat-10905	85	2	21	21	NUM
easat-10905	85	3	]	]	PUNCT
easat-10905	85	4	presented	present	VERB
easat-10905	85	5	an	an	DET
easat-10905	85	6	extensive	extensive	ADJ
easat-10905	85	7	overview	overview	NOUN
easat-10905	85	8	of	of	ADP
easat-10905	85	9	existing	exist	VERB
easat-10905	85	10	research	research	NOUN
easat-10905	85	11	on	on	ADP
easat-10905	85	12	continual	continual	ADJ
easat-10905	85	13	learning	learning	NOUN
easat-10905	85	14	in	in	ADP
easat-10905	85	15	nlp	nlp	NOUN
easat-10905	85	16	,	,	PUNCT
easat-10905	85	17	categorizing	categorize	VERB
easat-10905	85	18	various	various	ADJ
easat-10905	85	19	machine	machine	NOUN
easat-10905	85	20	learning	learning	NOUN
easat-10905	85	21	paradigms	paradigm	NOUN
easat-10905	85	22	and	and	CCONJ
easat-10905	85	23	methods	method	NOUN
easat-10905	85	24	designed	design	VERB
easat-10905	85	25	to	to	PART
easat-10905	85	26	mitigate	mitigate	VERB
easat-10905	85	27	cf	cf	NOUN
easat-10905	85	28	.	.	PUNCT
easat-10905	86	1	they	they	PRON
easat-10905	86	2	also	also	ADV
easat-10905	86	3	discussed	discuss	VERB
easat-10905	86	4	the	the	DET
easat-10905	86	5	application	application	NOUN
easat-10905	86	6	of	of	ADP
easat-10905	86	7	these	these	DET
easat-10905	86	8	methods	method	NOUN
easat-10905	86	9	to	to	ADP
easat-10905	86	10	different	different	ADJ
easat-10905	86	11	nlp	nlp	NOUN
easat-10905	86	12	tasks	task	NOUN
easat-10905	86	13	and	and	CCONJ
easat-10905	86	14	summarized	summarize	VERB
easat-10905	86	15	the	the	DET
easat-10905	86	16	available	available	ADJ
easat-10905	86	17	benchmark	benchmark	NOUN
easat-10905	86	18	datasets	dataset	NOUN
easat-10905	86	19	and	and	CCONJ
easat-10905	86	20	evaluation	evaluation	NOUN
easat-10905	86	21	approaches	approach	NOUN
easat-10905	86	22	.	.	PUNCT
easat-10905	87	1	yang	yang	PROPN
easat-10905	87	2	et	et	PROPN
easat-10905	87	3	al	al	PROPN
easat-10905	87	4	.	.	PUNCT
easat-10905	88	1	[	[	X
easat-10905	88	2	22	22	NUM
easat-10905	88	3	]	]	PUNCT
easat-10905	88	4	introduced	introduce	VERB
easat-10905	88	5	an	an	DET
easat-10905	88	6	innovative	innovative	ADJ
easat-10905	88	7	nlp	nlp	NOUN
easat-10905	88	8	transformer	transformer	NOUN
easat-10905	88	9	model	model	NOUN
easat-10905	88	10	designed	design	VERB
easat-10905	88	11	to	to	PART
easat-10905	88	12	reduce	reduce	VERB
easat-10905	88	13	cf	cf	NOUN
easat-10905	88	14	in	in	ADP
easat-10905	88	15	online	online	ADJ
easat-10905	88	16	continual	continual	ADJ
easat-10905	88	17	learning	learning	NOUN
easat-10905	88	18	through	through	ADP
easat-10905	88	19	attention	attention	NOUN
easat-10905	88	20	calibration	calibration	NOUN
easat-10905	88	21	.	.	PUNCT
easat-10905	89	1	this	this	DET
easat-10905	89	2	research	research	NOUN
easat-10905	89	3	proposed	propose	VERB
easat-10905	89	4	a	a	DET
easat-10905	89	5	general	general	ADJ
easat-10905	89	6	and	and	CCONJ
easat-10905	89	7	lightweight	lightweight	ADJ
easat-10905	89	8	feature	feature	NOUN
easat-10905	89	9	calibration	calibration	NOUN
easat-10905	89	10	method	method	NOUN
easat-10905	89	11	to	to	PART
easat-10905	89	12	address	address	VERB
easat-10905	89	13	task	task	NOUN
easat-10905	89	14	-	-	PUNCT
easat-10905	89	15	incremental	incremental	ADJ
easat-10905	89	16	continual	continual	ADJ
easat-10905	89	17	learning	learning	NOUN
easat-10905	89	18	issues	issue	NOUN
easat-10905	89	19	and	and	CCONJ
easat-10905	89	20	included	include	VERB
easat-10905	89	21	extensive	extensive	ADJ
easat-10905	89	22	empirical	empirical	ADJ
easat-10905	89	23	experiments	experiment	NOUN
easat-10905	89	24	on	on	ADP
easat-10905	89	25	seq2seq	seq2seq	NOUN
easat-10905	89	26	language	language	NOUN
easat-10905	89	27	generation	generation	NOUN
easat-10905	89	28	applications	application	NOUN
easat-10905	89	29	.	.	PUNCT
easat-10905	90	1	varshney	varshney	NOUN
easat-10905	90	2	et	et	PROPN
easat-10905	90	3	al	al	PROPN
easat-10905	90	4	.	.	PUNCT
easat-10905	91	1	[	[	X
easat-10905	91	2	23	23	NUM
easat-10905	91	3	]	]	PUNCT
easat-10905	91	4	proposed	propose	VERB
easat-10905	91	5	an	an	DET
easat-10905	91	6	incremental	incremental	ADJ
easat-10905	91	7	learning	learning	NOUN
easat-10905	91	8	method	method	NOUN
easat-10905	91	9	that	that	PRON
easat-10905	91	10	combines	combine	VERB
easat-10905	91	11	prompt	prompt	NOUN
easat-10905	91	12	-	-	PUNCT
easat-10905	91	13	based	base	VERB
easat-10905	91	14	classification	classification	NOUN
easat-10905	91	15	with	with	ADP
easat-10905	91	16	generative	generative	ADJ
easat-10905	91	17	replay	replay	NOUN
easat-10905	91	18	using	use	VERB
easat-10905	91	19	a	a	DET
easat-10905	91	20	pre	pre	ADJ
easat-10905	91	21	-	-	ADJ
easat-10905	91	22	trained	train	VERB
easat-10905	91	23	language	language	NOUN
easat-10905	91	24	model	model	NOUN
easat-10905	91	25	.	.	PUNCT
easat-10905	92	1	this	this	DET
easat-10905	92	2	approach	approach	NOUN
easat-10905	92	3	demonstrated	demonstrate	VERB
easat-10905	92	4	superior	superior	ADJ
easat-10905	92	5	performance	performance	NOUN
easat-10905	92	6	compared	compare	VERB
easat-10905	92	7	to	to	ADP
easat-10905	92	8	traditional	traditional	ADJ
easat-10905	92	9	replay	replay	NOUN
easat-10905	92	10	-	-	PUNCT
easat-10905	92	11	based	base	VERB
easat-10905	92	12	methods	method	NOUN
easat-10905	92	13	for	for	ADP
easat-10905	92	14	lifelong	lifelong	ADJ
easat-10905	92	15	intent	intent	NOUN
easat-10905	92	16	detection	detection	NOUN
easat-10905	92	17	across	across	ADP
easat-10905	92	18	three	three	NUM
easat-10905	92	19	public	public	ADJ
easat-10905	92	20	intent	intent	NOUN
easat-10905	92	21	classification	classification	NOUN
easat-10905	92	22	datasets	dataset	NOUN
easat-10905	92	23	,	,	PUNCT
easat-10905	92	24	a	a	DET
easat-10905	92	25	text	text	NOUN
easat-10905	92	26	classification	classification	NOUN
easat-10905	92	27	dataset	dataset	NOUN
easat-10905	92	28	,	,	PUNCT
easat-10905	92	29	and	and	CCONJ
easat-10905	92	30	two	two	NUM
easat-10905	92	31	public	public	ADJ
easat-10905	92	32	multi	multi	ADJ
easat-10905	92	33	-	-	ADJ
easat-10905	92	34	domain	domain	ADJ
easat-10905	92	35	dialog	dialog	NOUN
easat-10905	92	36	intent	intent	NOUN
easat-10905	92	37	detection	detection	NOUN
easat-10905	92	38	datasets	dataset	NOUN
easat-10905	92	39	.	.	PUNCT
easat-10905	93	1	wang	wang	PROPN
easat-10905	93	2	et	et	PROPN
easat-10905	93	3	al	al	PROPN
easat-10905	93	4	.	.	PUNCT
easat-10905	94	1	[	[	X
easat-10905	94	2	24	24	NUM
easat-10905	94	3	]	]	PUNCT
easat-10905	94	4	presented	present	VERB
easat-10905	94	5	the	the	DET
easat-10905	94	6	meta	meta	ADJ
easat-10905	94	7	lifelong	lifelong	ADJ
easat-10905	94	8	learning	learning	NOUN
easat-10905	94	9	(	(	PUNCT
easat-10905	94	10	mell	mell	PROPN
easat-10905	94	11	)	)	PUNCT
easat-10905	94	12	framework	framework	NOUN
easat-10905	94	13	to	to	PART
easat-10905	94	14	tackle	tackle	VERB
easat-10905	94	15	intent	intent	VERB
easat-10905	94	16	classification	classification	NOUN
easat-10905	94	17	tasks	task	NOUN
easat-10905	94	18	.	.	PUNCT
easat-10905	95	1	in	in	ADP
easat-10905	95	2	mell	mell	PROPN
easat-10905	95	3	,	,	PUNCT
easat-10905	95	4	a	a	DET
easat-10905	95	5	bert	bert	NOUN
easat-10905	95	6	-	-	PUNCT
easat-10905	95	7	based	base	VERB
easat-10905	95	8	text	text	NOUN
easat-10905	95	9	encoder	encoder	NOUN
easat-10905	95	10	is	be	AUX
easat-10905	95	11	used	use	VERB
easat-10905	95	12	to	to	PART
easat-10905	95	13	learn	learn	VERB
easat-10905	95	14	text	text	NOUN
easat-10905	95	15	representations	representation	NOUN
easat-10905	95	16	across	across	ADP
easat-10905	95	17	tasks	task	NOUN
easat-10905	95	18	,	,	PUNCT
easat-10905	95	19	which	which	PRON
easat-10905	95	20	is	be	AUX
easat-10905	95	21	incrementally	incrementally	ADV
easat-10905	95	22	updated	update	VERB
easat-10905	95	23	for	for	ADP
easat-10905	95	24	lifelong	lifelong	ADJ
easat-10905	95	25	learning	learning	NOUN
easat-10905	95	26	.	.	PUNCT
easat-10905	96	1	this	this	DET
easat-10905	96	2	framework	framework	NOUN
easat-10905	96	3	employs	employ	VERB
easat-10905	96	4	global	global	ADJ
easat-10905	96	5	and	and	CCONJ
easat-10905	96	6	local	local	ADJ
easat-10905	96	7	memory	memory	NOUN
easat-10905	96	8	networks	network	NOUN
easat-10905	96	9	to	to	PART
easat-10905	96	10	capture	capture	VERB
easat-10905	96	11	cross	cross	ADJ
easat-10905	96	12	-	-	ADJ
easat-10905	96	13	task	task	ADJ
easat-10905	96	14	prototype	prototype	NOUN
easat-10905	96	15	representations	representation	NOUN
easat-10905	96	16	of	of	ADP
easat-10905	96	17	different	different	ADJ
easat-10905	96	18	classes	class	NOUN
easat-10905	96	19	,	,	PUNCT
easat-10905	96	20	functioning	function	VERB
easat-10905	96	21	as	as	ADP
easat-10905	96	22	a	a	DET
easat-10905	96	23	meta	meta	ADJ
easat-10905	96	24	-	-	PUNCT
easat-10905	96	25	learner	learner	NOUN
easat-10905	96	26	that	that	PRON
easat-10905	96	27	quickly	quickly	ADV
easat-10905	96	28	adapts	adapt	VERB
easat-10905	96	29	to	to	ADP
easat-10905	96	30	various	various	ADJ
easat-10905	96	31	tasks	task	NOUN
easat-10905	96	32	.	.	PUNCT
easat-10905	97	1	the	the	DET
easat-10905	97	2	results	result	NOUN
easat-10905	97	3	demonstrate	demonstrate	VERB
easat-10905	97	4	that	that	SCONJ
easat-10905	97	5	mell	mell	PROPN
easat-10905	97	6	enhances	enhance	VERB
easat-10905	97	7	performance	performance	NOUN
easat-10905	97	8	compared	compare	VERB
easat-10905	97	9	to	to	ADP
easat-10905	97	10	strong	strong	ADJ
easat-10905	97	11	baselines	baseline	NOUN
easat-10905	97	12	while	while	SCONJ
easat-10905	97	13	also	also	ADV
easat-10905	97	14	reducing	reduce	VERB
easat-10905	97	15	total	total	ADJ
easat-10905	97	16	parameters	parameter	NOUN
easat-10905	97	17	.	.	PUNCT
easat-10905	98	1	qian	qian	PROPN
easat-10905	98	2	et	et	PROPN
easat-10905	98	3	al	al	PROPN
easat-10905	98	4	.	.	PUNCT
easat-10905	99	1	[	[	X
easat-10905	99	2	25	25	NUM
easat-10905	99	3	]	]	PUNCT
easat-10905	99	4	proposed	propose	VERB
easat-10905	99	5	a	a	DET
easat-10905	99	6	lifelong	lifelong	ADJ
easat-10905	99	7	learning	learning	NOUN
easat-10905	99	8	approach	approach	NOUN
easat-10905	99	9	for	for	ADP
easat-10905	99	10	hate	hate	NOUN
easat-10905	99	11	speech	speech	NOUN
easat-10905	99	12	classification	classification	NOUN
easat-10905	99	13	on	on	ADP
easat-10905	99	14	social	social	ADJ
easat-10905	99	15	media	medium	NOUN
easat-10905	99	16	.	.	PUNCT
easat-10905	100	1	to	to	PART
easat-10905	100	2	address	address	VERB
easat-10905	100	3	cf	cf	NOUN
easat-10905	100	4	,	,	PUNCT
easat-10905	100	5	they	they	PRON
easat-10905	100	6	utilized	utilize	VERB
easat-10905	100	7	variational	variational	ADJ
easat-10905	100	8	representation	representation	NOUN
easat-10905	100	9	learning	learn	VERB
easat-10905	100	10	along	along	ADP
easat-10905	100	11	with	with	ADP
easat-10905	100	12	a	a	DET
easat-10905	100	13	memory	memory	NOUN
easat-10905	100	14	module	module	NOUN
easat-10905	100	15	based	base	VERB
easat-10905	100	16	on	on	ADP
easat-10905	100	17	load	load	NOUN
easat-10905	100	18	-	-	PUNCT
easat-10905	100	19	balancing	balance	VERB
easat-10905	100	20	self	self	NOUN
easat-10905	100	21	-	-	PUNCT
easat-10905	100	22	organizing	organize	VERB
easat-10905	100	23	incremental	incremental	ADJ
easat-10905	100	24	neural	neural	ADJ
easat-10905	100	25	network	network	NOUN
easat-10905	100	26	(	(	PUNCT
easat-10905	100	27	lb	lb	ADP
easat-10905	100	28	soinn	soinn	NOUN
easat-10905	100	29	)	)	PUNCT
easat-10905	100	30	.	.	PUNCT
easat-10905	101	1	their	their	PRON
easat-10905	101	2	research	research	NOUN
easat-10905	101	3	demonstrated	demonstrate	VERB
easat-10905	101	4	that	that	SCONJ
easat-10905	101	5	the	the	DET
easat-10905	101	6	combination	combination	NOUN
easat-10905	101	7	of	of	ADP
easat-10905	101	8	variational	variational	ADJ
easat-10905	101	9	representation	representation	NOUN
easat-10905	101	10	learning	learning	NOUN
easat-10905	101	11	and	and	CCONJ
easat-10905	101	12	the	the	DET
easat-10905	101	13	lb	lb	NUM
easat-10905	101	14	soinn	soinn	NOUN
easat-10905	101	15	memory	memory	NOUN
easat-10905	101	16	module	module	NOUN
easat-10905	101	17	outperforms	outperform	NOUN
easat-10905	101	18	commonly	commonly	ADV
easat-10905	101	19	used	use	VERB
easat-10905	101	20	lifelong	lifelong	ADJ
easat-10905	101	21	learning	learning	NOUN
easat-10905	101	22	techniques	technique	NOUN
easat-10905	101	23	.	.	PUNCT
easat-10905	102	1	a	a	DET
easat-10905	102	2	lifelong	lifelong	ADJ
easat-10905	102	3	learning	learning	NOUN
easat-10905	102	4	method	method	NOUN
easat-10905	102	5	based	base	VERB
easat-10905	102	6	on	on	ADP
easat-10905	102	7	a	a	DET
easat-10905	102	8	bayesian	bayesian	NOUN
easat-10905	102	9	optimization	optimization	NOUN
easat-10905	102	10	framework	framework	NOUN
easat-10905	102	11	utilizing	utilize	VERB
easat-10905	102	12	stochastic	stochastic	ADJ
easat-10905	102	13	gradient	gradient	ADJ
easat-10905	102	14	descent	descent	NOUN
easat-10905	102	15	was	be	AUX
easat-10905	102	16	introduced	introduce	VERB
easat-10905	102	17	by	by	ADP
easat-10905	102	18	chen	chen	PROPN
easat-10905	102	19	et	et	PROPN
easat-10905	102	20	al	al	PROPN
easat-10905	102	21	.	.	PUNCT
easat-10905	103	1	[	[	X
easat-10905	103	2	26	26	NUM
easat-10905	103	3	]	]	PUNCT
easat-10905	103	4	.	.	PUNCT
easat-10905	104	1	their	their	PRON
easat-10905	104	2	experimental	experimental	ADJ
easat-10905	104	3	results	result	NOUN
easat-10905	104	4	showed	show	VERB
easat-10905	104	5	that	that	SCONJ
easat-10905	104	6	this	this	DET
easat-10905	104	7	method	method	NOUN
easat-10905	104	8	significantly	significantly	ADV
easat-10905	104	9	outperforms	outperform	VERB
easat-10905	104	10	baseline	baseline	ADJ
easat-10905	104	11	methods	method	NOUN
easat-10905	104	12	,	,	PUNCT
easat-10905	104	13	highlighting	highlight	VERB
easat-10905	104	14	lifelong	lifelong	ADJ
easat-10905	104	15	learning	learning	NOUN
easat-10905	104	16	as	as	ADP
easat-10905	104	17	a	a	DET
easat-10905	104	18	promising	promising	ADJ
easat-10905	104	19	research	research	NOUN
easat-10905	104	20	direction	direction	NOUN
easat-10905	104	21	.	.	PUNCT
easat-10905	105	1	the	the	DET
easat-10905	105	2	authors	author	NOUN
easat-10905	105	3	of	of	ADP
easat-10905	105	4	de	de	PROPN
easat-10905	105	5	 	 	SPACE
easat-10905	105	6	masson	masson	PROPN
easat-10905	105	7	 	 	SPACE
easat-10905	105	8	d’autume	d’autume	PROPN
easat-10905	105	9	et	et	PROPN
easat-10905	105	10	al	al	PROPN
easat-10905	105	11	.	.	PUNCT
easat-10905	106	1	[	[	X
easat-10905	106	2	27	27	NUM
easat-10905	106	3	]	]	PUNCT
easat-10905	106	4	introduced	introduce	VERB
easat-10905	106	5	a	a	DET
easat-10905	106	6	lifelong	lifelong	ADJ
easat-10905	106	7	language	language	NOUN
easat-10905	106	8	learning	learn	VERB
easat-10905	106	9	framework	framework	NOUN
easat-10905	106	10	where	where	SCONJ
easat-10905	106	11	the	the	DET
easat-10905	106	12	model	model	NOUN
easat-10905	106	13	continuously	continuously	ADV
easat-10905	106	14	learns	learn	VERB
easat-10905	106	15	from	from	ADP
easat-10905	106	16	a	a	DET
easat-10905	106	17	stream	stream	NOUN
easat-10905	106	18	of	of	ADP
easat-10905	106	19	text	text	NOUN
easat-10905	106	20	examples	example	NOUN
easat-10905	106	21	without	without	ADP
easat-10905	106	22	relying	rely	VERB
easat-10905	106	23	on	on	ADP
easat-10905	106	24	dataset	dataset	ADJ
easat-10905	106	25	identifiers	identifier	NOUN
easat-10905	106	26	.	.	PUNCT
easat-10905	107	1	they	they	PRON
easat-10905	107	2	proposed	propose	VERB
easat-10905	107	3	an	an	DET
easat-10905	107	4	episodic	episodic	ADJ
easat-10905	107	5	memory	memory	NOUN
easat-10905	107	6	model	model	NOUN
easat-10905	107	7	that	that	PRON
easat-10905	107	8	uses	use	VERB
easat-10905	107	9	sparse	sparse	ADJ
easat-10905	107	10	experience	experience	NOUN
easat-10905	107	11	replay	replay	NOUN
easat-10905	107	12	and	and	CCONJ
easat-10905	107	13	local	local	ADJ
easat-10905	107	14	adaptation	adaptation	NOUN
easat-10905	107	15	to	to	PART
easat-10905	107	16	reduce	reduce	VERB
easat-10905	107	17	catastrophic	catastrophic	ADJ
easat-10905	107	18	forgetting	forgetting	NOUN
easat-10905	107	19	(	(	PUNCT
easat-10905	107	20	cf	cf	NOUN
easat-10905	107	21	)	)	PUNCT
easat-10905	107	22	.	.	PUNCT
easat-10905	108	1	experiments	experiment	NOUN
easat-10905	108	2	conducted	conduct	VERB
easat-10905	108	3	on	on	ADP
easat-10905	108	4	text	text	NOUN
easat-10905	108	5	451	451	NUM
easat-10905	108	6	edelweiss	edelweiss	PROPN
easat-10905	108	7	applied	apply	VERB
easat-10905	108	8	science	science	NOUN
easat-10905	108	9	and	and	CCONJ
easat-10905	108	10	technology	technology	NOUN
easat-10905	108	11	issn	issn	PROPN
easat-10905	108	12	:	:	PUNCT
easat-10905	108	13	2576	2576	NUM
easat-10905	108	14	-	-	SYM
easat-10905	108	15	8484	8484	NUM
easat-10905	108	16	vol	vol	NOUN
easat-10905	108	17	.	.	PROPN
easat-10905	109	1	9	9	NUM
easat-10905	109	2	,	,	PUNCT
easat-10905	109	3	no	no	INTJ
easat-10905	109	4	.	.	NOUN
easat-10905	109	5	11	11	NUM
easat-10905	109	6	:	:	PUNCT
easat-10905	109	7	448	448	NUM
easat-10905	109	8	-	-	SYM
easat-10905	109	9	466	466	NUM
easat-10905	109	10	,	,	PUNCT
easat-10905	109	11	2025	2025	NUM
easat-10905	109	12	doi	doi	NOUN
easat-10905	109	13	:	:	PUNCT
easat-10905	109	14	10.55214/2576	10.55214/2576	NUM
easat-10905	109	15	-	-	PUNCT
easat-10905	109	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	110	1	©	©	PROPN
easat-10905	110	2	2025	2025	NUM
easat-10905	110	3	by	by	ADP
easat-10905	110	4	the	the	DET
easat-10905	110	5	authors	author	NOUN
easat-10905	110	6	;	;	PUNCT
easat-10905	110	7	licensee	licensee	PROPN
easat-10905	110	8	learning	learn	VERB
easat-10905	110	9	gate	gate	NOUN
easat-10905	110	10	classification	classification	NOUN
easat-10905	110	11	and	and	CCONJ
easat-10905	110	12	question	question	NOUN
easat-10905	110	13	answering	answering	NOUN
easat-10905	110	14	tasks	task	NOUN
easat-10905	110	15	demonstrated	demonstrate	VERB
easat-10905	110	16	that	that	SCONJ
easat-10905	110	17	sparse	sparse	ADJ
easat-10905	110	18	experience	experience	NOUN
easat-10905	110	19	replay	replay	NOUN
easat-10905	110	20	and	and	CCONJ
easat-10905	110	21	local	local	ADJ
easat-10905	110	22	adaptation	adaptation	NOUN
easat-10905	110	23	helped	help	VERB
easat-10905	110	24	the	the	DET
easat-10905	110	25	model	model	NOUN
easat-10905	110	26	effectively	effectively	ADV
easat-10905	110	27	learn	learn	VERB
easat-10905	110	28	from	from	ADP
easat-10905	110	29	new	new	ADJ
easat-10905	110	30	datasets	dataset	NOUN
easat-10905	110	31	over	over	ADP
easat-10905	110	32	time	time	NOUN
easat-10905	110	33	.	.	PUNCT
easat-10905	111	1	additionally	additionally	ADV
easat-10905	111	2	,	,	PUNCT
easat-10905	111	3	several	several	ADJ
easat-10905	111	4	studies	study	NOUN
easat-10905	111	5	have	have	AUX
easat-10905	111	6	focused	focus	VERB
easat-10905	111	7	on	on	ADP
easat-10905	111	8	continual	continual	ADJ
easat-10905	111	9	learning	learning	NOUN
easat-10905	111	10	in	in	ADP
easat-10905	111	11	computer	computer	NOUN
easat-10905	111	12	vision	vision	NOUN
easat-10905	111	13	.	.	PUNCT
easat-10905	112	1	li	li	PROPN
easat-10905	112	2	et	et	PROPN
easat-10905	112	3	al	al	PROPN
easat-10905	112	4	.	.	PUNCT
easat-10905	113	1	[	[	X
easat-10905	113	2	28	28	NUM
easat-10905	113	3	]	]	PUNCT
easat-10905	113	4	tackled	tackle	VERB
easat-10905	113	5	the	the	DET
easat-10905	113	6	challenge	challenge	NOUN
easat-10905	113	7	of	of	ADP
easat-10905	113	8	adapting	adapt	VERB
easat-10905	113	9	a	a	DET
easat-10905	113	10	vision	vision	NOUN
easat-10905	113	11	system	system	NOUN
easat-10905	113	12	to	to	ADP
easat-10905	113	13	new	new	ADJ
easat-10905	113	14	tasks	task	NOUN
easat-10905	113	15	while	while	SCONJ
easat-10905	113	16	maintaining	maintain	VERB
easat-10905	113	17	performance	performance	NOUN
easat-10905	113	18	on	on	ADP
easat-10905	113	19	original	original	ADJ
easat-10905	113	20	tasks	task	NOUN
easat-10905	113	21	without	without	ADP
easat-10905	113	22	access	access	NOUN
easat-10905	113	23	to	to	ADP
easat-10905	113	24	the	the	DET
easat-10905	113	25	original	original	ADJ
easat-10905	113	26	training	training	NOUN
easat-10905	113	27	data	datum	NOUN
easat-10905	113	28	.	.	PUNCT
easat-10905	114	1	they	they	PRON
easat-10905	114	2	proposed	propose	VERB
easat-10905	114	3	the	the	DET
easat-10905	114	4	learning	learning	NOUN
easat-10905	114	5	without	without	ADP
easat-10905	114	6	forgetting	forget	VERB
easat-10905	114	7	(	(	PUNCT
easat-10905	114	8	lwf	lwf	PROPN
easat-10905	114	9	)	)	PUNCT
easat-10905	114	10	method	method	NOUN
easat-10905	114	11	for	for	ADP
easat-10905	114	12	convolutional	convolutional	ADJ
easat-10905	114	13	neural	neural	ADJ
easat-10905	114	14	networks	network	NOUN
easat-10905	114	15	,	,	PUNCT
easat-10905	114	16	which	which	PRON
easat-10905	114	17	combines	combine	VERB
easat-10905	114	18	knowledge	knowledge	NOUN
easat-10905	114	19	distillation	distillation	NOUN
easat-10905	114	20	and	and	CCONJ
easat-10905	114	21	fine	fine	ADV
easat-10905	114	22	-	-	PUNCT
easat-10905	114	23	tuning	tuning	NOUN
easat-10905	114	24	to	to	PART
easat-10905	114	25	learn	learn	VERB
easat-10905	114	26	parameters	parameter	NOUN
easat-10905	114	27	that	that	PRON
easat-10905	114	28	are	be	AUX
easat-10905	114	29	discriminative	discriminative	NOUN
easat-10905	114	30	for	for	ADP
easat-10905	114	31	the	the	DET
easat-10905	114	32	new	new	ADJ
easat-10905	114	33	task	task	NOUN
easat-10905	114	34	while	while	SCONJ
easat-10905	114	35	preserving	preserve	VERB
easat-10905	114	36	the	the	DET
easat-10905	114	37	knowledge	knowledge	NOUN
easat-10905	114	38	of	of	ADP
easat-10905	114	39	the	the	DET
easat-10905	114	40	original	original	ADJ
easat-10905	114	41	tasks	task	NOUN
easat-10905	114	42	.	.	PUNCT
easat-10905	115	1	the	the	DET
easat-10905	115	2	effectiveness	effectiveness	NOUN
easat-10905	115	3	of	of	ADP
easat-10905	115	4	their	their	PRON
easat-10905	115	5	method	method	NOUN
easat-10905	115	6	was	be	AUX
easat-10905	115	7	demonstrated	demonstrate	VERB
easat-10905	115	8	on	on	ADP
easat-10905	115	9	several	several	ADJ
easat-10905	115	10	classification	classification	NOUN
easat-10905	115	11	tasks	task	NOUN
easat-10905	115	12	.	.	PUNCT
easat-10905	116	1	a	a	DET
easat-10905	116	2	study	study	NOUN
easat-10905	116	3	by	by	ADP
easat-10905	116	4	huang	huang	PROPN
easat-10905	116	5	and	and	CCONJ
easat-10905	116	6	asemi	asemi	PROPN
easat-10905	116	7	[	[	X
easat-10905	116	8	29	29	NUM
easat-10905	116	9	]	]	PUNCT
easat-10905	116	10	introduced	introduce	VERB
easat-10905	116	11	dllagpt	dllagpt	NOUN
easat-10905	116	12	(	(	PUNCT
easat-10905	116	13	dynamic	dynamic	ADJ
easat-10905	116	14	lifelong	lifelong	ADJ
easat-10905	116	15	learning	learning	NOUN
easat-10905	116	16	aspect	aspect	NOUN
easat-10905	116	17	-	-	PUNCT
easat-10905	116	18	based	base	VERB
easat-10905	116	19	sentiment	sentiment	NOUN
easat-10905	116	20	analysis	analysis	NOUN
easat-10905	116	21	gpt	gpt	NOUN
easat-10905	116	22	)	)	PUNCT
easat-10905	116	23	,	,	PUNCT
easat-10905	116	24	a	a	DET
easat-10905	116	25	transformer	transformer	NOUN
easat-10905	116	26	-	-	PUNCT
easat-10905	116	27	based	base	VERB
easat-10905	116	28	model	model	NOUN
easat-10905	116	29	capable	capable	ADJ
easat-10905	116	30	of	of	ADP
easat-10905	116	31	learning	learn	VERB
easat-10905	116	32	across	across	ADP
easat-10905	116	33	multiple	multiple	ADJ
easat-10905	116	34	domains	domain	NOUN
easat-10905	116	35	while	while	SCONJ
easat-10905	116	36	reducing	reduce	VERB
easat-10905	116	37	catastrophic	catastrophic	ADJ
easat-10905	116	38	forgetting	forgetting	NOUN
easat-10905	116	39	.	.	PUNCT
easat-10905	117	1	the	the	DET
easat-10905	117	2	model	model	NOUN
easat-10905	117	3	is	be	AUX
easat-10905	117	4	sequentially	sequentially	ADV
easat-10905	117	5	fine	fine	ADV
easat-10905	117	6	-	-	PUNCT
easat-10905	117	7	tuned	tune	VERB
easat-10905	117	8	on	on	ADP
easat-10905	117	9	laptops	laptop	NOUN
easat-10905	117	10	,	,	PUNCT
easat-10905	117	11	restaurants	restaurant	NOUN
easat-10905	117	12	,	,	PUNCT
easat-10905	117	13	tweets	tweet	NOUN
easat-10905	117	14	,	,	PUNCT
easat-10905	117	15	and	and	CCONJ
easat-10905	117	16	finance	finance	NOUN
easat-10905	117	17	datasets	dataset	NOUN
easat-10905	117	18	using	use	VERB
easat-10905	117	19	a	a	DET
easat-10905	117	20	data	data	NOUN
easat-10905	117	21	retention	retention	NOUN
easat-10905	117	22	mechanism	mechanism	NOUN
easat-10905	117	23	that	that	PRON
easat-10905	117	24	preserves	preserve	VERB
easat-10905	117	25	prior	prior	ADJ
easat-10905	117	26	domain	domain	NOUN
easat-10905	117	27	knowledge	knowledge	NOUN
easat-10905	117	28	during	during	ADP
easat-10905	117	29	new	new	ADJ
easat-10905	117	30	task	task	NOUN
easat-10905	117	31	learning	learning	NOUN
easat-10905	117	32	.	.	PUNCT
easat-10905	118	1	experimental	experimental	ADJ
easat-10905	118	2	results	result	NOUN
easat-10905	118	3	show	show	VERB
easat-10905	118	4	that	that	SCONJ
easat-10905	118	5	dllagpt	dllagpt	NOUN
easat-10905	118	6	achieves	achieve	VERB
easat-10905	118	7	an	an	DET
easat-10905	118	8	average	average	ADJ
easat-10905	118	9	accuracy	accuracy	NOUN
easat-10905	118	10	of	of	ADP
easat-10905	118	11	0.85	0.85	NUM
easat-10905	118	12	and	and	CCONJ
easat-10905	118	13	a	a	DET
easat-10905	118	14	backward	backward	ADJ
easat-10905	118	15	transfer	transfer	NOUN
easat-10905	118	16	score	score	NOUN
easat-10905	118	17	of	of	ADP
easat-10905	118	18	-0.09	-0.09	NUM
easat-10905	118	19	,	,	PUNCT
easat-10905	118	20	confirming	confirm	VERB
easat-10905	118	21	its	its	PRON
easat-10905	118	22	effectiveness	effectiveness	NOUN
easat-10905	118	23	and	and	CCONJ
easat-10905	118	24	stability	stability	NOUN
easat-10905	118	25	in	in	ADP
easat-10905	118	26	continual	continual	ADJ
easat-10905	118	27	,	,	PUNCT
easat-10905	118	28	multi	multi	ADJ
easat-10905	118	29	-	-	ADJ
easat-10905	118	30	domain	domain	ADJ
easat-10905	118	31	sentiment	sentiment	NOUN
easat-10905	118	32	analysis	analysis	NOUN
easat-10905	118	33	.	.	PUNCT
easat-10905	119	1	jain	jain	PROPN
easat-10905	119	2	et	et	PROPN
easat-10905	119	3	al	al	PROPN
easat-10905	119	4	.	.	PUNCT
easat-10905	120	1	[	[	X
easat-10905	120	2	30	30	NUM
easat-10905	120	3	]	]	PUNCT
easat-10905	120	4	proposed	propose	VERB
easat-10905	120	5	an	an	DET
easat-10905	120	6	incremental	incremental	ADJ
easat-10905	120	7	learning	learning	NOUN
easat-10905	120	8	approach	approach	NOUN
easat-10905	120	9	for	for	ADP
easat-10905	120	10	continuously	continuously	ADV
easat-10905	120	11	updating	update	VERB
easat-10905	120	12	domain	domain	NOUN
easat-10905	120	13	-	-	PUNCT
easat-10905	120	14	specific	specific	ADJ
easat-10905	120	15	vocabularies	vocabulary	NOUN
easat-10905	120	16	.	.	PUNCT
easat-10905	121	1	the	the	DET
easat-10905	121	2	proposed	propose	VERB
easat-10905	121	3	doclib	doclib	PROPN
easat-10905	121	4	archive	archive	NOUN
easat-10905	121	5	efficiently	efficiently	ADV
easat-10905	121	6	stores	store	VERB
easat-10905	121	7	essential	essential	ADJ
easat-10905	121	8	information	information	NOUN
easat-10905	121	9	from	from	ADP
easat-10905	121	10	earlier	early	ADJ
easat-10905	121	11	data	datum	NOUN
easat-10905	121	12	to	to	PART
easat-10905	121	13	support	support	VERB
easat-10905	121	14	ongoing	ongoing	ADJ
easat-10905	121	15	vocabulary	vocabulary	ADJ
easat-10905	121	16	refinement	refinement	NOUN
easat-10905	121	17	.	.	PUNCT
easat-10905	122	1	its	its	PRON
easat-10905	122	2	performance	performance	NOUN
easat-10905	122	3	is	be	AUX
easat-10905	122	4	evaluated	evaluate	VERB
easat-10905	122	5	through	through	ADP
easat-10905	122	6	a	a	DET
easat-10905	122	7	text	text	NOUN
easat-10905	122	8	classification	classification	NOUN
easat-10905	122	9	task	task	NOUN
easat-10905	122	10	,	,	PUNCT
easat-10905	122	11	where	where	SCONJ
easat-10905	122	12	classification	classification	NOUN
easat-10905	122	13	accuracy	accuracy	NOUN
easat-10905	122	14	indicates	indicate	VERB
easat-10905	122	15	the	the	DET
easat-10905	122	16	vocabulary	vocabulary	NOUN
easat-10905	122	17	’s	’s	PART
easat-10905	122	18	effectiveness	effectiveness	NOUN
easat-10905	122	19	in	in	ADP
easat-10905	122	20	identifying	identify	VERB
easat-10905	122	21	new	new	ADJ
easat-10905	122	22	,	,	PUNCT
easat-10905	122	23	domainrelevant	domainrelevant	ADJ
easat-10905	122	24	content	content	NOUN
easat-10905	122	25	.	.	PUNCT
easat-10905	123	1	experimental	experimental	ADJ
easat-10905	123	2	results	result	NOUN
easat-10905	123	3	confirm	confirm	VERB
easat-10905	123	4	that	that	SCONJ
easat-10905	123	5	incremental	incremental	ADJ
easat-10905	123	6	updates	update	NOUN
easat-10905	123	7	sustain	sustain	VERB
easat-10905	123	8	both	both	CCONJ
easat-10905	123	9	the	the	DET
easat-10905	123	10	relevance	relevance	NOUN
easat-10905	123	11	and	and	CCONJ
easat-10905	123	12	discriminative	discriminative	NOUN
easat-10905	123	13	power	power	NOUN
easat-10905	123	14	of	of	ADP
easat-10905	123	15	the	the	DET
easat-10905	123	16	vocabulary	vocabulary	NOUN
easat-10905	123	17	without	without	ADP
easat-10905	123	18	degrading	degrade	VERB
easat-10905	123	19	its	its	PRON
easat-10905	123	20	overall	overall	ADJ
easat-10905	123	21	performance	performance	NOUN
easat-10905	123	22	.	.	PUNCT
easat-10905	124	1	to	to	PART
easat-10905	124	2	summarize	summarize	VERB
easat-10905	124	3	,	,	PUNCT
easat-10905	124	4	incremental	incremental	ADJ
easat-10905	124	5	learning	learning	NOUN
easat-10905	124	6	faces	face	VERB
easat-10905	124	7	the	the	DET
easat-10905	124	8	significant	significant	ADJ
easat-10905	124	9	challenge	challenge	NOUN
easat-10905	124	10	of	of	ADP
easat-10905	124	11	cf	cf	NOUN
easat-10905	124	12	,	,	PUNCT
easat-10905	124	13	where	where	SCONJ
easat-10905	124	14	models	model	NOUN
easat-10905	124	15	lose	lose	VERB
easat-10905	124	16	previously	previously	ADV
easat-10905	124	17	learned	learn	VERB
easat-10905	124	18	knowledge	knowledge	NOUN
easat-10905	124	19	when	when	SCONJ
easat-10905	124	20	acquiring	acquire	VERB
easat-10905	124	21	new	new	ADJ
easat-10905	124	22	tasks	task	NOUN
easat-10905	124	23	.	.	PUNCT
easat-10905	125	1	continual	continual	ADJ
easat-10905	125	2	learning	learning	NOUN
easat-10905	125	3	strategies	strategy	NOUN
easat-10905	125	4	aim	aim	VERB
easat-10905	125	5	to	to	PART
easat-10905	125	6	mitigate	mitigate	VERB
easat-10905	125	7	this	this	DET
easat-10905	125	8	issue	issue	NOUN
easat-10905	125	9	,	,	PUNCT
easat-10905	125	10	ensuring	ensure	VERB
easat-10905	125	11	that	that	SCONJ
easat-10905	125	12	models	model	NOUN
easat-10905	125	13	retain	retain	VERB
easat-10905	125	14	essential	essential	ADJ
easat-10905	125	15	information	information	NOUN
easat-10905	125	16	from	from	ADP
easat-10905	125	17	prior	prior	ADJ
easat-10905	125	18	tasks	task	NOUN
easat-10905	125	19	.	.	PUNCT
easat-10905	126	1	from	from	ADP
easat-10905	126	2	the	the	DET
easat-10905	126	3	literature	literature	NOUN
easat-10905	126	4	review	review	NOUN
easat-10905	126	5	,	,	PUNCT
easat-10905	126	6	we	we	PRON
easat-10905	126	7	find	find	VERB
easat-10905	126	8	that	that	SCONJ
easat-10905	126	9	there	there	PRON
easat-10905	126	10	are	be	VERB
easat-10905	126	11	a	a	DET
easat-10905	126	12	few	few	ADJ
easat-10905	126	13	works	work	NOUN
easat-10905	126	14	focusing	focus	VERB
easat-10905	126	15	on	on	ADP
easat-10905	126	16	continual	continual	ADJ
easat-10905	126	17	learning	learning	NOUN
easat-10905	126	18	of	of	ADP
easat-10905	126	19	sentiment	sentiment	NOUN
easat-10905	126	20	classification	classification	NOUN
easat-10905	126	21	,	,	PUNCT
easat-10905	126	22	intent	intent	ADJ
easat-10905	126	23	classification	classification	NOUN
easat-10905	126	24	,	,	PUNCT
easat-10905	126	25	hate	hate	VERB
easat-10905	126	26	speech	speech	NOUN
easat-10905	126	27	classification	classification	NOUN
easat-10905	126	28	,	,	PUNCT
easat-10905	126	29	question	question	NOUN
easat-10905	126	30	-	-	PUNCT
easat-10905	126	31	answering	answering	NOUN
easat-10905	126	32	,	,	PUNCT
easat-10905	126	33	etc	etc	X
easat-10905	126	34	.	.	X
easat-10905	126	35	,	,	PUNCT
easat-10905	126	36	but	but	CCONJ
easat-10905	126	37	there	there	PRON
easat-10905	126	38	is	be	VERB
easat-10905	126	39	no	no	DET
easat-10905	126	40	work	work	NOUN
easat-10905	126	41	focusing	focus	VERB
easat-10905	126	42	on	on	ADP
easat-10905	126	43	continual	continual	ADJ
easat-10905	126	44	learning	learning	NOUN
easat-10905	126	45	of	of	ADP
easat-10905	126	46	abusive	abusive	ADJ
easat-10905	126	47	and	and	CCONJ
easat-10905	126	48	offensive	offensive	ADJ
easat-10905	126	49	texts	text	NOUN
easat-10905	126	50	on	on	ADP
easat-10905	126	51	social	social	ADJ
easat-10905	126	52	media	medium	NOUN
easat-10905	126	53	comments	comment	NOUN
easat-10905	126	54	.	.	PUNCT
easat-10905	127	1	developing	develop	VERB
easat-10905	127	2	models	model	NOUN
easat-10905	127	3	to	to	PART
easat-10905	127	4	detect	detect	VERB
easat-10905	127	5	abusive	abusive	ADJ
easat-10905	127	6	and	and	CCONJ
easat-10905	127	7	offensive	offensive	ADJ
easat-10905	127	8	comments	comment	NOUN
easat-10905	127	9	on	on	ADP
easat-10905	127	10	social	social	ADJ
easat-10905	127	11	media	medium	NOUN
easat-10905	127	12	is	be	AUX
easat-10905	127	13	essential	essential	ADJ
easat-10905	127	14	for	for	ADP
easat-10905	127	15	promoting	promote	VERB
easat-10905	127	16	a	a	DET
easat-10905	127	17	safer	safe	ADJ
easat-10905	127	18	online	online	ADJ
easat-10905	127	19	environment	environment	NOUN
easat-10905	127	20	and	and	CCONJ
easat-10905	127	21	protecting	protect	VERB
easat-10905	127	22	the	the	DET
easat-10905	127	23	mental	mental	ADJ
easat-10905	127	24	health	health	NOUN
easat-10905	127	25	of	of	ADP
easat-10905	127	26	users	user	NOUN
easat-10905	127	27	.	.	PUNCT
easat-10905	128	1	this	this	PRON
easat-10905	128	2	necessitates	necessitate	VERB
easat-10905	128	3	the	the	DET
easat-10905	128	4	need	need	NOUN
easat-10905	128	5	for	for	ADP
easat-10905	128	6	models	model	NOUN
easat-10905	128	7	that	that	PRON
easat-10905	128	8	use	use	VERB
easat-10905	128	9	continual	continual	ADJ
easat-10905	128	10	learning	learning	NOUN
easat-10905	128	11	to	to	PART
easat-10905	128	12	provide	provide	VERB
easat-10905	128	13	a	a	DET
easat-10905	128	14	scalable	scalable	ADJ
easat-10905	128	15	solution	solution	NOUN
easat-10905	128	16	,	,	PUNCT
easat-10905	128	17	ensuring	ensure	VERB
easat-10905	128	18	the	the	DET
easat-10905	128	19	models	model	NOUN
easat-10905	128	20	can	can	AUX
easat-10905	128	21	process	process	VERB
easat-10905	128	22	and	and	CCONJ
easat-10905	128	23	learn	learn	VERB
easat-10905	128	24	from	from	ADP
easat-10905	128	25	new	new	ADJ
easat-10905	128	26	data	datum	NOUN
easat-10905	128	27	consistently	consistently	ADV
easat-10905	128	28	.	.	PUNCT
easat-10905	129	1	such	such	ADJ
easat-10905	129	2	models	model	NOUN
easat-10905	129	3	can	can	AUX
easat-10905	129	4	handle	handle	VERB
easat-10905	129	5	a	a	DET
easat-10905	129	6	sequence	sequence	NOUN
easat-10905	129	7	of	of	ADP
easat-10905	129	8	tasks	task	NOUN
easat-10905	129	9	within	within	ADP
easat-10905	129	10	a	a	DET
easat-10905	129	11	unified	unified	ADJ
easat-10905	129	12	architecture	architecture	NOUN
easat-10905	129	13	,	,	PUNCT
easat-10905	129	14	leveraging	leverage	VERB
easat-10905	129	15	shared	share	VERB
easat-10905	129	16	features	feature	NOUN
easat-10905	129	17	and	and	CCONJ
easat-10905	129	18	improving	improve	VERB
easat-10905	129	19	overall	overall	ADJ
easat-10905	129	20	performance	performance	NOUN
easat-10905	129	21	,	,	PUNCT
easat-10905	129	22	making	make	VERB
easat-10905	129	23	them	they	PRON
easat-10905	129	24	well	well	ADV
easat-10905	129	25	-	-	PUNCT
easat-10905	129	26	suited	suit	VERB
easat-10905	129	27	for	for	ADP
easat-10905	129	28	the	the	DET
easat-10905	129	29	dynamic	dynamic	ADJ
easat-10905	129	30	and	and	CCONJ
easat-10905	129	31	evolving	evolve	VERB
easat-10905	129	32	nature	nature	NOUN
easat-10905	129	33	of	of	ADP
easat-10905	129	34	social	social	ADJ
easat-10905	129	35	media	medium	NOUN
easat-10905	129	36	content	content	NOUN
easat-10905	129	37	.	.	PUNCT
easat-10905	130	1	3	3	X
easat-10905	130	2	.	.	X
easat-10905	130	3	materials	material	NOUN
easat-10905	130	4	and	and	CCONJ
easat-10905	130	5	methods	method	NOUN
easat-10905	130	6	3.1	3.1	NUM
easat-10905	130	7	.	.	PUNCT
easat-10905	131	1	dataset	dataset	ADJ
easat-10905	131	2	description	description	NOUN
easat-10905	131	3	the	the	DET
easat-10905	131	4	dataset	dataset	NOUN
easat-10905	131	5	used	use	VERB
easat-10905	131	6	for	for	ADP
easat-10905	131	7	this	this	DET
easat-10905	131	8	research	research	NOUN
easat-10905	131	9	has	have	AUX
easat-10905	131	10	been	be	AUX
easat-10905	131	11	collected	collect	VERB
easat-10905	131	12	from	from	ADP
easat-10905	131	13	the	the	DET
easat-10905	131	14	social	social	ADJ
easat-10905	131	15	platforms1	platforms1	NOUN
easat-10905	131	16	such	such	ADJ
easat-10905	131	17	as	as	ADP
easat-10905	131	18	twitter	twitter	NOUN
easat-10905	131	19	and	and	CCONJ
easat-10905	131	20	facebook	facebook	NOUN
easat-10905	131	21	activities	activity	NOUN
easat-10905	131	22	are	be	AUX
easat-10905	131	23	often	often	ADV
easat-10905	131	24	based	base	VERB
easat-10905	131	25	on	on	ADP
easat-10905	131	26	suspicious	suspicious	ADJ
easat-10905	131	27	behaviors	behavior	NOUN
easat-10905	131	28	such	such	ADJ
easat-10905	131	29	as	as	ADP
easat-10905	131	30	racism	racism	NOUN
easat-10905	131	31	,	,	PUNCT
easat-10905	131	32	discrimination	discrimination	NOUN
easat-10905	131	33	,	,	PUNCT
easat-10905	131	34	abusive	abusive	ADJ
easat-10905	131	35	language	language	NOUN
easat-10905	131	36	,	,	PUNCT
easat-10905	131	37	and	and	CCONJ
easat-10905	131	38	threats	threat	NOUN
easat-10905	131	39	,	,	PUNCT
easat-10905	131	40	which	which	PRON
easat-10905	131	41	are	be	AUX
easat-10905	131	42	commonly	commonly	ADV
easat-10905	131	43	associated	associate	VERB
easat-10905	131	44	with	with	ADP
easat-10905	131	45	cyberbullying	cyberbullye	VERB
easat-10905	131	46	.	.	PUNCT
easat-10905	132	1	the	the	DET
easat-10905	132	2	dataset	dataset	NOUN
easat-10905	132	3	has	have	AUX
easat-10905	132	4	been	be	AUX
easat-10905	132	5	labeled	label	VERB
easat-10905	132	6	according	accord	VERB
easat-10905	132	7	to	to	ADP
easat-10905	132	8	the	the	DET
easat-10905	132	9	presence	presence	NOUN
easat-10905	132	10	of	of	ADP
easat-10905	132	11	suspicious	suspicious	ADJ
easat-10905	132	12	words	word	NOUN
easat-10905	132	13	used	use	VERB
easat-10905	132	14	in	in	ADP
easat-10905	132	15	tweets	tweet	NOUN
easat-10905	132	16	and	and	CCONJ
easat-10905	132	17	comments	comment	NOUN
easat-10905	132	18	.	.	PUNCT
easat-10905	133	1	suspicious	suspicious	ADJ
easat-10905	133	2	data	datum	NOUN
easat-10905	133	3	is	be	AUX
easat-10905	133	4	tagged	tag	VERB
easat-10905	133	5	with	with	ADP
easat-10905	133	6	1	1	NUM
easat-10905	133	7	,	,	PUNCT
easat-10905	133	8	while	while	SCONJ
easat-10905	133	9	non	non	ADJ
easat-10905	133	10	-	-	ADJ
easat-10905	133	11	suspicious	suspicious	ADJ
easat-10905	133	12	data	data	NOUN
easat-10905	133	13	is	be	AUX
easat-10905	133	14	tagged	tag	VERB
easat-10905	133	15	with	with	ADP
easat-10905	133	16	0	0	NUM
easat-10905	133	17	.	.	PUNCT
easat-10905	134	1	the	the	DET
easat-10905	134	2	dataset	dataset	NOUN
easat-10905	134	3	comprises	comprise	VERB
easat-10905	134	4	up	up	ADP
easat-10905	134	5	to	to	PART
easat-10905	134	6	20,001	20,001	NUM
easat-10905	134	7	rows	row	NOUN
easat-10905	134	8	of	of	ADP
easat-10905	134	9	sentiment	sentiment	NOUN
easat-10905	134	10	data	datum	NOUN
easat-10905	134	11	,	,	PUNCT
easat-10905	134	12	with	with	ADP
easat-10905	134	13	12,179	12,179	NUM
easat-10905	134	14	rows	row	NOUN
easat-10905	134	15	labeled	label	VERB
easat-10905	134	16	as	as	ADP
easat-10905	134	17	negative	negative	ADJ
easat-10905	134	18	sentiments	sentiment	NOUN
easat-10905	134	19	,	,	PUNCT
easat-10905	134	20	including	include	VERB
easat-10905	134	21	racism	racism	NOUN
easat-10905	134	22	,	,	PUNCT
easat-10905	134	23	discrimination	discrimination	NOUN
easat-10905	134	24	,	,	PUNCT
easat-10905	134	25	and	and	CCONJ
easat-10905	134	26	abuse	abuse	NOUN
easat-10905	134	27	.	.	PUNCT
easat-10905	135	1	the	the	DET
easat-10905	135	2	remaining	remain	VERB
easat-10905	135	3	7,822	7,822	NUM
easat-10905	135	4	rows	row	NOUN
easat-10905	135	5	are	be	AUX
easat-10905	135	6	marked	mark	VERB
easat-10905	135	7	with	with	ADP
easat-10905	135	8	positive	positive	ADJ
easat-10905	135	9	or	or	CCONJ
easat-10905	135	10	neutral	neutral	ADJ
easat-10905	135	11	sentiments	sentiment	NOUN
easat-10905	135	12	,	,	PUNCT
easat-10905	135	13	indicating	indicate	VERB
easat-10905	135	14	that	that	SCONJ
easat-10905	135	15	the	the	DET
easat-10905	135	16	comments	comment	NOUN
easat-10905	135	17	are	be	AUX
easat-10905	135	18	not	not	PART
easat-10905	135	19	suspicious	suspicious	ADJ
easat-10905	135	20	.	.	PUNCT
easat-10905	136	1	a	a	DET
easat-10905	136	2	sample	sample	NOUN
easat-10905	136	3	of	of	ADP
easat-10905	136	4	texts	text	NOUN
easat-10905	136	5	from	from	ADP
easat-10905	136	6	the	the	DET
easat-10905	136	7	dataset	dataset	NOUN
easat-10905	136	8	is	be	AUX
easat-10905	136	9	presented	present	VERB
easat-10905	136	10	in	in	ADP
easat-10905	136	11	table	table	NOUN
easat-10905	136	12	1	1	NUM
easat-10905	136	13	.	.	NOUN
easat-10905	136	14	1	1	NUM
easat-10905	136	15	https://www.kaggle.com/datasets/syedabbasraza/suspicious-communication-on-social-platforms	https://www.kaggle.com/datasets/syedabbasraza/suspicious-communication-on-social-platforms	ADP
easat-10905	136	16	452	452	NUM
easat-10905	136	17	edelweiss	edelweiss	PROPN
easat-10905	136	18	applied	apply	VERB
easat-10905	136	19	science	science	NOUN
easat-10905	136	20	and	and	CCONJ
easat-10905	136	21	technology	technology	NOUN
easat-10905	136	22	issn	issn	PROPN
easat-10905	136	23	:	:	PUNCT
easat-10905	136	24	2576	2576	NUM
easat-10905	136	25	-	-	SYM
easat-10905	136	26	8484	8484	NUM
easat-10905	136	27	vol	vol	NOUN
easat-10905	136	28	.	.	PROPN
easat-10905	137	1	9	9	NUM
easat-10905	137	2	,	,	PUNCT
easat-10905	137	3	no	no	INTJ
easat-10905	137	4	.	.	NOUN
easat-10905	137	5	11	11	NUM
easat-10905	137	6	:	:	PUNCT
easat-10905	137	7	448	448	NUM
easat-10905	137	8	-	-	SYM
easat-10905	137	9	466	466	NUM
easat-10905	137	10	,	,	PUNCT
easat-10905	137	11	2025	2025	NUM
easat-10905	137	12	doi	doi	NOUN
easat-10905	137	13	:	:	PUNCT
easat-10905	137	14	10.55214/2576	10.55214/2576	NUM
easat-10905	137	15	-	-	PUNCT
easat-10905	137	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	138	1	©	©	PROPN
easat-10905	138	2	2025	2025	NUM
easat-10905	138	3	by	by	ADP
easat-10905	138	4	the	the	DET
easat-10905	138	5	authors	author	NOUN
easat-10905	138	6	;	;	PUNCT
easat-10905	138	7	licensee	licensee	PROPN
easat-10905	138	8	learning	learn	VERB
easat-10905	138	9	gate	gate	NOUN
easat-10905	138	10	table	table	NOUN
easat-10905	138	11	1	1	NUM
easat-10905	138	12	.	.	PUNCT
easat-10905	138	13	sample	sample	NOUN
easat-10905	138	14	abusive	abusive	ADJ
easat-10905	138	15	texts	text	NOUN
easat-10905	138	16	.	.	PUNCT
easat-10905	139	1	texts	text	NOUN
easat-10905	139	2	category	category	NOUN
easat-10905	139	3	i	i	PRON
easat-10905	139	4	loved	love	VERB
easat-10905	139	5	it	it	PRON
easat-10905	139	6	,	,	PUNCT
easat-10905	139	7	i	i	PRON
easat-10905	139	8	laughed	laugh	VERB
easat-10905	139	9	my	my	PRON
easat-10905	139	10	ass	ass	NOUN
easat-10905	139	11	off	off	ADV
easat-10905	139	12	1	1	NUM
easat-10905	139	13	ahh	ahh	NOUN
easat-10905	139	14	,	,	PUNCT
easat-10905	139	15	you	you	PRON
easat-10905	139	16	might	might	AUX
easat-10905	139	17	be	be	AUX
easat-10905	139	18	on	on	ADP
easat-10905	139	19	to	to	ADP
easat-10905	139	20	something	something	PRON
easat-10905	139	21	...	...	PUNCT
easat-10905	140	1	damn	damn	ADJ
easat-10905	140	2	techies	techie	NOUN
easat-10905	140	3	1	1	NUM
easat-10905	140	4	what	what	PRON
easat-10905	140	5	is	be	AUX
easat-10905	140	6	your	your	PRON
easat-10905	140	7	guilty	guilty	ADJ
easat-10905	140	8	pleasure	pleasure	NOUN
easat-10905	140	9	disney	disney	NOUN
easat-10905	140	10	movie	movie	NOUN
easat-10905	140	11	?	?	PUNCT
easat-10905	140	12	0	0	PUNCT
easat-10905	141	1	yeah	yeah	INTJ
easat-10905	141	2	,	,	PUNCT
easat-10905	141	3	if	if	SCONJ
easat-10905	141	4	i	i	PRON
easat-10905	141	5	tried	try	VERB
easat-10905	141	6	.	.	PUNCT
easat-10905	142	1	0	0	NUM
easat-10905	143	1	3.2	3.2	NUM
easat-10905	143	2	.	.	PUNCT
easat-10905	144	1	classifiers	classifier	NOUN
easat-10905	144	2	in	in	ADP
easat-10905	144	3	this	this	DET
easat-10905	144	4	section	section	NOUN
easat-10905	144	5	,	,	PUNCT
easat-10905	144	6	we	we	PRON
easat-10905	144	7	present	present	VERB
easat-10905	144	8	a	a	DET
easat-10905	144	9	brief	brief	ADJ
easat-10905	144	10	overview	overview	NOUN
easat-10905	144	11	of	of	ADP
easat-10905	144	12	the	the	DET
easat-10905	144	13	pretrained	pretraine	VERB
easat-10905	144	14	models	model	NOUN
easat-10905	144	15	that	that	PRON
easat-10905	144	16	are	be	AUX
easat-10905	144	17	used	use	VERB
easat-10905	144	18	in	in	ADP
easat-10905	144	19	our	our	PRON
easat-10905	144	20	work	work	NOUN
easat-10905	144	21	.	.	PUNCT
easat-10905	145	1	3.2.1	3.2.1	X
easat-10905	145	2	.	.	PUNCT
easat-10905	145	3	bidirectional	bidirectional	ADJ
easat-10905	145	4	encoder	encoder	NOUN
easat-10905	145	5	representations	representation	VERB
easat-10905	145	6	from	from	ADP
easat-10905	145	7	transformers	transformer	NOUN
easat-10905	145	8	(	(	PUNCT
easat-10905	145	9	bert	bert	PROPN
easat-10905	145	10	)	)	PUNCT
easat-10905	145	11	bert	bert	PROPN
easat-10905	145	12	understands	understand	VERB
easat-10905	145	13	word	word	NOUN
easat-10905	145	14	context	context	NOUN
easat-10905	145	15	by	by	ADP
easat-10905	145	16	considering	consider	VERB
easat-10905	145	17	both	both	CCONJ
easat-10905	145	18	preceding	precede	VERB
easat-10905	145	19	and	and	CCONJ
easat-10905	145	20	following	follow	VERB
easat-10905	145	21	words	word	NOUN
easat-10905	145	22	in	in	ADP
easat-10905	145	23	a	a	DET
easat-10905	145	24	sentence	sentence	NOUN
easat-10905	145	25	,	,	PUNCT
easat-10905	145	26	making	make	VERB
easat-10905	145	27	it	it	PRON
easat-10905	145	28	bidirectional	bidirectional	ADJ
easat-10905	145	29	[	[	X
easat-10905	145	30	31	31	NUM
easat-10905	145	31	]	]	PUNCT
easat-10905	145	32	.	.	PUNCT
easat-10905	146	1	bert	bert	PROPN
easat-10905	146	2	is	be	AUX
easat-10905	146	3	pre	pre	ADJ
easat-10905	146	4	-	-	VERB
easat-10905	146	5	trained	train	VERB
easat-10905	146	6	using	use	VERB
easat-10905	146	7	two	two	NUM
easat-10905	146	8	unsupervised	unsupervised	ADJ
easat-10905	146	9	tasks	task	NOUN
easat-10905	146	10	:	:	PUNCT
easat-10905	146	11	masked	mask	VERB
easat-10905	146	12	language	language	NOUN
easat-10905	146	13	modeling	modeling	NOUN
easat-10905	146	14	and	and	CCONJ
easat-10905	146	15	next	next	ADJ
easat-10905	146	16	sentence	sentence	NOUN
easat-10905	146	17	prediction	prediction	NOUN
easat-10905	146	18	.	.	PUNCT
easat-10905	147	1	pre	pre	VERB
easat-10905	147	2	-	-	ADJ
easat-10905	147	3	trained	train	VERB
easat-10905	147	4	bert	bert	NOUN
easat-10905	147	5	can	can	AUX
easat-10905	147	6	be	be	AUX
easat-10905	147	7	fine	fine	ADV
easat-10905	147	8	-	-	PUNCT
easat-10905	147	9	tuned	tune	VERB
easat-10905	147	10	on	on	ADP
easat-10905	147	11	specific	specific	ADJ
easat-10905	147	12	tasks	task	NOUN
easat-10905	147	13	with	with	ADP
easat-10905	147	14	labeled	label	VERB
easat-10905	147	15	data	datum	NOUN
easat-10905	147	16	.	.	PUNCT
easat-10905	148	1	the	the	DET
easat-10905	148	2	bert	bert	PROPN
easat-10905	148	3	architecture	architecture	NOUN
easat-10905	148	4	is	be	AUX
easat-10905	148	5	based	base	VERB
easat-10905	148	6	on	on	ADP
easat-10905	148	7	the	the	DET
easat-10905	148	8	multilayer	multilayer	ADJ
easat-10905	148	9	bidirectional	bidirectional	NOUN
easat-10905	148	10	transformer	transformer	NOUN
easat-10905	148	11	described	describe	VERB
easat-10905	148	12	by	by	ADP
easat-10905	148	13	vaswani	vaswani	PROPN
easat-10905	148	14	et	et	PROPN
easat-10905	148	15	al	al	PROPN
easat-10905	148	16	.	.	PUNCT
easat-10905	149	1	[	[	X
easat-10905	149	2	32	32	NUM
easat-10905	149	3	]	]	PUNCT
easat-10905	149	4	.	.	PUNCT
easat-10905	150	1	by	by	ADP
easat-10905	150	2	adding	add	VERB
easat-10905	150	3	an	an	DET
easat-10905	150	4	output	output	NOUN
easat-10905	150	5	layer	layer	NOUN
easat-10905	150	6	,	,	PUNCT
easat-10905	150	7	bert	bert	PROPN
easat-10905	150	8	may	may	AUX
easat-10905	150	9	be	be	AUX
easat-10905	150	10	optimized	optimize	VERB
easat-10905	150	11	to	to	PART
easat-10905	150	12	provide	provide	VERB
easat-10905	150	13	state	state	NOUN
easat-10905	150	14	-	-	PUNCT
easat-10905	150	15	of	of	ADP
easat-10905	150	16	-	-	PUNCT
easat-10905	150	17	the	the	DET
easat-10905	150	18	-	-	PUNCT
easat-10905	150	19	art	art	NOUN
easat-10905	150	20	models	model	NOUN
easat-10905	150	21	for	for	ADP
easat-10905	150	22	a	a	DET
easat-10905	150	23	variety	variety	NOUN
easat-10905	150	24	of	of	ADP
easat-10905	150	25	applications	application	NOUN
easat-10905	150	26	,	,	PUNCT
easat-10905	150	27	including	include	VERB
easat-10905	150	28	language	language	NOUN
easat-10905	150	29	inference	inference	NOUN
easat-10905	150	30	and	and	CCONJ
easat-10905	150	31	question	question	NOUN
easat-10905	150	32	answering	answering	NOUN
easat-10905	150	33	,	,	PUNCT
easat-10905	150	34	without	without	ADP
easat-10905	150	35	requiring	require	VERB
easat-10905	150	36	significant	significant	ADJ
easat-10905	150	37	changes	change	NOUN
easat-10905	150	38	to	to	ADP
easat-10905	150	39	the	the	DET
easat-10905	150	40	task	task	NOUN
easat-10905	150	41	-	-	PUNCT
easat-10905	150	42	specific	specific	ADJ
easat-10905	150	43	architecture	architecture	NOUN
easat-10905	150	44	.	.	PUNCT
easat-10905	151	1	3.2.2	3.2.2	NUM
easat-10905	151	2	.	.	PUNCT
easat-10905	151	3	distilbert	distilbert	PROPN
easat-10905	151	4	distilbert	distilbert	PROPN
easat-10905	151	5	,	,	PUNCT
easat-10905	151	6	a	a	DET
easat-10905	151	7	variant	variant	NOUN
easat-10905	151	8	of	of	ADP
easat-10905	151	9	bert	bert	PROPN
easat-10905	151	10	,	,	PUNCT
easat-10905	151	11	offers	offer	VERB
easat-10905	151	12	a	a	DET
easat-10905	151	13	faster	fast	ADJ
easat-10905	151	14	and	and	CCONJ
easat-10905	151	15	more	more	ADV
easat-10905	151	16	lightweight	lightweight	ADJ
easat-10905	151	17	model	model	NOUN
easat-10905	151	18	while	while	SCONJ
easat-10905	151	19	preserving	preserve	VERB
easat-10905	151	20	much	much	ADJ
easat-10905	151	21	of	of	ADP
easat-10905	151	22	bert	bert	PROPN
easat-10905	151	23	’s	’s	PART
easat-10905	151	24	performance	performance	NOUN
easat-10905	151	25	through	through	ADP
easat-10905	151	26	knowledge	knowledge	NOUN
easat-10905	151	27	distillation	distillation	NOUN
easat-10905	151	28	[	[	X
easat-10905	151	29	33	33	NUM
easat-10905	151	30	]	]	PUNCT
easat-10905	151	31	.	.	PUNCT
easat-10905	152	1	distilbert	distilbert	PROPN
easat-10905	152	2	undergoes	undergoe	NOUN
easat-10905	152	3	pre	pre	NOUN
easat-10905	153	1	-	-	VERB
easat-10905	153	2	training	train	VERB
easat-10905	153	3	on	on	ADP
easat-10905	153	4	extensive	extensive	ADJ
easat-10905	153	5	text	text	NOUN
easat-10905	153	6	datasets	dataset	NOUN
easat-10905	153	7	using	use	VERB
easat-10905	153	8	masked	mask	VERB
easat-10905	153	9	language	language	NOUN
easat-10905	153	10	modeling	modeling	NOUN
easat-10905	153	11	to	to	PART
easat-10905	153	12	develop	develop	VERB
easat-10905	153	13	broad	broad	ADJ
easat-10905	153	14	language	language	NOUN
easat-10905	153	15	understanding	understanding	NOUN
easat-10905	153	16	.	.	PUNCT
easat-10905	154	1	it	it	PRON
easat-10905	154	2	can	can	AUX
easat-10905	154	3	then	then	ADV
easat-10905	154	4	be	be	AUX
easat-10905	154	5	fine	fine	ADV
easat-10905	154	6	-	-	PUNCT
easat-10905	154	7	tuned	tune	VERB
easat-10905	154	8	on	on	ADP
easat-10905	154	9	specific	specific	ADJ
easat-10905	154	10	text	text	NOUN
easat-10905	154	11	classification	classification	NOUN
easat-10905	154	12	tasks	task	NOUN
easat-10905	154	13	with	with	ADP
easat-10905	154	14	labeled	label	VERB
easat-10905	154	15	data	datum	NOUN
easat-10905	154	16	,	,	PUNCT
easat-10905	154	17	allowing	allow	VERB
easat-10905	154	18	it	it	PRON
easat-10905	154	19	to	to	PART
easat-10905	154	20	adapt	adapt	VERB
easat-10905	154	21	to	to	ADP
easat-10905	154	22	the	the	DET
easat-10905	154	23	needs	need	NOUN
easat-10905	154	24	of	of	ADP
easat-10905	154	25	these	these	DET
easat-10905	154	26	tasks	task	NOUN
easat-10905	154	27	.	.	PUNCT
easat-10905	155	1	its	its	PRON
easat-10905	155	2	compact	compact	ADJ
easat-10905	155	3	design	design	NOUN
easat-10905	155	4	makes	make	VERB
easat-10905	155	5	it	it	PRON
easat-10905	155	6	well	well	ADV
easat-10905	155	7	-	-	PUNCT
easat-10905	155	8	suited	suit	VERB
easat-10905	155	9	for	for	ADP
easat-10905	155	10	various	various	ADJ
easat-10905	155	11	nlp	nlp	NOUN
easat-10905	155	12	applications	application	NOUN
easat-10905	155	13	,	,	PUNCT
easat-10905	155	14	including	include	VERB
easat-10905	155	15	text	text	NOUN
easat-10905	155	16	classification	classification	NOUN
easat-10905	155	17	,	,	PUNCT
easat-10905	155	18	where	where	SCONJ
easat-10905	155	19	it	it	PRON
easat-10905	155	20	provides	provide	VERB
easat-10905	155	21	both	both	DET
easat-10905	155	22	high	high	ADJ
easat-10905	155	23	accuracy	accuracy	NOUN
easat-10905	155	24	and	and	CCONJ
easat-10905	155	25	efficiency	efficiency	NOUN
easat-10905	155	26	.	.	PUNCT
easat-10905	156	1	3.2.3	3.2.3	X
easat-10905	156	2	.	.	PUNCT
easat-10905	156	3	a	a	DET
easat-10905	156	4	lite	lite	ADJ
easat-10905	156	5	bert	bert	PROPN
easat-10905	156	6	(	(	PUNCT
easat-10905	156	7	albert	albert	PROPN
easat-10905	156	8	)	)	PUNCT
easat-10905	156	9	albert	albert	PROPN
easat-10905	156	10	is	be	AUX
easat-10905	156	11	a	a	DET
easat-10905	156	12	smaller	small	ADJ
easat-10905	156	13	variant	variant	NOUN
easat-10905	156	14	of	of	ADP
easat-10905	156	15	the	the	DET
easat-10905	156	16	bert	bert	PROPN
easat-10905	156	17	model	model	NOUN
easat-10905	156	18	,	,	PUNCT
easat-10905	156	19	intended	intend	VERB
easat-10905	156	20	to	to	PART
easat-10905	156	21	reduce	reduce	VERB
easat-10905	156	22	size	size	NOUN
easat-10905	156	23	and	and	CCONJ
easat-10905	156	24	computing	computing	NOUN
easat-10905	156	25	requirements	requirement	NOUN
easat-10905	156	26	while	while	SCONJ
easat-10905	156	27	preserving	preserve	VERB
easat-10905	156	28	performance	performance	NOUN
easat-10905	156	29	[	[	X
easat-10905	156	30	34	34	NUM
easat-10905	156	31	]	]	PUNCT
easat-10905	156	32	.	.	PUNCT
easat-10905	157	1	albert	albert	PROPN
easat-10905	157	2	accomplishes	accomplish	VERB
easat-10905	157	3	this	this	PRON
easat-10905	157	4	by	by	ADP
easat-10905	157	5	employing	employ	VERB
easat-10905	157	6	factorized	factorize	VERB
easat-10905	157	7	embedding	embed	VERB
easat-10905	157	8	parameterization	parameterization	NOUN
easat-10905	157	9	and	and	CCONJ
easat-10905	157	10	cross	cross	ADJ
easat-10905	157	11	-	-	ADJ
easat-10905	157	12	layer	layer	ADJ
easat-10905	157	13	parameter	parameter	NOUN
easat-10905	157	14	sharing	sharing	NOUN
easat-10905	157	15	.	.	PUNCT
easat-10905	158	1	albert	albert	PROPN
easat-10905	158	2	divides	divide	VERB
easat-10905	158	3	the	the	DET
easat-10905	158	4	size	size	NOUN
easat-10905	158	5	of	of	ADP
easat-10905	158	6	the	the	DET
easat-10905	158	7	embedding	embed	VERB
easat-10905	158	8	matrix	matrix	NOUN
easat-10905	158	9	into	into	ADP
easat-10905	158	10	two	two	NUM
easat-10905	158	11	smaller	small	ADJ
easat-10905	158	12	ones	one	NOUN
easat-10905	158	13	.	.	PUNCT
easat-10905	159	1	this	this	PRON
easat-10905	159	2	minimizes	minimize	VERB
easat-10905	159	3	the	the	DET
easat-10905	159	4	number	number	NOUN
easat-10905	159	5	of	of	ADP
easat-10905	159	6	parameters	parameter	NOUN
easat-10905	159	7	while	while	SCONJ
easat-10905	159	8	still	still	ADV
easat-10905	159	9	providing	provide	VERB
easat-10905	159	10	a	a	DET
easat-10905	159	11	high	high	ADJ
easat-10905	159	12	-	-	PUNCT
easat-10905	159	13	level	level	NOUN
easat-10905	159	14	representation	representation	NOUN
easat-10905	159	15	of	of	ADP
easat-10905	159	16	the	the	DET
easat-10905	159	17	input	input	NOUN
easat-10905	159	18	task	task	NOUN
easat-10905	159	19	.	.	PUNCT
easat-10905	160	1	instead	instead	ADV
easat-10905	160	2	of	of	ADP
easat-10905	160	3	having	have	VERB
easat-10905	160	4	separate	separate	ADJ
easat-10905	160	5	settings	setting	NOUN
easat-10905	160	6	for	for	ADP
easat-10905	160	7	each	each	DET
easat-10905	160	8	layer	layer	NOUN
easat-10905	160	9	,	,	PUNCT
easat-10905	160	10	albert	albert	PROPN
easat-10905	160	11	spreads	spread	VERB
easat-10905	160	12	them	they	PRON
easat-10905	160	13	across	across	ADP
easat-10905	160	14	several	several	ADJ
easat-10905	160	15	layers	layer	NOUN
easat-10905	160	16	.	.	PUNCT
easat-10905	161	1	this	this	PRON
easat-10905	161	2	reduces	reduce	VERB
easat-10905	161	3	the	the	DET
easat-10905	161	4	overall	overall	ADJ
easat-10905	161	5	number	number	NOUN
easat-10905	161	6	of	of	ADP
easat-10905	161	7	parameters	parameter	NOUN
easat-10905	161	8	,	,	PUNCT
easat-10905	161	9	making	make	VERB
easat-10905	161	10	the	the	DET
easat-10905	161	11	model	model	NOUN
easat-10905	161	12	lighter	light	ADJ
easat-10905	161	13	.	.	PUNCT
easat-10905	162	1	such	such	ADJ
easat-10905	162	2	improvements	improvement	NOUN
easat-10905	162	3	increase	increase	VERB
easat-10905	162	4	efficiency	efficiency	NOUN
easat-10905	162	5	,	,	PUNCT
easat-10905	162	6	allowing	allow	VERB
easat-10905	162	7	it	it	PRON
easat-10905	162	8	to	to	PART
easat-10905	162	9	handle	handle	VERB
easat-10905	162	10	nlp	nlp	ADJ
easat-10905	162	11	tasks	task	NOUN
easat-10905	162	12	with	with	ADP
easat-10905	162	13	fewer	few	ADJ
easat-10905	162	14	resources	resource	NOUN
easat-10905	162	15	and	and	CCONJ
easat-10905	162	16	faster	fast	ADJ
easat-10905	162	17	processing	processing	NOUN
easat-10905	162	18	times	time	NOUN
easat-10905	162	19	while	while	SCONJ
easat-10905	162	20	still	still	ADV
easat-10905	162	21	producing	produce	VERB
easat-10905	162	22	good	good	ADJ
easat-10905	162	23	results	result	NOUN
easat-10905	162	24	.	.	PUNCT
easat-10905	163	1	3.2.4	3.2.4	X
easat-10905	163	2	.	.	PUNCT
easat-10905	163	3	robustly	robustly	ADV
easat-10905	163	4	optimized	optimize	VERB
easat-10905	163	5	bert	bert	NOUN
easat-10905	163	6	approach	approach	NOUN
easat-10905	163	7	(	(	PUNCT
easat-10905	163	8	roberta	roberta	PROPN
easat-10905	163	9	)	)	PUNCT
easat-10905	163	10	by	by	ADP
easat-10905	163	11	improving	improve	VERB
easat-10905	163	12	its	its	PRON
easat-10905	163	13	pre	pre	ADJ
easat-10905	163	14	-	-	ADJ
easat-10905	163	15	training	training	ADJ
easat-10905	163	16	strategy	strategy	NOUN
easat-10905	163	17	,	,	PUNCT
easat-10905	163	18	roberta	roberta	PROPN
easat-10905	163	19	expands	expand	VERB
easat-10905	163	20	on	on	ADP
easat-10905	163	21	bert	bert	NOUN
easat-10905	163	22	and	and	CCONJ
easat-10905	163	23	achieves	achieve	VERB
easat-10905	163	24	better	well	ADJ
easat-10905	163	25	results	result	NOUN
easat-10905	163	26	in	in	ADP
easat-10905	163	27	a	a	DET
easat-10905	163	28	range	range	NOUN
easat-10905	163	29	of	of	ADP
easat-10905	163	30	nlp	nlp	NOUN
easat-10905	163	31	tasks	task	NOUN
easat-10905	163	32	.	.	PUNCT
easat-10905	164	1	it	it	PRON
easat-10905	164	2	uses	use	VERB
easat-10905	164	3	self	self	NOUN
easat-10905	164	4	-	-	PUNCT
easat-10905	164	5	attention	attention	NOUN
easat-10905	164	6	techniques	technique	NOUN
easat-10905	164	7	and	and	CCONJ
easat-10905	164	8	a	a	DET
easat-10905	164	9	stack	stack	NOUN
easat-10905	164	10	of	of	ADP
easat-10905	164	11	encoders	encoder	NOUN
easat-10905	164	12	in	in	ADP
easat-10905	164	13	the	the	DET
easat-10905	164	14	same	same	ADJ
easat-10905	164	15	transformerbased	transformerbase	VERB
easat-10905	164	16	encoder	encoder	NOUN
easat-10905	164	17	design	design	NOUN
easat-10905	164	18	as	as	ADP
easat-10905	164	19	bert	bert	PROPN
easat-10905	164	20	[	[	X
easat-10905	164	21	35	35	NUM
easat-10905	164	22	]	]	PUNCT
easat-10905	164	23	.	.	PUNCT
easat-10905	165	1	like	like	ADP
easat-10905	165	2	bert	bert	PROPN
easat-10905	165	3	,	,	PUNCT
easat-10905	165	4	roberta	roberta	PROPN
easat-10905	165	5	analyzes	analyze	VERB
easat-10905	165	6	text	text	NOUN
easat-10905	165	7	in	in	ADP
easat-10905	165	8	a	a	DET
easat-10905	165	9	bidirectional	bidirectional	ADJ
easat-10905	165	10	manner	manner	NOUN
easat-10905	165	11	,	,	PUNCT
easat-10905	165	12	considering	consider	VERB
easat-10905	165	13	context	context	NOUN
easat-10905	165	14	from	from	ADP
easat-10905	165	15	the	the	DET
easat-10905	165	16	left	left	ADJ
easat-10905	165	17	and	and	CCONJ
easat-10905	165	18	right	right	ADJ
easat-10905	165	19	sides	side	NOUN
easat-10905	165	20	.	.	PUNCT
easat-10905	166	1	roberta	roberta	PROPN
easat-10905	166	2	employs	employ	VERB
easat-10905	166	3	dynamic	dynamic	ADJ
easat-10905	166	4	masking	masking	NOUN
easat-10905	166	5	,	,	PUNCT
easat-10905	166	6	which	which	PRON
easat-10905	166	7	modifies	modify	VERB
easat-10905	166	8	the	the	DET
easat-10905	166	9	masked	mask	VERB
easat-10905	166	10	tokens	token	NOUN
easat-10905	166	11	in	in	ADP
easat-10905	166	12	each	each	DET
easat-10905	166	13	epoch	epoch	NOUN
easat-10905	166	14	to	to	PART
easat-10905	166	15	expose	expose	VERB
easat-10905	166	16	the	the	DET
easat-10905	166	17	model	model	NOUN
easat-10905	166	18	to	to	ADP
easat-10905	166	19	a	a	DET
easat-10905	166	20	wider	wide	ADJ
easat-10905	166	21	range	range	NOUN
easat-10905	166	22	of	of	ADP
easat-10905	166	23	text	text	NOUN
easat-10905	166	24	variations	variation	NOUN
easat-10905	166	25	.	.	PUNCT
easat-10905	167	1	in	in	ADP
easat-10905	167	2	addition	addition	NOUN
easat-10905	167	3	,	,	PUNCT
easat-10905	167	4	roberta	roberta	PROPN
easat-10905	167	5	only	only	ADV
easat-10905	167	6	focuses	focus	VERB
easat-10905	167	7	on	on	ADP
easat-10905	167	8	masked	mask	VERB
easat-10905	167	9	language	language	NOUN
easat-10905	167	10	modeling	modeling	NOUN
easat-10905	167	11	and	and	CCONJ
easat-10905	167	12	can	can	AUX
easat-10905	167	13	be	be	AUX
easat-10905	167	14	further	far	ADV
easat-10905	167	15	refined	refine	VERB
easat-10905	167	16	on	on	ADP
easat-10905	167	17	labeled	label	VERB
easat-10905	167	18	datasets	dataset	NOUN
easat-10905	167	19	for	for	ADP
easat-10905	167	20	specialized	specialized	ADJ
easat-10905	167	21	text	text	NOUN
easat-10905	167	22	classification	classification	NOUN
easat-10905	167	23	tasks	task	NOUN
easat-10905	167	24	,	,	PUNCT
easat-10905	167	25	modifying	modify	VERB
easat-10905	167	26	the	the	DET
easat-10905	167	27	pre	pre	ADJ
easat-10905	167	28	-	-	ADJ
easat-10905	167	29	trained	train	VERB
easat-10905	167	30	model	model	NOUN
easat-10905	167	31	to	to	PART
easat-10905	167	32	fit	fit	VERB
easat-10905	167	33	the	the	DET
easat-10905	167	34	requirements	requirement	NOUN
easat-10905	167	35	of	of	ADP
easat-10905	167	36	the	the	DET
easat-10905	167	37	tasks	task	NOUN
easat-10905	167	38	.	.	PUNCT
easat-10905	168	1	roberta	roberta	PROPN
easat-10905	168	2	has	have	AUX
easat-10905	168	3	been	be	AUX
easat-10905	168	4	extensively	extensively	ADV
easat-10905	168	5	pre	pre	VERB
easat-10905	168	6	-	-	VERB
easat-10905	168	7	trained	trained	ADJ
easat-10905	168	8	and	and	CCONJ
easat-10905	168	9	fine	fine	ADV
easat-10905	168	10	-	-	PUNCT
easat-10905	168	11	tuned	tune	VERB
easat-10905	168	12	,	,	PUNCT
easat-10905	168	13	which	which	PRON
easat-10905	168	14	makes	make	VERB
easat-10905	168	15	it	it	PRON
easat-10905	168	16	highly	highly	ADV
easat-10905	168	17	effective	effective	ADJ
easat-10905	168	18	for	for	ADP
easat-10905	168	19	text	text	NOUN
easat-10905	168	20	classification	classification	NOUN
easat-10905	168	21	and	and	CCONJ
easat-10905	168	22	other	other	ADJ
easat-10905	168	23	nlp	nlp	NOUN
easat-10905	168	24	applications	application	NOUN
easat-10905	168	25	.	.	PUNCT
easat-10905	169	1	453	453	NUM
easat-10905	169	2	edelweiss	edelweiss	PROPN
easat-10905	169	3	applied	apply	VERB
easat-10905	169	4	science	science	NOUN
easat-10905	169	5	and	and	CCONJ
easat-10905	169	6	technology	technology	NOUN
easat-10905	169	7	issn	issn	PROPN
easat-10905	169	8	:	:	PUNCT
easat-10905	169	9	2576	2576	NUM
easat-10905	169	10	-	-	SYM
easat-10905	169	11	8484	8484	NUM
easat-10905	169	12	vol	vol	NOUN
easat-10905	169	13	.	.	PROPN
easat-10905	170	1	9	9	NUM
easat-10905	170	2	,	,	PUNCT
easat-10905	170	3	no	no	INTJ
easat-10905	170	4	.	.	NOUN
easat-10905	170	5	11	11	NUM
easat-10905	170	6	:	:	PUNCT
easat-10905	170	7	448	448	NUM
easat-10905	170	8	-	-	SYM
easat-10905	170	9	466	466	NUM
easat-10905	170	10	,	,	PUNCT
easat-10905	170	11	2025	2025	NUM
easat-10905	170	12	doi	doi	NOUN
easat-10905	170	13	:	:	PUNCT
easat-10905	170	14	10.55214/2576	10.55214/2576	NUM
easat-10905	170	15	-	-	PUNCT
easat-10905	170	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	171	1	©	©	PROPN
easat-10905	171	2	2025	2025	NUM
easat-10905	171	3	by	by	ADP
easat-10905	171	4	the	the	DET
easat-10905	171	5	authors	author	NOUN
easat-10905	171	6	;	;	PUNCT
easat-10905	171	7	licensee	licensee	PROPN
easat-10905	171	8	learning	learning	NOUN
easat-10905	171	9	gate	gate	NOUN
easat-10905	171	10	3.3	3.3	NUM
easat-10905	171	11	.	.	PUNCT
easat-10905	172	1	continual	continual	ADJ
easat-10905	172	2	learning	learn	VERB
easat-10905	172	3	it	it	PRON
easat-10905	172	4	refers	refer	VERB
easat-10905	172	5	to	to	ADP
easat-10905	172	6	the	the	DET
easat-10905	172	7	ability	ability	NOUN
easat-10905	172	8	of	of	ADP
easat-10905	172	9	a	a	DET
easat-10905	172	10	model	model	NOUN
easat-10905	172	11	to	to	PART
easat-10905	172	12	learn	learn	VERB
easat-10905	172	13	and	and	CCONJ
easat-10905	172	14	adapt	adapt	VERB
easat-10905	172	15	to	to	ADP
easat-10905	172	16	new	new	ADJ
easat-10905	172	17	tasks	task	NOUN
easat-10905	172	18	over	over	ADP
easat-10905	172	19	time	time	NOUN
easat-10905	172	20	without	without	ADP
easat-10905	172	21	forgetting	forget	VERB
easat-10905	172	22	previously	previously	ADV
easat-10905	172	23	acquired	acquire	VERB
easat-10905	172	24	knowledge	knowledge	NOUN
easat-10905	172	25	[	[	X
easat-10905	172	26	18	18	NUM
easat-10905	172	27	,	,	PUNCT
easat-10905	172	28	36	36	NUM
easat-10905	172	29	]	]	PUNCT
easat-10905	172	30	.	.	PUNCT
easat-10905	173	1	this	this	PRON
easat-10905	173	2	is	be	AUX
easat-10905	173	3	necessary	necessary	ADJ
easat-10905	173	4	in	in	ADP
easat-10905	173	5	dynamic	dynamic	ADJ
easat-10905	173	6	environments	environment	NOUN
easat-10905	173	7	where	where	SCONJ
easat-10905	173	8	data	datum	NOUN
easat-10905	173	9	characteristics	characteristic	NOUN
easat-10905	173	10	or	or	CCONJ
easat-10905	173	11	user	user	NOUN
easat-10905	173	12	needs	need	VERB
easat-10905	173	13	change	change	NOUN
easat-10905	173	14	over	over	ADP
easat-10905	173	15	time	time	NOUN
easat-10905	173	16	.	.	PUNCT
easat-10905	174	1	continual	continual	ADJ
easat-10905	174	2	learning	learning	NOUN
easat-10905	174	3	techniques	technique	NOUN
easat-10905	174	4	aim	aim	VERB
easat-10905	174	5	to	to	PART
easat-10905	174	6	mitigate	mitigate	VERB
easat-10905	174	7	cf	cf	NOUN
easat-10905	174	8	,	,	PUNCT
easat-10905	174	9	where	where	SCONJ
easat-10905	174	10	learning	learn	VERB
easat-10905	174	11	new	new	ADJ
easat-10905	174	12	tasks	task	NOUN
easat-10905	174	13	causes	cause	VERB
easat-10905	174	14	the	the	DET
easat-10905	174	15	model	model	NOUN
easat-10905	174	16	to	to	PART
easat-10905	174	17	lose	lose	VERB
easat-10905	174	18	knowledge	knowledge	NOUN
easat-10905	174	19	of	of	ADP
easat-10905	174	20	previously	previously	ADV
easat-10905	174	21	learned	learn	VERB
easat-10905	174	22	tasks	task	NOUN
easat-10905	174	23	[	[	X
easat-10905	174	24	37	37	NUM
easat-10905	174	25	]	]	PUNCT
easat-10905	174	26	.	.	PUNCT
easat-10905	175	1	there	there	PRON
easat-10905	175	2	are	be	VERB
easat-10905	175	3	three	three	NUM
easat-10905	175	4	types	type	NOUN
easat-10905	175	5	of	of	ADP
easat-10905	175	6	continual	continual	ADJ
easat-10905	175	7	learning	learning	NOUN
easat-10905	175	8	methods	method	NOUN
easat-10905	175	9	,	,	PUNCT
easat-10905	175	10	namely	namely	ADV
easat-10905	175	11	task	task	NOUN
easat-10905	175	12	incremental	incremental	ADJ
easat-10905	175	13	learning	learning	NOUN
easat-10905	175	14	,	,	PUNCT
easat-10905	175	15	domain	domain	VERB
easat-10905	175	16	incremental	incremental	ADJ
easat-10905	175	17	learning	learning	NOUN
easat-10905	175	18	,	,	PUNCT
easat-10905	175	19	and	and	CCONJ
easat-10905	175	20	class	class	NOUN
easat-10905	175	21	incremental	incremental	ADJ
easat-10905	175	22	learning	learning	NOUN
easat-10905	176	1	[	[	X
easat-10905	176	2	16	16	NUM
easat-10905	176	3	,	,	PUNCT
easat-10905	176	4	36	36	NUM
easat-10905	176	5	,	,	PUNCT
easat-10905	176	6	38	38	NUM
easat-10905	176	7	]	]	PUNCT
easat-10905	176	8	,	,	PUNCT
easat-10905	176	9	which	which	PRON
easat-10905	176	10	are	be	AUX
easat-10905	176	11	briefly	briefly	ADV
easat-10905	176	12	introduced	introduce	VERB
easat-10905	176	13	below	below	ADV
easat-10905	176	14	.	.	PUNCT
easat-10905	177	1	task	task	PROPN
easat-10905	177	2	incremental	incremental	ADJ
easat-10905	177	3	learning	learning	NOUN
easat-10905	177	4	:	:	PUNCT
easat-10905	177	5	in	in	ADP
easat-10905	177	6	this	this	DET
easat-10905	177	7	setting	setting	NOUN
easat-10905	177	8	,	,	PUNCT
easat-10905	177	9	tasks	task	NOUN
easat-10905	177	10	are	be	AUX
easat-10905	177	11	learned	learn	VERB
easat-10905	177	12	sequentially	sequentially	ADV
easat-10905	177	13	,	,	PUNCT
easat-10905	177	14	and	and	CCONJ
easat-10905	177	15	the	the	DET
easat-10905	177	16	models	model	NOUN
easat-10905	177	17	are	be	AUX
easat-10905	177	18	aware	aware	ADJ
easat-10905	177	19	of	of	ADP
easat-10905	177	20	the	the	DET
easat-10905	177	21	task	task	NOUN
easat-10905	177	22	they	they	PRON
easat-10905	177	23	are	be	AUX
easat-10905	177	24	currently	currently	ADV
easat-10905	177	25	working	work	VERB
easat-10905	177	26	on	on	ADP
easat-10905	177	27	[	[	X
easat-10905	177	28	20	20	NUM
easat-10905	177	29	,	,	PUNCT
easat-10905	177	30	38	38	NUM
easat-10905	177	31	-	-	SYM
easat-10905	177	32	40	40	NUM
easat-10905	177	33	]	]	PUNCT
easat-10905	177	34	.	.	PUNCT
easat-10905	178	1	task	task	NOUN
easat-10905	178	2	labels	label	NOUN
easat-10905	178	3	are	be	AUX
easat-10905	178	4	provided	provide	VERB
easat-10905	178	5	during	during	ADP
easat-10905	178	6	both	both	DET
easat-10905	178	7	training	training	NOUN
easat-10905	178	8	and	and	CCONJ
easat-10905	178	9	testing	testing	NOUN
easat-10905	178	10	.	.	PUNCT
easat-10905	179	1	for	for	ADP
easat-10905	179	2	instance	instance	NOUN
easat-10905	179	3	,	,	PUNCT
easat-10905	179	4	a	a	DET
easat-10905	179	5	model	model	NOUN
easat-10905	179	6	first	first	ADV
easat-10905	179	7	learns	learn	VERB
easat-10905	179	8	to	to	PART
easat-10905	179	9	classify	classify	VERB
easat-10905	179	10	sentiments	sentiment	NOUN
easat-10905	179	11	in	in	ADP
easat-10905	179	12	movie	movie	NOUN
easat-10905	179	13	reviews	review	NOUN
easat-10905	179	14	,	,	PUNCT
easat-10905	179	15	then	then	ADV
easat-10905	179	16	learns	learn	VERB
easat-10905	179	17	to	to	PART
easat-10905	179	18	classify	classify	VERB
easat-10905	179	19	news	news	NOUN
easat-10905	179	20	articles	article	NOUN
easat-10905	179	21	by	by	ADP
easat-10905	179	22	topic	topic	NOUN
easat-10905	179	23	,	,	PUNCT
easat-10905	179	24	and	and	CCONJ
easat-10905	179	25	finally	finally	ADV
easat-10905	179	26	learns	learn	VERB
easat-10905	179	27	to	to	PART
easat-10905	179	28	detect	detect	VERB
easat-10905	179	29	spam	spam	NOUN
easat-10905	179	30	in	in	ADP
easat-10905	179	31	emails	email	NOUN
easat-10905	179	32	.	.	PUNCT
easat-10905	180	1	domain	domain	NOUN
easat-10905	180	2	incremental	incremental	ADJ
easat-10905	180	3	learning	learning	NOUN
easat-10905	180	4	:	:	PUNCT
easat-10905	180	5	in	in	ADP
easat-10905	180	6	this	this	DET
easat-10905	180	7	learning	learning	NOUN
easat-10905	180	8	,	,	PUNCT
easat-10905	180	9	the	the	DET
easat-10905	180	10	same	same	ADJ
easat-10905	180	11	task	task	NOUN
easat-10905	180	12	is	be	AUX
easat-10905	180	13	presented	present	VERB
easat-10905	180	14	in	in	ADP
easat-10905	180	15	different	different	ADJ
easat-10905	180	16	domains	domain	NOUN
easat-10905	180	17	,	,	PUNCT
easat-10905	180	18	but	but	CCONJ
easat-10905	180	19	the	the	DET
easat-10905	180	20	models	model	NOUN
easat-10905	180	21	do	do	AUX
easat-10905	180	22	not	not	PART
easat-10905	180	23	know	know	VERB
easat-10905	180	24	which	which	DET
easat-10905	180	25	domain	domain	NOUN
easat-10905	180	26	they	they	PRON
easat-10905	180	27	are	be	AUX
easat-10905	180	28	currently	currently	ADV
easat-10905	180	29	classifying	classify	VERB
easat-10905	180	30	[	[	X
easat-10905	180	31	38	38	NUM
easat-10905	180	32	,	,	PUNCT
easat-10905	180	33	41	41	NUM
easat-10905	180	34	]	]	PUNCT
easat-10905	180	35	.	.	PUNCT
easat-10905	181	1	the	the	DET
easat-10905	181	2	model	model	NOUN
easat-10905	181	3	must	must	AUX
easat-10905	181	4	generalize	generalize	VERB
easat-10905	181	5	across	across	ADP
easat-10905	181	6	domains	domain	NOUN
easat-10905	181	7	while	while	SCONJ
easat-10905	181	8	learning	learn	VERB
easat-10905	181	9	to	to	PART
easat-10905	181	10	differentiate	differentiate	VERB
easat-10905	181	11	between	between	ADP
easat-10905	181	12	them	they	PRON
easat-10905	181	13	.	.	PUNCT
easat-10905	182	1	for	for	ADP
easat-10905	182	2	example	example	NOUN
easat-10905	182	3	,	,	PUNCT
easat-10905	182	4	a	a	DET
easat-10905	182	5	sentiment	sentiment	NOUN
easat-10905	182	6	analysis	analysis	NOUN
easat-10905	182	7	model	model	NOUN
easat-10905	182	8	learns	learn	VERB
easat-10905	182	9	to	to	PART
easat-10905	182	10	classify	classify	VERB
easat-10905	182	11	text	text	NOUN
easat-10905	182	12	as	as	ADP
easat-10905	182	13	positive	positive	ADJ
easat-10905	182	14	or	or	CCONJ
easat-10905	182	15	negative	negative	ADJ
easat-10905	182	16	across	across	ADP
easat-10905	182	17	different	different	ADJ
easat-10905	182	18	domains	domain	NOUN
easat-10905	182	19	,	,	PUNCT
easat-10905	182	20	such	such	ADJ
easat-10905	182	21	as	as	ADP
easat-10905	182	22	product	product	NOUN
easat-10905	182	23	reviews	review	NOUN
easat-10905	182	24	,	,	PUNCT
easat-10905	182	25	movie	movie	NOUN
easat-10905	182	26	reviews	review	NOUN
easat-10905	182	27	,	,	PUNCT
easat-10905	182	28	and	and	CCONJ
easat-10905	182	29	social	social	ADJ
easat-10905	182	30	media	medium	NOUN
easat-10905	182	31	posts	post	NOUN
easat-10905	182	32	.	.	PUNCT
easat-10905	183	1	class	class	NOUN
easat-10905	183	2	incremental	incremental	ADJ
easat-10905	183	3	learning	learning	NOUN
easat-10905	183	4	:	:	PUNCT
easat-10905	183	5	in	in	ADP
easat-10905	183	6	class	class	NOUN
easat-10905	183	7	incremental	incremental	ADJ
easat-10905	183	8	learning	learning	NOUN
easat-10905	183	9	,	,	PUNCT
easat-10905	183	10	new	new	ADJ
easat-10905	183	11	classes	class	NOUN
easat-10905	183	12	are	be	AUX
easat-10905	183	13	introduced	introduce	VERB
easat-10905	183	14	sequentially	sequentially	ADV
easat-10905	183	15	,	,	PUNCT
easat-10905	183	16	and	and	CCONJ
easat-10905	183	17	the	the	DET
easat-10905	183	18	models	model	NOUN
easat-10905	183	19	must	must	AUX
easat-10905	183	20	learn	learn	VERB
easat-10905	183	21	to	to	PART
easat-10905	183	22	classify	classify	VERB
easat-10905	183	23	them	they	PRON
easat-10905	183	24	alongside	alongside	ADP
easat-10905	183	25	previously	previously	ADV
easat-10905	183	26	learned	learn	VERB
easat-10905	183	27	classes	class	NOUN
easat-10905	183	28	[	[	X
easat-10905	183	29	36	36	NUM
easat-10905	183	30	,	,	PUNCT
easat-10905	183	31	38	38	NUM
easat-10905	183	32	,	,	PUNCT
easat-10905	183	33	40	40	NUM
easat-10905	183	34	]	]	PUNCT
easat-10905	183	35	.	.	PUNCT
easat-10905	184	1	a	a	DET
easat-10905	184	2	model	model	NOUN
easat-10905	184	3	learns	learn	VERB
easat-10905	184	4	to	to	PART
easat-10905	184	5	classify	classify	VERB
easat-10905	184	6	animals	animal	NOUN
easat-10905	184	7	in	in	ADP
easat-10905	184	8	a	a	DET
easat-10905	184	9	sequence	sequence	NOUN
easat-10905	184	10	,	,	PUNCT
easat-10905	184	11	first	first	ADV
easat-10905	184	12	learning	learn	VERB
easat-10905	184	13	to	to	PART
easat-10905	184	14	identify	identify	VERB
easat-10905	184	15	cats	cat	NOUN
easat-10905	184	16	and	and	CCONJ
easat-10905	184	17	dogs	dog	NOUN
easat-10905	184	18	,	,	PUNCT
easat-10905	184	19	then	then	ADV
easat-10905	184	20	learning	learn	VERB
easat-10905	184	21	to	to	PART
easat-10905	184	22	identify	identify	VERB
easat-10905	184	23	tigers	tiger	NOUN
easat-10905	184	24	and	and	CCONJ
easat-10905	184	25	so	so	ADV
easat-10905	184	26	on	on	ADV
easat-10905	184	27	.	.	PUNCT
easat-10905	185	1	in	in	ADP
easat-10905	185	2	our	our	PRON
easat-10905	185	3	work	work	NOUN
easat-10905	185	4	,	,	PUNCT
easat-10905	185	5	we	we	PRON
easat-10905	185	6	employ	employ	VERB
easat-10905	185	7	task	task	NOUN
easat-10905	185	8	incremental	incremental	ADJ
easat-10905	185	9	learning	learning	NOUN
easat-10905	185	10	to	to	PART
easat-10905	185	11	first	first	ADV
easat-10905	185	12	learn	learn	VERB
easat-10905	185	13	hope	hope	NOUN
easat-10905	185	14	speech	speech	NOUN
easat-10905	185	15	,	,	PUNCT
easat-10905	185	16	followed	follow	VERB
easat-10905	185	17	by	by	ADP
easat-10905	185	18	hate	hate	NOUN
easat-10905	185	19	speech	speech	NOUN
easat-10905	185	20	sequentially	sequentially	ADV
easat-10905	185	21	.	.	PUNCT
easat-10905	186	1	the	the	DET
easat-10905	186	2	main	main	ADJ
easat-10905	186	3	features	feature	NOUN
easat-10905	186	4	of	of	ADP
easat-10905	186	5	task	task	NOUN
easat-10905	186	6	incremental	incremental	ADJ
easat-10905	186	7	learning	learning	NOUN
easat-10905	186	8	are	be	AUX
easat-10905	186	9	outlined	outline	VERB
easat-10905	186	10	as	as	SCONJ
easat-10905	186	11	follows	follow	VERB
easat-10905	186	12	:	:	PUNCT
easat-10905	186	13	•	•	NUM
easat-10905	186	14	sequential	sequential	ADJ
easat-10905	186	15	learning	learning	NOUN
easat-10905	186	16	:	:	PUNCT
easat-10905	186	17	the	the	DET
easat-10905	186	18	model	model	NOUN
easat-10905	186	19	is	be	AUX
easat-10905	186	20	trained	train	VERB
easat-10905	186	21	on	on	ADP
easat-10905	186	22	tasks	task	NOUN
easat-10905	186	23	in	in	ADP
easat-10905	186	24	a	a	DET
easat-10905	186	25	sequence	sequence	NOUN
easat-10905	186	26	,	,	PUNCT
easat-10905	186	27	where	where	SCONJ
easat-10905	186	28	each	each	DET
easat-10905	186	29	task	task	NOUN
easat-10905	186	30	comprises	comprise	VERB
easat-10905	186	31	a	a	DET
easat-10905	186	32	distinct	distinct	ADJ
easat-10905	186	33	set	set	NOUN
easat-10905	186	34	of	of	ADP
easat-10905	186	35	classes	class	NOUN
easat-10905	186	36	.	.	PUNCT
easat-10905	187	1	•	•	NUM
easat-10905	187	2	task	task	NOUN
easat-10905	187	3	identifiers	identifier	NOUN
easat-10905	187	4	(	(	PUNCT
easat-10905	187	5	task	task	NOUN
easat-10905	187	6	-	-	PUNCT
easat-10905	187	7	ids	id	NOUN
easat-10905	187	8	):	):	PUNCT
easat-10905	187	9	unique	unique	ADJ
easat-10905	187	10	identifiers	identifier	NOUN
easat-10905	187	11	,	,	PUNCT
easat-10905	187	12	known	know	VERB
easat-10905	187	13	as	as	ADP
easat-10905	187	14	task	task	NOUN
easat-10905	187	15	-	-	PUNCT
easat-10905	187	16	ids	id	NOUN
easat-10905	187	17	,	,	PUNCT
easat-10905	187	18	are	be	AUX
easat-10905	187	19	assigned	assign	VERB
easat-10905	187	20	to	to	ADP
easat-10905	187	21	each	each	DET
easat-10905	187	22	task	task	NOUN
easat-10905	187	23	.	.	PUNCT
easat-10905	188	1	these	these	DET
easat-10905	188	2	identifiers	identifier	NOUN
easat-10905	188	3	are	be	AUX
easat-10905	188	4	accessible	accessible	ADJ
easat-10905	188	5	during	during	ADP
easat-10905	188	6	both	both	CCONJ
easat-10905	188	7	the	the	DET
easat-10905	188	8	training	training	NOUN
easat-10905	188	9	and	and	CCONJ
easat-10905	188	10	testing	testing	NOUN
easat-10905	188	11	phases	phase	NOUN
easat-10905	188	12	,	,	PUNCT
easat-10905	188	13	and	and	CCONJ
easat-10905	188	14	the	the	DET
easat-10905	188	15	classes	class	NOUN
easat-10905	188	16	across	across	ADP
easat-10905	188	17	tasks	task	NOUN
easat-10905	188	18	may	may	AUX
easat-10905	188	19	or	or	CCONJ
easat-10905	188	20	may	may	AUX
easat-10905	188	21	not	not	PART
easat-10905	188	22	overlap	overlap	VERB
easat-10905	188	23	.	.	PUNCT
easat-10905	189	1	•	•	NUM
easat-10905	189	2	training	training	NOUN
easat-10905	189	3	and	and	CCONJ
easat-10905	189	4	testing	testing	NOUN
easat-10905	189	5	:	:	PUNCT
easat-10905	189	6	during	during	ADP
easat-10905	189	7	training	training	NOUN
easat-10905	189	8	,	,	PUNCT
easat-10905	189	9	the	the	DET
easat-10905	189	10	model	model	NOUN
easat-10905	189	11	sequentially	sequentially	ADV
easat-10905	189	12	learns	learn	VERB
easat-10905	189	13	from	from	ADP
easat-10905	189	14	the	the	DET
easat-10905	189	15	data	datum	NOUN
easat-10905	189	16	associated	associate	VERB
easat-10905	189	17	with	with	ADP
easat-10905	189	18	each	each	DET
easat-10905	189	19	task	task	NOUN
easat-10905	189	20	.	.	PUNCT
easat-10905	190	1	during	during	ADP
easat-10905	190	2	testing	testing	NOUN
easat-10905	190	3	,	,	PUNCT
easat-10905	190	4	the	the	DET
easat-10905	190	5	model	model	NOUN
easat-10905	190	6	uses	use	VERB
easat-10905	190	7	the	the	DET
easat-10905	190	8	provided	provide	VERB
easat-10905	190	9	task	task	NOUN
easat-10905	190	10	-	-	PUNCT
easat-10905	190	11	id	id	NOUN
easat-10905	190	12	to	to	PART
easat-10905	190	13	classify	classify	VERB
easat-10905	190	14	the	the	DET
easat-10905	190	15	relevant	relevant	ADJ
easat-10905	190	16	task	task	NOUN
easat-10905	190	17	.	.	PUNCT
easat-10905	191	1	•	•	NUM
easat-10905	191	2	class	class	NOUN
easat-10905	191	3	overlap	overlap	NOUN
easat-10905	191	4	:	:	PUNCT
easat-10905	191	5	the	the	DET
easat-10905	191	6	classes	class	NOUN
easat-10905	191	7	across	across	ADP
easat-10905	191	8	different	different	ADJ
easat-10905	191	9	tasks	task	NOUN
easat-10905	191	10	can	can	AUX
easat-10905	191	11	either	either	CCONJ
easat-10905	191	12	be	be	AUX
easat-10905	191	13	disjoint	disjoint	ADJ
easat-10905	191	14	or	or	CCONJ
easat-10905	191	15	overlapping	overlap	VERB
easat-10905	191	16	.	.	PUNCT
easat-10905	192	1	disjoint	disjoint	NOUN
easat-10905	192	2	tasks	task	NOUN
easat-10905	192	3	have	have	VERB
easat-10905	192	4	entirely	entirely	ADV
easat-10905	192	5	separate	separate	ADJ
easat-10905	192	6	classes	class	NOUN
easat-10905	192	7	,	,	PUNCT
easat-10905	192	8	while	while	SCONJ
easat-10905	192	9	non	non	ADJ
easat-10905	192	10	-	-	ADJ
easat-10905	192	11	disjoint	disjoint	ADJ
easat-10905	192	12	tasks	task	NOUN
easat-10905	192	13	may	may	AUX
easat-10905	192	14	share	share	VERB
easat-10905	192	15	some	some	DET
easat-10905	192	16	classes	class	NOUN
easat-10905	192	17	.	.	PUNCT
easat-10905	193	1	3.3.1	3.3.1	X
easat-10905	193	2	.	.	PUNCT
easat-10905	193	3	task	task	NOUN
easat-10905	193	4	incremental	incremental	ADJ
easat-10905	193	5	learning	learning	NOUN
easat-10905	193	6	in	in	ADP
easat-10905	193	7	task	task	NOUN
easat-10905	193	8	incremental	incremental	ADJ
easat-10905	193	9	learning	learning	NOUN
easat-10905	193	10	,	,	PUNCT
easat-10905	193	11	cf	cf	NOUN
easat-10905	193	12	refers	refer	VERB
easat-10905	193	13	to	to	ADP
easat-10905	193	14	the	the	DET
easat-10905	193	15	phenomenon	phenomenon	NOUN
easat-10905	193	16	where	where	SCONJ
easat-10905	193	17	a	a	DET
easat-10905	193	18	model	model	NOUN
easat-10905	193	19	,	,	PUNCT
easat-10905	193	20	when	when	SCONJ
easat-10905	193	21	trained	train	VERB
easat-10905	193	22	on	on	ADP
easat-10905	193	23	new	new	ADJ
easat-10905	193	24	tasks	task	NOUN
easat-10905	193	25	,	,	PUNCT
easat-10905	193	26	forgets	forget	VERB
easat-10905	193	27	the	the	DET
easat-10905	193	28	knowledge	knowledge	NOUN
easat-10905	193	29	it	it	PRON
easat-10905	193	30	acquired	acquire	VERB
easat-10905	193	31	from	from	ADP
easat-10905	193	32	previous	previous	ADJ
easat-10905	193	33	tasks	task	NOUN
easat-10905	193	34	.	.	PUNCT
easat-10905	194	1	this	this	PRON
easat-10905	194	2	occurs	occur	VERB
easat-10905	194	3	because	because	SCONJ
easat-10905	194	4	the	the	DET
easat-10905	194	5	model	model	NOUN
easat-10905	194	6	updates	update	VERB
easat-10905	194	7	its	its	PRON
easat-10905	194	8	parameters	parameter	NOUN
easat-10905	194	9	to	to	PART
easat-10905	194	10	learn	learn	VERB
easat-10905	194	11	for	for	ADP
easat-10905	194	12	the	the	DET
easat-10905	194	13	new	new	ADJ
easat-10905	194	14	task	task	NOUN
easat-10905	194	15	,	,	PUNCT
easat-10905	194	16	at	at	ADP
easat-10905	194	17	the	the	DET
easat-10905	194	18	cost	cost	NOUN
easat-10905	194	19	of	of	ADP
easat-10905	194	20	performance	performance	NOUN
easat-10905	194	21	on	on	ADP
easat-10905	194	22	earlier	early	ADJ
easat-10905	194	23	tasks	task	NOUN
easat-10905	194	24	.	.	PUNCT
easat-10905	195	1	cf	cf	NOUN
easat-10905	195	2	prevention	prevention	NOUN
easat-10905	195	3	involves	involve	VERB
easat-10905	195	4	decreasing	decrease	VERB
easat-10905	195	5	the	the	DET
easat-10905	195	6	amount	amount	NOUN
easat-10905	195	7	of	of	ADP
easat-10905	195	8	parameter	parameter	NOUN
easat-10905	195	9	sharing	sharing	NOUN
easat-10905	195	10	between	between	ADP
easat-10905	195	11	tasks	task	NOUN
easat-10905	195	12	or	or	CCONJ
easat-10905	195	13	avoiding	avoid	VERB
easat-10905	195	14	significant	significant	ADJ
easat-10905	195	15	changes	change	NOUN
easat-10905	195	16	to	to	ADP
easat-10905	195	17	previously	previously	ADV
easat-10905	195	18	learned	learn	VERB
easat-10905	195	19	parameters	parameter	NOUN
easat-10905	195	20	.	.	PUNCT
easat-10905	196	1	addressing	address	VERB
easat-10905	196	2	cf	cf	NOUN
easat-10905	196	3	is	be	AUX
easat-10905	196	4	important	important	ADJ
easat-10905	196	5	in	in	ADP
easat-10905	196	6	task	task	NOUN
easat-10905	196	7	incremental	incremental	ADJ
easat-10905	196	8	learning	learning	NOUN
easat-10905	196	9	because	because	SCONJ
easat-10905	196	10	it	it	PRON
easat-10905	196	11	ensures	ensure	VERB
easat-10905	196	12	that	that	SCONJ
easat-10905	196	13	the	the	DET
easat-10905	196	14	model	model	NOUN
easat-10905	196	15	retains	retain	VERB
easat-10905	196	16	knowledge	knowledge	NOUN
easat-10905	196	17	from	from	ADP
easat-10905	196	18	all	all	DET
easat-10905	196	19	previously	previously	ADV
easat-10905	196	20	learned	learn	VERB
easat-10905	196	21	tasks	task	NOUN
easat-10905	196	22	.	.	PUNCT
easat-10905	197	1	regularization	regularization	NOUN
easat-10905	197	2	,	,	PUNCT
easat-10905	197	3	replay	replay	NOUN
easat-10905	197	4	approaches	approach	NOUN
easat-10905	197	5	,	,	PUNCT
easat-10905	197	6	and	and	CCONJ
easat-10905	197	7	parameter	parameter	NOUN
easat-10905	197	8	isolation	isolation	NOUN
easat-10905	197	9	are	be	AUX
easat-10905	197	10	generally	generally	ADV
easat-10905	197	11	used	use	VERB
easat-10905	197	12	to	to	PART
easat-10905	197	13	prevent	prevent	VERB
easat-10905	197	14	cf	cf	NOUN
easat-10905	197	15	,	,	PUNCT
easat-10905	197	16	allowing	allow	VERB
easat-10905	197	17	the	the	DET
easat-10905	197	18	model	model	NOUN
easat-10905	197	19	to	to	PART
easat-10905	197	20	learn	learn	VERB
easat-10905	197	21	new	new	ADJ
easat-10905	197	22	tasks	task	NOUN
easat-10905	197	23	while	while	SCONJ
easat-10905	197	24	maintaining	maintain	VERB
easat-10905	197	25	its	its	PRON
easat-10905	197	26	ability	ability	NOUN
easat-10905	197	27	to	to	PART
easat-10905	197	28	perform	perform	VERB
easat-10905	197	29	well	well	ADV
easat-10905	197	30	on	on	ADP
easat-10905	197	31	previously	previously	ADV
easat-10905	197	32	learned	learn	VERB
easat-10905	197	33	tasks	task	NOUN
easat-10905	198	1	[	[	X
easat-10905	198	2	16	16	NUM
easat-10905	198	3	,	,	PUNCT
easat-10905	198	4	37	37	NUM
easat-10905	198	5	,	,	PUNCT
easat-10905	198	6	42	42	NUM
easat-10905	198	7	,	,	PUNCT
easat-10905	198	8	43	43	NUM
easat-10905	198	9	]	]	PUNCT
easat-10905	198	10	.	.	PUNCT
easat-10905	199	1	in	in	ADP
easat-10905	199	2	this	this	DET
easat-10905	199	3	work	work	NOUN
easat-10905	199	4	,	,	PUNCT
easat-10905	199	5	we	we	PRON
easat-10905	199	6	propose	propose	VERB
easat-10905	199	7	using	use	VERB
easat-10905	199	8	two	two	NUM
easat-10905	199	9	task	task	NOUN
easat-10905	199	10	incremental	incremental	ADJ
easat-10905	199	11	learning	learning	NOUN
easat-10905	199	12	approaches	approach	NOUN
easat-10905	199	13	,	,	PUNCT
easat-10905	199	14	namely	namely	ADV
easat-10905	199	15	,	,	PUNCT
easat-10905	199	16	replay	replay	NOUN
easat-10905	199	17	-	-	PUNCT
easat-10905	199	18	based	base	VERB
easat-10905	199	19	and	and	CCONJ
easat-10905	199	20	regularization	regularization	NOUN
easat-10905	199	21	-	-	PUNCT
easat-10905	199	22	based	base	VERB
easat-10905	199	23	approaches	approach	NOUN
easat-10905	199	24	,	,	PUNCT
easat-10905	199	25	to	to	PART
easat-10905	199	26	address	address	VERB
easat-10905	199	27	cf	cf	NOUN
easat-10905	199	28	in	in	ADP
easat-10905	199	29	the	the	DET
easat-10905	199	30	developed	develop	VERB
easat-10905	199	31	models	model	NOUN
easat-10905	199	32	.	.	PUNCT
easat-10905	200	1	to	to	PART
easat-10905	200	2	test	test	VERB
easat-10905	200	3	the	the	DET
easat-10905	200	4	incremental	incremental	ADJ
easat-10905	200	5	learning	learning	NOUN
easat-10905	200	6	capability	capability	NOUN
easat-10905	200	7	of	of	ADP
easat-10905	200	8	the	the	DET
easat-10905	200	9	models	model	NOUN
easat-10905	200	10	,	,	PUNCT
easat-10905	200	11	we	we	PRON
easat-10905	200	12	sequentially	sequentially	ADV
easat-10905	200	13	introduce	introduce	VERB
easat-10905	200	14	two	two	NUM
easat-10905	200	15	tasks	task	NOUN
easat-10905	200	16	:	:	PUNCT
easat-10905	200	17	hope	hope	NOUN
easat-10905	200	18	speech	speech	NOUN
easat-10905	200	19	,	,	PUNCT
easat-10905	200	20	followed	follow	VERB
easat-10905	200	21	by	by	ADP
easat-10905	200	22	hate	hate	NOUN
easat-10905	200	23	speech	speech	NOUN
easat-10905	200	24	.	.	PUNCT
easat-10905	201	1	below	below	ADV
easat-10905	201	2	,	,	PUNCT
easat-10905	201	3	we	we	PRON
easat-10905	201	4	summarize	summarize	VERB
easat-10905	201	5	the	the	DET
easat-10905	201	6	steps	step	NOUN
easat-10905	201	7	in	in	ADP
easat-10905	201	8	task	task	NOUN
easat-10905	201	9	incremental	incremental	ADJ
easat-10905	201	10	learning	learning	NOUN
easat-10905	201	11	approaches	approach	NOUN
easat-10905	201	12	.	.	PUNCT
easat-10905	202	1	3.3.2	3.3.2	X
easat-10905	202	2	.	.	PUNCT
easat-10905	202	3	replay	replay	NOUN
easat-10905	202	4	-	-	PUNCT
easat-10905	202	5	based	base	VERB
easat-10905	202	6	task	task	NOUN
easat-10905	202	7	incremental	incremental	ADJ
easat-10905	202	8	learning	learning	NOUN
easat-10905	202	9	this	this	PRON
easat-10905	202	10	is	be	AUX
easat-10905	202	11	an	an	DET
easat-10905	202	12	approach	approach	NOUN
easat-10905	202	13	to	to	ADP
easat-10905	202	14	continual	continual	ADJ
easat-10905	202	15	learning	learning	NOUN
easat-10905	202	16	where	where	SCONJ
easat-10905	202	17	a	a	DET
easat-10905	202	18	model	model	NOUN
easat-10905	202	19	is	be	AUX
easat-10905	202	20	trained	train	VERB
easat-10905	202	21	on	on	ADP
easat-10905	202	22	new	new	ADJ
easat-10905	202	23	tasks	task	NOUN
easat-10905	202	24	while	while	SCONJ
easat-10905	202	25	simultaneously	simultaneously	ADV
easat-10905	202	26	using	use	VERB
easat-10905	202	27	or	or	CCONJ
easat-10905	202	28	replaying	replay	VERB
easat-10905	202	29	data	datum	NOUN
easat-10905	202	30	from	from	ADP
easat-10905	202	31	previously	previously	ADV
easat-10905	202	32	learned	learn	VERB
easat-10905	202	33	tasks	task	NOUN
easat-10905	202	34	.	.	PUNCT
easat-10905	203	1	there	there	PRON
easat-10905	203	2	are	be	VERB
easat-10905	203	3	two	two	NUM
easat-10905	203	4	types	type	NOUN
easat-10905	203	5	of	of	ADP
easat-10905	203	6	replaybased	replaybase	VERB
easat-10905	203	7	methods	method	NOUN
easat-10905	203	8	,	,	PUNCT
easat-10905	203	9	namely	namely	ADV
easat-10905	203	10	rehearsal	rehearsal	NOUN
easat-10905	203	11	and	and	CCONJ
easat-10905	203	12	pseudo	pseudo	NOUN
easat-10905	203	13	-	-	NOUN
easat-10905	203	14	rehearsal	rehearsal	NOUN
easat-10905	203	15	.	.	PUNCT
easat-10905	204	1	454	454	NUM
easat-10905	204	2	edelweiss	edelweiss	PROPN
easat-10905	204	3	applied	apply	VERB
easat-10905	204	4	science	science	NOUN
easat-10905	204	5	and	and	CCONJ
easat-10905	204	6	technology	technology	NOUN
easat-10905	204	7	issn	issn	PROPN
easat-10905	204	8	:	:	PUNCT
easat-10905	204	9	2576	2576	NUM
easat-10905	204	10	-	-	SYM
easat-10905	204	11	8484	8484	NUM
easat-10905	204	12	vol	vol	NOUN
easat-10905	204	13	.	.	PROPN
easat-10905	205	1	9	9	NUM
easat-10905	205	2	,	,	PUNCT
easat-10905	205	3	no	no	INTJ
easat-10905	205	4	.	.	NOUN
easat-10905	205	5	11	11	NUM
easat-10905	205	6	:	:	PUNCT
easat-10905	205	7	448	448	NUM
easat-10905	205	8	-	-	SYM
easat-10905	205	9	466	466	NUM
easat-10905	205	10	,	,	PUNCT
easat-10905	205	11	2025	2025	NUM
easat-10905	205	12	doi	doi	NOUN
easat-10905	205	13	:	:	PUNCT
easat-10905	205	14	10.55214/2576	10.55214/2576	NUM
easat-10905	205	15	-	-	PUNCT
easat-10905	205	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	206	1	©	©	PROPN
easat-10905	206	2	2025	2025	NUM
easat-10905	206	3	by	by	ADP
easat-10905	206	4	the	the	DET
easat-10905	206	5	authors	author	NOUN
easat-10905	206	6	;	;	PUNCT
easat-10905	206	7	licensee	licensee	PROPN
easat-10905	206	8	learning	learn	VERB
easat-10905	206	9	gate	gate	NOUN
easat-10905	206	10	3.3.2.1	3.3.2.1	NUM
easat-10905	206	11	.	.	PUNCT
easat-10905	207	1	rehearsal	rehearsal	NOUN
easat-10905	207	2	approach	approach	NOUN
easat-10905	207	3	for	for	ADP
easat-10905	207	4	task	task	NOUN
easat-10905	207	5	incremental	incremental	ADJ
easat-10905	207	6	learning	learn	VERB
easat-10905	207	7	the	the	DET
easat-10905	207	8	rehearsal	rehearsal	NOUN
easat-10905	207	9	approach	approach	NOUN
easat-10905	207	10	for	for	ADP
easat-10905	207	11	task	task	NOUN
easat-10905	207	12	incremental	incremental	ADJ
easat-10905	207	13	learning	learning	NOUN
easat-10905	207	14	is	be	AUX
easat-10905	207	15	designed	design	VERB
easat-10905	207	16	to	to	PART
easat-10905	207	17	reduce	reduce	VERB
easat-10905	207	18	catastrophic	catastrophic	ADJ
easat-10905	207	19	forgetting	forgetting	NOUN
easat-10905	207	20	(	(	PUNCT
easat-10905	207	21	cf	cf	NOUN
easat-10905	207	22	)	)	PUNCT
easat-10905	207	23	by	by	ADP
easat-10905	207	24	revisiting	revisit	VERB
easat-10905	207	25	a	a	DET
easat-10905	207	26	subset	subset	NOUN
easat-10905	207	27	of	of	ADP
easat-10905	207	28	data	datum	NOUN
easat-10905	207	29	from	from	ADP
easat-10905	207	30	previously	previously	ADV
easat-10905	207	31	learned	learn	VERB
easat-10905	207	32	tasks	task	NOUN
easat-10905	207	33	.	.	PUNCT
easat-10905	208	1	the	the	DET
easat-10905	208	2	objective	objective	NOUN
easat-10905	208	3	is	be	AUX
easat-10905	208	4	to	to	PART
easat-10905	208	5	combine	combine	VERB
easat-10905	208	6	the	the	DET
easat-10905	208	7	loss	loss	NOUN
easat-10905	208	8	from	from	ADP
easat-10905	208	9	the	the	DET
easat-10905	208	10	current	current	ADJ
easat-10905	208	11	task	task	NOUN
easat-10905	208	12	and	and	CCONJ
easat-10905	208	13	the	the	DET
easat-10905	208	14	loss	loss	NOUN
easat-10905	208	15	from	from	ADP
easat-10905	208	16	the	the	DET
easat-10905	208	17	replayed	replayed	ADJ
easat-10905	208	18	data	datum	NOUN
easat-10905	208	19	.	.	PUNCT
easat-10905	209	1	the	the	DET
easat-10905	209	2	significant	significant	ADJ
easat-10905	209	3	components	component	NOUN
easat-10905	209	4	of	of	ADP
easat-10905	209	5	this	this	DET
easat-10905	209	6	approach	approach	NOUN
easat-10905	209	7	include	include	VERB
easat-10905	209	8	:	:	PUNCT
easat-10905	209	9	storing	store	VERB
easat-10905	209	10	of	of	ADP
easat-10905	209	11	data	datum	NOUN
easat-10905	209	12	:	:	PUNCT
easat-10905	209	13	a	a	DET
easat-10905	209	14	small	small	ADJ
easat-10905	209	15	buffer	buffer	NOUN
easat-10905	209	16	stores	store	NOUN
easat-10905	209	17	a	a	DET
easat-10905	209	18	sample	sample	NOUN
easat-10905	209	19	of	of	ADP
easat-10905	209	20	data	datum	NOUN
easat-10905	209	21	from	from	ADP
easat-10905	209	22	each	each	DET
easat-10905	209	23	task	task	NOUN
easat-10905	209	24	the	the	DET
easat-10905	209	25	model	model	NOUN
easat-10905	209	26	has	have	AUX
easat-10905	209	27	already	already	ADV
easat-10905	209	28	learned	learn	VERB
easat-10905	209	29	.	.	PUNCT
easat-10905	210	1	this	this	DET
easat-10905	210	2	sample	sample	NOUN
easat-10905	210	3	is	be	AUX
easat-10905	210	4	carefully	carefully	ADV
easat-10905	210	5	chosen	choose	VERB
easat-10905	210	6	so	so	SCONJ
easat-10905	210	7	that	that	SCONJ
easat-10905	210	8	it	it	PRON
easat-10905	210	9	represents	represent	VERB
easat-10905	210	10	the	the	DET
easat-10905	210	11	most	most	ADV
easat-10905	210	12	essential	essential	ADJ
easat-10905	210	13	aspects	aspect	NOUN
easat-10905	210	14	of	of	ADP
easat-10905	210	15	each	each	DET
easat-10905	210	16	task	task	NOUN
easat-10905	210	17	.	.	PUNCT
easat-10905	211	1	periodic	periodic	ADJ
easat-10905	211	2	rehearsal	rehearsal	NOUN
easat-10905	211	3	:	:	PUNCT
easat-10905	211	4	while	while	SCONJ
easat-10905	211	5	training	training	NOUN
easat-10905	211	6	on	on	ADP
easat-10905	211	7	new	new	ADJ
easat-10905	211	8	tasks	task	NOUN
easat-10905	211	9	,	,	PUNCT
easat-10905	211	10	the	the	DET
easat-10905	211	11	model	model	NOUN
easat-10905	211	12	periodically	periodically	ADV
easat-10905	211	13	rehearses	rehearse	VERB
easat-10905	211	14	by	by	ADP
easat-10905	211	15	retraining	retrain	VERB
easat-10905	211	16	on	on	ADP
easat-10905	211	17	stored	store	VERB
easat-10905	211	18	data	datum	NOUN
easat-10905	211	19	from	from	ADP
easat-10905	211	20	previous	previous	ADJ
easat-10905	211	21	tasks	task	NOUN
easat-10905	211	22	.	.	PUNCT
easat-10905	212	1	by	by	ADP
easat-10905	212	2	integrating	integrate	VERB
easat-10905	212	3	this	this	DET
easat-10905	212	4	rehearsal	rehearsal	NOUN
easat-10905	212	5	data	datum	NOUN
easat-10905	212	6	with	with	ADP
easat-10905	212	7	the	the	DET
easat-10905	212	8	new	new	ADJ
easat-10905	212	9	task	task	NOUN
easat-10905	212	10	data	data	PROPN
easat-10905	212	11	,	,	PUNCT
easat-10905	212	12	the	the	DET
easat-10905	212	13	model	model	NOUN
easat-10905	212	14	strengthens	strengthen	VERB
easat-10905	212	15	its	its	PRON
easat-10905	212	16	understanding	understanding	NOUN
easat-10905	212	17	of	of	ADP
easat-10905	212	18	earlier	early	ADJ
easat-10905	212	19	tasks	task	NOUN
easat-10905	212	20	while	while	SCONJ
easat-10905	212	21	simultaneously	simultaneously	ADV
easat-10905	212	22	learning	learn	VERB
easat-10905	212	23	the	the	DET
easat-10905	212	24	new	new	ADJ
easat-10905	212	25	ones	one	NOUN
easat-10905	212	26	.	.	PUNCT
easat-10905	213	1	preserving	preserve	VERB
easat-10905	213	2	knowledge	knowledge	NOUN
easat-10905	213	3	on	on	ADP
easat-10905	213	4	previous	previous	ADJ
easat-10905	213	5	tasks	task	NOUN
easat-10905	213	6	:	:	PUNCT
easat-10905	213	7	the	the	DET
easat-10905	213	8	rehearsal	rehearsal	NOUN
easat-10905	213	9	approach	approach	NOUN
easat-10905	213	10	helps	help	VERB
easat-10905	213	11	the	the	DET
easat-10905	213	12	models	model	NOUN
easat-10905	213	13	in	in	ADP
easat-10905	213	14	preserving	preserve	VERB
easat-10905	213	15	knowledge	knowledge	NOUN
easat-10905	213	16	from	from	ADP
easat-10905	213	17	previous	previous	ADJ
easat-10905	213	18	tasks	task	NOUN
easat-10905	213	19	by	by	ADP
easat-10905	213	20	regularly	regularly	ADV
easat-10905	213	21	revisiting	revisit	VERB
easat-10905	213	22	stored	store	VERB
easat-10905	213	23	data	datum	NOUN
easat-10905	213	24	.	.	PUNCT
easat-10905	214	1	this	this	PRON
easat-10905	214	2	minimizes	minimize	VERB
easat-10905	214	3	the	the	DET
easat-10905	214	4	forgetting	forgetting	NOUN
easat-10905	214	5	of	of	ADP
easat-10905	214	6	the	the	DET
easat-10905	214	7	parameters	parameter	NOUN
easat-10905	214	8	of	of	ADP
easat-10905	214	9	the	the	DET
easat-10905	214	10	models	model	NOUN
easat-10905	214	11	when	when	SCONJ
easat-10905	214	12	learning	learn	VERB
easat-10905	214	13	new	new	ADJ
easat-10905	214	14	tasks	task	NOUN
easat-10905	214	15	,	,	PUNCT
easat-10905	214	16	thereby	thereby	ADV
easat-10905	214	17	helping	help	VERB
easat-10905	214	18	to	to	PART
easat-10905	214	19	prevent	prevent	VERB
easat-10905	214	20	cf	cf	NOUN
easat-10905	214	21	.	.	PUNCT
easat-10905	215	1	the	the	DET
easat-10905	215	2	steps	step	NOUN
easat-10905	215	3	in	in	ADP
easat-10905	215	4	rehearsal	rehearsal	NOUN
easat-10905	215	5	-	-	PUNCT
easat-10905	215	6	based	base	VERB
easat-10905	215	7	approach	approach	NOUN
easat-10905	215	8	are	be	AUX
easat-10905	215	9	listed	list	VERB
easat-10905	215	10	below	below	ADV
easat-10905	215	11	:	:	PUNCT
easat-10905	215	12	step	step	NOUN
easat-10905	215	13	1	1	NUM
easat-10905	215	14	:	:	PUNCT
easat-10905	215	15	train	train	VERB
easat-10905	215	16	the	the	DET
easat-10905	215	17	models	model	NOUN
easat-10905	215	18	on	on	ADP
easat-10905	215	19	hope	hope	NOUN
easat-10905	215	20	speech	speech	NOUN
easat-10905	215	21	dataset	dataset	NOUN
easat-10905	215	22	.	.	PUNCT
easat-10905	216	1	step	step	NOUN
easat-10905	216	2	2	2	NUM
easat-10905	216	3	:	:	PUNCT
easat-10905	216	4	initialize	initialize	VERB
easat-10905	216	5	the	the	DET
easat-10905	216	6	buffer(m	buffer(m	NOUN
easat-10905	216	7	)	)	PUNCT
easat-10905	216	8	with	with	ADP
easat-10905	216	9	a	a	DET
easat-10905	216	10	subset	subset	NOUN
easat-10905	216	11	of	of	ADP
easat-10905	216	12	hope	hope	NOUN
easat-10905	216	13	speech	speech	NOUN
easat-10905	216	14	comments	comment	NOUN
easat-10905	216	15	step	step	VERB
easat-10905	216	16	3	3	NUM
easat-10905	216	17	:	:	PUNCT
easat-10905	216	18	train	train	VERB
easat-10905	216	19	the	the	DET
easat-10905	216	20	models	model	NOUN
easat-10905	216	21	on	on	ADP
easat-10905	216	22	offensive	offensive	ADJ
easat-10905	216	23	comments	comment	NOUN
easat-10905	216	24	(	(	PUNCT
easat-10905	216	25	𝐷𝑂𝑓𝑓	𝐷𝑂𝑓𝑓	PROPN
easat-10905	216	26	)	)	PUNCT
easat-10905	216	27	combined	combine	VERB
easat-10905	216	28	with	with	ADP
easat-10905	216	29	replayed	replayed	ADJ
easat-10905	216	30	data	datum	NOUN
easat-10905	216	31	from	from	ADP
easat-10905	216	32	the	the	DET
easat-10905	216	33	buffer	buffer	NOUN
easat-10905	216	34	step	step	NOUN
easat-10905	216	35	4	4	NUM
easat-10905	216	36	:	:	PUNCT
easat-10905	216	37	compute	compute	VERB
easat-10905	216	38	the	the	DET
easat-10905	216	39	loss	loss	NOUN
easat-10905	216	40	on	on	ADP
easat-10905	216	41	task	task	NOUN
easat-10905	216	42	2	2	NUM
easat-10905	216	43	,	,	PUNCT
easat-10905	216	44	𝐿𝑂𝑓𝑓	𝐿𝑂𝑓𝑓	PROPN
easat-10905	216	45	and	and	CCONJ
easat-10905	216	46	the	the	DET
easat-10905	216	47	loss	loss	NOUN
easat-10905	216	48	on	on	ADP
easat-10905	216	49	replayed	replay	VERB
easat-10905	216	50	data	datum	NOUN
easat-10905	216	51	,	,	PUNCT
easat-10905	216	52	𝐿𝐴𝑏𝑢	𝐿𝐴𝑏𝑢	PROPN
easat-10905	216	53	step	step	NOUN
easat-10905	216	54	5	5	NUM
easat-10905	216	55	:	:	PUNCT
easat-10905	216	56	combine	combine	VERB
easat-10905	216	57	both	both	DET
easat-10905	216	58	losses	loss	NOUN
easat-10905	216	59	as	as	ADP
easat-10905	216	60	in	in	ADP
easat-10905	216	61	equation	equation	NOUN
easat-10905	216	62	(	(	PUNCT
easat-10905	216	63	1	1	NUM
easat-10905	216	64	):	):	PUNCT
easat-10905	216	65	𝐿(𝜃	𝐿(𝜃	NUM
easat-10905	216	66	)	)	PUNCT
easat-10905	216	67	=	=	SYM
easat-10905	217	1	𝜆𝐿𝑂𝑓𝑓	𝜆𝐿𝑂𝑓𝑓	PROPN
easat-10905	218	1	+	+	CCONJ
easat-10905	219	1	(	(	PUNCT
easat-10905	219	2	1	1	NUM
easat-10905	219	3	−	−	NOUN
easat-10905	219	4	𝜆)𝐿𝐴𝑏𝑢(𝜃	𝜆)𝐿𝐴𝑏𝑢(𝜃	NOUN
easat-10905	219	5	)	)	PUNCT
easat-10905	219	6	(	(	PUNCT
easat-10905	219	7	1	1	X
easat-10905	219	8	)	)	PUNCT
easat-10905	219	9	where	where	SCONJ
easat-10905	219	10	𝜆	𝜆	NOUN
easat-10905	219	11	is	be	AUX
easat-10905	219	12	a	a	DET
easat-10905	219	13	hyperparameter	hyperparameter	NOUN
easat-10905	219	14	balancing	balance	VERB
easat-10905	219	15	the	the	DET
easat-10905	219	16	importance	importance	NOUN
easat-10905	219	17	of	of	ADP
easat-10905	219	18	new	new	ADJ
easat-10905	219	19	and	and	CCONJ
easat-10905	219	20	replayed	replay	VERB
easat-10905	219	21	data	datum	NOUN
easat-10905	219	22	and	and	CCONJ
easat-10905	219	23	𝜃	𝜃	NOUN
easat-10905	219	24	are	be	AUX
easat-10905	219	25	the	the	DET
easat-10905	219	26	parameters	parameter	NOUN
easat-10905	219	27	being	be	AUX
easat-10905	219	28	updated	update	VERB
easat-10905	219	29	.	.	PUNCT
easat-10905	220	1	and	and	CCONJ
easat-10905	220	2	𝐿𝑂𝑓𝑓	𝐿𝑂𝑓𝑓	PROPN
easat-10905	220	3	and	and	CCONJ
easat-10905	220	4	𝐿𝐴𝑏𝑢	𝐿𝐴𝑏𝑢	PROPN
easat-10905	220	5	are	be	AUX
easat-10905	220	6	calculated	calculate	VERB
easat-10905	220	7	as	as	ADP
easat-10905	220	8	equation	equation	NOUN
easat-10905	220	9	(	(	PUNCT
easat-10905	220	10	2	2	NUM
easat-10905	220	11	)	)	PUNCT
easat-10905	220	12	and	and	CCONJ
easat-10905	220	13	equation	equation	NOUN
easat-10905	220	14	(	(	PUNCT
easat-10905	220	15	3	3	NUM
easat-10905	220	16	)	)	PUNCT
easat-10905	220	17	𝐿𝑂𝑓𝑓(𝜃	𝐿𝑂𝑓𝑓(𝜃	NOUN
easat-10905	220	18	)	)	PUNCT
easat-10905	220	19	=	=	PUNCT
easat-10905	220	20	(	(	PUNCT
easat-10905	220	21	1	1	NUM
easat-10905	220	22	|𝐷𝑂𝑓𝑓|	|𝐷𝑂𝑓𝑓|	NOUN
easat-10905	220	23	)	)	PUNCT
easat-10905	220	24	∑	∑	PUNCT
easat-10905	220	25	𝐿(𝑓(𝑥	𝐿(𝑓(𝑥	NOUN
easat-10905	220	26	;	;	PUNCT
easat-10905	220	27	𝜃	𝜃	X
easat-10905	220	28	)	)	PUNCT
easat-10905	220	29	,	,	PUNCT
easat-10905	220	30	𝑦)(𝑥,𝑦)𝜖𝐷𝑂𝑓𝑓	𝑦)(𝑥,𝑦)𝜖𝐷𝑂𝑓𝑓	PROPN
easat-10905	220	31	(	(	PUNCT
easat-10905	220	32	2	2	NUM
easat-10905	220	33	)	)	PUNCT
easat-10905	220	34	𝐿𝐴𝑏𝑢(𝜃	𝐿𝐴𝑏𝑢(𝜃	NOUN
easat-10905	220	35	)	)	PUNCT
easat-10905	220	36	=	=	SYM
easat-10905	220	37	1	1	NUM
easat-10905	220	38	∣𝑀∣	∣𝑀∣	PROPN
easat-10905	220	39	∑	∑	PUNCT
easat-10905	220	40	𝐿(𝑓(𝑥	𝐿(𝑓(𝑥	PROPN
easat-10905	220	41	;	;	PUNCT
easat-10905	220	42	𝜃	𝜃	X
easat-10905	220	43	)	)	PUNCT
easat-10905	220	44	,	,	PUNCT
easat-10905	220	45	𝑦)(𝑥,𝑦)𝜖𝑀	𝑦)(𝑥,𝑦)𝜖𝑀	NOUN
easat-10905	220	46	(	(	PUNCT
easat-10905	220	47	3	3	X
easat-10905	220	48	)	)	PUNCT
easat-10905	220	49	step	step	NOUN
easat-10905	220	50	6	6	NUM
easat-10905	220	51	:	:	PUNCT
easat-10905	220	52	update	update	NOUN
easat-10905	220	53	𝜃	𝜃	NOUN
easat-10905	220	54	by	by	ADP
easat-10905	220	55	minimizing	minimize	VERB
easat-10905	220	56	the	the	DET
easat-10905	220	57	combined	combine	VERB
easat-10905	220	58	loss	loss	NOUN
easat-10905	220	59	𝐿(𝜃	𝐿(𝜃	NUM
easat-10905	220	60	)	)	PUNCT
easat-10905	220	61	.	.	PUNCT
easat-10905	221	1	step	step	NOUN
easat-10905	221	2	7	7	NUM
easat-10905	221	3	:	:	PUNCT
easat-10905	221	4	update	update	VERB
easat-10905	221	5	the	the	DET
easat-10905	221	6	memory	memory	NOUN
easat-10905	221	7	buffer	buffer	NOUN
easat-10905	221	8	with	with	ADP
easat-10905	221	9	a	a	DET
easat-10905	221	10	subset	subset	NOUN
easat-10905	221	11	of	of	ADP
easat-10905	221	12	the	the	DET
easat-10905	221	13	previous	previous	ADJ
easat-10905	221	14	task	task	NOUN
easat-10905	221	15	data	datum	NOUN
easat-10905	221	16	by	by	ADP
easat-10905	221	17	replacing	replace	VERB
easat-10905	221	18	older	old	ADJ
easat-10905	221	19	samples	sample	NOUN
easat-10905	221	20	.	.	PUNCT
easat-10905	222	1	3.3.2.2	3.3.2.2	X
easat-10905	222	2	.	.	X
easat-10905	222	3	pseudo	pseudo	NOUN
easat-10905	222	4	-	-	ADJ
easat-10905	222	5	rehearsal	rehearsal	NOUN
easat-10905	222	6	method	method	NOUN
easat-10905	222	7	since	since	SCONJ
easat-10905	222	8	the	the	DET
easat-10905	222	9	rehearsal	rehearsal	NOUN
easat-10905	222	10	-	-	PUNCT
easat-10905	222	11	based	base	VERB
easat-10905	222	12	approach	approach	NOUN
easat-10905	222	13	involves	involve	VERB
easat-10905	222	14	storing	store	VERB
easat-10905	222	15	a	a	DET
easat-10905	222	16	small	small	ADJ
easat-10905	222	17	part	part	NOUN
easat-10905	222	18	of	of	ADP
easat-10905	222	19	real	real	ADJ
easat-10905	222	20	data	datum	NOUN
easat-10905	222	21	from	from	ADP
easat-10905	222	22	each	each	DET
easat-10905	222	23	task	task	NOUN
easat-10905	222	24	,	,	PUNCT
easat-10905	222	25	which	which	PRON
easat-10905	222	26	can	can	AUX
easat-10905	222	27	become	become	VERB
easat-10905	222	28	memory	memory	NOUN
easat-10905	222	29	-	-	PUNCT
easat-10905	222	30	intensive	intensive	ADJ
easat-10905	222	31	as	as	SCONJ
easat-10905	222	32	the	the	DET
easat-10905	222	33	number	number	NOUN
easat-10905	222	34	of	of	ADP
easat-10905	222	35	tasks	task	NOUN
easat-10905	222	36	grows	grow	VERB
easat-10905	222	37	,	,	PUNCT
easat-10905	222	38	we	we	PRON
easat-10905	222	39	have	have	AUX
easat-10905	222	40	used	use	VERB
easat-10905	222	41	the	the	DET
easat-10905	222	42	pseudo	pseudo	NOUN
easat-10905	222	43	-	-	ADJ
easat-10905	222	44	rehearsal	rehearsal	ADJ
easat-10905	222	45	approach	approach	NOUN
easat-10905	222	46	instead	instead	ADV
easat-10905	222	47	.	.	PUNCT
easat-10905	223	1	this	this	DET
easat-10905	223	2	method	method	NOUN
easat-10905	223	3	generates	generate	VERB
easat-10905	223	4	synthetic	synthetic	ADJ
easat-10905	223	5	data	datum	NOUN
easat-10905	223	6	on	on	ADP
easat-10905	223	7	the	the	DET
easat-10905	223	8	fly	fly	NOUN
easat-10905	223	9	,	,	PUNCT
easat-10905	223	10	reducing	reduce	VERB
easat-10905	223	11	the	the	DET
easat-10905	223	12	need	need	NOUN
easat-10905	223	13	for	for	ADP
easat-10905	223	14	extensive	extensive	ADJ
easat-10905	223	15	memory	memory	NOUN
easat-10905	223	16	storage	storage	NOUN
easat-10905	223	17	.	.	PUNCT
easat-10905	224	1	in	in	ADP
easat-10905	224	2	this	this	DET
easat-10905	224	3	case	case	NOUN
easat-10905	224	4	,	,	PUNCT
easat-10905	224	5	the	the	DET
easat-10905	224	6	loss	loss	NOUN
easat-10905	224	7	for	for	ADP
easat-10905	224	8	both	both	DET
easat-10905	224	9	tasks	task	NOUN
easat-10905	224	10	is	be	AUX
easat-10905	224	11	computed	compute	VERB
easat-10905	224	12	similarly	similarly	ADV
easat-10905	224	13	to	to	ADP
easat-10905	224	14	the	the	DET
easat-10905	224	15	rehearsal	rehearsal	NOUN
easat-10905	224	16	approach	approach	NOUN
easat-10905	224	17	,	,	PUNCT
easat-10905	224	18	with	with	ADP
easat-10905	224	19	a	a	DET
easat-10905	224	20	small	small	ADJ
easat-10905	224	21	change	change	NOUN
easat-10905	224	22	in	in	ADP
easat-10905	224	23	how	how	SCONJ
easat-10905	224	24	the	the	DET
easat-10905	224	25	loss	loss	NOUN
easat-10905	224	26	on	on	ADP
easat-10905	224	27	pseudo	pseudo	NOUN
easat-10905	224	28	-	-	PUNCT
easat-10905	224	29	replayed	replay	VERB
easat-10905	224	30	data	datum	NOUN
easat-10905	224	31	is	be	AUX
easat-10905	224	32	calculated	calculate	VERB
easat-10905	224	33	,	,	PUNCT
easat-10905	224	34	as	as	SCONJ
easat-10905	224	35	shown	show	VERB
easat-10905	224	36	in	in	ADP
easat-10905	224	37	equation	equation	NOUN
easat-10905	224	38	(	(	PUNCT
easat-10905	224	39	4	4	NUM
easat-10905	224	40	)	)	PUNCT
easat-10905	224	41	.	.	PUNCT
easat-10905	225	1	𝐿𝐴𝑏𝑢(𝜃	𝐿𝐴𝑏𝑢(𝜃	NOUN
easat-10905	225	2	)	)	PUNCT
easat-10905	225	3	=	=	SYM
easat-10905	225	4	1	1	NUM
easat-10905	225	5	∣𝐷𝐺∣	∣𝐷𝐺∣	PROPN
easat-10905	225	6	∑	∑	PUNCT
easat-10905	225	7	𝐿(𝑓(𝑥	𝐿(𝑓(𝑥	PROPN
easat-10905	225	8	;	;	PUNCT
easat-10905	225	9	𝜃	𝜃	X
easat-10905	225	10	)	)	PUNCT
easat-10905	225	11	,	,	PUNCT
easat-10905	225	12	𝑦)(𝑥,𝑦)𝜖𝐷𝐺	𝑦)(𝑥,𝑦)𝜖𝐷𝐺	X
easat-10905	225	13	(	(	PUNCT
easat-10905	225	14	4	4	NUM
easat-10905	225	15	)	)	PUNCT
easat-10905	225	16	where	where	SCONJ
easat-10905	225	17	𝐺	𝐺	PROPN
easat-10905	225	18	is	be	AUX
easat-10905	225	19	the	the	DET
easat-10905	225	20	model	model	NOUN
easat-10905	225	21	that	that	PRON
easat-10905	225	22	generates	generate	VERB
easat-10905	225	23	synthetic	synthetic	ADJ
easat-10905	225	24	samples	sample	NOUN
easat-10905	225	25	from	from	ADP
easat-10905	225	26	the	the	DET
easat-10905	225	27	previous	previous	ADJ
easat-10905	225	28	tasks	task	NOUN
easat-10905	225	29	,	,	PUNCT
easat-10905	225	30	and	and	CCONJ
easat-10905	225	31	dg	dg	NOUN
easat-10905	225	32	is	be	AUX
easat-10905	225	33	the	the	DET
easat-10905	225	34	data	datum	NOUN
easat-10905	225	35	generated	generate	VERB
easat-10905	225	36	by	by	ADP
easat-10905	225	37	the	the	DET
easat-10905	225	38	generative	generative	ADJ
easat-10905	225	39	model	model	NOUN
easat-10905	225	40	.	.	PUNCT
easat-10905	226	1	both	both	DET
easat-10905	226	2	methods	method	NOUN
easat-10905	226	3	aim	aim	VERB
easat-10905	226	4	to	to	PART
easat-10905	226	5	balance	balance	VERB
easat-10905	226	6	learning	learn	VERB
easat-10905	226	7	new	new	ADJ
easat-10905	226	8	tasks	task	NOUN
easat-10905	226	9	while	while	SCONJ
easat-10905	226	10	retaining	retain	VERB
easat-10905	226	11	knowledge	knowledge	NOUN
easat-10905	226	12	from	from	ADP
easat-10905	226	13	previous	previous	ADJ
easat-10905	226	14	tasks	task	NOUN
easat-10905	226	15	,	,	PUNCT
easat-10905	226	16	thus	thus	ADV
easat-10905	226	17	addressing	address	VERB
easat-10905	226	18	the	the	DET
easat-10905	226	19	problem	problem	NOUN
easat-10905	226	20	of	of	ADP
easat-10905	226	21	cf	cf	NOUN
easat-10905	226	22	in	in	ADP
easat-10905	226	23	continual	continual	ADJ
easat-10905	226	24	learning	learn	VERB
easat-10905	226	25	scenarios	scenario	NOUN
easat-10905	226	26	.	.	PUNCT
easat-10905	227	1	3.3.3	3.3.3	X
easat-10905	227	2	.	.	PUNCT
easat-10905	227	3	regularization	regularization	NOUN
easat-10905	227	4	-	-	PUNCT
easat-10905	227	5	based	base	VERB
easat-10905	227	6	task	task	NOUN
easat-10905	227	7	incremental	incremental	ADJ
easat-10905	227	8	learning	learning	NOUN
easat-10905	227	9	regularization	regularization	NOUN
easat-10905	227	10	methods	method	NOUN
easat-10905	227	11	do	do	AUX
easat-10905	227	12	not	not	PART
easat-10905	227	13	require	require	VERB
easat-10905	227	14	storing	store	VERB
easat-10905	227	15	data	datum	NOUN
easat-10905	227	16	from	from	ADP
easat-10905	227	17	previous	previous	ADJ
easat-10905	227	18	tasks	task	NOUN
easat-10905	227	19	,	,	PUNCT
easat-10905	227	20	unlike	unlike	ADP
easat-10905	227	21	rehearsal	rehearsal	NOUN
easat-10905	227	22	and	and	CCONJ
easat-10905	227	23	pseudo	pseudo	NOUN
easat-10905	227	24	-	-	ADJ
easat-10905	227	25	rehearsal	rehearsal	NOUN
easat-10905	227	26	methods	method	NOUN
easat-10905	227	27	.	.	PUNCT
easat-10905	228	1	this	this	PRON
easat-10905	228	2	makes	make	VERB
easat-10905	228	3	them	they	PRON
easat-10905	228	4	more	more	ADJ
easat-10905	228	5	memory	memory	NOUN
easat-10905	228	6	-	-	PUNCT
easat-10905	228	7	efficient	efficient	ADJ
easat-10905	228	8	,	,	PUNCT
easat-10905	228	9	especially	especially	ADV
easat-10905	228	10	as	as	ADP
easat-10905	228	11	the	the	DET
easat-10905	228	12	number	number	NOUN
easat-10905	228	13	of	of	ADP
easat-10905	228	14	tasks	task	NOUN
easat-10905	228	15	increases	increase	NOUN
easat-10905	228	16	.	.	PUNCT
easat-10905	229	1	regularization	regularization	NOUN
easat-10905	229	2	methods	method	NOUN
easat-10905	229	3	work	work	VERB
easat-10905	229	4	by	by	ADP
easat-10905	229	5	adding	add	VERB
easat-10905	229	6	regularization	regularization	NOUN
easat-10905	229	7	terms	term	NOUN
easat-10905	229	8	to	to	ADP
easat-10905	229	9	the	the	DET
easat-10905	229	10	loss	loss	NOUN
easat-10905	229	11	function	function	NOUN
easat-10905	229	12	that	that	PRON
easat-10905	229	13	penalize	penalize	VERB
easat-10905	229	14	changes	change	NOUN
easat-10905	229	15	to	to	ADP
easat-10905	229	16	important	important	ADJ
easat-10905	229	17	parameters	parameter	NOUN
easat-10905	229	18	.	.	PUNCT
easat-10905	230	1	in	in	ADP
easat-10905	230	2	this	this	DET
easat-10905	230	3	work	work	NOUN
easat-10905	230	4	,	,	PUNCT
easat-10905	230	5	we	we	PRON
easat-10905	230	6	use	use	VERB
easat-10905	230	7	a	a	DET
easat-10905	230	8	regularization	regularization	NOUN
easat-10905	230	9	-	-	PUNCT
easat-10905	230	10	based	base	VERB
easat-10905	230	11	approach	approach	NOUN
easat-10905	230	12	called	call	VERB
easat-10905	230	13	learning	learn	VERB
easat-10905	230	14	without	without	ADP
easat-10905	230	15	forgetting	forget	VERB
easat-10905	230	16	(	(	PUNCT
easat-10905	230	17	lwf	lwf	PROPN
easat-10905	230	18	)	)	PUNCT
easat-10905	230	19	.	.	PUNCT
easat-10905	231	1	the	the	DET
easat-10905	231	2	details	detail	NOUN
easat-10905	231	3	of	of	ADP
easat-10905	231	4	lwf	lwf	PROPN
easat-10905	231	5	are	be	AUX
easat-10905	231	6	given	give	VERB
easat-10905	231	7	below	below	ADP
easat-10905	231	8	.	.	PUNCT
easat-10905	232	1	lwf	lwf	PROPN
easat-10905	232	2	:	:	PUNCT
easat-10905	232	3	it	it	PRON
easat-10905	232	4	is	be	AUX
easat-10905	232	5	a	a	DET
easat-10905	232	6	regularization	regularization	NOUN
easat-10905	232	7	-	-	PUNCT
easat-10905	232	8	based	base	VERB
easat-10905	232	9	approach	approach	NOUN
easat-10905	232	10	for	for	ADP
easat-10905	232	11	task	task	NOUN
easat-10905	232	12	-	-	PUNCT
easat-10905	232	13	incremental	incremental	ADJ
easat-10905	232	14	learning	learning	NOUN
easat-10905	232	15	that	that	PRON
easat-10905	232	16	mitigates	mitigate	VERB
easat-10905	232	17	catastrophic	catastrophic	ADJ
easat-10905	232	18	forgetting	forgetting	NOUN
easat-10905	232	19	(	(	PUNCT
easat-10905	232	20	cf	cf	NOUN
easat-10905	232	21	)	)	PUNCT
easat-10905	232	22	by	by	ADP
easat-10905	232	23	preserving	preserve	VERB
easat-10905	232	24	the	the	DET
easat-10905	232	25	knowledge	knowledge	NOUN
easat-10905	232	26	of	of	ADP
easat-10905	232	27	previously	previously	ADV
easat-10905	232	28	learned	learn	VERB
easat-10905	232	29	tasks	task	NOUN
easat-10905	232	30	while	while	SCONJ
easat-10905	232	31	learning	learn	VERB
easat-10905	232	32	new	new	ADJ
easat-10905	232	33	ones	one	NOUN
easat-10905	232	34	.	.	PUNCT
easat-10905	233	1	it	it	PRON
easat-10905	233	2	455	455	NUM
easat-10905	233	3	edelweiss	edelweiss	PROPN
easat-10905	233	4	applied	apply	VERB
easat-10905	233	5	science	science	NOUN
easat-10905	233	6	and	and	CCONJ
easat-10905	233	7	technology	technology	NOUN
easat-10905	233	8	issn	issn	PROPN
easat-10905	233	9	:	:	PUNCT
easat-10905	233	10	2576	2576	NUM
easat-10905	233	11	-	-	SYM
easat-10905	233	12	8484	8484	NUM
easat-10905	233	13	vol	vol	NOUN
easat-10905	233	14	.	.	PROPN
easat-10905	234	1	9	9	NUM
easat-10905	234	2	,	,	PUNCT
easat-10905	234	3	no	no	INTJ
easat-10905	234	4	.	.	NOUN
easat-10905	234	5	11	11	NUM
easat-10905	234	6	:	:	PUNCT
easat-10905	234	7	448	448	NUM
easat-10905	234	8	-	-	SYM
easat-10905	234	9	466	466	NUM
easat-10905	234	10	,	,	PUNCT
easat-10905	234	11	2025	2025	NUM
easat-10905	234	12	doi	doi	NOUN
easat-10905	234	13	:	:	PUNCT
easat-10905	234	14	10.55214/2576	10.55214/2576	NUM
easat-10905	234	15	-	-	PUNCT
easat-10905	234	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	235	1	©	©	PROPN
easat-10905	235	2	2025	2025	NUM
easat-10905	235	3	by	by	ADP
easat-10905	235	4	the	the	DET
easat-10905	235	5	authors	author	NOUN
easat-10905	235	6	;	;	PUNCT
easat-10905	235	7	licensee	licensee	PROPN
easat-10905	235	8	learning	learning	PROPN
easat-10905	235	9	gate	gate	PROPN
easat-10905	235	10	works	work	VERB
easat-10905	235	11	on	on	ADP
easat-10905	235	12	the	the	DET
easat-10905	235	13	principle	principle	NOUN
easat-10905	235	14	of	of	ADP
easat-10905	235	15	using	use	VERB
easat-10905	235	16	the	the	DET
easat-10905	235	17	predictions	prediction	NOUN
easat-10905	235	18	of	of	ADP
easat-10905	235	19	the	the	DET
easat-10905	235	20	models	model	NOUN
easat-10905	235	21	on	on	ADP
easat-10905	235	22	previous	previous	ADJ
easat-10905	235	23	tasks	task	NOUN
easat-10905	235	24	as	as	ADP
easat-10905	235	25	a	a	DET
easat-10905	235	26	form	form	NOUN
easat-10905	235	27	of	of	ADP
easat-10905	235	28	soft	soft	ADJ
easat-10905	235	29	target	target	NOUN
easat-10905	235	30	regularization	regularization	NOUN
easat-10905	235	31	while	while	SCONJ
easat-10905	235	32	training	training	NOUN
easat-10905	235	33	on	on	ADP
easat-10905	235	34	new	new	ADJ
easat-10905	235	35	tasks	task	NOUN
easat-10905	235	36	.	.	PUNCT
easat-10905	236	1	the	the	DET
easat-10905	236	2	steps	step	NOUN
easat-10905	236	3	involved	involve	VERB
easat-10905	236	4	in	in	ADP
easat-10905	236	5	the	the	DET
easat-10905	236	6	lwf	lwf	PROPN
easat-10905	236	7	approach	approach	NOUN
easat-10905	236	8	are	be	AUX
easat-10905	236	9	as	as	SCONJ
easat-10905	236	10	follows	follow	VERB
easat-10905	236	11	:	:	PUNCT
easat-10905	236	12	the	the	DET
easat-10905	236	13	following	follow	VERB
easat-10905	236	14	notations	notation	NOUN
easat-10905	236	15	are	be	AUX
easat-10905	236	16	used	use	VERB
easat-10905	236	17	•	•	NUM
easat-10905	236	18	𝑥	𝑥	PROPN
easat-10905	236	19	is	be	AUX
easat-10905	236	20	the	the	DET
easat-10905	236	21	input	input	NOUN
easat-10905	236	22	data	datum	NOUN
easat-10905	236	23	•	•	NOUN
easat-10905	236	24	𝑦1	𝑦1	PROPN
easat-10905	236	25	and	and	CCONJ
easat-10905	236	26	y2	y2	NOUN
easat-10905	236	27	are	be	AUX
easat-10905	236	28	the	the	DET
easat-10905	236	29	true	true	ADJ
easat-10905	236	30	labels	label	NOUN
easat-10905	236	31	of	of	ADP
easat-10905	236	32	the	the	DET
easat-10905	236	33	abusive	abusive	ADJ
easat-10905	236	34	and	and	CCONJ
easat-10905	236	35	offensive	offensive	ADJ
easat-10905	236	36	tasks	task	NOUN
easat-10905	236	37	,	,	PUNCT
easat-10905	236	38	respectively	respectively	ADV
easat-10905	236	39	.	.	PUNCT
easat-10905	237	1	step	step	NOUN
easat-10905	237	2	1	1	NUM
easat-10905	237	3	:	:	PUNCT
easat-10905	237	4	train	train	VERB
easat-10905	237	5	the	the	DET
easat-10905	237	6	abusive	abusive	ADJ
easat-10905	237	7	texts	text	NOUN
easat-10905	237	8	i.	i.	NOUN
easat-10905	237	9	train	train	VERB
easat-10905	237	10	the	the	DET
easat-10905	237	11	models	model	NOUN
easat-10905	237	12	on	on	ADP
easat-10905	237	13	the	the	DET
easat-10905	237	14	abusive	abusive	ADJ
easat-10905	237	15	texts	text	NOUN
easat-10905	237	16	and	and	CCONJ
easat-10905	237	17	calculate	calculate	VERB
easat-10905	237	18	the	the	DET
easat-10905	237	19	loss	loss	NOUN
easat-10905	237	20	as	as	ADP
easat-10905	237	21	in	in	ADP
easat-10905	237	22	equation	equation	NOUN
easat-10905	237	23	(	(	PUNCT
easat-10905	237	24	5	5	NUM
easat-10905	237	25	)	)	PUNCT
easat-10905	237	26	𝐿𝐴𝑏𝑢(𝜃	𝐿𝐴𝑏𝑢(𝜃	NOUN
easat-10905	237	27	)	)	PUNCT
easat-10905	238	1	=	=	SYM
easat-10905	238	2	𝐶𝑟𝑜𝑠𝑠𝐸𝑛𝑡𝑟𝑜𝑝𝑦𝐿𝑜𝑠𝑠(𝑓	𝐶𝑟𝑜𝑠𝑠𝐸𝑛𝑡𝑟𝑜𝑝𝑦𝐿𝑜𝑠𝑠(𝑓	PROPN
easat-10905	238	3	θ(𝑥	θ(𝑥	PROPN
easat-10905	238	4	)	)	PUNCT
easat-10905	238	5	,	,	PUNCT
easat-10905	238	6	𝑦1	𝑦1	PROPN
easat-10905	238	7	)	)	PUNCT
easat-10905	238	8	(	(	PUNCT
easat-10905	238	9	5	5	X
easat-10905	238	10	)	)	PUNCT
easat-10905	238	11	where	where	SCONJ
easat-10905	238	12	𝑓	𝑓	DET
easat-10905	238	13	θ(𝑥	θ(𝑥	NOUN
easat-10905	238	14	)	)	PUNCT
easat-10905	238	15	is	be	AUX
easat-10905	238	16	the	the	DET
easat-10905	238	17	model	model	NOUN
easat-10905	238	18	output	output	NOUN
easat-10905	238	19	with	with	ADP
easat-10905	238	20	parameters	parameter	NOUN
easat-10905	238	21	𝜃.	𝜃.	PROPN
easat-10905	238	22	ii	ii	PROPN
easat-10905	238	23	.	.	PUNCT
easat-10905	239	1	save	save	VERB
easat-10905	239	2	the	the	DET
easat-10905	239	3	parameters	parameter	NOUN
easat-10905	239	4	such	such	ADJ
easat-10905	239	5	as	as	ADP
easat-10905	239	6	weights	weight	NOUN
easat-10905	239	7	and	and	CCONJ
easat-10905	239	8	biases	bias	NOUN
easat-10905	239	9	of	of	ADP
easat-10905	239	10	the	the	DET
easat-10905	239	11	developed	develop	VERB
easat-10905	239	12	models	model	NOUN
easat-10905	239	13	and	and	CCONJ
easat-10905	239	14	their	their	PRON
easat-10905	239	15	outputs	output	NOUN
easat-10905	239	16	of	of	ADP
easat-10905	239	17	the	the	DET
easat-10905	239	18	initial	initial	ADJ
easat-10905	239	19	task	task	NOUN
easat-10905	239	20	(	(	PUNCT
easat-10905	239	21	abusive	abusive	ADJ
easat-10905	239	22	task	task	NOUN
easat-10905	239	23	)	)	PUNCT
easat-10905	239	24	.	.	PUNCT
easat-10905	240	1	these	these	DET
easat-10905	240	2	outputs	output	NOUN
easat-10905	240	3	will	will	AUX
easat-10905	240	4	serve	serve	VERB
easat-10905	240	5	as	as	ADP
easat-10905	240	6	soft	soft	ADJ
easat-10905	240	7	targets	target	NOUN
easat-10905	240	8	when	when	SCONJ
easat-10905	240	9	training	train	VERB
easat-10905	240	10	on	on	ADP
easat-10905	240	11	the	the	DET
easat-10905	240	12	next	next	ADJ
easat-10905	240	13	(	(	PUNCT
easat-10905	240	14	offensive	offensive	ADJ
easat-10905	240	15	texts	text	NOUN
easat-10905	240	16	)	)	PUNCT
easat-10905	240	17	task	task	NOUN
easat-10905	240	18	.	.	PUNCT
easat-10905	241	1	then	then	ADV
easat-10905	241	2	,	,	PUNCT
easat-10905	241	3	compute	compute	VERB
easat-10905	241	4	the	the	DET
easat-10905	241	5	soft	soft	ADJ
easat-10905	241	6	targets	target	NOUN
easat-10905	241	7	,	,	PUNCT
easat-10905	241	8	𝑍𝐴𝑏𝑢	𝑍𝐴𝑏𝑢	PROPN
easat-10905	241	9	for	for	ADP
easat-10905	241	10	an	an	DET
easat-10905	241	11	abusive	abusive	ADJ
easat-10905	241	12	task	task	NOUN
easat-10905	241	13	using	use	VERB
easat-10905	241	14	the	the	DET
easat-10905	241	15	model	model	NOUN
easat-10905	241	16	trained	train	VERB
easat-10905	241	17	on	on	ADP
easat-10905	241	18	the	the	DET
easat-10905	241	19	abusive	abusive	ADJ
easat-10905	241	20	task	task	NOUN
easat-10905	241	21	as	as	SCONJ
easat-10905	241	22	given	give	VERB
easat-10905	241	23	in	in	ADP
easat-10905	241	24	equation	equation	NOUN
easat-10905	241	25	(	(	PUNCT
easat-10905	241	26	6	6	NUM
easat-10905	241	27	)	)	PUNCT
easat-10905	241	28	𝑍𝐴𝑏𝑢	𝑍𝐴𝑏𝑢	NOUN
easat-10905	242	1	=	=	SYM
easat-10905	242	2	𝑓θ𝐴𝑏𝑢	𝑓θ𝐴𝑏𝑢	VERB
easat-10905	242	3	∗	∗	NOUN
easat-10905	242	4	(	(	PUNCT
easat-10905	242	5	𝑥	𝑥	NOUN
easat-10905	242	6	)	)	PUNCT
easat-10905	242	7	(	(	PUNCT
easat-10905	242	8	6	6	NUM
easat-10905	242	9	)	)	PUNCT
easat-10905	242	10	where	where	SCONJ
easat-10905	242	11	θ𝐴𝑏𝑢	θ𝐴𝑏𝑢	PROPN
easat-10905	242	12	∗	∗	NOUN
easat-10905	242	13	are	be	AUX
easat-10905	242	14	the	the	DET
easat-10905	242	15	optimal	optimal	ADJ
easat-10905	242	16	parameters	parameter	NOUN
easat-10905	242	17	after	after	ADP
easat-10905	242	18	training	train	VERB
easat-10905	242	19	on	on	ADP
easat-10905	242	20	the	the	DET
easat-10905	242	21	abusive	abusive	ADJ
easat-10905	242	22	task	task	NOUN
easat-10905	242	23	.	.	PUNCT
easat-10905	243	1	step	step	NOUN
easat-10905	243	2	2	2	NUM
easat-10905	243	3	:	:	PUNCT
easat-10905	243	4	train	train	VERB
easat-10905	243	5	on	on	ADP
easat-10905	243	6	offensive	offensive	ADJ
easat-10905	243	7	texts	text	NOUN
easat-10905	243	8	using	use	VERB
easat-10905	243	9	regularization	regularization	NOUN
easat-10905	243	10	with	with	ADP
easat-10905	243	11	lwf	lwf	PROPN
easat-10905	243	12	i.	i.	PROPN
easat-10905	243	13	initialize	initialize	VERB
easat-10905	243	14	the	the	DET
easat-10905	243	15	training	training	NOUN
easat-10905	243	16	with	with	ADP
easat-10905	243	17	the	the	DET
easat-10905	243	18	previously	previously	ADV
easat-10905	243	19	learned	learn	VERB
easat-10905	243	20	model	model	NOUN
easat-10905	243	21	parameters	parameter	NOUN
easat-10905	243	22	θ𝐴𝑏𝑢	θ𝐴𝑏𝑢	PROPN
easat-10905	243	23	∗	∗	NOUN
easat-10905	243	24	and	and	CCONJ
easat-10905	243	25	train	train	VERB
easat-10905	243	26	the	the	DET
easat-10905	243	27	models	model	NOUN
easat-10905	243	28	using	use	VERB
easat-10905	243	29	offensive	offensive	ADJ
easat-10905	243	30	texts	text	NOUN
easat-10905	243	31	ii	ii	PROPN
easat-10905	243	32	.	.	PUNCT
easat-10905	243	33	compute	compute	VERB
easat-10905	243	34	the	the	DET
easat-10905	243	35	loss	loss	NOUN
easat-10905	243	36	for	for	ADP
easat-10905	243	37	the	the	DET
easat-10905	243	38	offensive	offensive	ADJ
easat-10905	243	39	comments	comment	NOUN
easat-10905	243	40	as	as	SCONJ
easat-10905	243	41	shown	show	VERB
easat-10905	243	42	in	in	ADP
easat-10905	243	43	equation	equation	NOUN
easat-10905	243	44	(	(	PUNCT
easat-10905	243	45	7	7	NUM
easat-10905	243	46	)	)	PUNCT
easat-10905	243	47	𝐿𝑂𝑓𝑓(𝜃	𝐿𝑂𝑓𝑓(𝜃	NOUN
easat-10905	243	48	)	)	PUNCT
easat-10905	243	49	=	=	PROPN
easat-10905	243	50	𝐶𝑟𝑜𝑠𝑠𝐸𝑛𝑡𝑟𝑜𝑝𝑦𝐿𝑜𝑠𝑠(𝑓	𝐶𝑟𝑜𝑠𝑠𝐸𝑛𝑡𝑟𝑜𝑝𝑦𝐿𝑜𝑠𝑠(𝑓	PROPN
easat-10905	243	51	θ(𝑥	θ(𝑥	PROPN
easat-10905	243	52	)	)	PUNCT
easat-10905	243	53	,	,	PUNCT
easat-10905	243	54	𝑦2	𝑦2	PROPN
easat-10905	243	55	)	)	PUNCT
easat-10905	243	56	(	(	PUNCT
easat-10905	243	57	7	7	X
easat-10905	243	58	)	)	PUNCT
easat-10905	243	59	iii	iii	PROPN
easat-10905	243	60	.	.	PUNCT
easat-10905	244	1	use	use	NOUN
easat-10905	244	2	equation	equation	NOUN
easat-10905	244	3	(	(	PUNCT
easat-10905	244	4	8)	8)	NUM
easat-10905	244	5	to	to	PART
easat-10905	244	6	compute	compute	VERB
easat-10905	244	7	the	the	DET
easat-10905	244	8	loss	loss	NOUN
easat-10905	244	9	on	on	ADP
easat-10905	244	10	the	the	DET
easat-10905	244	11	soft	soft	ADJ
easat-10905	244	12	targets	target	NOUN
easat-10905	244	13	from	from	ADP
easat-10905	244	14	the	the	DET
easat-10905	244	15	previous	previous	ADJ
easat-10905	244	16	tasks	task	NOUN
easat-10905	244	17	using	use	VERB
easat-10905	244	18	the	the	DET
easat-10905	244	19	saved	save	VERB
easat-10905	244	20	outputs	output	NOUN
easat-10905	244	21	.	.	PUNCT
easat-10905	245	1	𝐿𝐾𝐷(𝜃	𝐿𝐾𝐷(𝜃	NUM
easat-10905	245	2	)	)	PUNCT
easat-10905	245	3	=	=	PUNCT
easat-10905	246	1	𝐾𝐿𝐷𝑖𝑣𝑒𝑟𝑔𝑒𝑛𝑐𝑒(𝑆𝑜𝑓𝑡𝑚𝑎𝑥(𝑓	𝐾𝐿𝐷𝑖𝑣𝑒𝑟𝑔𝑒𝑛𝑐𝑒(𝑆𝑜𝑓𝑡𝑚𝑎𝑥(𝑓	PROPN
easat-10905	246	2	θ(𝑥)/𝑇	θ(𝑥)/𝑇	PROPN
easat-10905	246	3	)	)	PUNCT
easat-10905	246	4	,	,	PUNCT
easat-10905	246	5	𝑆𝑜𝑓𝑡𝑚𝑎𝑥(𝑍𝐴𝑏𝑢	𝑆𝑜𝑓𝑡𝑚𝑎𝑥(𝑍𝐴𝑏𝑢	PROPN
easat-10905	246	6	/𝑇	/𝑇	PUNCT
easat-10905	246	7	)	)	PUNCT
easat-10905	246	8	)	)	PUNCT
easat-10905	247	1	(	(	PUNCT
easat-10905	247	2	8)	8)	NUM
easat-10905	247	3	where	where	SCONJ
easat-10905	247	4	t	t	PROPN
easat-10905	247	5	is	be	AUX
easat-10905	247	6	the	the	DET
easat-10905	247	7	temperature	temperature	NOUN
easat-10905	247	8	parameter	parameter	NOUN
easat-10905	247	9	used	use	VERB
easat-10905	247	10	to	to	PART
easat-10905	247	11	soften	soften	VERB
easat-10905	247	12	the	the	DET
easat-10905	247	13	probability	probability	NOUN
easat-10905	247	14	distributions	distribution	NOUN
easat-10905	247	15	.	.	PUNCT
easat-10905	248	1	higher	high	ADJ
easat-10905	248	2	temperatures	temperature	NOUN
easat-10905	248	3	produce	produce	VERB
easat-10905	248	4	softer	soft	ADJ
easat-10905	248	5	probability	probability	NOUN
easat-10905	248	6	distributions	distribution	NOUN
easat-10905	248	7	,	,	PUNCT
easat-10905	248	8	making	make	VERB
easat-10905	248	9	it	it	PRON
easat-10905	248	10	easier	easy	ADJ
easat-10905	248	11	to	to	PART
easat-10905	248	12	transfer	transfer	VERB
easat-10905	248	13	knowledge	knowledge	NOUN
easat-10905	248	14	.	.	PUNCT
easat-10905	249	1	step	step	NOUN
easat-10905	249	2	3	3	NUM
easat-10905	249	3	:	:	PUNCT
easat-10905	249	4	calculate	calculate	VERB
easat-10905	249	5	the	the	DET
easat-10905	249	6	total	total	ADJ
easat-10905	249	7	loss	loss	NOUN
easat-10905	249	8	i.	i.	NOUN
easat-10905	249	9	combine	combine	VERB
easat-10905	249	10	the	the	DET
easat-10905	249	11	loss	loss	NOUN
easat-10905	249	12	on	on	ADP
easat-10905	249	13	the	the	DET
easat-10905	249	14	offensive	offensive	ADJ
easat-10905	249	15	task	task	NOUN
easat-10905	249	16	and	and	CCONJ
easat-10905	249	17	the	the	DET
easat-10905	249	18	knowledge	knowledge	NOUN
easat-10905	249	19	distillation	distillation	NOUN
easat-10905	249	20	loss	loss	NOUN
easat-10905	249	21	from	from	ADP
easat-10905	249	22	the	the	DET
easat-10905	249	23	abusive	abusive	ADJ
easat-10905	249	24	task	task	NOUN
easat-10905	249	25	to	to	PART
easat-10905	249	26	find	find	VERB
easat-10905	249	27	the	the	DET
easat-10905	249	28	total	total	ADJ
easat-10905	249	29	loss	loss	NOUN
easat-10905	249	30	.	.	PUNCT
easat-10905	250	1	(	(	PUNCT
easat-10905	250	2	the	the	DET
easat-10905	250	3	knowledge	knowledge	NOUN
easat-10905	250	4	distillation	distillation	NOUN
easat-10905	250	5	loss	loss	NOUN
easat-10905	250	6	helps	help	VERB
easat-10905	250	7	to	to	PART
easat-10905	250	8	retain	retain	VERB
easat-10905	250	9	the	the	DET
easat-10905	250	10	information	information	NOUN
easat-10905	250	11	from	from	ADP
easat-10905	250	12	previous	previous	ADJ
easat-10905	250	13	tasks	task	NOUN
easat-10905	250	14	)	)	PUNCT
easat-10905	250	15	.	.	PUNCT
easat-10905	251	1	this	this	PRON
easat-10905	251	2	is	be	AUX
easat-10905	251	3	done	do	VERB
easat-10905	251	4	using	use	VERB
easat-10905	251	5	equation	equation	NOUN
easat-10905	251	6	(	(	PUNCT
easat-10905	251	7	9	9	NUM
easat-10905	251	8	)	)	PUNCT
easat-10905	251	9	.	.	PUNCT
easat-10905	252	1	𝐿𝑡𝑜𝑡𝑎𝑙(𝜃	𝐿𝑡𝑜𝑡𝑎𝑙(𝜃	NOUN
easat-10905	252	2	)	)	PUNCT
easat-10905	253	1	=	=	SYM
easat-10905	253	2	𝐿𝑂𝑓𝑓(𝜃	𝐿𝑂𝑓𝑓(𝜃	X
easat-10905	253	3	)	)	PUNCT
easat-10905	253	4	+	+	CCONJ
easat-10905	253	5	𝜆𝐿𝐾𝐷(𝜃	𝜆𝐿𝐾𝐷(𝜃	ADJ
easat-10905	253	6	)	)	PUNCT
easat-10905	253	7	(	(	PUNCT
easat-10905	253	8	9	9	NUM
easat-10905	253	9	)	)	PUNCT
easat-10905	253	10	where	where	SCONJ
easat-10905	253	11	𝜆	𝜆	NOUN
easat-10905	253	12	is	be	AUX
easat-10905	253	13	the	the	DET
easat-10905	253	14	balancing	balancing	NOUN
easat-10905	253	15	parameter	parameter	NOUN
easat-10905	253	16	used	use	VERB
easat-10905	253	17	to	to	PART
easat-10905	253	18	control	control	VERB
easat-10905	253	19	the	the	DET
easat-10905	253	20	trade	trade	NOUN
easat-10905	253	21	-	-	PUNCT
easat-10905	253	22	off	off	NOUN
easat-10905	253	23	between	between	ADP
easat-10905	253	24	retaining	retain	VERB
easat-10905	253	25	old	old	ADJ
easat-10905	253	26	knowledge	knowledge	NOUN
easat-10905	253	27	on	on	ADP
easat-10905	253	28	abusive	abusive	ADJ
easat-10905	253	29	texts	text	NOUN
easat-10905	253	30	and	and	CCONJ
easat-10905	253	31	learning	learn	VERB
easat-10905	253	32	new	new	ADJ
easat-10905	253	33	information	information	NOUN
easat-10905	253	34	from	from	ADP
easat-10905	253	35	offensive	offensive	ADJ
easat-10905	253	36	texts	text	NOUN
easat-10905	253	37	.	.	PUNCT
easat-10905	254	1	step	step	NOUN
easat-10905	254	2	4	4	NUM
easat-10905	254	3	:	:	PUNCT
easat-10905	254	4	update	update	VERB
easat-10905	254	5	the	the	DET
easat-10905	254	6	parameters	parameter	NOUN
easat-10905	254	7	of	of	ADP
easat-10905	254	8	the	the	DET
easat-10905	254	9	models	model	NOUN
easat-10905	254	10	i.	i.	NOUN
easat-10905	254	11	update	update	VERB
easat-10905	254	12	the	the	DET
easat-10905	254	13	model	model	NOUN
easat-10905	254	14	parameters	parameter	NOUN
easat-10905	254	15	as	as	ADP
easat-10905	254	16	in	in	ADP
easat-10905	254	17	equation	equation	NOUN
easat-10905	254	18	(	(	PUNCT
easat-10905	254	19	10	10	NUM
easat-10905	254	20	)	)	PUNCT
easat-10905	254	21	by	by	ADP
easat-10905	254	22	minimizing	minimize	VERB
easat-10905	254	23	the	the	DET
easat-10905	254	24	combined	combine	VERB
easat-10905	254	25	loss	loss	NOUN
easat-10905	254	26	to	to	PART
easat-10905	254	27	ensure	ensure	VERB
easat-10905	254	28	that	that	SCONJ
easat-10905	254	29	while	while	SCONJ
easat-10905	254	30	the	the	DET
easat-10905	254	31	model	model	NOUN
easat-10905	254	32	learns	learn	VERB
easat-10905	254	33	the	the	DET
easat-10905	254	34	offensive	offensive	ADJ
easat-10905	254	35	task	task	NOUN
easat-10905	254	36	,	,	PUNCT
easat-10905	254	37	it	it	PRON
easat-10905	254	38	does	do	AUX
easat-10905	254	39	not	not	PART
easat-10905	254	40	forget	forget	VERB
easat-10905	254	41	the	the	DET
easat-10905	254	42	knowledge	knowledge	NOUN
easat-10905	254	43	acquired	acquire	VERB
easat-10905	254	44	from	from	ADP
easat-10905	254	45	the	the	DET
easat-10905	254	46	abusive	abusive	ADJ
easat-10905	254	47	task	task	NOUN
easat-10905	254	48	.	.	PUNCT
easat-10905	255	1	𝜃∗	𝜃∗	PROPN
easat-10905	255	2	=	=	PUNCT
easat-10905	255	3	𝑎𝑟𝑔𝑚𝑖𝑛𝜃𝐿𝑡𝑜𝑡𝑎𝑙(𝜃	𝑎𝑟𝑔𝑚𝑖𝑛𝜃𝐿𝑡𝑜𝑡𝑎𝑙(𝜃	PROPN
easat-10905	255	4	)	)	PUNCT
easat-10905	255	5	(	(	PUNCT
easat-10905	255	6	10	10	NUM
easat-10905	255	7	)	)	PUNCT
easat-10905	256	1	the	the	DET
easat-10905	256	2	workflow	workflow	NOUN
easat-10905	256	3	of	of	ADP
easat-10905	256	4	the	the	DET
easat-10905	256	5	proposed	propose	VERB
easat-10905	256	6	research	research	NOUN
easat-10905	256	7	is	be	AUX
easat-10905	256	8	shown	show	VERB
easat-10905	256	9	in	in	ADP
easat-10905	256	10	figure	figure	NOUN
easat-10905	256	11	1	1	NUM
easat-10905	256	12	.	.	NOUN
easat-10905	256	13	4	4	NUM
easat-10905	256	14	.	.	X
easat-10905	256	15	experimental	experimental	ADJ
easat-10905	256	16	setup	setup	NOUN
easat-10905	256	17	and	and	CCONJ
easat-10905	256	18	results	result	VERB
easat-10905	256	19	4.1	4.1	NUM
easat-10905	256	20	.	.	PUNCT
easat-10905	257	1	experimental	experimental	ADJ
easat-10905	257	2	setup	setup	NOUN
easat-10905	257	3	to	to	PART
easat-10905	257	4	implement	implement	VERB
easat-10905	257	5	and	and	CCONJ
easat-10905	257	6	assess	assess	VERB
easat-10905	257	7	the	the	DET
easat-10905	257	8	performance	performance	NOUN
easat-10905	257	9	of	of	ADP
easat-10905	257	10	continual	continual	ADJ
easat-10905	257	11	learning	learning	NOUN
easat-10905	257	12	approaches	approach	NOUN
easat-10905	257	13	in	in	ADP
easat-10905	257	14	comparison	comparison	NOUN
easat-10905	257	15	to	to	ADP
easat-10905	257	16	pretrained	pretraine	VERB
easat-10905	257	17	models	model	NOUN
easat-10905	258	1	,	,	PUNCT
easat-10905	258	2	we	we	PRON
easat-10905	258	3	adapted	adapt	VERB
easat-10905	258	4	the	the	DET
easat-10905	258	5	continual	continual	ADJ
easat-10905	258	6	hyperparameter	hyperparameter	NOUN
easat-10905	258	7	selection	selection	NOUN
easat-10905	258	8	framework	framework	NOUN
easat-10905	258	9	introduced	introduce	VERB
easat-10905	258	10	in	in	ADP
easat-10905	258	11	ritter	ritter	PROPN
easat-10905	258	12	et	et	PROPN
easat-10905	258	13	al	al	PROPN
easat-10905	258	14	.	.	PROPN
easat-10905	259	1	456	456	PROPN
easat-10905	259	2	edelweiss	edelweiss	PROPN
easat-10905	259	3	applied	apply	VERB
easat-10905	259	4	science	science	NOUN
easat-10905	259	5	and	and	CCONJ
easat-10905	259	6	technology	technology	NOUN
easat-10905	259	7	issn	issn	PROPN
easat-10905	259	8	:	:	PUNCT
easat-10905	259	9	2576	2576	NUM
easat-10905	259	10	-	-	SYM
easat-10905	259	11	8484	8484	NUM
easat-10905	259	12	vol	vol	NOUN
easat-10905	259	13	.	.	PROPN
easat-10905	260	1	9	9	NUM
easat-10905	260	2	,	,	PUNCT
easat-10905	260	3	no	no	INTJ
easat-10905	260	4	.	.	NOUN
easat-10905	260	5	11	11	NUM
easat-10905	260	6	:	:	PUNCT
easat-10905	260	7	448	448	NUM
easat-10905	260	8	-	-	SYM
easat-10905	260	9	466	466	NUM
easat-10905	260	10	,	,	PUNCT
easat-10905	260	11	2025	2025	NUM
easat-10905	260	12	doi	doi	NOUN
easat-10905	260	13	:	:	PUNCT
easat-10905	260	14	10.55214/2576	10.55214/2576	NUM
easat-10905	260	15	-	-	PUNCT
easat-10905	260	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	261	1	©	©	PROPN
easat-10905	261	2	2025	2025	NUM
easat-10905	261	3	by	by	ADP
easat-10905	261	4	the	the	DET
easat-10905	261	5	authors	author	NOUN
easat-10905	261	6	;	;	PUNCT
easat-10905	261	7	licensee	licensee	PROPN
easat-10905	261	8	learning	learning	NOUN
easat-10905	261	9	gate	gate	NOUN
easat-10905	262	1	[	[	X
easat-10905	262	2	3	3	NUM
easat-10905	262	3	]	]	PUNCT
easat-10905	262	4	to	to	PART
easat-10905	262	5	ensure	ensure	VERB
easat-10905	262	6	a	a	DET
easat-10905	262	7	fair	fair	ADJ
easat-10905	262	8	comparison	comparison	NOUN
easat-10905	262	9	with	with	ADP
easat-10905	262	10	existing	exist	VERB
easat-10905	262	11	methods	method	NOUN
easat-10905	262	12	.	.	PUNCT
easat-10905	263	1	the	the	DET
easat-10905	263	2	experimental	experimental	ADJ
easat-10905	263	3	setup	setup	NOUN
easat-10905	263	4	utilized	utilize	VERB
easat-10905	263	5	in	in	ADP
easat-10905	263	6	this	this	DET
easat-10905	263	7	study	study	NOUN
easat-10905	263	8	is	be	AUX
easat-10905	263	9	detailed	detail	VERB
easat-10905	263	10	in	in	ADP
easat-10905	263	11	table	table	NOUN
easat-10905	263	12	1	1	NUM
easat-10905	263	13	.	.	PUNCT
easat-10905	264	1	all	all	DET
easat-10905	264	2	models	model	NOUN
easat-10905	264	3	were	be	AUX
easat-10905	264	4	developed	develop	VERB
easat-10905	264	5	using	use	VERB
easat-10905	264	6	the	the	DET
easat-10905	264	7	pytorch	pytorch	NOUN
easat-10905	264	8	library	library	NOUN
easat-10905	264	9	.	.	PUNCT
easat-10905	265	1	table	table	NOUN
easat-10905	265	2	1	1	NUM
easat-10905	265	3	.	.	PUNCT
easat-10905	265	4	experimental	experimental	ADJ
easat-10905	265	5	setup	setup	NOUN
easat-10905	265	6	.	.	PUNCT
easat-10905	266	1	parameters	parameter	NOUN
easat-10905	266	2	values	value	NOUN
easat-10905	266	3	set	set	VERB
easat-10905	266	4	stochastic	stochastic	ADJ
easat-10905	266	5	gradient	gradient	ADJ
easat-10905	266	6	descent	descent	NOUN
easat-10905	266	7	with	with	ADP
easat-10905	266	8	momentum	momentum	NOUN
easat-10905	266	9	as	as	ADP
easat-10905	266	10	an	an	DET
easat-10905	266	11	optimizer	optimizer	NOUN
easat-10905	266	12	0.8	0.8	NUM
easat-10905	266	13	training	training	NOUN
easat-10905	266	14	epochs	epoch	NOUN
easat-10905	266	15	100	100	NUM
easat-10905	266	16	early	early	ADV
easat-10905	266	17	stopping	stop	VERB
easat-10905	266	18	5	5	NUM
easat-10905	266	19	batch	batch	NOUN
easat-10905	266	20	size	size	NOUN
easat-10905	266	21	128	128	NUM
easat-10905	266	22	softmax	softmax	NOUN
easat-10905	266	23	temperature	temperature	NOUN
easat-10905	266	24	(	(	PUNCT
easat-10905	266	25	t	t	NOUN
easat-10905	266	26	)	)	PUNCT
easat-10905	266	27	2	2	NUM
easat-10905	266	28	knowledge	knowledge	NOUN
easat-10905	266	29	distillation	distillation	NOUN
easat-10905	266	30	strength	strength	NOUN
easat-10905	266	31	(	(	PUNCT
easat-10905	266	32	𝜆	𝜆	NOUN
easat-10905	266	33	)	)	PUNCT
easat-10905	266	34	0.5	0.5	NUM
easat-10905	266	35	figure	figure	NOUN
easat-10905	266	36	1	1	NUM
easat-10905	266	37	.	.	PUNCT
easat-10905	266	38	proposed	propose	VERB
easat-10905	266	39	workflow	workflow	NOUN
easat-10905	266	40	.	.	PUNCT
easat-10905	267	1	our	our	PRON
easat-10905	267	2	work	work	NOUN
easat-10905	267	3	utilizes	utilize	VERB
easat-10905	267	4	gradient	gradient	ADJ
easat-10905	267	5	episodic	episodic	ADJ
easat-10905	267	6	memory	memory	NOUN
easat-10905	267	7	(	(	PUNCT
easat-10905	267	8	gem	gem	NOUN
easat-10905	267	9	)	)	PUNCT
easat-10905	267	10	for	for	ADP
easat-10905	267	11	replay	replay	NOUN
easat-10905	267	12	-	-	PUNCT
easat-10905	267	13	based	base	VERB
easat-10905	267	14	task	task	NOUN
easat-10905	267	15	incremental	incremental	ADJ
easat-10905	267	16	learning	learning	NOUN
easat-10905	267	17	and	and	CCONJ
easat-10905	267	18	implements	implement	VERB
easat-10905	267	19	learning	learn	VERB
easat-10905	267	20	without	without	ADP
easat-10905	267	21	forgetting	forget	VERB
easat-10905	267	22	(	(	PUNCT
easat-10905	267	23	lwf	lwf	PROPN
easat-10905	267	24	)	)	PUNCT
easat-10905	267	25	from	from	ADP
easat-10905	267	26	scratch	scratch	NOUN
easat-10905	267	27	using	use	VERB
easat-10905	267	28	python	python	NOUN
easat-10905	267	29	.	.	PUNCT
easat-10905	268	1	a	a	DET
easat-10905	268	2	multi	multi	ADJ
easat-10905	268	3	-	-	ADJ
easat-10905	268	4	head	head	ADJ
easat-10905	268	5	configuration	configuration	NOUN
easat-10905	268	6	is	be	AUX
easat-10905	268	7	employed	employ	VERB
easat-10905	268	8	,	,	PUNCT
easat-10905	268	9	where	where	SCONJ
easat-10905	268	10	the	the	DET
easat-10905	268	11	model	model	NOUN
easat-10905	268	12	shares	share	VERB
easat-10905	268	13	hidden	hide	VERB
easat-10905	268	14	layers	layer	NOUN
easat-10905	268	15	but	but	CCONJ
easat-10905	268	16	has	have	VERB
easat-10905	268	17	separate	separate	ADJ
easat-10905	268	18	output	output	NOUN
easat-10905	268	19	heads	head	NOUN
easat-10905	268	20	for	for	ADP
easat-10905	268	21	different	different	ADJ
easat-10905	268	22	tasks	task	NOUN
easat-10905	268	23	.	.	PUNCT
easat-10905	269	1	specifically	specifically	ADV
easat-10905	269	2	,	,	PUNCT
easat-10905	269	3	a	a	DET
easat-10905	269	4	two	two	NUM
easat-10905	269	5	-	-	PUNCT
easat-10905	269	6	head	head	NOUN
easat-10905	269	7	configuration	configuration	NOUN
easat-10905	269	8	is	be	AUX
easat-10905	269	9	implemented	implement	VERB
easat-10905	269	10	,	,	PUNCT
easat-10905	269	11	with	with	ADP
easat-10905	269	12	one	one	NUM
easat-10905	269	13	head	head	NOUN
easat-10905	269	14	classifying	classify	VERB
easat-10905	269	15	abusive	abusive	ADJ
easat-10905	269	16	comments	comment	NOUN
easat-10905	269	17	and	and	CCONJ
easat-10905	269	18	the	the	DET
easat-10905	269	19	other	other	ADJ
easat-10905	269	20	classifying	classify	VERB
easat-10905	269	21	offensive	offensive	ADJ
easat-10905	269	22	comments	comment	NOUN
easat-10905	269	23	.	.	PUNCT
easat-10905	270	1	457	457	NUM
easat-10905	270	2	edelweiss	edelweiss	PROPN
easat-10905	270	3	applied	apply	VERB
easat-10905	270	4	science	science	NOUN
easat-10905	270	5	and	and	CCONJ
easat-10905	270	6	technology	technology	NOUN
easat-10905	270	7	issn	issn	PROPN
easat-10905	270	8	:	:	PUNCT
easat-10905	270	9	2576	2576	NUM
easat-10905	270	10	-	-	SYM
easat-10905	270	11	8484	8484	NUM
easat-10905	270	12	vol	vol	NOUN
easat-10905	270	13	.	.	PROPN
easat-10905	271	1	9	9	NUM
easat-10905	271	2	,	,	PUNCT
easat-10905	271	3	no	no	INTJ
easat-10905	271	4	.	.	NOUN
easat-10905	271	5	11	11	NUM
easat-10905	271	6	:	:	PUNCT
easat-10905	271	7	448	448	NUM
easat-10905	271	8	-	-	SYM
easat-10905	271	9	466	466	NUM
easat-10905	271	10	,	,	PUNCT
easat-10905	271	11	2025	2025	NUM
easat-10905	271	12	doi	doi	NOUN
easat-10905	271	13	:	:	PUNCT
easat-10905	271	14	10.55214/2576	10.55214/2576	NUM
easat-10905	271	15	-	-	PUNCT
easat-10905	271	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	272	1	©	©	PROPN
easat-10905	272	2	2025	2025	NUM
easat-10905	272	3	by	by	ADP
easat-10905	272	4	the	the	DET
easat-10905	272	5	authors	author	NOUN
easat-10905	272	6	;	;	PUNCT
easat-10905	272	7	licensee	licensee	PROPN
easat-10905	272	8	learning	learning	NOUN
easat-10905	272	9	gate	gate	NOUN
easat-10905	272	10	4.2	4.2	NUM
easat-10905	272	11	.	.	PUNCT
easat-10905	273	1	experimental	experimental	ADJ
easat-10905	273	2	results	result	NOUN
easat-10905	273	3	we	we	PRON
easat-10905	273	4	then	then	ADV
easat-10905	273	5	present	present	VERB
easat-10905	273	6	the	the	DET
easat-10905	273	7	results	result	NOUN
easat-10905	273	8	from	from	ADP
easat-10905	273	9	a	a	DET
easat-10905	273	10	series	series	NOUN
easat-10905	273	11	of	of	ADP
easat-10905	273	12	experiments	experiment	NOUN
easat-10905	273	13	conducted	conduct	VERB
easat-10905	273	14	on	on	ADP
easat-10905	273	15	the	the	DET
easat-10905	273	16	developed	develop	VERB
easat-10905	273	17	models	model	NOUN
easat-10905	273	18	.	.	PUNCT
easat-10905	274	1	initially	initially	ADV
easat-10905	274	2	,	,	PUNCT
easat-10905	274	3	the	the	DET
easat-10905	274	4	pre	pre	ADJ
easat-10905	274	5	-	-	ADJ
easat-10905	274	6	trained	train	VERB
easat-10905	274	7	models	model	NOUN
easat-10905	274	8	were	be	AUX
easat-10905	274	9	tested	test	VERB
easat-10905	274	10	on	on	ADP
easat-10905	274	11	abusive	abusive	ADJ
easat-10905	274	12	texts	text	NOUN
easat-10905	274	13	,	,	PUNCT
easat-10905	274	14	followed	follow	VERB
easat-10905	274	15	by	by	ADP
easat-10905	274	16	offensive	offensive	ADJ
easat-10905	274	17	texts	text	NOUN
easat-10905	274	18	,	,	PUNCT
easat-10905	274	19	without	without	ADP
easat-10905	274	20	any	any	DET
easat-10905	274	21	retraining	retraining	NOUN
easat-10905	274	22	.	.	PUNCT
easat-10905	275	1	subsequently	subsequently	ADV
easat-10905	275	2	,	,	PUNCT
easat-10905	275	3	we	we	PRON
easat-10905	275	4	implemented	implement	VERB
easat-10905	275	5	task	task	NOUN
easat-10905	275	6	incremental	incremental	ADJ
easat-10905	275	7	learning	learning	NOUN
easat-10905	275	8	using	use	VERB
easat-10905	275	9	both	both	CCONJ
easat-10905	275	10	replay	replay	NOUN
easat-10905	275	11	-	-	PUNCT
easat-10905	275	12	based	base	VERB
easat-10905	275	13	and	and	CCONJ
easat-10905	275	14	regularization	regularization	NOUN
easat-10905	275	15	-	-	PUNCT
easat-10905	275	16	based	base	VERB
easat-10905	275	17	approaches	approach	NOUN
easat-10905	275	18	.	.	PUNCT
easat-10905	276	1	table	table	NOUN
easat-10905	276	2	2	2	NUM
easat-10905	276	3	shows	show	VERB
easat-10905	276	4	the	the	DET
easat-10905	276	5	performance	performance	NOUN
easat-10905	276	6	of	of	ADP
easat-10905	276	7	the	the	DET
easat-10905	276	8	models	model	NOUN
easat-10905	276	9	on	on	ADP
easat-10905	276	10	abusive	abusive	ADJ
easat-10905	276	11	comments	comment	NOUN
easat-10905	276	12	.	.	PUNCT
easat-10905	277	1	table	table	NOUN
easat-10905	277	2	2	2	NUM
easat-10905	277	3	.	.	PUNCT
easat-10905	277	4	performance	performance	NOUN
easat-10905	277	5	of	of	ADP
easat-10905	277	6	the	the	DET
easat-10905	277	7	classifiers	classifier	NOUN
easat-10905	277	8	on	on	ADP
easat-10905	277	9	abusive	abusive	ADJ
easat-10905	277	10	comments	comment	NOUN
easat-10905	277	11	.	.	PUNCT
easat-10905	278	1	classifiers	classifier	NOUN
easat-10905	278	2	class	class	NOUN
easat-10905	278	3	accuracy	accuracy	NOUN
easat-10905	278	4	recall	recall	PROPN
easat-10905	278	5	precision	precision	PROPN
easat-10905	278	6	f1	f1	PROPN
easat-10905	278	7	score	score	NOUN
easat-10905	278	8	bert	bert	NOUN
easat-10905	278	9	abusive	abusive	ADJ
easat-10905	278	10	91.33	91.33	NUM
easat-10905	278	11	89.83	89.83	NUM
easat-10905	278	12	95.56	95.56	NUM
easat-10905	278	13	92.61	92.61	NUM
easat-10905	278	14	non	non	ADJ
easat-10905	278	15	-	-	ADJ
easat-10905	278	16	abusive	abusive	ADJ
easat-10905	278	17	93.62	93.62	NUM
easat-10905	278	18	85.76	85.76	NUM
easat-10905	278	19	89.51	89.51	NUM
easat-10905	278	20	macro	macro	ADJ
easat-10905	278	21	f1	f1	NOUN
easat-10905	278	22	91.06	91.06	NUM
easat-10905	278	23	weighted	weight	VERB
easat-10905	278	24	f1	f1	NOUN
easat-10905	278	25	91.38	91.38	NUM
easat-10905	278	26	albert	albert	NOUN
easat-10905	278	27	abusive	abusive	NOUN
easat-10905	278	28	77.76	77.76	NUM
easat-10905	278	29	74.00	74.00	NUM
easat-10905	278	30	87.27	87.27	NUM
easat-10905	278	31	80.09	80.09	NUM
easat-10905	278	32	non	non	ADJ
easat-10905	278	33	-	-	ADJ
easat-10905	278	34	abusive	abusive	ADJ
easat-10905	278	35	83.50	83.50	NUM
easat-10905	278	36	67.74	67.74	NUM
easat-10905	278	37	74.80	74.80	NUM
easat-10905	278	38	macro	macro	NOUN
easat-10905	278	39	f1	f1	NOUN
easat-10905	278	40	77.45	77.45	NUM
easat-10905	278	41	weighted	weight	VERB
easat-10905	278	42	f1	f1	NOUN
easat-10905	278	43	78.00	78.00	NUM
easat-10905	278	44	roberta	roberta	PROPN
easat-10905	278	45	abusive	abusive	PROPN
easat-10905	278	46	90.98	90.98	NUM
easat-10905	278	47	95.51	95.51	NUM
easat-10905	278	48	83.90	83.90	NUM
easat-10905	278	49	89.33	89.33	NUM
easat-10905	278	50	non	non	ADJ
easat-10905	278	51	-	-	ADJ
easat-10905	278	52	abusive	abusive	ADJ
easat-10905	278	53	88.01	88.01	NUM
easat-10905	278	54	96.77	96.77	NUM
easat-10905	278	55	92.18	92.18	NUM
easat-10905	278	56	macro	macro	ADJ
easat-10905	278	57	f1	f1	NOUN
easat-10905	278	58	90.76	90.76	NUM
easat-10905	278	59	weighted	weight	VERB
easat-10905	278	60	f1	f1	PROPN
easat-10905	278	61	91.06	91.06	NUM
easat-10905	278	62	distilbert	distilbert	PROPN
easat-10905	278	63	abusive	abusive	ADJ
easat-10905	278	64	92.03	92.03	NUM
easat-10905	278	65	92.90	92.90	NUM
easat-10905	278	66	93.88	93.88	NUM
easat-10905	278	67	93.39	93.39	NUM
easat-10905	278	68	non	non	ADJ
easat-10905	278	69	-	-	ADJ
easat-10905	278	70	abusive	abusive	ADJ
easat-10905	278	71	90.70	90.70	NUM
easat-10905	278	72	89.27	89.27	NUM
easat-10905	278	73	89.98	89.98	NUM
easat-10905	278	74	macro	macro	NOUN
easat-10905	278	75	f1	f1	NOUN
easat-10905	278	76	91.68	91.68	NUM
easat-10905	278	77	weighted	weight	VERB
easat-10905	278	78	f1	f1	NOUN
easat-10905	278	79	92.05	92.05	NUM
easat-10905	278	80	next	next	ADV
easat-10905	278	81	,	,	PUNCT
easat-10905	278	82	we	we	PRON
easat-10905	278	83	evaluated	evaluate	VERB
easat-10905	278	84	the	the	DET
easat-10905	278	85	models	model	NOUN
easat-10905	278	86	on	on	ADP
easat-10905	278	87	an	an	DET
easat-10905	278	88	offensive	offensive	ADJ
easat-10905	278	89	dataset	dataset	NOUN
easat-10905	278	90	to	to	PART
easat-10905	278	91	assess	assess	VERB
easat-10905	278	92	their	their	PRON
easat-10905	278	93	continual	continual	ADJ
easat-10905	278	94	learning	learning	NOUN
easat-10905	278	95	capability	capability	NOUN
easat-10905	278	96	.	.	PUNCT
easat-10905	279	1	the	the	DET
easat-10905	279	2	offensive	offensive	ADJ
easat-10905	279	3	comments	comment	NOUN
easat-10905	279	4	dataset	dataset	VERB
easat-10905	279	5	,	,	PUNCT
easat-10905	279	6	sourced	source	VERB
easat-10905	279	7	from	from	ADP
easat-10905	279	8	huggingface2	huggingface2	NOUN
easat-10905	279	9	contains	contain	VERB
easat-10905	279	10	18,858	18,858	NUM
easat-10905	279	11	rows	row	NOUN
easat-10905	279	12	labeled	label	VERB
easat-10905	279	13	as	as	ADP
easat-10905	279	14	offensive	offensive	ADJ
easat-10905	279	15	and	and	CCONJ
easat-10905	279	16	10,364	10,364	NUM
easat-10905	279	17	rows	row	NOUN
easat-10905	279	18	labeled	label	VERB
easat-10905	279	19	as	as	ADP
easat-10905	279	20	non	non	ADJ
easat-10905	279	21	-	-	ADJ
easat-10905	279	22	offensive	offensive	ADJ
easat-10905	279	23	.	.	PUNCT
easat-10905	280	1	initially	initially	ADV
easat-10905	280	2	,	,	PUNCT
easat-10905	280	3	the	the	DET
easat-10905	280	4	models	model	NOUN
easat-10905	280	5	were	be	AUX
easat-10905	280	6	tested	test	VERB
easat-10905	280	7	without	without	ADP
easat-10905	280	8	any	any	DET
easat-10905	280	9	retraining	retraining	ADJ
easat-10905	280	10	or	or	CCONJ
easat-10905	280	11	fine	fine	ADV
easat-10905	280	12	-	-	PUNCT
easat-10905	280	13	tuning	tuning	NOUN
easat-10905	280	14	,	,	PUNCT
easat-10905	280	15	and	and	CCONJ
easat-10905	280	16	their	their	PRON
easat-10905	280	17	performance	performance	NOUN
easat-10905	280	18	is	be	AUX
easat-10905	280	19	detailed	detail	VERB
easat-10905	280	20	in	in	ADP
easat-10905	280	21	table	table	NOUN
easat-10905	280	22	3	3	NUM
easat-10905	280	23	.	.	PUNCT
easat-10905	280	24	table	table	NOUN
easat-10905	280	25	3	3	NUM
easat-10905	280	26	.	.	PUNCT
easat-10905	281	1	performance	performance	NOUN
easat-10905	281	2	of	of	ADP
easat-10905	281	3	the	the	DET
easat-10905	281	4	classifiers	classifier	NOUN
easat-10905	281	5	on	on	ADP
easat-10905	281	6	offensive	offensive	ADJ
easat-10905	281	7	comments	comment	NOUN
easat-10905	281	8	without	without	ADP
easat-10905	281	9	retraining	retrain	VERB
easat-10905	281	10	.	.	PUNCT
easat-10905	282	1	classifiers	classifier	NOUN
easat-10905	282	2	class	class	NOUN
easat-10905	282	3	accuracy	accuracy	NOUN
easat-10905	282	4	recall	recall	PROPN
easat-10905	282	5	precision	precision	PROPN
easat-10905	282	6	f1	f1	PROPN
easat-10905	282	7	score	score	NOUN
easat-10905	282	8	bert	bert	PROPN
easat-10905	282	9	offensive	offensive	VERB
easat-10905	282	10	62.24	62.24	NUM
easat-10905	282	11	45.90	45.90	NUM
easat-10905	282	12	92.14	92.14	NUM
easat-10905	282	13	61.27	61.27	NUM
easat-10905	282	14	not	not	PART
easat-10905	282	15	offensive	offensive	VERB
easat-10905	282	16	92.70	92.70	NUM
easat-10905	282	17	47.90	47.90	NUM
easat-10905	282	18	63.16	63.16	NUM
easat-10905	282	19	macro	macro	ADJ
easat-10905	282	20	f1	f1	NOUN
easat-10905	282	21	62.22	62.22	NUM
easat-10905	282	22	weighted	weight	VERB
easat-10905	282	23	f1	f1	PROPN
easat-10905	282	24	61.93	61.93	NUM
easat-10905	282	25	albert	albert	PROPN
easat-10905	282	26	offensive	offensive	PROPN
easat-10905	282	27	73.19	73.19	NUM
easat-10905	282	28	66.48	66.48	NUM
easat-10905	282	29	89.65	89.65	NUM
easat-10905	282	30	76.35	76.35	NUM
easat-10905	282	31	not	not	PART
easat-10905	282	32	offensive	offensive	ADJ
easat-10905	282	33	85.69	85.69	NUM
easat-10905	282	34	57.84	57.84	NUM
easat-10905	282	35	69.06	69.06	NUM
easat-10905	282	36	macro	macro	NOUN
easat-10905	282	37	f1	f1	PROPN
easat-10905	282	38	72.70	72.70	NUM
easat-10905	282	39	weighted	weight	VERB
easat-10905	282	40	f1	f1	PROPN
easat-10905	282	41	73.80	73.80	NUM
easat-10905	282	42	roberta	roberta	PROPN
easat-10905	282	43	offensive	offensive	ADJ
easat-10905	282	44	69.24	69.24	NUM
easat-10905	282	45	69.87	69.87	NUM
easat-10905	282	46	80.30	80.30	NUM
easat-10905	282	47	74.73	74.73	NUM
easat-10905	282	48	not	not	PART
easat-10905	282	49	offensive	offensive	ADJ
easat-10905	282	50	68.05	68.05	NUM
easat-10905	282	51	54.79	54.79	NUM
easat-10905	282	52	60.71	60.71	NUM
easat-10905	282	53	macro	macro	NOUN
easat-10905	282	54	f1	f1	NOUN
easat-10905	282	55	67.72	67.72	NUM
easat-10905	282	56	weighted	weight	VERB
easat-10905	282	57	f1	f1	PROPN
easat-10905	282	58	69.83	69.83	NUM
easat-10905	282	59	distilbert	distilbert	NOUN
easat-10905	282	60	offensive	offensive	ADJ
easat-10905	282	61	60.17	60.17	NUM
easat-10905	282	62	43.69	43.69	NUM
easat-10905	282	63	89.94	89.94	NUM
easat-10905	282	64	58.81	58.81	NUM
easat-10905	282	65	not	not	PART
easat-10905	282	66	offensive	offensive	ADJ
easat-10905	282	67	90.89	90.89	NUM
easat-10905	282	68	46.41	46.41	NUM
easat-10905	282	69	61.44	61.44	NUM
easat-10905	282	70	macro	macro	ADJ
easat-10905	282	71	f1	f1	PROPN
easat-10905	282	72	60.13	60.13	NUM
easat-10905	282	73	weighted	weight	VERB
easat-10905	282	74	f1	f1	PROPN
easat-10905	282	75	59.73	59.73	NUM
easat-10905	282	76	as	as	SCONJ
easat-10905	282	77	seen	see	VERB
easat-10905	282	78	in	in	ADP
easat-10905	282	79	table	table	NOUN
easat-10905	282	80	3	3	NUM
easat-10905	282	81	,	,	PUNCT
easat-10905	282	82	the	the	DET
easat-10905	282	83	performance	performance	NOUN
easat-10905	282	84	of	of	ADP
easat-10905	282	85	the	the	DET
easat-10905	282	86	models	model	NOUN
easat-10905	282	87	on	on	ADP
easat-10905	282	88	the	the	DET
easat-10905	282	89	offensive	offensive	ADJ
easat-10905	282	90	dataset	dataset	NOUN
easat-10905	282	91	is	be	AUX
easat-10905	282	92	relatively	relatively	ADV
easat-10905	282	93	lower	low	ADJ
easat-10905	282	94	compared	compare	VERB
easat-10905	282	95	to	to	ADP
easat-10905	282	96	the	the	DET
easat-10905	282	97	abusive	abusive	ADJ
easat-10905	282	98	dataset	dataset	NOUN
easat-10905	282	99	.	.	PUNCT
easat-10905	283	1	this	this	DET
easat-10905	283	2	difference	difference	NOUN
easat-10905	283	3	arises	arise	VERB
easat-10905	283	4	because	because	SCONJ
easat-10905	283	5	the	the	DET
easat-10905	283	6	models	model	NOUN
easat-10905	283	7	were	be	AUX
easat-10905	283	8	trained	train	VERB
easat-10905	283	9	on	on	ADP
easat-10905	283	10	abusive	abusive	ADJ
easat-10905	283	11	https://huggingface.co/datasets/christinacdl/offensive_language_dataset2	https://huggingface.co/datasets/christinacdl/offensive_language_dataset2	NOUN
easat-10905	283	12	458	458	NUM
easat-10905	283	13	edelweiss	edelweiss	PROPN
easat-10905	283	14	applied	apply	VERB
easat-10905	283	15	science	science	NOUN
easat-10905	283	16	and	and	CCONJ
easat-10905	283	17	technology	technology	NOUN
easat-10905	283	18	issn	issn	PROPN
easat-10905	283	19	:	:	PUNCT
easat-10905	283	20	2576	2576	NUM
easat-10905	283	21	-	-	SYM
easat-10905	283	22	8484	8484	NUM
easat-10905	283	23	vol	vol	NOUN
easat-10905	283	24	.	.	PROPN
easat-10905	284	1	9	9	NUM
easat-10905	284	2	,	,	PUNCT
easat-10905	284	3	no	no	INTJ
easat-10905	284	4	.	.	NOUN
easat-10905	284	5	11	11	NUM
easat-10905	284	6	:	:	PUNCT
easat-10905	284	7	448	448	NUM
easat-10905	284	8	-	-	SYM
easat-10905	284	9	466	466	NUM
easat-10905	284	10	,	,	PUNCT
easat-10905	284	11	2025	2025	NUM
easat-10905	284	12	doi	doi	NOUN
easat-10905	284	13	:	:	PUNCT
easat-10905	284	14	10.55214/2576	10.55214/2576	NUM
easat-10905	284	15	-	-	PUNCT
easat-10905	284	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	285	1	©	©	PROPN
easat-10905	285	2	2025	2025	NUM
easat-10905	285	3	by	by	ADP
easat-10905	285	4	the	the	DET
easat-10905	285	5	authors	author	NOUN
easat-10905	285	6	;	;	PUNCT
easat-10905	285	7	licensee	licensee	PROPN
easat-10905	285	8	learning	learn	VERB
easat-10905	285	9	gate	gate	NOUN
easat-10905	285	10	comments	comment	NOUN
easat-10905	285	11	and	and	CCONJ
easat-10905	285	12	therefore	therefore	ADV
easat-10905	285	13	learned	learn	VERB
easat-10905	285	14	parameters	parameter	NOUN
easat-10905	285	15	specific	specific	ADJ
easat-10905	285	16	to	to	ADP
easat-10905	285	17	that	that	DET
easat-10905	285	18	dataset	dataset	NOUN
easat-10905	285	19	.	.	PUNCT
easat-10905	286	1	when	when	SCONJ
easat-10905	286	2	applied	apply	VERB
easat-10905	286	3	to	to	ADP
easat-10905	286	4	the	the	DET
easat-10905	286	5	offensive	offensive	ADJ
easat-10905	286	6	dataset	dataset	NOUN
easat-10905	286	7	,	,	PUNCT
easat-10905	286	8	the	the	DET
easat-10905	286	9	knowledge	knowledge	NOUN
easat-10905	286	10	gained	gain	VERB
easat-10905	286	11	from	from	ADP
easat-10905	286	12	the	the	DET
easat-10905	286	13	previous	previous	ADJ
easat-10905	286	14	task	task	NOUN
easat-10905	286	15	did	do	AUX
easat-10905	286	16	not	not	PART
easat-10905	286	17	significantly	significantly	ADV
easat-10905	286	18	enhance	enhance	VERB
easat-10905	286	19	the	the	DET
easat-10905	286	20	classification	classification	NOUN
easat-10905	286	21	accuracy	accuracy	NOUN
easat-10905	286	22	.	.	PUNCT
easat-10905	287	1	therefore	therefore	ADV
easat-10905	287	2	,	,	PUNCT
easat-10905	287	3	we	we	PRON
easat-10905	287	4	incorporated	incorporate	VERB
easat-10905	287	5	continual	continual	ADJ
easat-10905	287	6	learning	learning	NOUN
easat-10905	287	7	approaches	approach	NOUN
easat-10905	287	8	into	into	ADP
easat-10905	287	9	the	the	DET
easat-10905	287	10	developed	develop	VERB
easat-10905	287	11	models	model	NOUN
easat-10905	287	12	to	to	PART
easat-10905	287	13	preserve	preserve	VERB
easat-10905	287	14	existing	exist	VERB
easat-10905	287	15	knowledge	knowledge	NOUN
easat-10905	287	16	while	while	SCONJ
easat-10905	287	17	acquiring	acquire	VERB
easat-10905	287	18	new	new	ADJ
easat-10905	287	19	knowledge	knowledge	NOUN
easat-10905	287	20	for	for	ADP
easat-10905	287	21	different	different	ADJ
easat-10905	287	22	tasks	task	NOUN
easat-10905	287	23	.	.	PUNCT
easat-10905	288	1	first	first	ADV
easat-10905	288	2	,	,	PUNCT
easat-10905	288	3	we	we	PRON
easat-10905	288	4	implemented	implement	VERB
easat-10905	288	5	replay	replay	NOUN
easat-10905	288	6	-	-	PUNCT
easat-10905	288	7	based	base	VERB
easat-10905	288	8	approaches	approach	NOUN
easat-10905	288	9	,	,	PUNCT
easat-10905	288	10	namely	namely	ADV
easat-10905	288	11	rehearsal	rehearsal	NOUN
easat-10905	288	12	and	and	CCONJ
easat-10905	288	13	pseudo	pseudo	NOUN
easat-10905	288	14	-	-	ADJ
easat-10905	288	15	rehearsal	rehearsal	NOUN
easat-10905	288	16	methods	method	NOUN
easat-10905	288	17	,	,	PUNCT
easat-10905	288	18	as	as	SCONJ
easat-10905	288	19	discussed	discuss	VERB
easat-10905	288	20	in	in	ADP
easat-10905	288	21	section	section	NOUN
easat-10905	288	22	3.3	3.3	NUM
easat-10905	288	23	.	.	PUNCT
easat-10905	289	1	the	the	DET
easat-10905	289	2	performance	performance	NOUN
easat-10905	289	3	of	of	ADP
easat-10905	289	4	the	the	DET
easat-10905	289	5	models	model	NOUN
easat-10905	289	6	using	use	VERB
easat-10905	289	7	replay	replay	NOUN
easat-10905	289	8	-	-	PUNCT
easat-10905	289	9	based	base	VERB
easat-10905	289	10	approaches	approach	NOUN
easat-10905	289	11	is	be	AUX
easat-10905	289	12	presented	present	VERB
easat-10905	289	13	in	in	ADP
easat-10905	289	14	tables	table	NOUN
easat-10905	289	15	4	4	NUM
easat-10905	289	16	and	and	CCONJ
easat-10905	289	17	5	5	NUM
easat-10905	289	18	.	.	PUNCT
easat-10905	290	1	these	these	DET
easat-10905	290	2	tables	table	NOUN
easat-10905	290	3	depict	depict	VERB
easat-10905	290	4	the	the	DET
easat-10905	290	5	performance	performance	NOUN
easat-10905	290	6	of	of	ADP
easat-10905	290	7	the	the	DET
easat-10905	290	8	models	model	NOUN
easat-10905	290	9	on	on	ADP
easat-10905	290	10	both	both	CCONJ
easat-10905	290	11	the	the	DET
easat-10905	290	12	old	old	ADJ
easat-10905	290	13	task	task	NOUN
easat-10905	290	14	(	(	PUNCT
easat-10905	290	15	abusive	abusive	ADJ
easat-10905	290	16	comments	comment	NOUN
easat-10905	290	17	)	)	PUNCT
easat-10905	290	18	and	and	CCONJ
easat-10905	290	19	the	the	DET
easat-10905	290	20	new	new	ADJ
easat-10905	290	21	task	task	NOUN
easat-10905	290	22	(	(	PUNCT
easat-10905	290	23	offensive	offensive	ADJ
easat-10905	290	24	comments	comment	NOUN
easat-10905	290	25	)	)	PUNCT
easat-10905	290	26	and	and	CCONJ
easat-10905	290	27	clearly	clearly	ADV
easat-10905	290	28	show	show	VERB
easat-10905	290	29	that	that	SCONJ
easat-10905	290	30	after	after	ADP
easat-10905	290	31	incorporating	incorporate	VERB
easat-10905	290	32	replay	replay	NOUN
easat-10905	290	33	-	-	PUNCT
easat-10905	290	34	based	base	VERB
easat-10905	290	35	approaches	approach	NOUN
easat-10905	290	36	,	,	PUNCT
easat-10905	290	37	the	the	DET
easat-10905	290	38	performance	performance	NOUN
easat-10905	290	39	on	on	ADP
easat-10905	290	40	the	the	DET
easat-10905	290	41	abusive	abusive	ADJ
easat-10905	290	42	dataset	dataset	NOUN
easat-10905	290	43	declined	decline	VERB
easat-10905	290	44	slightly	slightly	ADV
easat-10905	290	45	compared	compare	VERB
easat-10905	290	46	to	to	ADP
easat-10905	290	47	before	before	ADP
easat-10905	290	48	their	their	PRON
easat-10905	290	49	implementation	implementation	NOUN
easat-10905	290	50	,	,	PUNCT
easat-10905	290	51	while	while	SCONJ
easat-10905	290	52	performance	performance	NOUN
easat-10905	290	53	on	on	ADP
easat-10905	290	54	the	the	DET
easat-10905	290	55	offensive	offensive	ADJ
easat-10905	290	56	dataset	dataset	NOUN
easat-10905	290	57	improved	improve	VERB
easat-10905	290	58	significantly	significantly	ADV
easat-10905	290	59	.	.	PUNCT
easat-10905	291	1	this	this	PRON
easat-10905	291	2	indicates	indicate	VERB
easat-10905	291	3	that	that	SCONJ
easat-10905	291	4	the	the	DET
easat-10905	291	5	models	model	NOUN
easat-10905	291	6	are	be	AUX
easat-10905	291	7	attempting	attempt	VERB
easat-10905	291	8	to	to	PART
easat-10905	291	9	balance	balance	VERB
easat-10905	291	10	both	both	DET
easat-10905	291	11	datasets	dataset	NOUN
easat-10905	291	12	using	use	VERB
easat-10905	291	13	the	the	DET
easat-10905	291	14	replay	replay	NOUN
easat-10905	291	15	-	-	PUNCT
easat-10905	291	16	based	base	VERB
easat-10905	291	17	approach	approach	NOUN
easat-10905	291	18	.	.	PUNCT
easat-10905	292	1	notably	notably	ADV
easat-10905	292	2	,	,	PUNCT
easat-10905	292	3	pseudo	pseudo	NOUN
easat-10905	292	4	-	-	ADJ
easat-10905	292	5	rehearsal	rehearsal	NOUN
easat-10905	292	6	approaches	approach	NOUN
easat-10905	292	7	demonstrated	demonstrate	VERB
easat-10905	292	8	better	well	ADJ
easat-10905	292	9	performance	performance	NOUN
easat-10905	292	10	compared	compare	VERB
easat-10905	292	11	to	to	ADP
easat-10905	292	12	rehearsal	rehearsal	NOUN
easat-10905	292	13	approaches	approach	NOUN
easat-10905	292	14	.	.	PUNCT
easat-10905	293	1	table	table	NOUN
easat-10905	293	2	4	4	NUM
easat-10905	293	3	.	.	PUNCT
easat-10905	294	1	performance	performance	NOUN
easat-10905	294	2	of	of	ADP
easat-10905	294	3	the	the	DET
easat-10905	294	4	classifiers	classifier	NOUN
easat-10905	294	5	on	on	ADP
easat-10905	294	6	abusive	abusive	ADJ
easat-10905	294	7	comments	comment	NOUN
easat-10905	294	8	using	use	VERB
easat-10905	294	9	replay	replay	NOUN
easat-10905	294	10	-	-	PUNCT
easat-10905	294	11	based	base	VERB
easat-10905	294	12	approaches	approach	NOUN
easat-10905	294	13	.	.	PUNCT
easat-10905	295	1	a.	a.	PROPN
easat-10905	295	2	rehearsal	rehearsal	PROPN
easat-10905	295	3	-	-	PUNCT
easat-10905	295	4	based	base	VERB
easat-10905	295	5	approach	approach	NOUN
easat-10905	295	6	.	.	PUNCT
easat-10905	296	1	classifiers	classifier	NOUN
easat-10905	296	2	class	class	NOUN
easat-10905	296	3	accuracy	accuracy	NOUN
easat-10905	296	4	recall	recall	PROPN
easat-10905	296	5	precision	precision	PROPN
easat-10905	296	6	f1	f1	PROPN
easat-10905	296	7	score	score	NOUN
easat-10905	296	8	bert	bert	NOUN
easat-10905	296	9	abusive	abusive	PROPN
easat-10905	296	10	73.20	73.20	NUM
easat-10905	296	11	93.91	93.91	NUM
easat-10905	296	12	66.05	66.05	NUM
easat-10905	296	13	77.55	77.55	NUM
easat-10905	296	14	non	non	ADJ
easat-10905	296	15	-	-	ADJ
easat-10905	296	16	abusive	abusive	ADJ
easat-10905	296	17	53.06	53.06	NUM
easat-10905	296	18	89.97	89.97	NUM
easat-10905	296	19	66.75	66.75	NUM
easat-10905	296	20	macro	macro	NOUN
easat-10905	296	21	f1	f1	PROPN
easat-10905	296	22	72.15	72.15	NUM
easat-10905	296	23	weighted	weight	VERB
easat-10905	296	24	f1	f1	PROPN
easat-10905	296	25	72.08	72.08	NUM
easat-10905	296	26	albert	albert	PROPN
easat-10905	296	27	abusive	abusive	NOUN
easat-10905	296	28	73.50	73.50	NUM
easat-10905	296	29	87.63	87.63	NUM
easat-10905	296	30	67.92	67.92	NUM
easat-10905	296	31	76.53	76.53	NUM
easat-10905	296	32	non	non	ADJ
easat-10905	296	33	-	-	ADJ
easat-10905	296	34	abusive	abusive	ADJ
easat-10905	296	35	59.76	59.76	NUM
easat-10905	296	36	83.24	83.24	NUM
easat-10905	296	37	69.58	69.58	NUM
easat-10905	296	38	macro	macro	ADJ
easat-10905	296	39	f1	f1	PROPN
easat-10905	296	40	73.05	73.05	NUM
easat-10905	296	41	weighted	weight	VERB
easat-10905	296	42	f1	f1	NOUN
easat-10905	296	43	73.00	73.00	NUM
easat-10905	296	44	roberta	roberta	PROPN
easat-10905	296	45	abusive	abusive	NOUN
easat-10905	296	46	75.80	75.80	NUM
easat-10905	296	47	71.20	71.20	NUM
easat-10905	296	48	77.83	77.83	NUM
easat-10905	296	49	74.36	74.36	NUM
easat-10905	296	50	non	non	ADJ
easat-10905	296	51	-	-	ADJ
easat-10905	296	52	abusive	abusive	ADJ
easat-10905	296	53	80.28	80.28	NUM
easat-10905	296	54	74.13	74.13	NUM
easat-10905	296	55	77.08	77.08	NUM
easat-10905	296	56	macro	macro	ADJ
easat-10905	296	57	f1	f1	NOUN
easat-10905	296	58	75.72	75.72	NUM
easat-10905	296	59	weighted	weight	VERB
easat-10905	296	60	f1	f1	NOUN
easat-10905	296	61	75.74	75.74	NUM
easat-10905	296	62	distilbert	distilbert	NOUN
easat-10905	296	63	abusive	abusive	ADJ
easat-10905	296	64	76.10	76.10	NUM
easat-10905	296	65	75.74	75.74	NUM
easat-10905	296	66	75.40	75.40	NUM
easat-10905	296	67	75.93	75.93	NUM
easat-10905	296	68	non	non	ADJ
easat-10905	296	69	-	-	ADJ
easat-10905	296	70	abusive	abusive	ADJ
easat-10905	296	71	76.47	76.47	NUM
easat-10905	296	72	76.80	76.80	NUM
easat-10905	296	73	76.27	76.27	NUM
easat-10905	296	74	macro	macro	NOUN
easat-10905	296	75	f1	f1	NOUN
easat-10905	296	76	76.10	76.10	NUM
easat-10905	296	77	weighted	weight	VERB
easat-10905	296	78	f1	f1	PROPN
easat-10905	296	79	76.10	76.10	NUM
easat-10905	296	80	b.	b.	NOUN
easat-10905	296	81	pseudo	pseudo	NOUN
easat-10905	296	82	rehearsal	rehearsal	NOUN
easat-10905	296	83	approach	approach	NOUN
easat-10905	296	84	classifiers	classifier	NOUN
easat-10905	296	85	class	class	NOUN
easat-10905	296	86	accuracy	accuracy	NOUN
easat-10905	296	87	recall	recall	PROPN
easat-10905	296	88	precision	precision	PROPN
easat-10905	296	89	f1	f1	PROPN
easat-10905	296	90	score	score	NOUN
easat-10905	296	91	bert	bert	NOUN
easat-10905	296	92	abusive	abusive	ADJ
easat-10905	296	93	89.34	89.34	NUM
easat-10905	296	94	81.56	81.56	NUM
easat-10905	296	95	96.59	96.59	NUM
easat-10905	296	96	88.44	88.44	NUM
easat-10905	296	97	non	non	ADJ
easat-10905	296	98	-	-	ADJ
easat-10905	296	99	abusive	abusive	ADJ
easat-10905	296	100	97.12	97.12	NUM
easat-10905	296	101	84.04	84.04	NUM
easat-10905	296	102	90.11	90.11	NUM
easat-10905	296	103	macro	macro	NOUN
easat-10905	296	104	f1	f1	PROPN
easat-10905	296	105	89.28	89.28	NUM
easat-10905	296	106	weighted	weight	VERB
easat-10905	296	107	f1	f1	NOUN
easat-10905	296	108	89.28	89.28	NUM
easat-10905	296	109	albert	albert	NOUN
easat-10905	296	110	abusive	abusive	NOUN
easat-10905	296	111	78.00	78.00	NUM
easat-10905	296	112	76.00	76.00	NUM
easat-10905	296	113	80.00	80.00	NUM
easat-10905	296	114	78.00	78.00	NUM
easat-10905	296	115	non	non	ADJ
easat-10905	296	116	-	-	ADJ
easat-10905	296	117	abusive	abusive	ADJ
easat-10905	296	118	81.00	81.00	NUM
easat-10905	296	119	77.00	77.00	NUM
easat-10905	296	120	79.00	79.00	NUM
easat-10905	296	121	macro	macro	NOUN
easat-10905	296	122	f1	f1	NOUN
easat-10905	296	123	78.00	78.00	NUM
easat-10905	296	124	weighted	weight	VERB
easat-10905	296	125	f1	f1	NOUN
easat-10905	296	126	78.00	78.00	NUM
easat-10905	296	127	roberta	roberta	PROPN
easat-10905	296	128	abusive	abusive	PROPN
easat-10905	296	129	86.62	86.62	NUM
easat-10905	296	130	85.48	85.48	NUM
easat-10905	296	131	87.47	87.47	NUM
easat-10905	296	132	86.47	86.47	NUM
easat-10905	296	133	non	non	ADJ
easat-10905	296	134	-	-	ADJ
easat-10905	296	135	abusive	abusive	ADJ
easat-10905	296	136	87.76	87.76	NUM
easat-10905	296	137	85.80	85.80	NUM
easat-10905	296	138	86.77	86.77	NUM
easat-10905	296	139	macro	macro	ADJ
easat-10905	296	140	f1	f1	PROPN
easat-10905	296	141	86.62	86.62	NUM
easat-10905	296	142	weighted	weight	VERB
easat-10905	296	143	f1	f1	PROPN
easat-10905	296	144	86.62	86.62	NUM
easat-10905	296	145	distilbert	distilbert	PROPN
easat-10905	296	146	abusive	abusive	ADJ
easat-10905	296	147	90.72	90.72	NUM
easat-10905	296	148	98.36	98.36	NUM
easat-10905	296	149	85.32	85.32	NUM
easat-10905	296	150	91.38	91.38	NUM
easat-10905	296	151	non	non	ADJ
easat-10905	296	152	-	-	ADJ
easat-10905	296	153	abusive	abusive	ADJ
easat-10905	296	154	83.08	83.08	NUM
easat-10905	296	155	98.06	98.06	NUM
easat-10905	296	156	89.95	89.95	NUM
easat-10905	296	157	macro	macro	NOUN
easat-10905	296	158	f1	f1	PROPN
easat-10905	296	159	90.67	90.67	NUM
easat-10905	296	160	weighted	weight	VERB
easat-10905	296	161	f1	f1	NOUN
easat-10905	296	162	90.67	90.67	NUM
easat-10905	296	163	459	459	NUM
easat-10905	296	164	edelweiss	edelweiss	PROPN
easat-10905	296	165	applied	apply	VERB
easat-10905	296	166	science	science	NOUN
easat-10905	296	167	and	and	CCONJ
easat-10905	296	168	technology	technology	NOUN
easat-10905	296	169	issn	issn	PROPN
easat-10905	296	170	:	:	PUNCT
easat-10905	296	171	2576	2576	NUM
easat-10905	296	172	-	-	SYM
easat-10905	296	173	8484	8484	NUM
easat-10905	296	174	vol	vol	NOUN
easat-10905	296	175	.	.	PROPN
easat-10905	297	1	9	9	NUM
easat-10905	297	2	,	,	PUNCT
easat-10905	297	3	no	no	INTJ
easat-10905	297	4	.	.	NOUN
easat-10905	297	5	11	11	NUM
easat-10905	297	6	:	:	PUNCT
easat-10905	297	7	448	448	NUM
easat-10905	297	8	-	-	SYM
easat-10905	297	9	466	466	NUM
easat-10905	297	10	,	,	PUNCT
easat-10905	297	11	2025	2025	NUM
easat-10905	297	12	doi	doi	NOUN
easat-10905	297	13	:	:	PUNCT
easat-10905	297	14	10.55214/2576	10.55214/2576	NUM
easat-10905	297	15	-	-	PUNCT
easat-10905	297	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	298	1	©	©	PROPN
easat-10905	298	2	2025	2025	NUM
easat-10905	298	3	by	by	ADP
easat-10905	298	4	the	the	DET
easat-10905	298	5	authors	author	NOUN
easat-10905	298	6	;	;	PUNCT
easat-10905	298	7	licensee	licensee	PROPN
easat-10905	298	8	learning	learn	VERB
easat-10905	298	9	gate	gate	NOUN
easat-10905	298	10	table	table	NOUN
easat-10905	298	11	5	5	NUM
easat-10905	298	12	.	.	PUNCT
easat-10905	299	1	performance	performance	NOUN
easat-10905	299	2	of	of	ADP
easat-10905	299	3	the	the	DET
easat-10905	299	4	classifiers	classifier	NOUN
easat-10905	299	5	on	on	ADP
easat-10905	299	6	offensive	offensive	ADJ
easat-10905	299	7	comments	comment	NOUN
easat-10905	299	8	using	use	VERB
easat-10905	299	9	replay	replay	NOUN
easat-10905	299	10	-	-	PUNCT
easat-10905	299	11	based	base	VERB
easat-10905	299	12	approaches	approach	NOUN
easat-10905	299	13	.	.	PUNCT
easat-10905	300	1	a.	a.	PROPN
easat-10905	300	2	rehearsal	rehearsal	PROPN
easat-10905	300	3	-	-	PUNCT
easat-10905	300	4	based	base	VERB
easat-10905	300	5	approach	approach	NOUN
easat-10905	300	6	classifiers	classifier	NOUN
easat-10905	300	7	class	class	NOUN
easat-10905	300	8	accuracy	accuracy	NOUN
easat-10905	300	9	recall	recall	PROPN
easat-10905	300	10	precision	precision	PROPN
easat-10905	300	11	f1	f1	PROPN
easat-10905	300	12	score	score	NOUN
easat-10905	300	13	bert	bert	PROPN
easat-10905	300	14	offensive	offensive	VERB
easat-10905	300	15	87.60	87.60	NUM
easat-10905	300	16	92.05	92.05	NUM
easat-10905	300	17	83.65	83.65	NUM
easat-10905	300	18	87.65	87.65	NUM
easat-10905	300	19	not	not	PART
easat-10905	300	20	offensive	offensive	ADJ
easat-10905	300	21	83.52	83.52	NUM
easat-10905	300	22	91.98	91.98	NUM
easat-10905	300	23	87.55	87.55	NUM
easat-10905	300	24	macro	macro	ADJ
easat-10905	300	25	f1	f1	PROPN
easat-10905	300	26	87.60	87.60	NUM
easat-10905	300	27	weighted	weight	VERB
easat-10905	300	28	f1	f1	PROPN
easat-10905	300	29	87.60	87.60	NUM
easat-10905	300	30	albert	albert	PROPN
easat-10905	300	31	offensive	offensive	PROPN
easat-10905	300	32	73.40	73.40	NUM
easat-10905	300	33	94.56	94.56	NUM
easat-10905	300	34	65.32	65.32	NUM
easat-10905	300	35	77.26	77.26	NUM
easat-10905	300	36	not	not	PART
easat-10905	300	37	offensive	offensive	ADJ
easat-10905	300	38	54.02	54.02	NUM
easat-10905	300	39	91.56	91.56	NUM
easat-10905	300	40	67.95	67.95	NUM
easat-10905	300	41	macro	macro	ADJ
easat-10905	300	42	f1	f1	NOUN
easat-10905	300	43	72.61	72.61	NUM
easat-10905	300	44	weighted	weight	VERB
easat-10905	300	45	f1	f1	PROPN
easat-10905	300	46	72.40	72.40	NUM
easat-10905	300	47	roberta	roberta	PROPN
easat-10905	300	48	offensive	offensive	VERB
easat-10905	300	49	81.60	81.60	NUM
easat-10905	300	50	81.59	81.59	NUM
easat-10905	300	51	80.25	80.25	NUM
easat-10905	300	52	80.91	80.91	NUM
easat-10905	300	53	not	not	PART
easat-10905	300	54	offensive	offensive	ADJ
easat-10905	300	55	81.61	81.61	NUM
easat-10905	300	56	82.88	82.88	NUM
easat-10905	300	57	82.24	82.24	NUM
easat-10905	300	58	macro	macro	ADJ
easat-10905	300	59	f1	f1	PROPN
easat-10905	300	60	81.58	81.58	NUM
easat-10905	300	61	weighted	weight	VERB
easat-10905	300	62	f1	f1	PROPN
easat-10905	300	63	81.61	81.61	NUM
easat-10905	300	64	distilbert	distilbert	NOUN
easat-10905	300	65	offensive	offensive	ADJ
easat-10905	300	66	90.80	90.80	NUM
easat-10905	300	67	90.79	90.79	NUM
easat-10905	300	68	90.04	90.04	NUM
easat-10905	300	69	90.42	90.42	NUM
easat-10905	300	70	not	not	PART
easat-10905	300	71	offensive	offensive	VERB
easat-10905	300	72	90.80	90.80	NUM
easat-10905	300	73	91.51	91.51	NUM
easat-10905	300	74	91.15	91.15	NUM
easat-10905	300	75	macro	macro	ADJ
easat-10905	300	76	f1	f1	NOUN
easat-10905	300	77	90.79	90.79	NUM
easat-10905	300	78	weighted	weight	VERB
easat-10905	300	79	f1	f1	PROPN
easat-10905	300	80	90.80	90.80	NUM
easat-10905	300	81	b.	b.	PROPN
easat-10905	300	82	pseudo	pseudo	PROPN
easat-10905	300	83	rehearsal	rehearsal	NOUN
easat-10905	300	84	approach	approach	NOUN
easat-10905	300	85	classifiers	classifier	NOUN
easat-10905	300	86	class	class	NOUN
easat-10905	300	87	accuracy	accuracy	NOUN
easat-10905	300	88	recall	recall	PROPN
easat-10905	300	89	precision	precision	PROPN
easat-10905	300	90	f1	f1	PROPN
easat-10905	300	91	score	score	NOUN
easat-10905	300	92	bert	bert	PROPN
easat-10905	300	93	offensive	offensive	VERB
easat-10905	300	94	98.02	98.02	NUM
easat-10905	300	95	97.72	97.72	NUM
easat-10905	300	96	97.73	97.73	NUM
easat-10905	300	97	98.01	98.01	NUM
easat-10905	300	98	not	not	PART
easat-10905	300	99	offensive	offensive	ADJ
easat-10905	300	100	98.32	98.32	NUM
easat-10905	300	101	98.31	98.31	NUM
easat-10905	300	102	98.03	98.03	NUM
easat-10905	300	103	macro	macro	ADJ
easat-10905	300	104	f1	f1	NOUN
easat-10905	300	105	98.02	98.02	NUM
easat-10905	300	106	weighted	weight	VERB
easat-10905	300	107	f1	f1	PROPN
easat-10905	300	108	98.02	98.02	NUM
easat-10905	300	109	albert	albert	PROPN
easat-10905	300	110	offensive	offensive	PROPN
easat-10905	300	111	93.00	93.00	NUM
easat-10905	300	112	88.00	88.00	NUM
easat-10905	300	113	97.00	97.00	NUM
easat-10905	300	114	92.00	92.00	NUM
easat-10905	300	115	not	not	PART
easat-10905	300	116	offensive	offensive	ADJ
easat-10905	300	117	97.00	97.00	NUM
easat-10905	300	118	88.00	88.00	NUM
easat-10905	300	119	93.00	93.00	NUM
easat-10905	300	120	macro	macro	NOUN
easat-10905	300	121	f1	f1	PROPN
easat-10905	300	122	93.00	93.00	NUM
easat-10905	300	123	weighted	weight	VERB
easat-10905	300	124	f1	f1	PROPN
easat-10905	300	125	93.00	93.00	NUM
easat-10905	300	126	roberta	roberta	PROPN
easat-10905	300	127	offensive	offensive	PROPN
easat-10905	300	128	95.50	95.50	NUM
easat-10905	300	129	94.40	94.40	NUM
easat-10905	300	130	96.52	96.52	NUM
easat-10905	300	131	95.45	95.45	NUM
easat-10905	300	132	not	not	PART
easat-10905	300	133	offensive	offensive	ADJ
easat-10905	300	134	96.60	96.60	NUM
easat-10905	300	135	94.52	94.52	NUM
easat-10905	300	136	95.55	95.55	NUM
easat-10905	300	137	macro	macro	NOUN
easat-10905	300	138	f1	f1	PROPN
easat-10905	300	139	95.50	95.50	NUM
easat-10905	300	140	weighted	weight	VERB
easat-10905	300	141	f1	f1	PROPN
easat-10905	300	142	95.50	95.50	NUM
easat-10905	300	143	distilbert	distilbert	NOUN
easat-10905	300	144	offensive	offensive	ADJ
easat-10905	300	145	98.10	98.10	NUM
easat-10905	300	146	96.80	96.80	NUM
easat-10905	300	147	99.38	99.38	NUM
easat-10905	300	148	98.07	98.07	NUM
easat-10905	300	149	not	not	PART
easat-10905	300	150	offensive	offensive	ADJ
easat-10905	300	151	99.40	99.40	NUM
easat-10905	300	152	96.88	96.88	NUM
easat-10905	300	153	98.12	98.12	NUM
easat-10905	300	154	macro	macro	NOUN
easat-10905	300	155	f1	f1	NOUN
easat-10905	300	156	98.10	98.10	NUM
easat-10905	300	157	weighted	weight	VERB
easat-10905	300	158	f1	f1	NOUN
easat-10905	300	159	98.10	98.10	NUM
easat-10905	300	160	in	in	ADP
easat-10905	300	161	addition	addition	NOUN
easat-10905	300	162	to	to	ADP
easat-10905	300	163	replay	replay	NOUN
easat-10905	300	164	-	-	PUNCT
easat-10905	300	165	based	base	VERB
easat-10905	300	166	approaches	approach	NOUN
easat-10905	300	167	,	,	PUNCT
easat-10905	300	168	we	we	PRON
easat-10905	300	169	have	have	AUX
easat-10905	300	170	also	also	ADV
easat-10905	300	171	integrated	integrate	VERB
easat-10905	300	172	lwf	lwf	PROPN
easat-10905	300	173	into	into	ADP
easat-10905	300	174	the	the	DET
easat-10905	300	175	developed	develop	VERB
easat-10905	300	176	models	model	NOUN
easat-10905	300	177	,	,	PUNCT
easat-10905	300	178	and	and	CCONJ
easat-10905	300	179	the	the	DET
easat-10905	300	180	results	result	NOUN
easat-10905	300	181	are	be	AUX
easat-10905	300	182	shown	show	VERB
easat-10905	300	183	in	in	ADP
easat-10905	300	184	tables	table	NOUN
easat-10905	300	185	6	6	NUM
easat-10905	300	186	and	and	CCONJ
easat-10905	300	187	7	7	NUM
easat-10905	300	188	.	.	X
easat-10905	300	189	from	from	ADP
easat-10905	300	190	these	these	DET
easat-10905	300	191	tables	table	NOUN
easat-10905	300	192	,	,	PUNCT
easat-10905	300	193	we	we	PRON
easat-10905	300	194	infer	infer	VERB
easat-10905	300	195	that	that	SCONJ
easat-10905	300	196	the	the	DET
easat-10905	300	197	performance	performance	NOUN
easat-10905	300	198	of	of	ADP
easat-10905	300	199	the	the	DET
easat-10905	300	200	models	model	NOUN
easat-10905	300	201	significantly	significantly	ADV
easat-10905	300	202	increased	increase	VERB
easat-10905	300	203	after	after	ADP
easat-10905	300	204	integrating	integrate	VERB
easat-10905	300	205	lwf	lwf	PROPN
easat-10905	300	206	.	.	PUNCT
easat-10905	301	1	460	460	NUM
easat-10905	301	2	edelweiss	edelweiss	PROPN
easat-10905	301	3	applied	apply	VERB
easat-10905	301	4	science	science	NOUN
easat-10905	301	5	and	and	CCONJ
easat-10905	301	6	technology	technology	NOUN
easat-10905	301	7	issn	issn	PROPN
easat-10905	301	8	:	:	PUNCT
easat-10905	301	9	2576	2576	NUM
easat-10905	301	10	-	-	SYM
easat-10905	301	11	8484	8484	NUM
easat-10905	301	12	vol	vol	NOUN
easat-10905	301	13	.	.	PROPN
easat-10905	302	1	9	9	NUM
easat-10905	302	2	,	,	PUNCT
easat-10905	302	3	no	no	INTJ
easat-10905	302	4	.	.	NOUN
easat-10905	302	5	11	11	NUM
easat-10905	302	6	:	:	PUNCT
easat-10905	302	7	448	448	NUM
easat-10905	302	8	-	-	SYM
easat-10905	302	9	466	466	NUM
easat-10905	302	10	,	,	PUNCT
easat-10905	302	11	2025	2025	NUM
easat-10905	302	12	doi	doi	NOUN
easat-10905	302	13	:	:	PUNCT
easat-10905	302	14	10.55214/2576	10.55214/2576	NUM
easat-10905	302	15	-	-	PUNCT
easat-10905	302	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	303	1	©	©	PROPN
easat-10905	303	2	2025	2025	NUM
easat-10905	303	3	by	by	ADP
easat-10905	303	4	the	the	DET
easat-10905	303	5	authors	author	NOUN
easat-10905	303	6	;	;	PUNCT
easat-10905	303	7	licensee	licensee	PROPN
easat-10905	303	8	learning	learn	VERB
easat-10905	303	9	gate	gate	NOUN
easat-10905	303	10	table	table	NOUN
easat-10905	303	11	6	6	NUM
easat-10905	303	12	.	.	PUNCT
easat-10905	304	1	performance	performance	NOUN
easat-10905	304	2	of	of	ADP
easat-10905	304	3	the	the	DET
easat-10905	304	4	classifiers	classifier	NOUN
easat-10905	304	5	on	on	ADP
easat-10905	304	6	abusive	abusive	ADJ
easat-10905	304	7	comments	comment	NOUN
easat-10905	304	8	using	use	VERB
easat-10905	304	9	lwf	lwf	PROPN
easat-10905	304	10	.	.	PUNCT
easat-10905	305	1	classifiers	classifier	NOUN
easat-10905	305	2	class	class	NOUN
easat-10905	305	3	accuracy	accuracy	NOUN
easat-10905	305	4	recall	recall	PROPN
easat-10905	305	5	precision	precision	PROPN
easat-10905	305	6	f1	f1	PROPN
easat-10905	305	7	score	score	NOUN
easat-10905	305	8	bert	bert	NOUN
easat-10905	305	9	abusive	abusive	ADJ
easat-10905	305	10	91.06	91.06	NUM
easat-10905	305	11	97.52	97.52	NUM
easat-10905	305	12	86.36	86.36	NUM
easat-10905	305	13	91.60	91.60	NUM
easat-10905	305	14	non	non	ADJ
easat-10905	305	15	-	-	ADJ
easat-10905	305	16	abusive	abusive	ADJ
easat-10905	305	17	84.60	84.60	NUM
easat-10905	305	18	97.15	97.15	NUM
easat-10905	305	19	90.44	90.44	NUM
easat-10905	305	20	macro	macro	ADJ
easat-10905	305	21	f1	f1	NOUN
easat-10905	305	22	91.02	91.02	NUM
easat-10905	305	23	weighted	weight	VERB
easat-10905	305	24	f1	f1	PROPN
easat-10905	305	25	91.02	91.02	NUM
easat-10905	305	26	albert	albert	PROPN
easat-10905	305	27	abusive	abusive	NOUN
easat-10905	305	28	76.92	76.92	NUM
easat-10905	305	29	93.68	93.68	NUM
easat-10905	305	30	70.16	70.16	NUM
easat-10905	305	31	80.23	80.23	NUM
easat-10905	305	32	non	non	ADJ
easat-10905	305	33	-	-	ADJ
easat-10905	305	34	abusive	abusive	ADJ
easat-10905	305	35	60.16	60.16	NUM
easat-10905	305	36	90.49	90.49	NUM
easat-10905	305	37	72.27	72.27	NUM
easat-10905	305	38	macro	macro	NOUN
easat-10905	305	39	f1	f1	PROPN
easat-10905	305	40	76.25	76.25	NUM
easat-10905	305	41	weighted	weight	VERB
easat-10905	305	42	f1	f1	PROPN
easat-10905	305	43	76.25	76.25	NUM
easat-10905	305	44	roberta	roberta	PROPN
easat-10905	305	45	abusive	abusive	PROPN
easat-10905	305	46	80.90	80.90	NUM
easat-10905	305	47	93.08	93.08	NUM
easat-10905	305	48	74.85	74.85	NUM
easat-10905	305	49	82.97	82.97	NUM
easat-10905	305	50	non	non	ADJ
easat-10905	305	51	-	-	ADJ
easat-10905	305	52	abusive	abusive	ADJ
easat-10905	305	53	68.72	68.72	NUM
easat-10905	305	54	90.85	90.85	NUM
easat-10905	305	55	78.25	78.25	NUM
easat-10905	305	56	macro	macro	NOUN
easat-10905	305	57	f1	f1	NOUN
easat-10905	305	58	80.61	80.61	NUM
easat-10905	305	59	weighted	weight	VERB
easat-10905	305	60	f1	f1	NOUN
easat-10905	305	61	80.61	80.61	NUM
easat-10905	305	62	distilbert	distilbert	NOUN
easat-10905	305	63	abusive	abusive	ADJ
easat-10905	305	64	87.74	87.74	NUM
easat-10905	305	65	98.24	98.24	NUM
easat-10905	305	66	81.19	81.19	NUM
easat-10905	305	67	88.90	88.90	NUM
easat-10905	305	68	non	non	ADJ
easat-10905	305	69	-	-	ADJ
easat-10905	305	70	abusive	abusive	ADJ
easat-10905	305	71	77.24	77.24	NUM
easat-10905	305	72	97.77	97.77	NUM
easat-10905	305	73	86.30	86.30	NUM
easat-10905	305	74	macro	macro	ADJ
easat-10905	305	75	f1	f1	PROPN
easat-10905	305	76	87.60	87.60	NUM
easat-10905	305	77	weighted	weight	VERB
easat-10905	305	78	f1	f1	PROPN
easat-10905	305	79	87.60	87.60	NUM
easat-10905	305	80	table	table	NOUN
easat-10905	305	81	7	7	NUM
easat-10905	305	82	.	.	PUNCT
easat-10905	305	83	performance	performance	NOUN
easat-10905	305	84	of	of	ADP
easat-10905	305	85	the	the	DET
easat-10905	305	86	classifiers	classifier	NOUN
easat-10905	305	87	on	on	ADP
easat-10905	305	88	offensive	offensive	ADJ
easat-10905	305	89	comments	comment	NOUN
easat-10905	305	90	using	use	VERB
easat-10905	305	91	lwf	lwf	PROPN
easat-10905	305	92	.	.	PUNCT
easat-10905	306	1	classifiers	classifier	NOUN
easat-10905	306	2	class	class	NOUN
easat-10905	306	3	accuracy	accuracy	NOUN
easat-10905	306	4	recall	recall	PROPN
easat-10905	306	5	precision	precision	PROPN
easat-10905	306	6	f1	f1	PROPN
easat-10905	306	7	score	score	NOUN
easat-10905	306	8	bert	bert	PROPN
easat-10905	306	9	offensive	offensive	VERB
easat-10905	306	10	98.80	98.80	NUM
easat-10905	306	11	99.12	99.12	NUM
easat-10905	306	12	98.49	98.49	NUM
easat-10905	306	13	98.80	98.80	NUM
easat-10905	306	14	not	not	PART
easat-10905	306	15	offensive	offensive	ADJ
easat-10905	306	16	98.48	98.48	NUM
easat-10905	306	17	99.11	99.11	NUM
easat-10905	306	18	98.80	98.80	NUM
easat-10905	306	19	macro	macro	NOUN
easat-10905	306	20	f1	f1	PROPN
easat-10905	306	21	98.80	98.80	NUM
easat-10905	306	22	weighted	weight	VERB
easat-10905	306	23	f1	f1	PROPN
easat-10905	306	24	98.80	98.80	NUM
easat-10905	306	25	albert	albert	PROPN
easat-10905	306	26	offensive	offensive	NOUN
easat-10905	306	27	95.82	95.82	NUM
easat-10905	306	28	94.92	94.92	NUM
easat-10905	306	29	96.66	96.66	NUM
easat-10905	306	30	95.78	95.78	NUM
easat-10905	306	31	not	not	PART
easat-10905	306	32	offensive	offensive	ADJ
easat-10905	306	33	96.72	96.72	NUM
easat-10905	306	34	95.01	95.01	NUM
easat-10905	306	35	95.86	95.86	NUM
easat-10905	306	36	macro	macro	ADJ
easat-10905	306	37	f1	f1	NOUN
easat-10905	306	38	95.82	95.82	NUM
easat-10905	306	39	weighted	weight	VERB
easat-10905	306	40	f1	f1	NOUN
easat-10905	306	41	95.82	95.82	NUM
easat-10905	306	42	roberta	roberta	PROPN
easat-10905	306	43	offensive	offensive	ADJ
easat-10905	306	44	95.52	95.52	NUM
easat-10905	306	45	92.76	92.76	NUM
easat-10905	306	46	98.18	98.18	NUM
easat-10905	306	47	95.39	95.39	NUM
easat-10905	306	48	not	not	PART
easat-10905	306	49	offensive	offensive	ADJ
easat-10905	306	50	98.28	98.28	NUM
easat-10905	306	51	93.14	93.14	NUM
easat-10905	306	52	95.64	95.64	NUM
easat-10905	306	53	macro	macro	NOUN
easat-10905	306	54	f1	f1	NOUN
easat-10905	306	55	95.52	95.52	NUM
easat-10905	306	56	weighted	weight	VERB
easat-10905	306	57	f1	f1	NOUN
easat-10905	306	58	95.52	95.52	NUM
easat-10905	306	59	distilbert	distilbert	NOUN
easat-10905	306	60	offensive	offensive	ADJ
easat-10905	306	61	97.86	97.86	NUM
easat-10905	306	62	98.00	98.00	NUM
easat-10905	306	63	97.73	97.73	NUM
easat-10905	306	64	97.86	97.86	NUM
easat-10905	306	65	not	not	PART
easat-10905	306	66	offensive	offensive	ADJ
easat-10905	306	67	97.72	97.72	NUM
easat-10905	306	68	97.99	97.99	NUM
easat-10905	306	69	97.86	97.86	NUM
easat-10905	306	70	macro	macro	NOUN
easat-10905	306	71	f1	f1	PROPN
easat-10905	306	72	97.86	97.86	NUM
easat-10905	306	73	weighted	weight	VERB
easat-10905	306	74	f1	f1	PROPN
easat-10905	306	75	97.86	97.86	NUM
easat-10905	306	76	further	far	ADV
easat-10905	306	77	to	to	PART
easat-10905	306	78	evaluate	evaluate	VERB
easat-10905	306	79	the	the	DET
easat-10905	306	80	performance	performance	NOUN
easat-10905	306	81	of	of	ADP
easat-10905	306	82	the	the	DET
easat-10905	306	83	developed	develop	VERB
easat-10905	306	84	task	task	NOUN
easat-10905	306	85	incremental	incremental	ADJ
easat-10905	306	86	models	model	NOUN
easat-10905	306	87	,	,	PUNCT
easat-10905	306	88	we	we	PRON
easat-10905	306	89	adapt	adapt	VERB
easat-10905	306	90	the	the	DET
easat-10905	306	91	metrics	metric	NOUN
easat-10905	306	92	recommended	recommend	VERB
easat-10905	306	93	by	by	ADP
easat-10905	306	94	akhter	akhter	NOUN
easat-10905	306	95	et	et	PROPN
easat-10905	306	96	al	al	PROPN
easat-10905	306	97	.	.	PUNCT
easat-10905	307	1	[	[	X
easat-10905	307	2	2	2	NUM
easat-10905	307	3	]	]	PUNCT
easat-10905	307	4	,	,	PUNCT
easat-10905	307	5	such	such	ADJ
easat-10905	307	6	as	as	ADP
easat-10905	307	7	average	average	ADJ
easat-10905	307	8	accuracy	accuracy	NOUN
easat-10905	307	9	,	,	PUNCT
easat-10905	307	10	average	average	ADJ
easat-10905	307	11	incremental	incremental	ADJ
easat-10905	307	12	accuracy	accuracy	NOUN
easat-10905	307	13	,	,	PUNCT
easat-10905	307	14	forgetting	forget	VERB
easat-10905	307	15	rate	rate	NOUN
easat-10905	307	16	,	,	PUNCT
easat-10905	307	17	forward	forward	ADJ
easat-10905	307	18	transfer	transfer	NOUN
easat-10905	307	19	,	,	PUNCT
easat-10905	307	20	and	and	CCONJ
easat-10905	307	21	backward	backward	ADJ
easat-10905	307	22	transfer	transfer	NOUN
easat-10905	307	23	.	.	PUNCT
easat-10905	308	1	measuring	measure	VERB
easat-10905	308	2	these	these	DET
easat-10905	308	3	metrics	metric	NOUN
easat-10905	308	4	in	in	ADP
easat-10905	308	5	incremental	incremental	ADJ
easat-10905	308	6	learning	learning	NOUN
easat-10905	308	7	is	be	AUX
easat-10905	308	8	crucial	crucial	ADJ
easat-10905	308	9	for	for	ADP
easat-10905	308	10	a	a	DET
easat-10905	308	11	comprehensive	comprehensive	ADJ
easat-10905	308	12	evaluation	evaluation	NOUN
easat-10905	308	13	of	of	ADP
easat-10905	308	14	the	the	DET
easat-10905	308	15	models	model	NOUN
easat-10905	308	16	'	'	PART
easat-10905	308	17	performance	performance	NOUN
easat-10905	308	18	.	.	PUNCT
easat-10905	309	1	these	these	DET
easat-10905	309	2	metrics	metric	NOUN
easat-10905	309	3	assess	assess	VERB
easat-10905	309	4	the	the	DET
easat-10905	309	5	ability	ability	NOUN
easat-10905	309	6	of	of	ADP
easat-10905	309	7	the	the	DET
easat-10905	309	8	models	model	NOUN
easat-10905	309	9	to	to	PART
easat-10905	309	10	learn	learn	VERB
easat-10905	309	11	new	new	ADJ
easat-10905	309	12	tasks	task	NOUN
easat-10905	309	13	while	while	SCONJ
easat-10905	309	14	retaining	retain	VERB
easat-10905	309	15	knowledge	knowledge	NOUN
easat-10905	309	16	from	from	ADP
easat-10905	309	17	previous	previous	ADJ
easat-10905	309	18	ones	one	NOUN
easat-10905	309	19	,	,	PUNCT
easat-10905	309	20	balancing	balance	VERB
easat-10905	309	21	performance	performance	NOUN
easat-10905	309	22	across	across	ADP
easat-10905	309	23	tasks	task	NOUN
easat-10905	309	24	,	,	PUNCT
easat-10905	309	25	and	and	CCONJ
easat-10905	309	26	mitigating	mitigate	VERB
easat-10905	309	27	the	the	DET
easat-10905	309	28	effects	effect	NOUN
easat-10905	309	29	of	of	ADP
easat-10905	309	30	catastrophic	catastrophic	ADJ
easat-10905	309	31	forgetting	forgetting	NOUN
easat-10905	309	32	(	(	PUNCT
easat-10905	309	33	cf	cf	NOUN
easat-10905	309	34	)	)	PUNCT
easat-10905	309	35	.	.	PUNCT
easat-10905	310	1	they	they	PRON
easat-10905	310	2	also	also	ADV
easat-10905	310	3	help	help	VERB
easat-10905	310	4	evaluate	evaluate	VERB
easat-10905	310	5	how	how	SCONJ
easat-10905	310	6	well	well	ADJ
easat-10905	310	7	knowledge	knowledge	NOUN
easat-10905	310	8	is	be	AUX
easat-10905	310	9	transferred	transfer	VERB
easat-10905	310	10	between	between	ADP
easat-10905	310	11	tasks	task	NOUN
easat-10905	310	12	,	,	PUNCT
easat-10905	310	13	both	both	CCONJ
easat-10905	310	14	forward	forward	ADV
easat-10905	310	15	and	and	CCONJ
easat-10905	310	16	backward	backward	ADV
easat-10905	310	17	,	,	PUNCT
easat-10905	310	18	providing	provide	VERB
easat-10905	310	19	insights	insight	NOUN
easat-10905	310	20	into	into	ADP
easat-10905	310	21	the	the	DET
easat-10905	310	22	overall	overall	ADJ
easat-10905	310	23	effectiveness	effectiveness	NOUN
easat-10905	310	24	in	in	ADP
easat-10905	310	25	sequential	sequential	ADJ
easat-10905	310	26	and	and	CCONJ
easat-10905	310	27	continual	continual	ADJ
easat-10905	310	28	learning	learning	NOUN
easat-10905	310	29	settings	setting	NOUN
easat-10905	310	30	.	.	PUNCT
easat-10905	311	1	a	a	DET
easat-10905	311	2	brief	brief	ADJ
easat-10905	311	3	note	note	NOUN
easat-10905	311	4	on	on	ADP
easat-10905	311	5	these	these	DET
easat-10905	311	6	metrics	metric	NOUN
easat-10905	311	7	is	be	AUX
easat-10905	311	8	provided	provide	VERB
easat-10905	311	9	below	below	ADP
easat-10905	311	10	:	:	PUNCT
easat-10905	311	11	average	average	ADJ
easat-10905	311	12	accuracy	accuracy	NOUN
easat-10905	311	13	:	:	PUNCT
easat-10905	311	14	it	it	PRON
easat-10905	311	15	refers	refer	VERB
easat-10905	311	16	to	to	ADP
easat-10905	311	17	the	the	DET
easat-10905	311	18	mean	mean	ADJ
easat-10905	311	19	performance	performance	NOUN
easat-10905	311	20	of	of	ADP
easat-10905	311	21	a	a	DET
easat-10905	311	22	model	model	NOUN
easat-10905	311	23	across	across	ADP
easat-10905	311	24	all	all	DET
easat-10905	311	25	tasks	task	NOUN
easat-10905	311	26	in	in	ADP
easat-10905	311	27	a	a	DET
easat-10905	311	28	taskincremental	taskincremental	ADJ
easat-10905	311	29	learning	learning	NOUN
easat-10905	311	30	setting	set	VERB
easat-10905	311	31	.	.	PUNCT
easat-10905	312	1	after	after	SCONJ
easat-10905	312	2	the	the	DET
easat-10905	312	3	model	model	NOUN
easat-10905	312	4	has	have	AUX
easat-10905	312	5	learned	learn	VERB
easat-10905	312	6	all	all	DET
easat-10905	312	7	tasks	task	NOUN
easat-10905	312	8	,	,	PUNCT
easat-10905	312	9	average	average	ADJ
easat-10905	312	10	accuracy	accuracy	NOUN
easat-10905	312	11	is	be	AUX
easat-10905	312	12	calculated	calculate	VERB
easat-10905	312	13	by	by	ADP
easat-10905	312	14	evaluating	evaluate	VERB
easat-10905	312	15	the	the	DET
easat-10905	312	16	model	model	NOUN
easat-10905	312	17	on	on	ADP
easat-10905	312	18	each	each	DET
easat-10905	312	19	task	task	NOUN
easat-10905	312	20	and	and	CCONJ
easat-10905	312	21	computing	compute	VERB
easat-10905	312	22	the	the	DET
easat-10905	312	23	average	average	NOUN
easat-10905	312	24	of	of	ADP
easat-10905	312	25	these	these	DET
easat-10905	312	26	performance	performance	NOUN
easat-10905	312	27	scores	score	NOUN
easat-10905	312	28	.	.	PUNCT
easat-10905	313	1	it	it	PRON
easat-10905	313	2	is	be	AUX
easat-10905	313	3	calculated	calculate	VERB
easat-10905	313	4	as	as	ADP
easat-10905	313	5	given	give	VERB
easat-10905	313	6	in	in	ADP
easat-10905	313	7	equation	equation	NOUN
easat-10905	313	8	(	(	PUNCT
easat-10905	313	9	11	11	NUM
easat-10905	313	10	)	)	PUNCT
easat-10905	313	11	.	.	PUNCT
easat-10905	314	1	𝐴𝑇	𝐴𝑇	PROPN
easat-10905	314	2	=	=	SYM
easat-10905	314	3	1	1	NUM
easat-10905	314	4	𝑇	𝑇	NOUN
easat-10905	314	5	∑	∑	PUNCT
easat-10905	314	6	𝑎𝑇.𝑡	𝑎𝑇.𝑡	PROPN
easat-10905	314	7	𝑇	𝑇	PROPN
easat-10905	314	8	𝑡=1	𝑡=1	PROPN
easat-10905	314	9	(	(	PUNCT
easat-10905	314	10	11	11	NUM
easat-10905	314	11	)	)	PUNCT
easat-10905	314	12	461	461	NUM
easat-10905	314	13	edelweiss	edelweiss	PROPN
easat-10905	314	14	applied	apply	VERB
easat-10905	314	15	science	science	NOUN
easat-10905	314	16	and	and	CCONJ
easat-10905	314	17	technology	technology	NOUN
easat-10905	314	18	issn	issn	PROPN
easat-10905	314	19	:	:	PUNCT
easat-10905	314	20	2576	2576	NUM
easat-10905	314	21	-	-	SYM
easat-10905	314	22	8484	8484	NUM
easat-10905	314	23	vol	vol	NOUN
easat-10905	314	24	.	.	PROPN
easat-10905	315	1	9	9	NUM
easat-10905	315	2	,	,	PUNCT
easat-10905	315	3	no	no	INTJ
easat-10905	315	4	.	.	NOUN
easat-10905	315	5	11	11	NUM
easat-10905	315	6	:	:	PUNCT
easat-10905	315	7	448	448	NUM
easat-10905	315	8	-	-	SYM
easat-10905	315	9	466	466	NUM
easat-10905	315	10	,	,	PUNCT
easat-10905	315	11	2025	2025	NUM
easat-10905	315	12	doi	doi	NOUN
easat-10905	315	13	:	:	PUNCT
easat-10905	315	14	10.55214/2576	10.55214/2576	NUM
easat-10905	315	15	-	-	PUNCT
easat-10905	315	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	316	1	©	©	PROPN
easat-10905	316	2	2025	2025	NUM
easat-10905	316	3	by	by	ADP
easat-10905	316	4	the	the	DET
easat-10905	316	5	authors	author	NOUN
easat-10905	316	6	;	;	PUNCT
easat-10905	316	7	licensee	licensee	PROPN
easat-10905	316	8	learning	learning	NOUN
easat-10905	316	9	gate	gate	NOUN
easat-10905	316	10	where	where	SCONJ
easat-10905	316	11	t	t	PROPN
easat-10905	316	12	is	be	AUX
easat-10905	316	13	the	the	DET
easat-10905	316	14	number	number	NOUN
easat-10905	316	15	of	of	ADP
easat-10905	316	16	tasks	task	NOUN
easat-10905	316	17	and	and	CCONJ
easat-10905	316	18	𝑎𝑇.𝑡	𝑎𝑇.𝑡	NOUN
easat-10905	316	19	refers	refer	VERB
easat-10905	316	20	to	to	ADP
easat-10905	316	21	the	the	DET
easat-10905	316	22	performance	performance	NOUN
easat-10905	316	23	of	of	ADP
easat-10905	316	24	the	the	DET
easat-10905	316	25	model	model	NOUN
easat-10905	316	26	on	on	ADP
easat-10905	316	27	the	the	DET
easat-10905	316	28	testing	testing	NOUN
easat-10905	316	29	set	set	VERB
easat-10905	316	30	of	of	ADP
easat-10905	316	31	tasks	task	NOUN
easat-10905	316	32	after	after	SCONJ
easat-10905	316	33	the	the	DET
easat-10905	316	34	model	model	NOUN
easat-10905	316	35	has	have	AUX
easat-10905	316	36	continually	continually	ADV
easat-10905	316	37	trained	train	VERB
easat-10905	316	38	all	all	DET
easat-10905	316	39	t	t	NOUN
easat-10905	316	40	tasks	task	NOUN
easat-10905	316	41	,	,	PUNCT
easat-10905	316	42	that	that	ADV
easat-10905	316	43	is	is	ADV
easat-10905	316	44	,	,	PUNCT
easat-10905	316	45	the	the	DET
easat-10905	316	46	accuracy	accuracy	NOUN
easat-10905	316	47	on	on	ADP
easat-10905	316	48	task	task	PROPN
easat-10905	316	49	‘	'	PUNCT
easat-10905	316	50	t	t	NOUN
easat-10905	316	51	’	'	PUNCT
easat-10905	316	52	after	after	ADP
easat-10905	316	53	learning	learn	VERB
easat-10905	316	54	the	the	DET
easat-10905	316	55	tth	tth	PROPN
easat-10905	316	56	task	task	NOUN
easat-10905	316	57	.	.	PUNCT
easat-10905	317	1	average	average	ADJ
easat-10905	317	2	incremental	incremental	ADJ
easat-10905	317	3	accuracy	accuracy	NOUN
easat-10905	317	4	:	:	PUNCT
easat-10905	317	5	it	it	PRON
easat-10905	317	6	measures	measure	VERB
easat-10905	317	7	how	how	SCONJ
easat-10905	317	8	well	well	ADV
easat-10905	317	9	a	a	DET
easat-10905	317	10	model	model	NOUN
easat-10905	317	11	retains	retain	VERB
easat-10905	317	12	knowledge	knowledge	NOUN
easat-10905	317	13	of	of	ADP
easat-10905	317	14	previous	previous	ADJ
easat-10905	317	15	tasks	task	NOUN
easat-10905	317	16	while	while	SCONJ
easat-10905	317	17	learning	learn	VERB
easat-10905	317	18	new	new	ADJ
easat-10905	317	19	ones	one	NOUN
easat-10905	317	20	.	.	PUNCT
easat-10905	318	1	it	it	PRON
easat-10905	318	2	is	be	AUX
easat-10905	318	3	calculated	calculate	VERB
easat-10905	318	4	as	as	ADP
easat-10905	318	5	the	the	DET
easat-10905	318	6	average	average	ADJ
easat-10905	318	7	accuracy	accuracy	NOUN
easat-10905	318	8	of	of	ADP
easat-10905	318	9	the	the	DET
easat-10905	318	10	model	model	NOUN
easat-10905	318	11	on	on	ADP
easat-10905	318	12	each	each	DET
easat-10905	318	13	task	task	NOUN
easat-10905	318	14	at	at	ADP
easat-10905	318	15	the	the	DET
easat-10905	318	16	time	time	NOUN
easat-10905	318	17	that	that	SCONJ
easat-10905	318	18	task	task	NOUN
easat-10905	318	19	was	be	AUX
easat-10905	318	20	learned	learn	VERB
easat-10905	318	21	.	.	PUNCT
easat-10905	319	1	this	this	DET
easat-10905	319	2	metric	metric	ADJ
easat-10905	319	3	highlights	highlight	NOUN
easat-10905	319	4	the	the	DET
easat-10905	319	5	ability	ability	NOUN
easat-10905	319	6	of	of	ADP
easat-10905	319	7	the	the	DET
easat-10905	319	8	models	model	NOUN
easat-10905	319	9	to	to	PART
easat-10905	319	10	maintain	maintain	VERB
easat-10905	319	11	performance	performance	NOUN
easat-10905	319	12	on	on	ADP
easat-10905	319	13	earlier	early	ADJ
easat-10905	319	14	tasks	task	NOUN
easat-10905	319	15	as	as	SCONJ
easat-10905	319	16	it	it	PRON
easat-10905	319	17	progresses	progress	VERB
easat-10905	319	18	through	through	ADP
easat-10905	319	19	new	new	ADJ
easat-10905	319	20	tasks	task	NOUN
easat-10905	319	21	and	and	CCONJ
easat-10905	319	22	is	be	AUX
easat-10905	319	23	represented	represent	VERB
easat-10905	319	24	as	as	ADP
easat-10905	319	25	given	give	VERB
easat-10905	319	26	in	in	ADP
easat-10905	319	27	equation	equation	NOUN
easat-10905	319	28	(	(	PUNCT
easat-10905	319	29	12	12	NUM
easat-10905	319	30	)	)	PUNCT
easat-10905	319	31	.	.	PUNCT
easat-10905	320	1	𝐴𝐼𝐴	𝐴𝐼𝐴	PROPN
easat-10905	320	2	=	=	PUNCT
easat-10905	320	3	1	1	NUM
easat-10905	320	4	𝑇	𝑇	PROPN
easat-10905	320	5	∑	∑	PUNCT
easat-10905	320	6	𝑎𝑡.𝑡	𝑎𝑡.𝑡	PROPN
easat-10905	320	7	𝑇	𝑇	PROPN
easat-10905	320	8	𝑡=1	𝑡=1	PROPN
easat-10905	320	9	(	(	PUNCT
easat-10905	320	10	12	12	NUM
easat-10905	320	11	)	)	PUNCT
easat-10905	320	12	where	where	SCONJ
easat-10905	320	13	𝑎𝑡.𝑡	𝑎𝑡.𝑡	PROPN
easat-10905	320	14	is	be	AUX
easat-10905	320	15	the	the	DET
easat-10905	320	16	accuracy	accuracy	NOUN
easat-10905	320	17	on	on	ADP
easat-10905	320	18	task	task	PROPN
easat-10905	320	19	‘	'	PUNCT
easat-10905	320	20	t	t	NOUN
easat-10905	320	21	’	'	PUNCT
easat-10905	320	22	immediately	immediately	ADV
easat-10905	320	23	after	after	ADP
easat-10905	320	24	learning	learn	VERB
easat-10905	320	25	task	task	NOUN
easat-10905	320	26	‘	'	PUNCT
easat-10905	320	27	t	t	PROPN
easat-10905	320	28	’	'	PUNCT
easat-10905	320	29	.	.	PUNCT
easat-10905	321	1	4.3	4.3	NUM
easat-10905	321	2	.	.	PUNCT
easat-10905	321	3	forgetting	forget	VERB
easat-10905	321	4	rate	rate	NOUN
easat-10905	321	5	it	it	PRON
easat-10905	321	6	measures	measure	VERB
easat-10905	321	7	the	the	DET
easat-10905	321	8	extent	extent	NOUN
easat-10905	321	9	to	to	PART
easat-10905	321	10	which	which	PRON
easat-10905	321	11	the	the	DET
easat-10905	321	12	performance	performance	NOUN
easat-10905	321	13	of	of	ADP
easat-10905	321	14	a	a	DET
easat-10905	321	15	model	model	NOUN
easat-10905	321	16	on	on	ADP
easat-10905	321	17	previously	previously	ADV
easat-10905	321	18	learned	learn	VERB
easat-10905	321	19	tasks	task	NOUN
easat-10905	321	20	degrades	degrade	VERB
easat-10905	321	21	as	as	SCONJ
easat-10905	321	22	it	it	PRON
easat-10905	321	23	learns	learn	VERB
easat-10905	321	24	new	new	ADJ
easat-10905	321	25	tasks	task	NOUN
easat-10905	321	26	.	.	PUNCT
easat-10905	322	1	it	it	PRON
easat-10905	322	2	is	be	AUX
easat-10905	322	3	calculated	calculate	VERB
easat-10905	322	4	by	by	ADP
easat-10905	322	5	comparing	compare	VERB
easat-10905	322	6	the	the	DET
easat-10905	322	7	accuracy	accuracy	NOUN
easat-10905	322	8	of	of	ADP
easat-10905	322	9	the	the	DET
easat-10905	322	10	model	model	NOUN
easat-10905	322	11	on	on	ADP
easat-10905	322	12	a	a	DET
easat-10905	322	13	task	task	NOUN
easat-10905	322	14	immediately	immediately	ADV
easat-10905	322	15	after	after	SCONJ
easat-10905	322	16	it	it	PRON
easat-10905	322	17	was	be	AUX
easat-10905	322	18	learned	learn	VERB
easat-10905	322	19	to	to	ADP
easat-10905	322	20	its	its	PRON
easat-10905	322	21	accuracy	accuracy	NOUN
easat-10905	322	22	on	on	ADP
easat-10905	322	23	the	the	DET
easat-10905	322	24	same	same	ADJ
easat-10905	322	25	task	task	NOUN
easat-10905	322	26	after	after	ADP
easat-10905	322	27	learning	learn	VERB
easat-10905	322	28	subsequent	subsequent	ADJ
easat-10905	322	29	tasks	task	NOUN
easat-10905	322	30	.	.	PUNCT
easat-10905	323	1	a	a	DET
easat-10905	323	2	higher	high	ADJ
easat-10905	323	3	forgetting	forget	VERB
easat-10905	323	4	rate	rate	NOUN
easat-10905	323	5	indicates	indicate	VERB
easat-10905	323	6	more	more	ADV
easat-10905	323	7	significant	significant	ADJ
easat-10905	323	8	forgetting	forgetting	NOUN
easat-10905	323	9	,	,	PUNCT
easat-10905	323	10	which	which	PRON
easat-10905	323	11	is	be	AUX
easat-10905	323	12	a	a	DET
easat-10905	323	13	key	key	ADJ
easat-10905	323	14	challenge	challenge	NOUN
easat-10905	323	15	in	in	ADP
easat-10905	323	16	task	task	NOUN
easat-10905	323	17	incremental	incremental	ADJ
easat-10905	323	18	learning	learning	NOUN
easat-10905	323	19	,	,	PUNCT
easat-10905	323	20	and	and	CCONJ
easat-10905	323	21	is	be	AUX
easat-10905	323	22	calculated	calculate	VERB
easat-10905	323	23	as	as	SCONJ
easat-10905	323	24	follows	follow	VERB
easat-10905	323	25	in	in	ADP
easat-10905	323	26	equation	equation	NOUN
easat-10905	323	27	(	(	PUNCT
easat-10905	323	28	13	13	NUM
easat-10905	323	29	):	):	PUNCT
easat-10905	323	30	𝐹𝑇	𝐹𝑇	PROPN
easat-10905	323	31	=	=	SYM
easat-10905	323	32	1	1	NUM
easat-10905	323	33	𝑇−1	𝑇−1	PROPN
easat-10905	323	34	∑	∑	PUNCT
easat-10905	323	35	(	(	PUNCT
easat-10905	323	36	𝑎𝑡,𝑡	𝑎𝑡,𝑡	NOUN
easat-10905	324	1	𝑇−1	𝑇−1	PROPN
easat-10905	324	2	𝑡=1	𝑡=1	VERB
easat-10905	324	3	−	−	PROPN
easat-10905	324	4	𝑎𝑇,𝑡	𝑎𝑇,𝑡	PROPN
easat-10905	324	5	)	)	PUNCT
easat-10905	324	6	(	(	PUNCT
easat-10905	324	7	13	13	NUM
easat-10905	324	8	)	)	PUNCT
easat-10905	324	9	where	where	SCONJ
easat-10905	324	10	𝑎𝑡,𝑡	𝑎𝑡,𝑡	NOUN
easat-10905	324	11	represents	represent	VERB
easat-10905	324	12	the	the	DET
easat-10905	324	13	test	test	NOUN
easat-10905	324	14	accuracy	accuracy	NOUN
easat-10905	324	15	for	for	ADP
easat-10905	324	16	task	task	NOUN
easat-10905	324	17	t	t	PROPN
easat-10905	324	18	immediately	immediately	ADV
easat-10905	324	19	after	after	SCONJ
easat-10905	324	20	it	it	PRON
easat-10905	324	21	is	be	AUX
easat-10905	324	22	learned	learn	VERB
easat-10905	324	23	,	,	PUNCT
easat-10905	324	24	and	and	CCONJ
easat-10905	324	25	𝑎𝑇,𝑡	𝑎𝑇,𝑡	NOUN
easat-10905	324	26	represents	represent	VERB
easat-10905	324	27	the	the	DET
easat-10905	324	28	accuracy	accuracy	NOUN
easat-10905	324	29	for	for	ADP
easat-10905	324	30	task	task	NOUN
easat-10905	324	31	t	t	PROPN
easat-10905	324	32	after	after	ADP
easat-10905	324	33	training	train	VERB
easat-10905	324	34	on	on	ADP
easat-10905	324	35	the	the	DET
easat-10905	324	36	final	final	ADJ
easat-10905	324	37	task	task	NOUN
easat-10905	324	38	t.	t.	X
easat-10905	324	39	we	we	PRON
easat-10905	324	40	average	average	VERB
easat-10905	324	41	these	these	DET
easat-10905	324	42	values	value	NOUN
easat-10905	324	43	across	across	ADP
easat-10905	324	44	all	all	DET
easat-10905	324	45	trained	train	VERB
easat-10905	324	46	tasks	task	NOUN
easat-10905	324	47	except	except	SCONJ
easat-10905	324	48	the	the	DET
easat-10905	324	49	last	last	ADJ
easat-10905	324	50	one	one	NOUN
easat-10905	324	51	,	,	PUNCT
easat-10905	324	52	since	since	SCONJ
easat-10905	324	53	the	the	DET
easat-10905	324	54	final	final	ADJ
easat-10905	324	55	task	task	NOUN
easat-10905	324	56	does	do	AUX
easat-10905	324	57	not	not	PART
easat-10905	324	58	exhibit	exhibit	VERB
easat-10905	324	59	forgetting	forget	VERB
easat-10905	324	60	.	.	PUNCT
easat-10905	325	1	a	a	DET
easat-10905	325	2	lower	low	ADJ
easat-10905	325	3	forgetting	forget	VERB
easat-10905	325	4	rate	rate	NOUN
easat-10905	325	5	indicates	indicate	VERB
easat-10905	325	6	higher	high	ADJ
easat-10905	325	7	knowledge	knowledge	NOUN
easat-10905	325	8	preservation	preservation	NOUN
easat-10905	325	9	.	.	PUNCT
easat-10905	326	1	forward	forward	ADJ
easat-10905	326	2	transfer	transfer	NOUN
easat-10905	326	3	:	:	PUNCT
easat-10905	326	4	this	this	PRON
easat-10905	326	5	refers	refer	VERB
easat-10905	326	6	to	to	ADP
easat-10905	326	7	the	the	DET
easat-10905	326	8	ability	ability	NOUN
easat-10905	326	9	of	of	ADP
easat-10905	326	10	a	a	DET
easat-10905	326	11	model	model	NOUN
easat-10905	326	12	to	to	PART
easat-10905	326	13	use	use	VERB
easat-10905	326	14	knowledge	knowledge	NOUN
easat-10905	326	15	from	from	ADP
easat-10905	326	16	previously	previously	ADV
easat-10905	326	17	learned	learn	VERB
easat-10905	326	18	tasks	task	NOUN
easat-10905	326	19	to	to	PART
easat-10905	326	20	improve	improve	VERB
easat-10905	326	21	learning	learning	NOUN
easat-10905	326	22	and	and	CCONJ
easat-10905	326	23	performance	performance	NOUN
easat-10905	326	24	on	on	ADP
easat-10905	326	25	a	a	DET
easat-10905	326	26	new	new	ADJ
easat-10905	326	27	task	task	NOUN
easat-10905	326	28	.	.	PUNCT
easat-10905	327	1	positive	positive	ADJ
easat-10905	327	2	forward	forward	ADJ
easat-10905	327	3	transfer	transfer	NOUN
easat-10905	327	4	occurs	occur	VERB
easat-10905	327	5	when	when	SCONJ
easat-10905	327	6	knowledge	knowledge	NOUN
easat-10905	327	7	gained	gain	VERB
easat-10905	327	8	from	from	ADP
easat-10905	327	9	previous	previous	ADJ
easat-10905	327	10	tasks	task	NOUN
easat-10905	327	11	helps	help	VERB
easat-10905	327	12	improve	improve	VERB
easat-10905	327	13	performance	performance	NOUN
easat-10905	327	14	on	on	ADP
easat-10905	327	15	subsequent	subsequent	ADJ
easat-10905	327	16	tasks	task	NOUN
easat-10905	327	17	.	.	PUNCT
easat-10905	328	1	it	it	PRON
easat-10905	328	2	is	be	AUX
easat-10905	328	3	measured	measure	VERB
easat-10905	328	4	by	by	ADP
easat-10905	328	5	comparing	compare	VERB
easat-10905	328	6	the	the	DET
easat-10905	328	7	performance	performance	NOUN
easat-10905	328	8	on	on	ADP
easat-10905	328	9	a	a	DET
easat-10905	328	10	new	new	ADJ
easat-10905	328	11	task	task	NOUN
easat-10905	328	12	with	with	ADP
easat-10905	328	13	and	and	CCONJ
easat-10905	328	14	without	without	ADP
easat-10905	328	15	prior	prior	ADJ
easat-10905	328	16	task	task	NOUN
easat-10905	328	17	knowledge	knowledge	NOUN
easat-10905	328	18	.	.	PUNCT
easat-10905	329	1	it	it	PRON
easat-10905	329	2	is	be	AUX
easat-10905	329	3	given	give	VERB
easat-10905	329	4	as	as	SCONJ
easat-10905	329	5	mentioned	mention	VERB
easat-10905	329	6	in	in	ADP
easat-10905	329	7	equation	equation	NOUN
easat-10905	329	8	(	(	PUNCT
easat-10905	329	9	14	14	NUM
easat-10905	329	10	)	)	PUNCT
easat-10905	329	11	.	.	PUNCT
easat-10905	330	1	𝐹𝑊𝑇𝑇	𝐹𝑊𝑇𝑇	ADV
easat-10905	330	2	=	=	SYM
easat-10905	330	3	1	1	NUM
easat-10905	330	4	𝑇	𝑇	PROPN
easat-10905	330	5	∑	∑	PRON
easat-10905	330	6	(	(	PUNCT
easat-10905	330	7	𝑎𝑡,𝑡	𝑎𝑡,𝑡	NOUN
easat-10905	330	8	−	−	PROPN
easat-10905	330	9	𝑎0,𝑡)𝑇	𝑎0,𝑡)𝑇	PROPN
easat-10905	330	10	𝑡=1	𝑡=1	PROPN
easat-10905	330	11	(	(	PUNCT
easat-10905	330	12	14	14	NUM
easat-10905	330	13	)	)	PUNCT
easat-10905	330	14	where	where	SCONJ
easat-10905	330	15	𝑎0,𝑡	𝑎0,𝑡	PROPN
easat-10905	330	16	refers	refer	VERB
easat-10905	330	17	to	to	ADP
easat-10905	330	18	the	the	DET
easat-10905	330	19	performance	performance	NOUN
easat-10905	330	20	of	of	ADP
easat-10905	330	21	training	training	NOUN
easat-10905	330	22	task	task	NOUN
easat-10905	330	23	t	t	PROPN
easat-10905	330	24	when	when	SCONJ
easat-10905	330	25	it	it	PRON
easat-10905	330	26	is	be	AUX
easat-10905	330	27	trained	train	VERB
easat-10905	330	28	from	from	ADP
easat-10905	330	29	scratch	scratch	NOUN
easat-10905	330	30	individually	individually	ADV
easat-10905	330	31	and	and	CCONJ
easat-10905	330	32	𝑎𝑡,𝑡	𝑎𝑡,𝑡	NOUN
easat-10905	330	33	represents	represent	VERB
easat-10905	330	34	the	the	DET
easat-10905	330	35	test	test	NOUN
easat-10905	330	36	accuracy	accuracy	NOUN
easat-10905	330	37	for	for	ADP
easat-10905	330	38	task	task	NOUN
easat-10905	330	39	t	t	PROPN
easat-10905	330	40	immediately	immediately	ADV
easat-10905	330	41	after	after	SCONJ
easat-10905	330	42	it	it	PRON
easat-10905	330	43	is	be	AUX
easat-10905	330	44	learned	learn	VERB
easat-10905	330	45	.	.	PUNCT
easat-10905	331	1	backward	backward	ADJ
easat-10905	331	2	transfer	transfer	NOUN
easat-10905	331	3	:	:	PUNCT
easat-10905	331	4	it	it	PRON
easat-10905	331	5	evaluates	evaluate	VERB
easat-10905	331	6	the	the	DET
easat-10905	331	7	impact	impact	NOUN
easat-10905	331	8	that	that	PRON
easat-10905	331	9	learning	learn	VERB
easat-10905	331	10	new	new	ADJ
easat-10905	331	11	tasks	task	NOUN
easat-10905	331	12	has	have	VERB
easat-10905	331	13	on	on	ADP
easat-10905	331	14	the	the	DET
easat-10905	331	15	performance	performance	NOUN
easat-10905	331	16	of	of	ADP
easat-10905	331	17	the	the	DET
easat-10905	331	18	models	model	NOUN
easat-10905	331	19	on	on	ADP
easat-10905	331	20	previously	previously	ADV
easat-10905	331	21	learned	learn	VERB
easat-10905	331	22	tasks	task	NOUN
easat-10905	331	23	.	.	PUNCT
easat-10905	332	1	positive	positive	ADJ
easat-10905	332	2	backward	backward	ADJ
easat-10905	332	3	transfer	transfer	NOUN
easat-10905	332	4	occurs	occur	VERB
easat-10905	332	5	when	when	SCONJ
easat-10905	332	6	learning	learn	VERB
easat-10905	332	7	a	a	DET
easat-10905	332	8	new	new	ADJ
easat-10905	332	9	task	task	NOUN
easat-10905	332	10	enhances	enhance	VERB
easat-10905	332	11	the	the	DET
easat-10905	332	12	performance	performance	NOUN
easat-10905	332	13	on	on	ADP
easat-10905	332	14	earlier	early	ADJ
easat-10905	332	15	tasks	task	NOUN
easat-10905	332	16	,	,	PUNCT
easat-10905	332	17	while	while	SCONJ
easat-10905	332	18	negative	negative	ADJ
easat-10905	332	19	backward	backward	ADJ
easat-10905	332	20	transfer	transfer	NOUN
easat-10905	332	21	indicates	indicate	VERB
easat-10905	332	22	that	that	SCONJ
easat-10905	332	23	learning	learn	VERB
easat-10905	332	24	new	new	ADJ
easat-10905	332	25	tasks	task	NOUN
easat-10905	332	26	has	have	AUX
easat-10905	332	27	led	lead	VERB
easat-10905	332	28	to	to	ADP
easat-10905	332	29	a	a	DET
easat-10905	332	30	decline	decline	NOUN
easat-10905	332	31	in	in	ADP
easat-10905	332	32	performance	performance	NOUN
easat-10905	332	33	on	on	ADP
easat-10905	332	34	previous	previous	ADJ
easat-10905	332	35	tasks	task	NOUN
easat-10905	332	36	.	.	PUNCT
easat-10905	333	1	the	the	DET
easat-10905	333	2	formula	formula	NOUN
easat-10905	333	3	to	to	PART
easat-10905	333	4	find	find	VERB
easat-10905	333	5	the	the	DET
easat-10905	333	6	backward	backward	ADJ
easat-10905	333	7	transfer	transfer	NOUN
easat-10905	333	8	rate	rate	NOUN
easat-10905	333	9	is	be	AUX
easat-10905	333	10	given	give	VERB
easat-10905	333	11	in	in	ADP
easat-10905	333	12	equation	equation	NOUN
easat-10905	333	13	(	(	PUNCT
easat-10905	333	14	15	15	NUM
easat-10905	333	15	)	)	PUNCT
easat-10905	333	16	.	.	PUNCT
easat-10905	334	1	𝐵𝑊𝑇𝑇	𝐵𝑊𝑇𝑇	ADJ
easat-10905	334	2	=	=	PUNCT
easat-10905	334	3	1	1	NUM
easat-10905	334	4	𝑇−1	𝑇−1	X
easat-10905	334	5	∑	∑	PUNCT
easat-10905	334	6	(	(	PUNCT
easat-10905	334	7	𝑇−1	𝑇−1	PROPN
easat-10905	334	8	𝑡=1	𝑡=1	PROPN
easat-10905	334	9	𝑎𝑇,𝑡	𝑎𝑇,𝑡	PROPN
easat-10905	334	10	−	−	NOUN
easat-10905	334	11	𝑎𝑡,𝑡	𝑎𝑡,𝑡	NOUN
easat-10905	334	12	)	)	PUNCT
easat-10905	334	13	(	(	PUNCT
easat-10905	334	14	15	15	X
easat-10905	334	15	)	)	PUNCT
easat-10905	334	16	all	all	DET
easat-10905	334	17	the	the	DET
easat-10905	334	18	above	above	ADJ
easat-10905	334	19	metrics	metric	NOUN
easat-10905	334	20	have	have	AUX
easat-10905	334	21	been	be	AUX
easat-10905	334	22	measured	measure	VERB
easat-10905	334	23	,	,	PUNCT
easat-10905	334	24	and	and	CCONJ
easat-10905	334	25	these	these	DET
easat-10905	334	26	values	value	NOUN
easat-10905	334	27	are	be	AUX
easat-10905	334	28	presented	present	VERB
easat-10905	334	29	in	in	ADP
easat-10905	334	30	tables	table	NOUN
easat-10905	334	31	8	8	NUM
easat-10905	334	32	to	to	PART
easat-10905	334	33	10	10	NUM
easat-10905	334	34	.	.	PUNCT
easat-10905	335	1	table	table	NOUN
easat-10905	335	2	8	8	NUM
easat-10905	335	3	.	.	PUNCT
easat-10905	336	1	performance	performance	NOUN
easat-10905	336	2	of	of	ADP
easat-10905	336	3	the	the	DET
easat-10905	336	4	classifiers	classifier	NOUN
easat-10905	336	5	using	use	VERB
easat-10905	336	6	the	the	DET
easat-10905	336	7	reply	reply	NOUN
easat-10905	336	8	-	-	PUNCT
easat-10905	336	9	based	base	VERB
easat-10905	336	10	rehearsal	rehearsal	NOUN
easat-10905	336	11	approach	approach	NOUN
easat-10905	336	12	.	.	PUNCT
easat-10905	337	1	metrics/	metrics/	NUM
easat-10905	337	2	models	model	NOUN
easat-10905	337	3	average	average	ADJ
easat-10905	337	4	accuracy	accuracy	NOUN
easat-10905	337	5	average	average	ADJ
easat-10905	337	6	incremental	incremental	ADJ
easat-10905	337	7	accuracy	accuracy	NOUN
easat-10905	337	8	forgetting	forget	VERB
easat-10905	337	9	rate	rate	NOUN
easat-10905	337	10	forward	forward	ADJ
easat-10905	337	11	transfer	transfer	VERB
easat-10905	337	12	backward	backward	ADJ
easat-10905	337	13	transfer	transfer	NOUN
easat-10905	337	14	bert	bert	NOUN
easat-10905	337	15	80.4	80.4	NUM
easat-10905	337	16	87.6	87.6	NUM
easat-10905	337	17	0.2	0.2	NUM
easat-10905	337	18	4.01	4.01	NUM
easat-10905	337	19	0.01	0.01	NUM
easat-10905	337	20	albert	albert	PROPN
easat-10905	337	21	73.45	73.45	NUM
easat-10905	337	22	79.4	79.4	NUM
easat-10905	337	23	9.79	9.79	NUM
easat-10905	337	24	5.21	5.21	NUM
easat-10905	337	25	0.01	0.01	NUM
easat-10905	337	26	roberta	roberta	PROPN
easat-10905	337	27	78.7	78.7	NUM
easat-10905	337	28	81.6	81.6	NUM
easat-10905	337	29	0.01	0.01	NUM
easat-10905	337	30	8.59	8.59	NUM
easat-10905	337	31	7.43	7.43	NUM
easat-10905	337	32	distilbert	distilbert	NOUN
easat-10905	337	33	83.45	83.45	NUM
easat-10905	337	34	90.8	90.8	NUM
easat-10905	337	35	0.01	0.01	NUM
easat-10905	337	36	8.67	8.67	NUM
easat-10905	337	37	9.79	9.79	NUM
easat-10905	337	38	462	462	NUM
easat-10905	337	39	edelweiss	edelweiss	PROPN
easat-10905	337	40	applied	apply	VERB
easat-10905	337	41	science	science	NOUN
easat-10905	337	42	and	and	CCONJ
easat-10905	337	43	technology	technology	NOUN
easat-10905	337	44	issn	issn	PROPN
easat-10905	337	45	:	:	PUNCT
easat-10905	337	46	2576	2576	NUM
easat-10905	337	47	-	-	SYM
easat-10905	337	48	8484	8484	NUM
easat-10905	337	49	vol	vol	NOUN
easat-10905	337	50	.	.	PROPN
easat-10905	338	1	9	9	NUM
easat-10905	338	2	,	,	PUNCT
easat-10905	338	3	no	no	INTJ
easat-10905	338	4	.	.	NOUN
easat-10905	338	5	11	11	NUM
easat-10905	338	6	:	:	PUNCT
easat-10905	338	7	448	448	NUM
easat-10905	338	8	-	-	SYM
easat-10905	338	9	466	466	NUM
easat-10905	338	10	,	,	PUNCT
easat-10905	338	11	2025	2025	NUM
easat-10905	338	12	doi	doi	NOUN
easat-10905	338	13	:	:	PUNCT
easat-10905	338	14	10.55214/2576	10.55214/2576	NUM
easat-10905	338	15	-	-	PUNCT
easat-10905	338	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	339	1	©	©	PROPN
easat-10905	339	2	2025	2025	NUM
easat-10905	339	3	by	by	ADP
easat-10905	339	4	the	the	DET
easat-10905	339	5	authors	author	NOUN
easat-10905	339	6	;	;	PUNCT
easat-10905	339	7	licensee	licensee	PROPN
easat-10905	339	8	learning	learn	VERB
easat-10905	339	9	gate	gate	NOUN
easat-10905	339	10	table	table	NOUN
easat-10905	339	11	9	9	NUM
easat-10905	339	12	.	.	PUNCT
easat-10905	339	13	performance	performance	NOUN
easat-10905	339	14	of	of	ADP
easat-10905	339	15	the	the	DET
easat-10905	339	16	classifiers	classifier	NOUN
easat-10905	339	17	using	use	VERB
easat-10905	339	18	the	the	DET
easat-10905	339	19	reply	reply	NOUN
easat-10905	339	20	-	-	PUNCT
easat-10905	339	21	based	base	VERB
easat-10905	339	22	pseudo	pseudo	NOUN
easat-10905	339	23	rehearsal	rehearsal	NOUN
easat-10905	339	24	approach	approach	NOUN
easat-10905	339	25	.	.	PUNCT
easat-10905	340	1	metrics/	metrics/	NUM
easat-10905	340	2	models	model	NOUN
easat-10905	340	3	average	average	ADJ
easat-10905	340	4	accuracy	accuracy	NOUN
easat-10905	340	5	average	average	ADJ
easat-10905	340	6	incremental	incremental	ADJ
easat-10905	340	7	accuracy	accuracy	NOUN
easat-10905	340	8	forgetting	forget	VERB
easat-10905	340	9	rate	rate	NOUN
easat-10905	340	10	forward	forward	ADJ
easat-10905	340	11	transfer	transfer	VERB
easat-10905	340	12	backward	backward	ADJ
easat-10905	340	13	transfer	transfer	NOUN
easat-10905	340	14	bert	bert	NOUN
easat-10905	340	15	93.68	93.68	NUM
easat-10905	340	16	98.8	98.8	NUM
easat-10905	340	17	1.98	1.98	NUM
easat-10905	340	18	98.02	98.02	NUM
easat-10905	340	19	10.66	10.66	NUM
easat-10905	340	20	albert	albert	PROPN
easat-10905	340	21	85.53	85.53	NUM
easat-10905	340	22	96.32	96.32	NUM
easat-10905	340	23	7.36	7.36	NUM
easat-10905	340	24	92.64	92.64	NUM
easat-10905	340	25	21.58	21.58	NUM
easat-10905	340	26	roberta	roberta	PROPN
easat-10905	340	27	91.06	91.06	NUM
easat-10905	340	28	97.75	97.75	NUM
easat-10905	340	29	4.5	4.5	NUM
easat-10905	340	30	95.5	95.5	NUM
easat-10905	340	31	13.38	13.38	NUM
easat-10905	340	32	distilbert	distilbert	NOUN
easat-10905	340	33	94.41	94.41	NUM
easat-10905	340	34	99.05	99.05	NUM
easat-10905	340	35	1.9	1.9	NUM
easat-10905	340	36	98.1	98.1	NUM
easat-10905	340	37	9.28	9.28	NUM
easat-10905	340	38	table	table	NOUN
easat-10905	340	39	10	10	NUM
easat-10905	340	40	.	.	PUNCT
easat-10905	341	1	performance	performance	NOUN
easat-10905	341	2	of	of	ADP
easat-10905	341	3	the	the	DET
easat-10905	341	4	classifiers	classifier	NOUN
easat-10905	341	5	using	use	VERB
easat-10905	341	6	lwf	lwf	PROPN
easat-10905	341	7	.	.	PUNCT
easat-10905	342	1	metrics/	metrics/	NUM
easat-10905	342	2	models	model	NOUN
easat-10905	342	3	average	average	ADJ
easat-10905	342	4	accuracy	accuracy	NOUN
easat-10905	342	5	average	average	ADJ
easat-10905	342	6	incremental	incremental	ADJ
easat-10905	342	7	accuracy	accuracy	NOUN
easat-10905	342	8	forgetting	forget	VERB
easat-10905	342	9	rate	rate	NOUN
easat-10905	342	10	forward	forward	ADJ
easat-10905	342	11	transfer	transfer	VERB
easat-10905	342	12	backward	backward	ADJ
easat-10905	342	13	transfer	transfer	NOUN
easat-10905	342	14	bert	bert	PROPN
easat-10905	342	15	94.93	94.93	NUM
easat-10905	342	16	92.4	92.4	NUM
easat-10905	342	17	1.06	1.06	NUM
easat-10905	342	18	98.8	98.8	NUM
easat-10905	342	19	7.74	7.74	NUM
easat-10905	342	20	albert	albert	PROPN
easat-10905	342	21	86.37	86.37	NUM
easat-10905	342	22	84.98	84.98	NUM
easat-10905	342	23	2.78	2.78	NUM
easat-10905	342	24	95.82	95.82	NUM
easat-10905	342	25	18.9	18.9	NUM
easat-10905	342	26	roberta	roberta	PROPN
easat-10905	342	27	88.21	88.21	NUM
easat-10905	342	28	84.65	84.65	NUM
easat-10905	342	29	7.12	7.12	NUM
easat-10905	342	30	95.52	95.52	NUM
easat-10905	342	31	14.62	14.62	NUM
easat-10905	342	32	distilbert	distilbert	NOUN
easat-10905	342	33	92.8	92.8	NUM
easat-10905	342	34	90.87	90.87	NUM
easat-10905	342	35	3.86	3.86	NUM
easat-10905	342	36	97.86	97.86	NUM
easat-10905	342	37	10.12	10.12	NUM
easat-10905	342	38	5	5	NUM
easat-10905	342	39	.	.	PUNCT
easat-10905	342	40	findings	finding	NOUN
easat-10905	342	41	and	and	CCONJ
easat-10905	342	42	discussion	discussion	NOUN
easat-10905	342	43	this	this	DET
easat-10905	342	44	section	section	NOUN
easat-10905	342	45	presents	present	VERB
easat-10905	342	46	and	and	CCONJ
easat-10905	342	47	discusses	discuss	VERB
easat-10905	342	48	the	the	DET
easat-10905	342	49	impact	impact	NOUN
easat-10905	342	50	of	of	ADP
easat-10905	342	51	using	use	VERB
easat-10905	342	52	task	task	NOUN
easat-10905	342	53	incremental	incremental	ADJ
easat-10905	342	54	learning	learning	NOUN
easat-10905	342	55	for	for	ADP
easat-10905	342	56	the	the	DET
easat-10905	342	57	sequential	sequential	ADJ
easat-10905	342	58	classification	classification	NOUN
easat-10905	342	59	of	of	ADP
easat-10905	342	60	abusive	abusive	ADJ
easat-10905	342	61	and	and	CCONJ
easat-10905	342	62	offensive	offensive	ADJ
easat-10905	342	63	comments	comment	NOUN
easat-10905	342	64	.	.	PUNCT
easat-10905	343	1	the	the	DET
easat-10905	343	2	focus	focus	NOUN
easat-10905	343	3	of	of	ADP
easat-10905	343	4	this	this	DET
easat-10905	343	5	study	study	NOUN
easat-10905	343	6	is	be	AUX
easat-10905	343	7	to	to	PART
easat-10905	343	8	evaluate	evaluate	VERB
easat-10905	343	9	the	the	DET
easat-10905	343	10	effectiveness	effectiveness	NOUN
easat-10905	343	11	of	of	ADP
easat-10905	343	12	task	task	NOUN
easat-10905	343	13	incremental	incremental	ADJ
easat-10905	343	14	learning	learning	NOUN
easat-10905	343	15	in	in	ADP
easat-10905	343	16	enabling	enable	VERB
easat-10905	343	17	models	model	NOUN
easat-10905	343	18	to	to	PART
easat-10905	343	19	learn	learn	VERB
easat-10905	343	20	new	new	ADJ
easat-10905	343	21	tasks	task	NOUN
easat-10905	343	22	,	,	PUNCT
easat-10905	343	23	such	such	ADJ
easat-10905	343	24	as	as	ADP
easat-10905	343	25	offensive	offensive	ADJ
easat-10905	343	26	comment	comment	NOUN
easat-10905	343	27	detection	detection	NOUN
easat-10905	343	28	,	,	PUNCT
easat-10905	343	29	without	without	ADP
easat-10905	343	30	forgetting	forget	VERB
easat-10905	343	31	previously	previously	ADV
easat-10905	343	32	learned	learn	VERB
easat-10905	343	33	tasks	task	NOUN
easat-10905	343	34	,	,	PUNCT
easat-10905	343	35	like	like	ADP
easat-10905	343	36	abusive	abusive	ADJ
easat-10905	343	37	comment	comment	NOUN
easat-10905	343	38	detection	detection	NOUN
easat-10905	343	39	.	.	PUNCT
easat-10905	344	1	prior	prior	ADV
easat-10905	344	2	to	to	ADP
easat-10905	344	3	retraining	retrain	VERB
easat-10905	344	4	,	,	PUNCT
easat-10905	344	5	the	the	DET
easat-10905	344	6	models	model	NOUN
easat-10905	344	7	demonstrated	demonstrate	VERB
easat-10905	344	8	good	good	ADJ
easat-10905	344	9	performance	performance	NOUN
easat-10905	344	10	on	on	ADP
easat-10905	344	11	abusive	abusive	ADJ
easat-10905	344	12	comments	comment	NOUN
easat-10905	344	13	but	but	CCONJ
easat-10905	344	14	exhibited	exhibit	VERB
easat-10905	344	15	a	a	DET
easat-10905	344	16	decline	decline	NOUN
easat-10905	344	17	in	in	ADP
easat-10905	344	18	accuracy	accuracy	NOUN
easat-10905	344	19	when	when	SCONJ
easat-10905	344	20	applied	apply	VERB
easat-10905	344	21	to	to	ADP
easat-10905	344	22	offensive	offensive	ADJ
easat-10905	344	23	comments	comment	NOUN
easat-10905	344	24	.	.	PUNCT
easat-10905	345	1	this	this	DET
easat-10905	345	2	decrease	decrease	NOUN
easat-10905	345	3	is	be	AUX
easat-10905	345	4	attributed	attribute	VERB
easat-10905	345	5	to	to	ADP
easat-10905	345	6	the	the	DET
easat-10905	345	7	models	model	NOUN
easat-10905	345	8	having	having	AUX
easat-10905	345	9	optimized	optimize	VERB
easat-10905	345	10	their	their	PRON
easat-10905	345	11	parameters	parameter	NOUN
easat-10905	345	12	specifically	specifically	ADV
easat-10905	345	13	for	for	ADP
easat-10905	345	14	abusive	abusive	ADJ
easat-10905	345	15	text	text	NOUN
easat-10905	345	16	,	,	PUNCT
easat-10905	345	17	which	which	PRON
easat-10905	345	18	are	be	AUX
easat-10905	345	19	not	not	PART
easat-10905	345	20	effective	effective	ADJ
easat-10905	345	21	when	when	SCONJ
easat-10905	345	22	evaluating	evaluate	VERB
easat-10905	345	23	offensive	offensive	ADJ
easat-10905	345	24	comments	comment	NOUN
easat-10905	345	25	.	.	PUNCT
easat-10905	346	1	among	among	ADP
easat-10905	346	2	the	the	DET
easat-10905	346	3	models	model	NOUN
easat-10905	346	4	,	,	PUNCT
easat-10905	346	5	distilbert	distilbert	PROPN
easat-10905	346	6	achieved	achieve	VERB
easat-10905	346	7	an	an	DET
easat-10905	346	8	accuracy	accuracy	NOUN
easat-10905	346	9	of	of	ADP
easat-10905	346	10	92.03	92.03	NUM
easat-10905	346	11	%	%	NOUN
easat-10905	346	12	on	on	ADP
easat-10905	346	13	the	the	DET
easat-10905	346	14	abusive	abusive	ADJ
easat-10905	346	15	dataset	dataset	NOUN
easat-10905	346	16	,	,	PUNCT
easat-10905	346	17	while	while	SCONJ
easat-10905	346	18	only	only	ADV
easat-10905	346	19	60.17	60.17	NUM
easat-10905	346	20	%	%	NOUN
easat-10905	346	21	on	on	ADP
easat-10905	346	22	the	the	DET
easat-10905	346	23	offensive	offensive	ADJ
easat-10905	346	24	dataset	dataset	NOUN
easat-10905	346	25	.	.	PUNCT
easat-10905	347	1	following	follow	VERB
easat-10905	347	2	the	the	DET
easat-10905	347	3	implementation	implementation	NOUN
easat-10905	347	4	of	of	ADP
easat-10905	347	5	replay	replay	NOUN
easat-10905	347	6	-	-	PUNCT
easat-10905	347	7	based	base	VERB
easat-10905	347	8	approaches	approach	NOUN
easat-10905	347	9	,	,	PUNCT
easat-10905	347	10	we	we	PRON
easat-10905	347	11	observed	observe	VERB
easat-10905	347	12	a	a	DET
easat-10905	347	13	decline	decline	NOUN
easat-10905	347	14	in	in	ADP
easat-10905	347	15	the	the	DET
easat-10905	347	16	performance	performance	NOUN
easat-10905	347	17	of	of	ADP
easat-10905	347	18	the	the	DET
easat-10905	347	19	models	model	NOUN
easat-10905	347	20	on	on	ADP
easat-10905	347	21	abusive	abusive	ADJ
easat-10905	347	22	comments	comment	NOUN
easat-10905	347	23	.	.	PUNCT
easat-10905	348	1	in	in	ADP
easat-10905	348	2	particular	particular	ADJ
easat-10905	348	3	,	,	PUNCT
easat-10905	348	4	the	the	DET
easat-10905	348	5	rehearsal	rehearsal	NOUN
easat-10905	348	6	approach	approach	NOUN
easat-10905	348	7	led	lead	VERB
easat-10905	348	8	to	to	ADP
easat-10905	348	9	a	a	DET
easat-10905	348	10	high	high	ADJ
easat-10905	348	11	reduction	reduction	NOUN
easat-10905	348	12	in	in	ADP
easat-10905	348	13	accuracy	accuracy	NOUN
easat-10905	348	14	for	for	ADP
easat-10905	348	15	abusive	abusive	ADJ
easat-10905	348	16	text	text	NOUN
easat-10905	348	17	classification	classification	NOUN
easat-10905	348	18	,	,	PUNCT
easat-10905	348	19	whereas	whereas	SCONJ
easat-10905	348	20	the	the	DET
easat-10905	348	21	pseudo	pseudo	NOUN
easat-10905	348	22	-	-	ADJ
easat-10905	348	23	rehearsal	rehearsal	ADJ
easat-10905	348	24	approach	approach	NOUN
easat-10905	348	25	demonstrated	demonstrate	VERB
easat-10905	348	26	comparatively	comparatively	ADV
easat-10905	348	27	better	well	ADJ
easat-10905	348	28	results	result	NOUN
easat-10905	348	29	.	.	PUNCT
easat-10905	349	1	notably	notably	ADV
easat-10905	349	2	,	,	PUNCT
easat-10905	349	3	models	model	NOUN
easat-10905	349	4	like	like	ADP
easat-10905	349	5	bert	bert	PROPN
easat-10905	349	6	,	,	PUNCT
easat-10905	349	7	albert	albert	PROPN
easat-10905	349	8	,	,	PUNCT
easat-10905	349	9	and	and	CCONJ
easat-10905	349	10	distilbert	distilbert	PROPN
easat-10905	349	11	maintained	maintain	VERB
easat-10905	349	12	performance	performance	NOUN
easat-10905	349	13	levels	level	NOUN
easat-10905	349	14	on	on	ADP
easat-10905	349	15	abusive	abusive	ADJ
easat-10905	349	16	comments	comment	NOUN
easat-10905	349	17	that	that	PRON
easat-10905	349	18	were	be	AUX
easat-10905	349	19	nearly	nearly	ADV
easat-10905	349	20	consistent	consistent	ADJ
easat-10905	349	21	with	with	ADP
easat-10905	349	22	those	those	PRON
easat-10905	349	23	observed	observe	VERB
easat-10905	349	24	before	before	ADP
easat-10905	349	25	applying	apply	VERB
easat-10905	349	26	incremental	incremental	ADJ
easat-10905	349	27	learning	learning	NOUN
easat-10905	349	28	.	.	PUNCT
easat-10905	350	1	furthermore	furthermore	ADV
easat-10905	350	2	,	,	PUNCT
easat-10905	350	3	these	these	DET
easat-10905	350	4	models	model	NOUN
easat-10905	350	5	exhibited	exhibit	VERB
easat-10905	350	6	significantly	significantly	ADV
easat-10905	350	7	improved	improve	VERB
easat-10905	350	8	performance	performance	NOUN
easat-10905	350	9	on	on	ADP
easat-10905	350	10	offensive	offensive	ADJ
easat-10905	350	11	texts	text	NOUN
easat-10905	350	12	compared	compare	VERB
easat-10905	350	13	to	to	ADP
easat-10905	350	14	their	their	PRON
easat-10905	350	15	performance	performance	NOUN
easat-10905	350	16	prior	prior	ADV
easat-10905	350	17	to	to	ADP
easat-10905	350	18	retraining	retrain	VERB
easat-10905	350	19	.	.	PUNCT
easat-10905	351	1	unlike	unlike	ADP
easat-10905	351	2	rehearsal	rehearsal	NOUN
easat-10905	351	3	-	-	PUNCT
easat-10905	351	4	based	base	VERB
easat-10905	351	5	methods	method	NOUN
easat-10905	351	6	that	that	PRON
easat-10905	351	7	rely	rely	VERB
easat-10905	351	8	on	on	ADP
easat-10905	351	9	storing	store	VERB
easat-10905	351	10	and	and	CCONJ
easat-10905	351	11	replaying	replay	VERB
easat-10905	351	12	a	a	DET
easat-10905	351	13	limited	limited	ADJ
easat-10905	351	14	subset	subset	NOUN
easat-10905	351	15	of	of	ADP
easat-10905	351	16	actual	actual	ADJ
easat-10905	351	17	data	datum	NOUN
easat-10905	351	18	,	,	PUNCT
easat-10905	351	19	pseudo	pseudo	NOUN
easat-10905	351	20	-	-	NOUN
easat-10905	351	21	rehearsal	rehearsal	NOUN
easat-10905	351	22	generates	generate	VERB
easat-10905	351	23	synthetic	synthetic	ADJ
easat-10905	351	24	data	datum	NOUN
easat-10905	351	25	that	that	PRON
easat-10905	351	26	broadly	broadly	ADV
easat-10905	351	27	represents	represent	VERB
easat-10905	351	28	the	the	DET
easat-10905	351	29	distribution	distribution	NOUN
easat-10905	351	30	of	of	ADP
easat-10905	351	31	abusive	abusive	ADJ
easat-10905	351	32	tasks	task	NOUN
easat-10905	351	33	,	,	PUNCT
easat-10905	351	34	leading	lead	VERB
easat-10905	351	35	to	to	ADP
easat-10905	351	36	improved	improved	ADJ
easat-10905	351	37	results	result	NOUN
easat-10905	351	38	.	.	PUNCT
easat-10905	352	1	this	this	DET
easat-10905	352	2	synthetic	synthetic	ADJ
easat-10905	352	3	data	datum	NOUN
easat-10905	352	4	helps	help	VERB
easat-10905	352	5	prevent	prevent	VERB
easat-10905	352	6	overfitting	overfitting	NOUN
easat-10905	352	7	and	and	CCONJ
easat-10905	352	8	better	well	ADJ
easat-10905	352	9	preserves	preserve	NOUN
easat-10905	352	10	knowledge	knowledge	NOUN
easat-10905	352	11	,	,	PUNCT
easat-10905	352	12	which	which	PRON
easat-10905	352	13	enhances	enhance	VERB
easat-10905	352	14	the	the	DET
easat-10905	352	15	models	model	NOUN
easat-10905	352	16	'	'	PART
easat-10905	352	17	ability	ability	NOUN
easat-10905	352	18	to	to	PART
easat-10905	352	19	generalize	generalize	VERB
easat-10905	352	20	across	across	ADP
easat-10905	352	21	tasks	task	NOUN
easat-10905	352	22	.	.	PUNCT
easat-10905	353	1	additionally	additionally	ADV
easat-10905	353	2	,	,	PUNCT
easat-10905	353	3	pseudo	pseudo	NOUN
easat-10905	353	4	-	-	NOUN
easat-10905	353	5	rehearsal	rehearsal	NOUN
easat-10905	353	6	effectively	effectively	ADV
easat-10905	353	7	mitigates	mitigate	VERB
easat-10905	353	8	catastrophic	catastrophic	ADJ
easat-10905	353	9	forgetting	forgetting	NOUN
easat-10905	353	10	by	by	ADP
easat-10905	353	11	providing	provide	VERB
easat-10905	353	12	diverse	diverse	ADJ
easat-10905	353	13	and	and	CCONJ
easat-10905	353	14	representative	representative	ADJ
easat-10905	353	15	examples	example	NOUN
easat-10905	353	16	,	,	PUNCT
easat-10905	353	17	allowing	allow	VERB
easat-10905	353	18	the	the	DET
easat-10905	353	19	models	model	NOUN
easat-10905	353	20	to	to	PART
easat-10905	353	21	adapt	adapt	VERB
easat-10905	353	22	to	to	ADP
easat-10905	353	23	new	new	ADJ
easat-10905	353	24	tasks	task	NOUN
easat-10905	353	25	without	without	ADP
easat-10905	353	26	compromising	compromise	VERB
easat-10905	353	27	performance	performance	NOUN
easat-10905	353	28	on	on	ADP
easat-10905	353	29	earlier	early	ADJ
easat-10905	353	30	ones	one	NOUN
easat-10905	353	31	.	.	PUNCT
easat-10905	354	1	following	follow	VERB
easat-10905	354	2	the	the	DET
easat-10905	354	3	implementation	implementation	NOUN
easat-10905	354	4	of	of	ADP
easat-10905	354	5	lwf	lwf	PROPN
easat-10905	354	6	,	,	PUNCT
easat-10905	354	7	a	a	DET
easat-10905	354	8	notable	notable	ADJ
easat-10905	354	9	improvement	improvement	NOUN
easat-10905	354	10	in	in	ADP
easat-10905	354	11	performance	performance	NOUN
easat-10905	354	12	on	on	ADP
easat-10905	354	13	offensive	offensive	ADJ
easat-10905	354	14	comments	comment	NOUN
easat-10905	354	15	was	be	AUX
easat-10905	354	16	observed	observe	VERB
easat-10905	354	17	across	across	ADP
easat-10905	354	18	the	the	DET
easat-10905	354	19	models	model	NOUN
easat-10905	354	20	,	,	PUNCT
easat-10905	354	21	with	with	ADP
easat-10905	354	22	bert	bert	NOUN
easat-10905	354	23	achieving	achieve	VERB
easat-10905	354	24	98.8	98.8	NUM
easat-10905	354	25	%	%	NOUN
easat-10905	354	26	,	,	PUNCT
easat-10905	354	27	albert	albert	PROPN
easat-10905	354	28	95.82	95.82	NUM
easat-10905	354	29	%	%	NOUN
easat-10905	354	30	,	,	PUNCT
easat-10905	354	31	and	and	CCONJ
easat-10905	354	32	roberta	roberta	PROPN
easat-10905	354	33	95.52	95.52	NUM
easat-10905	354	34	%	%	NOUN
easat-10905	354	35	.	.	PUNCT
easat-10905	355	1	the	the	DET
easat-10905	355	2	increase	increase	NOUN
easat-10905	355	3	in	in	ADP
easat-10905	355	4	performance	performance	NOUN
easat-10905	355	5	on	on	ADP
easat-10905	355	6	offensive	offensive	ADJ
easat-10905	355	7	comments	comment	NOUN
easat-10905	355	8	with	with	ADP
easat-10905	355	9	lwf	lwf	PROPN
easat-10905	355	10	is	be	AUX
easat-10905	355	11	attributed	attribute	VERB
easat-10905	355	12	to	to	ADP
easat-10905	355	13	the	the	DET
easat-10905	355	14	ability	ability	NOUN
easat-10905	355	15	of	of	ADP
easat-10905	355	16	the	the	DET
easat-10905	355	17	models	model	NOUN
easat-10905	355	18	to	to	PART
easat-10905	355	19	adapt	adapt	VERB
easat-10905	355	20	effectively	effectively	ADV
easat-10905	355	21	to	to	ADP
easat-10905	355	22	new	new	ADJ
easat-10905	355	23	tasks	task	NOUN
easat-10905	355	24	while	while	SCONJ
easat-10905	355	25	preserving	preserve	VERB
easat-10905	355	26	knowledge	knowledge	NOUN
easat-10905	355	27	of	of	ADP
easat-10905	355	28	previous	previous	ADJ
easat-10905	355	29	tasks	task	NOUN
easat-10905	355	30	through	through	ADP
easat-10905	355	31	knowledge	knowledge	NOUN
easat-10905	355	32	distillation	distillation	NOUN
easat-10905	355	33	.	.	PUNCT
easat-10905	356	1	lwf	lwf	PROPN
easat-10905	356	2	regularizes	regularize	VERB
easat-10905	356	3	the	the	DET
easat-10905	356	4	learning	learning	NOUN
easat-10905	356	5	process	process	NOUN
easat-10905	356	6	,	,	PUNCT
easat-10905	356	7	ensuring	ensure	VERB
easat-10905	356	8	that	that	SCONJ
easat-10905	356	9	the	the	DET
easat-10905	356	10	models	model	NOUN
easat-10905	356	11	retain	retain	VERB
easat-10905	356	12	the	the	DET
easat-10905	356	13	required	require	VERB
easat-10905	356	14	knowledge	knowledge	NOUN
easat-10905	356	15	from	from	ADP
easat-10905	356	16	earlier	early	ADJ
easat-10905	356	17	tasks	task	NOUN
easat-10905	356	18	,	,	PUNCT
easat-10905	356	19	namely	namely	ADV
easat-10905	356	20	abusive	abusive	ADJ
easat-10905	356	21	comment	comment	NOUN
easat-10905	356	22	classification	classification	NOUN
easat-10905	356	23	,	,	PUNCT
easat-10905	356	24	while	while	SCONJ
easat-10905	356	25	learning	learn	VERB
easat-10905	356	26	to	to	PART
easat-10905	356	27	classify	classify	VERB
easat-10905	356	28	offensive	offensive	ADJ
easat-10905	356	29	comments	comment	NOUN
easat-10905	356	30	.	.	PUNCT
easat-10905	357	1	this	this	DET
easat-10905	357	2	regularization	regularization	NOUN
easat-10905	357	3	helps	help	VERB
easat-10905	357	4	prevent	prevent	VERB
easat-10905	357	5	catastrophic	catastrophic	ADJ
easat-10905	357	6	forgetting	forgetting	NOUN
easat-10905	357	7	(	(	PUNCT
easat-10905	357	8	cf	cf	NOUN
easat-10905	357	9	)	)	PUNCT
easat-10905	357	10	and	and	CCONJ
easat-10905	357	11	allows	allow	VERB
easat-10905	357	12	the	the	DET
easat-10905	357	13	models	model	NOUN
easat-10905	357	14	to	to	PART
easat-10905	357	15	perform	perform	VERB
easat-10905	357	16	well	well	ADV
easat-10905	357	17	on	on	ADP
easat-10905	357	18	the	the	DET
easat-10905	357	19	offensive	offensive	ADJ
easat-10905	357	20	dataset	dataset	NOUN
easat-10905	357	21	.	.	PUNCT
easat-10905	358	1	except	except	SCONJ
easat-10905	358	2	for	for	ADP
easat-10905	358	3	bert	bert	NOUN
easat-10905	358	4	,	,	PUNCT
easat-10905	358	5	all	all	DET
easat-10905	358	6	models	model	NOUN
easat-10905	358	7	experienced	experience	VERB
easat-10905	358	8	a	a	DET
easat-10905	358	9	decline	decline	NOUN
easat-10905	358	10	in	in	ADP
easat-10905	358	11	performance	performance	NOUN
easat-10905	358	12	on	on	ADP
easat-10905	358	13	abusive	abusive	ADJ
easat-10905	358	14	comments	comment	NOUN
easat-10905	358	15	when	when	SCONJ
easat-10905	358	16	using	use	VERB
easat-10905	358	17	lwf	lwf	NOUN
easat-10905	358	18	compared	compare	VERB
easat-10905	358	19	to	to	ADP
easat-10905	358	20	the	the	DET
easat-10905	358	21	pseudo	pseudo	NOUN
easat-10905	358	22	-	-	ADJ
easat-10905	358	23	rehearsal	rehearsal	NOUN
easat-10905	358	24	approach	approach	NOUN
easat-10905	358	25	,	,	PUNCT
easat-10905	358	26	with	with	ADP
easat-10905	358	27	accuracy	accuracy	NOUN
easat-10905	358	28	of	of	ADP
easat-10905	358	29	albert	albert	PROPN
easat-10905	358	30	dropping	drop	VERB
easat-10905	358	31	from	from	ADP
easat-10905	358	32	78	78	NUM
easat-10905	358	33	%	%	NOUN
easat-10905	358	34	to	to	ADP
easat-10905	358	35	76.92	76.92	NUM
easat-10905	358	36	%	%	NOUN
easat-10905	358	37	,	,	PUNCT
easat-10905	358	38	roberta	roberta	PROPN
easat-10905	358	39	’s	’s	PART
easat-10905	358	40	from	from	ADP
easat-10905	358	41	86.6	86.6	NUM
easat-10905	358	42	%	%	NOUN
easat-10905	358	43	to	to	ADP
easat-10905	358	44	80.9	80.9	NUM
easat-10905	358	45	%	%	NOUN
easat-10905	358	46	,	,	PUNCT
easat-10905	358	47	and	and	CCONJ
easat-10905	358	48	distilbert	distilbert	PROPN
easat-10905	358	49	’s	’s	ADV
easat-10905	358	50	from	from	ADP
easat-10905	358	51	90.72	90.72	NUM
easat-10905	358	52	%	%	NOUN
easat-10905	358	53	to	to	ADP
easat-10905	358	54	87.74	87.74	NUM
easat-10905	358	55	%	%	NOUN
easat-10905	358	56	.	.	PUNCT
easat-10905	359	1	these	these	DET
easat-10905	359	2	results	result	NOUN
easat-10905	359	3	indicate	indicate	VERB
easat-10905	359	4	that	that	SCONJ
easat-10905	359	5	bert	bert	NOUN
easat-10905	359	6	,	,	PUNCT
easat-10905	359	7	when	when	SCONJ
easat-10905	359	8	used	use	VERB
easat-10905	359	9	with	with	ADP
easat-10905	359	10	lwf	lwf	PROPN
easat-10905	359	11	,	,	PUNCT
easat-10905	359	12	surpasses	surpass	VERB
easat-10905	359	13	all	all	DET
easat-10905	359	14	other	other	ADJ
easat-10905	359	15	models	model	NOUN
easat-10905	359	16	in	in	ADP
easat-10905	359	17	accuracy	accuracy	NOUN
easat-10905	359	18	for	for	ADP
easat-10905	359	19	both	both	CCONJ
easat-10905	359	20	abusive	abusive	ADJ
easat-10905	359	21	and	and	CCONJ
easat-10905	359	22	offensive	offensive	ADJ
easat-10905	359	23	tasks	task	NOUN
easat-10905	359	24	.	.	PUNCT
easat-10905	360	1	the	the	DET
easat-10905	360	2	improved	improved	ADJ
easat-10905	360	3	performance	performance	NOUN
easat-10905	360	4	463	463	NUM
easat-10905	360	5	edelweiss	edelweiss	PROPN
easat-10905	360	6	applied	apply	VERB
easat-10905	360	7	science	science	NOUN
easat-10905	360	8	and	and	CCONJ
easat-10905	360	9	technology	technology	NOUN
easat-10905	360	10	issn	issn	PROPN
easat-10905	360	11	:	:	PUNCT
easat-10905	360	12	2576	2576	NUM
easat-10905	360	13	-	-	SYM
easat-10905	360	14	8484	8484	NUM
easat-10905	360	15	vol	vol	NOUN
easat-10905	360	16	.	.	PROPN
easat-10905	360	17	9	9	NUM
easat-10905	360	18	,	,	PUNCT
easat-10905	360	19	no	no	INTJ
easat-10905	360	20	.	.	NOUN
easat-10905	360	21	11	11	NUM
easat-10905	360	22	:	:	PUNCT
easat-10905	360	23	448	448	NUM
easat-10905	360	24	-	-	SYM
easat-10905	360	25	466	466	NUM
easat-10905	360	26	,	,	PUNCT
easat-10905	360	27	2025	2025	NUM
easat-10905	360	28	doi	doi	NOUN
easat-10905	360	29	:	:	PUNCT
easat-10905	360	30	10.55214/2576	10.55214/2576	NUM
easat-10905	360	31	-	-	PUNCT
easat-10905	360	32	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	360	33	©	©	PROPN
easat-10905	360	34	2025	2025	NUM
easat-10905	360	35	by	by	ADP
easat-10905	360	36	the	the	DET
easat-10905	360	37	authors	author	NOUN
easat-10905	360	38	;	;	PUNCT
easat-10905	360	39	licensee	licensee	PROPN
easat-10905	360	40	learning	learn	VERB
easat-10905	360	41	gate	gate	NOUN
easat-10905	360	42	of	of	ADP
easat-10905	360	43	bert	bert	PROPN
easat-10905	360	44	with	with	ADP
easat-10905	360	45	lwf	lwf	PROPN
easat-10905	360	46	for	for	ADP
easat-10905	360	47	both	both	CCONJ
easat-10905	360	48	abusive	abusive	ADJ
easat-10905	360	49	and	and	CCONJ
easat-10905	360	50	offensive	offensive	ADJ
easat-10905	360	51	datasets	dataset	NOUN
easat-10905	360	52	is	be	AUX
easat-10905	360	53	due	due	ADJ
easat-10905	360	54	to	to	ADP
easat-10905	360	55	its	its	PRON
easat-10905	360	56	robust	robust	ADJ
easat-10905	360	57	bidirectional	bidirectional	ADJ
easat-10905	360	58	architecture	architecture	NOUN
easat-10905	360	59	.	.	PUNCT
easat-10905	361	1	figure	figure	NOUN
easat-10905	361	2	2	2	NUM
easat-10905	361	3	comprehensively	comprehensively	ADV
easat-10905	361	4	compares	compare	VERB
easat-10905	361	5	the	the	DET
easat-10905	361	6	accuracy	accuracy	NOUN
easat-10905	361	7	achieved	achieve	VERB
easat-10905	361	8	by	by	ADP
easat-10905	361	9	all	all	DET
easat-10905	361	10	the	the	DET
easat-10905	361	11	evaluated	evaluated	ADJ
easat-10905	361	12	models	model	NOUN
easat-10905	361	13	.	.	PUNCT
easat-10905	362	1	it	it	PRON
easat-10905	362	2	demonstrates	demonstrate	VERB
easat-10905	362	3	that	that	SCONJ
easat-10905	362	4	the	the	DET
easat-10905	362	5	lwf	lwf	PROPN
easat-10905	362	6	consistently	consistently	ADV
easat-10905	362	7	delivers	deliver	VERB
easat-10905	362	8	superior	superior	ADJ
easat-10905	362	9	performance	performance	NOUN
easat-10905	362	10	across	across	ADP
easat-10905	362	11	both	both	DET
easat-10905	362	12	task	task	NOUN
easat-10905	362	13	types	type	NOUN
easat-10905	362	14	.	.	PUNCT
easat-10905	363	1	figure	figure	NOUN
easat-10905	363	2	2	2	NUM
easat-10905	363	3	.	.	PUNCT
easat-10905	363	4	comparison	comparison	NOUN
easat-10905	363	5	of	of	ADP
easat-10905	363	6	all	all	DET
easat-10905	363	7	models	model	NOUN
easat-10905	363	8	on	on	ADP
easat-10905	363	9	both	both	DET
easat-10905	363	10	tasks	task	NOUN
easat-10905	363	11	.	.	PUNCT
easat-10905	364	1	unlike	unlike	ADP
easat-10905	364	2	metrics	metric	NOUN
easat-10905	364	3	such	such	ADJ
easat-10905	364	4	as	as	ADP
easat-10905	364	5	accuracy	accuracy	NOUN
easat-10905	364	6	,	,	PUNCT
easat-10905	364	7	recall	recall	NOUN
easat-10905	364	8	,	,	PUNCT
easat-10905	364	9	precision	precision	NOUN
easat-10905	364	10	,	,	PUNCT
easat-10905	364	11	and	and	CCONJ
easat-10905	364	12	f1	f1	PROPN
easat-10905	364	13	score	score	NOUN
easat-10905	364	14	,	,	PUNCT
easat-10905	364	15	which	which	PRON
easat-10905	364	16	focus	focus	VERB
easat-10905	364	17	on	on	ADP
easat-10905	364	18	single	single	ADJ
easat-10905	364	19	-	-	PUNCT
easat-10905	364	20	task	task	NOUN
easat-10905	364	21	performance	performance	NOUN
easat-10905	364	22	,	,	PUNCT
easat-10905	364	23	metrics	metric	NOUN
easat-10905	364	24	such	such	ADJ
easat-10905	364	25	as	as	ADP
easat-10905	364	26	average	average	ADJ
easat-10905	364	27	accuracy	accuracy	NOUN
easat-10905	364	28	,	,	PUNCT
easat-10905	364	29	average	average	ADJ
easat-10905	364	30	incremental	incremental	ADJ
easat-10905	364	31	accuracy	accuracy	NOUN
easat-10905	364	32	,	,	PUNCT
easat-10905	364	33	forgetting	forget	VERB
easat-10905	364	34	rate	rate	NOUN
easat-10905	364	35	,	,	PUNCT
easat-10905	364	36	forward	forward	ADJ
easat-10905	364	37	transfer	transfer	NOUN
easat-10905	364	38	,	,	PUNCT
easat-10905	364	39	and	and	CCONJ
easat-10905	364	40	backward	backward	ADJ
easat-10905	364	41	transfer	transfer	NOUN
easat-10905	364	42	provide	provide	VERB
easat-10905	364	43	a	a	DET
easat-10905	364	44	comprehensive	comprehensive	ADJ
easat-10905	364	45	view	view	NOUN
easat-10905	364	46	of	of	ADP
easat-10905	364	47	how	how	SCONJ
easat-10905	364	48	well	well	ADV
easat-10905	364	49	models	model	NOUN
easat-10905	364	50	perform	perform	VERB
easat-10905	364	51	in	in	ADP
easat-10905	364	52	task	task	NOUN
easat-10905	364	53	incremental	incremental	ADJ
easat-10905	364	54	learning	learning	NOUN
easat-10905	364	55	by	by	ADP
easat-10905	364	56	assessing	assess	VERB
easat-10905	364	57	their	their	PRON
easat-10905	364	58	ability	ability	NOUN
easat-10905	364	59	to	to	PART
easat-10905	364	60	retain	retain	VERB
easat-10905	364	61	knowledge	knowledge	NOUN
easat-10905	364	62	,	,	PUNCT
easat-10905	364	63	adapt	adapt	VERB
easat-10905	364	64	to	to	ADP
easat-10905	364	65	new	new	ADJ
easat-10905	364	66	tasks	task	NOUN
easat-10905	364	67	,	,	PUNCT
easat-10905	364	68	and	and	CCONJ
easat-10905	364	69	manage	manage	VERB
easat-10905	364	70	performance	performance	NOUN
easat-10905	364	71	across	across	ADP
easat-10905	364	72	multiple	multiple	ADJ
easat-10905	364	73	tasks	task	NOUN
easat-10905	364	74	.	.	PUNCT
easat-10905	365	1	from	from	ADP
easat-10905	365	2	tables	table	NOUN
easat-10905	365	3	8	8	NUM
easat-10905	365	4	,	,	PUNCT
easat-10905	365	5	9	9	NUM
easat-10905	365	6	,	,	PUNCT
easat-10905	365	7	and	and	CCONJ
easat-10905	365	8	10	10	NUM
easat-10905	365	9	,	,	PUNCT
easat-10905	365	10	we	we	PRON
easat-10905	365	11	can	can	AUX
easat-10905	365	12	infer	infer	VERB
easat-10905	365	13	that	that	SCONJ
easat-10905	365	14	the	the	DET
easat-10905	365	15	average	average	ADJ
easat-10905	365	16	accuracy	accuracy	NOUN
easat-10905	365	17	and	and	CCONJ
easat-10905	365	18	average	average	ADJ
easat-10905	365	19	incremental	incremental	ADJ
easat-10905	365	20	accuracy	accuracy	NOUN
easat-10905	365	21	are	be	AUX
easat-10905	365	22	high	high	ADJ
easat-10905	365	23	for	for	ADP
easat-10905	365	24	bert	bert	PROPN
easat-10905	365	25	with	with	ADP
easat-10905	365	26	lwf	lwf	PROPN
easat-10905	365	27	compared	compare	VERB
easat-10905	365	28	to	to	ADP
easat-10905	365	29	other	other	ADJ
easat-10905	365	30	models	model	NOUN
easat-10905	365	31	.	.	PUNCT
easat-10905	366	1	high	high	ADJ
easat-10905	366	2	values	value	NOUN
easat-10905	366	3	of	of	ADP
easat-10905	366	4	these	these	DET
easat-10905	366	5	metrics	metric	NOUN
easat-10905	366	6	indicate	indicate	VERB
easat-10905	366	7	improved	improved	ADJ
easat-10905	366	8	performance	performance	NOUN
easat-10905	366	9	.	.	PUNCT
easat-10905	367	1	likewise	likewise	ADV
easat-10905	367	2	,	,	PUNCT
easat-10905	367	3	this	this	DET
easat-10905	367	4	model	model	NOUN
easat-10905	367	5	demonstrates	demonstrate	VERB
easat-10905	367	6	a	a	DET
easat-10905	367	7	lower	low	ADJ
easat-10905	367	8	forgetting	forget	VERB
easat-10905	367	9	rate	rate	NOUN
easat-10905	367	10	compared	compare	VERB
easat-10905	367	11	to	to	ADP
easat-10905	367	12	others	other	NOUN
easat-10905	367	13	,	,	PUNCT
easat-10905	367	14	indicating	indicate	VERB
easat-10905	367	15	that	that	SCONJ
easat-10905	367	16	bert	bert	PROPN
easat-10905	367	17	with	with	ADP
easat-10905	367	18	lwf	lwf	PROPN
easat-10905	367	19	effectively	effectively	ADV
easat-10905	367	20	maintains	maintain	VERB
easat-10905	367	21	performance	performance	NOUN
easat-10905	367	22	on	on	ADP
easat-10905	367	23	previously	previously	ADV
easat-10905	367	24	learned	learn	VERB
easat-10905	367	25	tasks	task	NOUN
easat-10905	367	26	as	as	SCONJ
easat-10905	367	27	new	new	ADJ
easat-10905	367	28	tasks	task	NOUN
easat-10905	367	29	are	be	AUX
easat-10905	367	30	introduced	introduce	VERB
easat-10905	367	31	,	,	PUNCT
easat-10905	367	32	thereby	thereby	ADV
easat-10905	367	33	reducing	reduce	VERB
easat-10905	367	34	cf	cf	NOUN
easat-10905	367	35	.	.	PUNCT
easat-10905	368	1	additionally	additionally	ADV
easat-10905	368	2	,	,	PUNCT
easat-10905	368	3	when	when	SCONJ
easat-10905	368	4	evaluating	evaluate	VERB
easat-10905	368	5	forward	forward	ADV
easat-10905	368	6	and	and	CCONJ
easat-10905	368	7	backward	backward	ADJ
easat-10905	368	8	transfer	transfer	NOUN
easat-10905	368	9	,	,	PUNCT
easat-10905	368	10	bert	bert	PROPN
easat-10905	368	11	with	with	ADP
easat-10905	368	12	lwf	lwf	PROPN
easat-10905	368	13	shows	show	VERB
easat-10905	368	14	high	high	ADJ
easat-10905	368	15	forward	forward	ADJ
easat-10905	368	16	transfer	transfer	NOUN
easat-10905	368	17	and	and	CCONJ
easat-10905	368	18	low	low	ADJ
easat-10905	368	19	backward	backward	ADJ
easat-10905	368	20	transfer	transfer	NOUN
easat-10905	368	21	compared	compare	VERB
easat-10905	368	22	to	to	ADP
easat-10905	368	23	the	the	DET
easat-10905	368	24	other	other	ADJ
easat-10905	368	25	models	model	NOUN
easat-10905	368	26	.	.	PUNCT
easat-10905	369	1	all	all	DET
easat-10905	369	2	these	these	DET
easat-10905	369	3	findings	finding	NOUN
easat-10905	369	4	support	support	VERB
easat-10905	369	5	our	our	PRON
easat-10905	369	6	results	result	NOUN
easat-10905	369	7	,	,	PUNCT
easat-10905	369	8	showing	show	VERB
easat-10905	369	9	that	that	SCONJ
easat-10905	369	10	bert	bert	NOUN
easat-10905	369	11	with	with	ADP
easat-10905	369	12	lwf	lwf	PROPN
easat-10905	369	13	outperforms	outperform	VERB
easat-10905	369	14	other	other	ADJ
easat-10905	369	15	models	model	NOUN
easat-10905	369	16	using	use	VERB
easat-10905	369	17	replay	replay	NOUN
easat-10905	369	18	-	-	PUNCT
easat-10905	369	19	based	base	VERB
easat-10905	369	20	approaches	approach	NOUN
easat-10905	369	21	.	.	PUNCT
easat-10905	370	1	to	to	PART
easat-10905	370	2	facilitate	facilitate	VERB
easat-10905	370	3	a	a	DET
easat-10905	370	4	comprehensive	comprehensive	ADJ
easat-10905	370	5	understanding	understanding	NOUN
easat-10905	370	6	of	of	ADP
easat-10905	370	7	the	the	DET
easat-10905	370	8	performance	performance	NOUN
easat-10905	370	9	of	of	ADP
easat-10905	370	10	the	the	DET
easat-10905	370	11	models	model	NOUN
easat-10905	370	12	in	in	ADP
easat-10905	370	13	incremental	incremental	ADJ
easat-10905	370	14	learning	learning	NOUN
easat-10905	370	15	,	,	PUNCT
easat-10905	370	16	figure	figure	VERB
easat-10905	370	17	3	3	NUM
easat-10905	370	18	presents	present	VERB
easat-10905	370	19	a	a	DET
easat-10905	370	20	comparative	comparative	ADJ
easat-10905	370	21	analysis	analysis	NOUN
easat-10905	370	22	of	of	ADP
easat-10905	370	23	all	all	DET
easat-10905	370	24	models	model	NOUN
easat-10905	370	25	within	within	ADP
easat-10905	370	26	a	a	DET
easat-10905	370	27	single	single	ADJ
easat-10905	370	28	diagram	diagram	NOUN
easat-10905	370	29	.	.	PUNCT
easat-10905	371	1	overall	overall	ADV
easat-10905	371	2	,	,	PUNCT
easat-10905	371	3	these	these	DET
easat-10905	371	4	findings	finding	NOUN
easat-10905	371	5	suggest	suggest	VERB
easat-10905	371	6	that	that	SCONJ
easat-10905	371	7	integrating	integrate	VERB
easat-10905	371	8	advanced	advanced	ADJ
easat-10905	371	9	models	model	NOUN
easat-10905	371	10	and	and	CCONJ
easat-10905	371	11	techniques	technique	NOUN
easat-10905	371	12	like	like	ADP
easat-10905	371	13	lwf	lwf	PROPN
easat-10905	371	14	significantly	significantly	ADV
easat-10905	371	15	enhances	enhance	VERB
easat-10905	371	16	performance	performance	NOUN
easat-10905	371	17	in	in	ADP
easat-10905	371	18	applications	application	NOUN
easat-10905	371	19	requiring	require	VERB
easat-10905	371	20	continuous	continuous	ADJ
easat-10905	371	21	learning	learning	NOUN
easat-10905	371	22	and	and	CCONJ
easat-10905	371	23	knowledge	knowledge	NOUN
easat-10905	371	24	preservation	preservation	NOUN
easat-10905	371	25	.	.	PUNCT
easat-10905	372	1	0	0	NUM
easat-10905	373	1	20	20	NUM
easat-10905	373	2	40	40	NUM
easat-10905	373	3	60	60	NUM
easat-10905	373	4	80	80	NUM
easat-10905	373	5	100	100	NUM
easat-10905	373	6	120	120	NUM
easat-10905	373	7	abusive	abusive	ADJ
easat-10905	373	8	offensive	offensive	ADJ
easat-10905	373	9	abusive	abusive	ADJ
easat-10905	373	10	offensive	offensive	ADJ
easat-10905	373	11	abusive	abusive	ADJ
easat-10905	373	12	offensive	offensive	ADJ
easat-10905	373	13	abusive	abusive	ADJ
easat-10905	373	14	offensive	offensive	NOUN
easat-10905	373	15	without	without	ADP
easat-10905	373	16	retraining	retrain	VERB
easat-10905	373	17	rehearsal	rehearsal	NOUN
easat-10905	373	18	pseudo	pseudo	NOUN
easat-10905	373	19	rehearsal	rehearsal	NOUN
easat-10905	373	20	lwf	lwf	VERB
easat-10905	373	21	comparison	comparison	NOUN
easat-10905	373	22	of	of	ADP
easat-10905	373	23	all	all	DET
easat-10905	373	24	models	model	NOUN
easat-10905	373	25	(	(	PUNCT
easat-10905	373	26	accuracy	accuracy	NOUN
easat-10905	373	27	)	)	PUNCT
easat-10905	373	28	bert	bert	PROPN
easat-10905	373	29	albert	albert	PROPN
easat-10905	373	30	roberta	roberta	PROPN
easat-10905	373	31	distilbert	distilbert	VERB
easat-10905	373	32	464	464	NUM
easat-10905	373	33	edelweiss	edelweiss	PROPN
easat-10905	373	34	applied	apply	VERB
easat-10905	373	35	science	science	NOUN
easat-10905	373	36	and	and	CCONJ
easat-10905	373	37	technology	technology	NOUN
easat-10905	373	38	issn	issn	PROPN
easat-10905	373	39	:	:	PUNCT
easat-10905	373	40	2576	2576	NUM
easat-10905	373	41	-	-	SYM
easat-10905	373	42	8484	8484	NUM
easat-10905	373	43	vol	vol	NOUN
easat-10905	373	44	.	.	PROPN
easat-10905	374	1	9	9	NUM
easat-10905	374	2	,	,	PUNCT
easat-10905	374	3	no	no	INTJ
easat-10905	374	4	.	.	NOUN
easat-10905	374	5	11	11	NUM
easat-10905	374	6	:	:	PUNCT
easat-10905	374	7	448	448	NUM
easat-10905	374	8	-	-	SYM
easat-10905	374	9	466	466	NUM
easat-10905	374	10	,	,	PUNCT
easat-10905	374	11	2025	2025	NUM
easat-10905	374	12	doi	doi	NOUN
easat-10905	374	13	:	:	PUNCT
easat-10905	374	14	10.55214/2576	10.55214/2576	NUM
easat-10905	374	15	-	-	PUNCT
easat-10905	374	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	375	1	©	©	PROPN
easat-10905	375	2	2025	2025	NUM
easat-10905	375	3	by	by	ADP
easat-10905	375	4	the	the	DET
easat-10905	375	5	authors	author	NOUN
easat-10905	375	6	;	;	PUNCT
easat-10905	375	7	licensee	licensee	PROPN
easat-10905	375	8	learning	learn	VERB
easat-10905	375	9	gate	gate	PROPN
easat-10905	375	10	figure	figure	NOUN
easat-10905	375	11	3	3	NUM
easat-10905	375	12	.	.	PUNCT
easat-10905	375	13	performance	performance	NOUN
easat-10905	375	14	of	of	ADP
easat-10905	375	15	all	all	DET
easat-10905	375	16	models	model	NOUN
easat-10905	375	17	on	on	ADP
easat-10905	375	18	incremental	incremental	ADJ
easat-10905	375	19	learning	learning	NOUN
easat-10905	375	20	.	.	PUNCT
easat-10905	376	1	6	6	X
easat-10905	376	2	.	.	X
easat-10905	376	3	conclusion	conclusion	NOUN
easat-10905	376	4	and	and	CCONJ
easat-10905	376	5	future	future	ADJ
easat-10905	376	6	work	work	NOUN
easat-10905	376	7	in	in	ADP
easat-10905	376	8	this	this	DET
easat-10905	376	9	study	study	NOUN
easat-10905	376	10	,	,	PUNCT
easat-10905	376	11	we	we	PRON
easat-10905	376	12	investigated	investigate	VERB
easat-10905	376	13	task	task	NOUN
easat-10905	376	14	incremental	incremental	ADJ
easat-10905	376	15	learning	learning	NOUN
easat-10905	376	16	for	for	ADP
easat-10905	376	17	sequential	sequential	ADJ
easat-10905	376	18	classification	classification	NOUN
easat-10905	376	19	of	of	ADP
easat-10905	376	20	abusive	abusive	ADJ
easat-10905	376	21	and	and	CCONJ
easat-10905	376	22	offensive	offensive	ADJ
easat-10905	376	23	texts	text	NOUN
easat-10905	376	24	using	use	VERB
easat-10905	376	25	bert	bert	PROPN
easat-10905	376	26	,	,	PUNCT
easat-10905	376	27	albert	albert	PROPN
easat-10905	376	28	,	,	PUNCT
easat-10905	376	29	distilbert	distilbert	PROPN
easat-10905	376	30	,	,	PUNCT
easat-10905	376	31	and	and	CCONJ
easat-10905	376	32	roberta	roberta	PROPN
easat-10905	376	33	models	model	NOUN
easat-10905	376	34	.	.	PUNCT
easat-10905	377	1	to	to	PART
easat-10905	377	2	add	add	VERB
easat-10905	377	3	incremental	incremental	ADJ
easat-10905	377	4	learning	learning	NOUN
easat-10905	377	5	capability	capability	NOUN
easat-10905	377	6	,	,	PUNCT
easat-10905	377	7	we	we	PRON
easat-10905	377	8	integrated	integrate	VERB
easat-10905	377	9	approaches	approach	NOUN
easat-10905	377	10	such	such	ADJ
easat-10905	377	11	as	as	ADP
easat-10905	377	12	rehearsal	rehearsal	NOUN
easat-10905	377	13	-	-	PUNCT
easat-10905	377	14	based	base	VERB
easat-10905	377	15	,	,	PUNCT
easat-10905	377	16	pseudo	pseudo	NOUN
easat-10905	377	17	-	-	NOUN
easat-10905	377	18	rehearsal	rehearsal	NOUN
easat-10905	377	19	,	,	PUNCT
easat-10905	377	20	and	and	CCONJ
easat-10905	377	21	lwf	lwf	VERB
easat-10905	377	22	approaches	approach	NOUN
easat-10905	377	23	.	.	PUNCT
easat-10905	378	1	the	the	DET
easat-10905	378	2	results	result	NOUN
easat-10905	378	3	demonstrate	demonstrate	VERB
easat-10905	378	4	that	that	SCONJ
easat-10905	378	5	bert	bert	PROPN
easat-10905	378	6	with	with	ADP
easat-10905	378	7	lwf	lwf	PROPN
easat-10905	378	8	provides	provide	VERB
easat-10905	378	9	better	well	ADJ
easat-10905	378	10	performance	performance	NOUN
easat-10905	378	11	compared	compare	VERB
easat-10905	378	12	to	to	ADP
easat-10905	378	13	the	the	DET
easat-10905	378	14	other	other	ADJ
easat-10905	378	15	models	model	NOUN
easat-10905	378	16	,	,	PUNCT
easat-10905	378	17	including	include	VERB
easat-10905	378	18	albert	albert	PROPN
easat-10905	378	19	,	,	PUNCT
easat-10905	378	20	distilbert	distilbert	PROPN
easat-10905	378	21	,	,	PUNCT
easat-10905	378	22	and	and	CCONJ
easat-10905	378	23	roberta	roberta	PROPN
easat-10905	378	24	.	.	PUNCT
easat-10905	379	1	bert	bert	PROPN
easat-10905	379	2	with	with	ADP
easat-10905	379	3	lwf	lwf	ADJ
easat-10905	379	4	excels	excel	NOUN
easat-10905	379	5	in	in	ADP
easat-10905	379	6	maintaining	maintain	VERB
easat-10905	379	7	high	high	ADJ
easat-10905	379	8	average	average	ADJ
easat-10905	379	9	accuracy	accuracy	NOUN
easat-10905	379	10	and	and	CCONJ
easat-10905	379	11	average	average	ADJ
easat-10905	379	12	incremental	incremental	ADJ
easat-10905	379	13	accuracy	accuracy	NOUN
easat-10905	379	14	,	,	PUNCT
easat-10905	379	15	while	while	SCONJ
easat-10905	379	16	also	also	ADV
easat-10905	379	17	demonstrating	demonstrate	VERB
easat-10905	379	18	a	a	DET
easat-10905	379	19	low	low	ADJ
easat-10905	379	20	forgetting	forgetting	NOUN
easat-10905	379	21	rate	rate	NOUN
easat-10905	379	22	and	and	CCONJ
easat-10905	379	23	high	high	ADJ
easat-10905	379	24	forward	forward	ADJ
easat-10905	379	25	transfer	transfer	NOUN
easat-10905	379	26	.	.	PUNCT
easat-10905	380	1	these	these	DET
easat-10905	380	2	results	result	NOUN
easat-10905	380	3	highlight	highlight	VERB
easat-10905	380	4	the	the	DET
easat-10905	380	5	effectiveness	effectiveness	NOUN
easat-10905	380	6	of	of	ADP
easat-10905	380	7	bert	bert	PROPN
easat-10905	380	8	with	with	ADP
easat-10905	380	9	lwf	lwf	PROPN
easat-10905	380	10	in	in	ADP
easat-10905	380	11	preserving	preserve	VERB
easat-10905	380	12	knowledge	knowledge	NOUN
easat-10905	380	13	of	of	ADP
easat-10905	380	14	previously	previously	ADV
easat-10905	380	15	learned	learn	VERB
easat-10905	380	16	tasks	task	NOUN
easat-10905	380	17	and	and	CCONJ
easat-10905	380	18	adapting	adapt	VERB
easat-10905	380	19	efficiently	efficiently	ADV
easat-10905	380	20	to	to	ADP
easat-10905	380	21	new	new	ADJ
easat-10905	380	22	tasks	task	NOUN
easat-10905	380	23	.	.	PUNCT
easat-10905	381	1	furthermore	furthermore	ADV
easat-10905	381	2	,	,	PUNCT
easat-10905	381	3	it	it	PRON
easat-10905	381	4	is	be	AUX
easat-10905	381	5	understood	understand	VERB
easat-10905	381	6	that	that	SCONJ
easat-10905	381	7	the	the	DET
easat-10905	381	8	performance	performance	NOUN
easat-10905	381	9	of	of	ADP
easat-10905	381	10	sequential	sequential	ADJ
easat-10905	381	11	learning	learning	NOUN
easat-10905	381	12	applications	application	NOUN
easat-10905	381	13	can	can	AUX
easat-10905	381	14	be	be	AUX
easat-10905	381	15	significantly	significantly	ADV
easat-10905	381	16	improved	improve	VERB
easat-10905	381	17	by	by	ADP
easat-10905	381	18	integrating	integrate	VERB
easat-10905	381	19	incremental	incremental	ADJ
easat-10905	381	20	learning	learning	NOUN
easat-10905	381	21	approaches	approach	NOUN
easat-10905	381	22	into	into	ADP
easat-10905	381	23	transformer	transformer	NOUN
easat-10905	381	24	models	model	NOUN
easat-10905	381	25	,	,	PUNCT
easat-10905	381	26	where	where	SCONJ
easat-10905	381	27	a	a	DET
easat-10905	381	28	series	series	NOUN
easat-10905	381	29	of	of	ADP
easat-10905	381	30	related	related	ADJ
easat-10905	381	31	tasks	task	NOUN
easat-10905	381	32	need	need	VERB
easat-10905	381	33	to	to	PART
easat-10905	381	34	be	be	AUX
easat-10905	381	35	learned	learn	VERB
easat-10905	381	36	and	and	CCONJ
easat-10905	381	37	adapted	adapt	VERB
easat-10905	381	38	over	over	ADP
easat-10905	381	39	time	time	NOUN
easat-10905	381	40	.	.	PUNCT
easat-10905	382	1	in	in	ADP
easat-10905	382	2	the	the	DET
easat-10905	382	3	future	future	NOUN
easat-10905	382	4	,	,	PUNCT
easat-10905	382	5	we	we	PRON
easat-10905	382	6	intend	intend	VERB
easat-10905	382	7	to	to	PART
easat-10905	382	8	explore	explore	VERB
easat-10905	382	9	various	various	ADJ
easat-10905	382	10	approaches	approach	NOUN
easat-10905	382	11	to	to	PART
easat-10905	382	12	further	far	ADV
easat-10905	382	13	advance	advance	VERB
easat-10905	382	14	task	task	PROPN
easat-10905	382	15	incremental	incremental	ADJ
easat-10905	382	16	learning	learning	NOUN
easat-10905	382	17	in	in	ADP
easat-10905	382	18	sequential	sequential	ADJ
easat-10905	382	19	applications	application	NOUN
easat-10905	382	20	.	.	PUNCT
easat-10905	383	1	examining	examine	VERB
easat-10905	383	2	the	the	DET
easat-10905	383	3	integration	integration	NOUN
easat-10905	383	4	of	of	ADP
easat-10905	383	5	advanced	advanced	ADJ
easat-10905	383	6	techniques	technique	NOUN
easat-10905	383	7	,	,	PUNCT
easat-10905	383	8	such	such	ADJ
easat-10905	383	9	as	as	ADP
easat-10905	383	10	meta	meta	NOUN
easat-10905	383	11	-	-	PUNCT
easat-10905	383	12	learning	learning	NOUN
easat-10905	383	13	or	or	CCONJ
easat-10905	383	14	continual	continual	ADJ
easat-10905	383	15	learning	learning	NOUN
easat-10905	383	16	frameworks	framework	NOUN
easat-10905	383	17	,	,	PUNCT
easat-10905	383	18	could	could	AUX
easat-10905	383	19	yield	yield	VERB
easat-10905	383	20	further	further	ADJ
easat-10905	383	21	improvements	improvement	NOUN
easat-10905	383	22	in	in	ADP
easat-10905	383	23	managing	manage	VERB
easat-10905	383	24	cf	cf	NOUN
easat-10905	383	25	and	and	CCONJ
easat-10905	383	26	enhancing	enhance	VERB
easat-10905	383	27	forward	forward	ADJ
easat-10905	383	28	transfer	transfer	NOUN
easat-10905	383	29	.	.	PUNCT
easat-10905	384	1	moreover	moreover	ADV
easat-10905	384	2	,	,	PUNCT
easat-10905	384	3	experimenting	experiment	VERB
easat-10905	384	4	with	with	ADP
easat-10905	384	5	a	a	DET
easat-10905	384	6	wider	wide	ADJ
easat-10905	384	7	range	range	NOUN
easat-10905	384	8	of	of	ADP
easat-10905	384	9	tasks	task	NOUN
easat-10905	384	10	beyond	beyond	ADP
easat-10905	384	11	abusive	abusive	ADJ
easat-10905	384	12	and	and	CCONJ
easat-10905	384	13	offensive	offensive	ADJ
easat-10905	384	14	text	text	NOUN
easat-10905	384	15	classification	classification	NOUN
easat-10905	384	16	may	may	AUX
easat-10905	384	17	offer	offer	VERB
easat-10905	384	18	insights	insight	NOUN
easat-10905	384	19	into	into	ADP
easat-10905	384	20	the	the	DET
easat-10905	384	21	general	general	ADJ
easat-10905	384	22	applicability	applicability	NOUN
easat-10905	384	23	of	of	ADP
easat-10905	384	24	our	our	PRON
easat-10905	384	25	findings	finding	NOUN
easat-10905	384	26	.	.	PUNCT
easat-10905	385	1	lastly	lastly	ADV
easat-10905	385	2	,	,	PUNCT
easat-10905	385	3	investigating	investigate	VERB
easat-10905	385	4	the	the	DET
easat-10905	385	5	effects	effect	NOUN
easat-10905	385	6	of	of	ADP
easat-10905	385	7	different	different	ADJ
easat-10905	385	8	hyperparameters	hyperparameter	NOUN
easat-10905	385	9	and	and	CCONJ
easat-10905	385	10	fine	fine	ADV
easat-10905	385	11	-	-	PUNCT
easat-10905	385	12	tuning	tune	VERB
easat-10905	385	13	strategies	strategy	NOUN
easat-10905	385	14	on	on	ADP
easat-10905	385	15	the	the	DET
easat-10905	385	16	performance	performance	NOUN
easat-10905	385	17	of	of	ADP
easat-10905	385	18	task	task	NOUN
easat-10905	385	19	incremental	incremental	ADJ
easat-10905	385	20	learning	learning	NOUN
easat-10905	385	21	models	model	NOUN
easat-10905	385	22	could	could	AUX
easat-10905	385	23	provide	provide	VERB
easat-10905	385	24	valuable	valuable	ADJ
easat-10905	385	25	guidance	guidance	NOUN
easat-10905	385	26	for	for	ADP
easat-10905	385	27	optimizing	optimize	VERB
easat-10905	385	28	their	their	PRON
easat-10905	385	29	effectiveness	effectiveness	NOUN
easat-10905	385	30	in	in	ADP
easat-10905	385	31	realworld	realworld	PROPN
easat-10905	385	32	scenarios	scenario	NOUN
easat-10905	385	33	.	.	PUNCT
easat-10905	386	1	0	0	NUM
easat-10905	387	1	20	20	NUM
easat-10905	387	2	40	40	NUM
easat-10905	387	3	60	60	NUM
easat-10905	387	4	80	80	NUM
easat-10905	387	5	100	100	NUM
easat-10905	387	6	r	r	NOUN
easat-10905	387	7	e	e	NOUN
easat-10905	387	8	h	h	NOUN
easat-10905	387	9	ea	ea	NOUN
easat-10905	387	10	rs	rs	X
easat-10905	387	11	a	a	DET
easat-10905	387	12	l	l	NOUN
easat-10905	387	13	p	p	X
easat-10905	387	14	se	se	X
easat-10905	387	15	u	u	PROPN
easat-10905	388	1	d	d	X
easat-10905	388	2	o	o	NOUN
easat-10905	388	3	r	r	NOUN
easat-10905	389	1	e	e	NOUN
easat-10905	389	2	h	h	NOUN
easat-10905	389	3	ea	ea	INTJ
easat-10905	389	4	rs	rs	X
easat-10905	389	5	a	a	DET
easat-10905	389	6	l	l	NOUN
easat-10905	389	7	l	l	NOUN
easat-10905	390	1	w	w	NOUN
easat-10905	390	2	f	f	NOUN
easat-10905	390	3	r	r	NOUN
easat-10905	390	4	e	e	NOUN
easat-10905	390	5	h	h	NOUN
easat-10905	390	6	ea	ea	INTJ
easat-10905	390	7	rs	rs	X
easat-10905	390	8	a	a	DET
easat-10905	390	9	l	l	NOUN
easat-10905	390	10	p	p	X
easat-10905	390	11	se	se	X
easat-10905	390	12	u	u	PROPN
easat-10905	391	1	d	d	X
easat-10905	391	2	o	o	NOUN
easat-10905	391	3	r	r	NOUN
easat-10905	392	1	e	e	NOUN
easat-10905	392	2	h	h	NOUN
easat-10905	392	3	ea	ea	INTJ
easat-10905	392	4	rs	rs	X
easat-10905	392	5	a	a	DET
easat-10905	392	6	l	l	NOUN
easat-10905	392	7	l	l	NOUN
easat-10905	393	1	w	w	NOUN
easat-10905	393	2	f	f	NOUN
easat-10905	393	3	r	r	NOUN
easat-10905	393	4	e	e	NOUN
easat-10905	393	5	h	h	NOUN
easat-10905	393	6	ea	ea	INTJ
easat-10905	393	7	rs	rs	X
easat-10905	393	8	a	a	DET
easat-10905	393	9	l	l	NOUN
easat-10905	393	10	p	p	X
easat-10905	393	11	se	se	X
easat-10905	393	12	u	u	PROPN
easat-10905	394	1	d	d	X
easat-10905	394	2	o	o	NOUN
easat-10905	394	3	r	r	NOUN
easat-10905	395	1	e	e	NOUN
easat-10905	395	2	h	h	NOUN
easat-10905	395	3	ea	ea	INTJ
easat-10905	395	4	rs	rs	X
easat-10905	395	5	a	a	DET
easat-10905	395	6	l	l	NOUN
easat-10905	395	7	l	l	NOUN
easat-10905	396	1	w	w	NOUN
easat-10905	396	2	f	f	NOUN
easat-10905	396	3	r	r	NOUN
easat-10905	396	4	e	e	NOUN
easat-10905	396	5	h	h	NOUN
easat-10905	396	6	ea	ea	INTJ
easat-10905	396	7	rs	rs	X
easat-10905	396	8	a	a	DET
easat-10905	396	9	l	l	NOUN
easat-10905	396	10	p	p	X
easat-10905	396	11	se	se	X
easat-10905	396	12	u	u	PROPN
easat-10905	397	1	d	d	X
easat-10905	397	2	o	o	NOUN
easat-10905	397	3	r	r	NOUN
easat-10905	398	1	e	e	NOUN
easat-10905	398	2	h	h	NOUN
easat-10905	398	3	ea	ea	INTJ
easat-10905	398	4	rs	rs	X
easat-10905	398	5	a	a	DET
easat-10905	398	6	l	l	NOUN
easat-10905	398	7	l	l	NOUN
easat-10905	399	1	w	w	NOUN
easat-10905	399	2	f	f	NOUN
easat-10905	399	3	r	r	NOUN
easat-10905	399	4	e	e	NOUN
easat-10905	399	5	h	h	NOUN
easat-10905	399	6	ea	ea	INTJ
easat-10905	399	7	rs	rs	X
easat-10905	399	8	a	a	DET
easat-10905	399	9	l	l	NOUN
easat-10905	399	10	p	p	X
easat-10905	399	11	se	se	X
easat-10905	399	12	u	u	PROPN
easat-10905	400	1	d	d	X
easat-10905	400	2	o	o	NOUN
easat-10905	400	3	r	r	NOUN
easat-10905	401	1	e	e	NOUN
easat-10905	401	2	h	h	NOUN
easat-10905	401	3	ea	ea	INTJ
easat-10905	401	4	rs	rs	X
easat-10905	401	5	a	a	DET
easat-10905	401	6	l	l	NOUN
easat-10905	401	7	l	l	NOUN
easat-10905	401	8	w	w	NOUN
easat-10905	401	9	f	f	PROPN
easat-10905	401	10	average	average	ADJ
easat-10905	401	11	accuracy	accuracy	NOUN
easat-10905	401	12	average	average	ADJ
easat-10905	401	13	incremental	incremental	ADJ
easat-10905	401	14	accuracy	accuracy	NOUN
easat-10905	401	15	forgetting	forget	VERB
easat-10905	401	16	rate	rate	NOUN
easat-10905	401	17	forward	forward	ADJ
easat-10905	401	18	transfer	transfer	VERB
easat-10905	401	19	backward	backward	ADJ
easat-10905	401	20	transfer	transfer	NOUN
easat-10905	401	21	performance	performance	NOUN
easat-10905	401	22	on	on	ADP
easat-10905	401	23	incremental	incremental	ADJ
easat-10905	401	24	learning	learn	VERB
easat-10905	401	25	bert	bert	PROPN
easat-10905	401	26	albert	albert	PROPN
easat-10905	401	27	roberta	roberta	PROPN
easat-10905	401	28	distilbert	distilbert	VERB
easat-10905	401	29	465	465	NUM
easat-10905	401	30	edelweiss	edelweiss	PROPN
easat-10905	401	31	applied	apply	VERB
easat-10905	401	32	science	science	NOUN
easat-10905	401	33	and	and	CCONJ
easat-10905	401	34	technology	technology	NOUN
easat-10905	401	35	issn	issn	PROPN
easat-10905	401	36	:	:	PUNCT
easat-10905	401	37	2576	2576	NUM
easat-10905	401	38	-	-	SYM
easat-10905	401	39	8484	8484	NUM
easat-10905	401	40	vol	vol	NOUN
easat-10905	401	41	.	.	PROPN
easat-10905	402	1	9	9	NUM
easat-10905	402	2	,	,	PUNCT
easat-10905	402	3	no	no	INTJ
easat-10905	402	4	.	.	NOUN
easat-10905	402	5	11	11	NUM
easat-10905	402	6	:	:	PUNCT
easat-10905	402	7	448	448	NUM
easat-10905	402	8	-	-	SYM
easat-10905	402	9	466	466	NUM
easat-10905	402	10	,	,	PUNCT
easat-10905	402	11	2025	2025	NUM
easat-10905	402	12	doi	doi	NOUN
easat-10905	402	13	:	:	PUNCT
easat-10905	402	14	10.55214/2576	10.55214/2576	NUM
easat-10905	402	15	-	-	PUNCT
easat-10905	402	16	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	403	1	©	©	PROPN
easat-10905	403	2	2025	2025	NUM
easat-10905	403	3	by	by	ADP
easat-10905	403	4	the	the	DET
easat-10905	403	5	authors	author	NOUN
easat-10905	403	6	;	;	PUNCT
easat-10905	403	7	licensee	licensee	PROPN
easat-10905	403	8	learning	learn	VERB
easat-10905	403	9	gate	gate	NOUN
easat-10905	403	10	transparency	transparency	NOUN
easat-10905	403	11	:	:	PUNCT
easat-10905	403	12	the	the	DET
easat-10905	403	13	authors	author	NOUN
easat-10905	403	14	confirm	confirm	VERB
easat-10905	403	15	that	that	SCONJ
easat-10905	403	16	the	the	DET
easat-10905	403	17	manuscript	manuscript	NOUN
easat-10905	403	18	is	be	AUX
easat-10905	403	19	an	an	DET
easat-10905	403	20	honest	honest	ADJ
easat-10905	403	21	,	,	PUNCT
easat-10905	403	22	accurate	accurate	ADJ
easat-10905	403	23	,	,	PUNCT
easat-10905	403	24	and	and	CCONJ
easat-10905	403	25	transparent	transparent	ADJ
easat-10905	403	26	account	account	NOUN
easat-10905	403	27	of	of	ADP
easat-10905	403	28	the	the	DET
easat-10905	403	29	study	study	NOUN
easat-10905	403	30	;	;	PUNCT
easat-10905	403	31	that	that	SCONJ
easat-10905	403	32	no	no	DET
easat-10905	403	33	vital	vital	ADJ
easat-10905	403	34	features	feature	NOUN
easat-10905	403	35	of	of	ADP
easat-10905	403	36	the	the	DET
easat-10905	403	37	study	study	NOUN
easat-10905	403	38	have	have	AUX
easat-10905	403	39	been	be	AUX
easat-10905	403	40	omitted	omit	VERB
easat-10905	403	41	;	;	PUNCT
easat-10905	403	42	and	and	CCONJ
easat-10905	403	43	that	that	SCONJ
easat-10905	403	44	any	any	DET
easat-10905	403	45	discrepancies	discrepancy	NOUN
easat-10905	403	46	from	from	ADP
easat-10905	403	47	the	the	DET
easat-10905	403	48	study	study	NOUN
easat-10905	403	49	as	as	SCONJ
easat-10905	403	50	planned	plan	VERB
easat-10905	403	51	have	have	AUX
easat-10905	403	52	been	be	AUX
easat-10905	403	53	explained	explain	VERB
easat-10905	403	54	.	.	PUNCT
easat-10905	404	1	this	this	DET
easat-10905	404	2	study	study	NOUN
easat-10905	404	3	followed	follow	VERB
easat-10905	404	4	all	all	DET
easat-10905	404	5	ethical	ethical	ADJ
easat-10905	404	6	practices	practice	NOUN
easat-10905	404	7	during	during	ADP
easat-10905	404	8	writing	writing	NOUN
easat-10905	404	9	.	.	PUNCT
easat-10905	405	1	copyright	copyright	NOUN
easat-10905	405	2	:	:	PUNCT
easat-10905	405	3	©	©	PROPN
easat-10905	405	4	2025	2025	NUM
easat-10905	405	5	by	by	ADP
easat-10905	405	6	the	the	DET
easat-10905	405	7	authors	author	NOUN
easat-10905	405	8	.	.	PUNCT
easat-10905	406	1	this	this	DET
easat-10905	406	2	article	article	NOUN
easat-10905	406	3	is	be	AUX
easat-10905	406	4	an	an	DET
easat-10905	406	5	open	open	ADJ
easat-10905	406	6	-	-	PUNCT
easat-10905	406	7	access	access	NOUN
easat-10905	406	8	article	article	NOUN
easat-10905	406	9	distributed	distribute	VERB
easat-10905	406	10	under	under	ADP
easat-10905	406	11	the	the	DET
easat-10905	406	12	terms	term	NOUN
easat-10905	406	13	and	and	CCONJ
easat-10905	406	14	conditions	condition	NOUN
easat-10905	406	15	of	of	ADP
easat-10905	406	16	the	the	DET
easat-10905	406	17	creative	creative	ADJ
easat-10905	406	18	commons	common	NOUN
easat-10905	406	19	attribution	attribution	NOUN
easat-10905	406	20	(	(	PUNCT
easat-10905	406	21	cc	cc	NOUN
easat-10905	406	22	by	by	ADP
easat-10905	406	23	)	)	PUNCT
easat-10905	406	24	license	license	NOUN
easat-10905	406	25	(	(	PUNCT
easat-10905	406	26	https://creativecommons.org/licenses/by/4.0/	https://creativecommons.org/licenses/by/4.0/	PROPN
easat-10905	406	27	)	)	PUNCT
easat-10905	406	28	.	.	PUNCT
easat-10905	407	1	references	reference	NOUN
easat-10905	407	2	[	[	X
easat-10905	407	3	1	1	X
easat-10905	407	4	]	]	PUNCT
easat-10905	407	5	s.	s.	PROPN
easat-10905	407	6	a.	a.	PROPN
easat-10905	407	7	castaño	castaño	PROPN
easat-10905	407	8	-	-	PUNCT
easat-10905	407	9	pulgarín	pulgarín	PROPN
easat-10905	407	10	,	,	PUNCT
easat-10905	407	11	n.	n.	PROPN
easat-10905	407	12	suárez	suárez	NOUN
easat-10905	407	13	-	-	PUNCT
easat-10905	407	14	betancur	betancur	NOUN
easat-10905	407	15	,	,	PUNCT
easat-10905	407	16	l.	l.	PROPN
easat-10905	407	17	m.	m.	PROPN
easat-10905	407	18	t.	t.	PROPN
easat-10905	407	19	vega	vega	PROPN
easat-10905	407	20	,	,	PUNCT
easat-10905	407	21	and	and	CCONJ
easat-10905	407	22	h.	h.	PROPN
easat-10905	407	23	m.	m.	PROPN
easat-10905	407	24	h.	h.	PROPN
easat-10905	407	25	lópez	lópez	PROPN
easat-10905	407	26	,	,	PUNCT
easat-10905	407	27	"	"	PUNCT
easat-10905	407	28	internet	internet	NOUN
easat-10905	407	29	,	,	PUNCT
easat-10905	407	30	social	social	ADJ
easat-10905	407	31	media	medium	NOUN
easat-10905	407	32	and	and	CCONJ
easat-10905	407	33	online	online	ADJ
easat-10905	407	34	hate	hate	NOUN
easat-10905	407	35	speech	speech	NOUN
easat-10905	407	36	.	.	PUNCT
easat-10905	408	1	systematic	systematic	ADJ
easat-10905	408	2	review	review	NOUN
easat-10905	408	3	,	,	PUNCT
easat-10905	408	4	"	"	PUNCT
easat-10905	408	5	aggression	aggression	NOUN
easat-10905	408	6	and	and	CCONJ
easat-10905	408	7	violent	violent	ADJ
easat-10905	408	8	behavior	behavior	NOUN
easat-10905	408	9	,	,	PUNCT
easat-10905	408	10	vol	vol	NOUN
easat-10905	408	11	.	.	PROPN
easat-10905	408	12	58	58	NUM
easat-10905	408	13	,	,	PUNCT
easat-10905	408	14	p.	p.	NOUN
easat-10905	408	15	101608	101608	NUM
easat-10905	408	16	,	,	PUNCT
easat-10905	408	17	2021	2021	NUM
easat-10905	408	18	.	.	PUNCT
easat-10905	409	1	https://doi.org/10.1016/j.avb.2021.101608	https://doi.org/10.1016/j.avb.2021.101608	PROPN
easat-10905	410	1	[	[	X
easat-10905	410	2	2	2	X
easat-10905	410	3	]	]	PUNCT
easat-10905	410	4	m.	m.	NOUN
easat-10905	410	5	p.	p.	NOUN
easat-10905	410	6	akhter	akhter	PROPN
easat-10905	410	7	,	,	PUNCT
easat-10905	410	8	z.	z.	PROPN
easat-10905	410	9	jiangbin	jiangbin	PROPN
easat-10905	410	10	,	,	PUNCT
easat-10905	410	11	i.	i.	PROPN
easat-10905	410	12	r.	r.	PROPN
easat-10905	410	13	naqvi	naqvi	PROPN
easat-10905	410	14	,	,	PUNCT
easat-10905	410	15	m.	m.	NOUN
easat-10905	410	16	abdelmajeed	abdelmajeed	NOUN
easat-10905	410	17	,	,	PUNCT
easat-10905	410	18	and	and	CCONJ
easat-10905	410	19	t.	t.	PROPN
easat-10905	410	20	zia	zia	PROPN
easat-10905	410	21	,	,	PUNCT
easat-10905	410	22	"	"	PUNCT
easat-10905	410	23	abusive	abusive	ADJ
easat-10905	410	24	language	language	NOUN
easat-10905	410	25	detection	detection	NOUN
easat-10905	410	26	from	from	ADP
easat-10905	410	27	social	social	ADJ
easat-10905	410	28	media	medium	NOUN
easat-10905	410	29	comments	comment	NOUN
easat-10905	410	30	using	use	VERB
easat-10905	410	31	conventional	conventional	ADJ
easat-10905	410	32	machine	machine	NOUN
easat-10905	410	33	learning	learning	NOUN
easat-10905	410	34	and	and	CCONJ
easat-10905	410	35	deep	deep	ADJ
easat-10905	410	36	learning	learning	NOUN
easat-10905	410	37	approaches	approach	NOUN
easat-10905	410	38	,	,	PUNCT
easat-10905	410	39	"	"	PUNCT
easat-10905	410	40	multimedia	multimedia	NOUN
easat-10905	410	41	systems	system	NOUN
easat-10905	410	42	,	,	PUNCT
easat-10905	410	43	vol	vol	NOUN
easat-10905	410	44	.	.	PROPN
easat-10905	410	45	28	28	NUM
easat-10905	410	46	,	,	PUNCT
easat-10905	410	47	pp	pp	ADJ
easat-10905	410	48	.	.	PUNCT
easat-10905	410	49	19251940	19251940	NUM
easat-10905	410	50	,	,	PUNCT
easat-10905	410	51	2022	2022	NUM
easat-10905	410	52	.	.	PUNCT
easat-10905	411	1	https://doi.org/10.1007/s00530-021-00784-8	https://doi.org/10.1007/s00530-021-00784-8	NOUN
easat-10905	412	1	[	[	X
easat-10905	412	2	3	3	NUM
easat-10905	412	3	]	]	PUNCT
easat-10905	412	4	a.	a.	NOUN
easat-10905	412	5	ritter	ritter	PROPN
easat-10905	412	6	,	,	PUNCT
easat-10905	412	7	c.	c.	PROPN
easat-10905	412	8	cherry	cherry	PROPN
easat-10905	412	9	,	,	PUNCT
easat-10905	412	10	and	and	CCONJ
easat-10905	412	11	w.	w.	PROPN
easat-10905	412	12	b.	b.	PROPN
easat-10905	412	13	dolan	dolan	PROPN
easat-10905	412	14	,	,	PUNCT
easat-10905	412	15	"	"	PUNCT
easat-10905	412	16	data	data	NOUN
easat-10905	412	17	-	-	PUNCT
easat-10905	412	18	driven	drive	VERB
easat-10905	412	19	response	response	NOUN
easat-10905	412	20	generation	generation	NOUN
easat-10905	412	21	in	in	ADP
easat-10905	412	22	social	social	ADJ
easat-10905	412	23	media	medium	NOUN
easat-10905	412	24	,	,	PUNCT
easat-10905	412	25	"	"	PUNCT
easat-10905	412	26	in	in	ADP
easat-10905	412	27	proceedings	proceeding	NOUN
easat-10905	412	28	of	of	ADP
easat-10905	412	29	the	the	DET
easat-10905	412	30	2011	2011	NUM
easat-10905	412	31	conference	conference	NOUN
easat-10905	412	32	on	on	ADP
easat-10905	412	33	empirical	empirical	ADJ
easat-10905	412	34	methods	method	NOUN
easat-10905	412	35	in	in	ADP
easat-10905	412	36	natural	natural	ADJ
easat-10905	412	37	language	language	NOUN
easat-10905	412	38	processing	processing	NOUN
easat-10905	412	39	(	(	PUNCT
easat-10905	412	40	emnlp	emnlp	ADJ
easat-10905	412	41	)	)	PUNCT
easat-10905	412	42	(	(	PUNCT
easat-10905	412	43	pp	pp	ADP
easat-10905	412	44	.	.	PUNCT
easat-10905	413	1	583–593	583–593	NUM
easat-10905	413	2	)	)	PUNCT
easat-10905	413	3	.	.	PUNCT
easat-10905	414	1	stroudsburg	stroudsburg	PROPN
easat-10905	414	2	,	,	PUNCT
easat-10905	414	3	pa	pa	PROPN
easat-10905	414	4	:	:	PUNCT
easat-10905	414	5	association	association	NOUN
easat-10905	414	6	for	for	ADP
easat-10905	414	7	computational	computational	ADJ
easat-10905	414	8	linguistics	linguistic	NOUN
easat-10905	414	9	,	,	PUNCT
easat-10905	414	10	2011	2011	NUM
easat-10905	414	11	.	.	PUNCT
easat-10905	415	1	[	[	X
easat-10905	415	2	4	4	X
easat-10905	415	3	]	]	PUNCT
easat-10905	415	4	b.	b.	PROPN
easat-10905	415	5	r.	r.	PROPN
easat-10905	415	6	chakravarthi	chakravarthi	PROPN
easat-10905	415	7	et	et	PROPN
easat-10905	415	8	al	al	PROPN
easat-10905	415	9	.	.	PROPN
easat-10905	415	10	,	,	PUNCT
easat-10905	415	11	"	"	PUNCT
easat-10905	415	12	detecting	detect	VERB
easat-10905	415	13	abusive	abusive	ADJ
easat-10905	415	14	comments	comment	NOUN
easat-10905	415	15	at	at	ADP
easat-10905	415	16	a	a	DET
easat-10905	415	17	fine	fine	ADV
easat-10905	415	18	-	-	PUNCT
easat-10905	415	19	grained	grain	VERB
easat-10905	415	20	level	level	NOUN
easat-10905	415	21	in	in	ADP
easat-10905	415	22	a	a	DET
easat-10905	415	23	low	low	ADJ
easat-10905	415	24	-	-	PUNCT
easat-10905	415	25	resource	resource	NOUN
easat-10905	415	26	language	language	NOUN
easat-10905	415	27	,	,	PUNCT
easat-10905	415	28	"	"	PUNCT
easat-10905	415	29	natural	natural	ADJ
easat-10905	415	30	language	language	NOUN
easat-10905	415	31	processing	processing	NOUN
easat-10905	415	32	journal	journal	NOUN
easat-10905	415	33	,	,	PUNCT
easat-10905	415	34	vol	vol	NOUN
easat-10905	415	35	.	.	PROPN
easat-10905	415	36	3	3	NUM
easat-10905	415	37	,	,	PUNCT
easat-10905	415	38	p.	p.	NOUN
easat-10905	415	39	100006	100006	NUM
easat-10905	415	40	,	,	PUNCT
easat-10905	415	41	2023	2023	NUM
easat-10905	415	42	.	.	PUNCT
easat-10905	415	43	https://doi.org/10.1016/j.nlp.2023.100006	https://doi.org/10.1016/j.nlp.2023.100006	X
easat-10905	416	1	[	[	X
easat-10905	416	2	5	5	X
easat-10905	416	3	]	]	PUNCT
easat-10905	416	4	f.	f.	PROPN
easat-10905	416	5	poletto	poletto	PROPN
easat-10905	416	6	,	,	PUNCT
easat-10905	416	7	v.	v.	CCONJ
easat-10905	416	8	basile	basile	PROPN
easat-10905	416	9	,	,	PUNCT
easat-10905	416	10	m.	m.	NOUN
easat-10905	416	11	sanguinetti	sanguinetti	PROPN
easat-10905	416	12	,	,	PUNCT
easat-10905	416	13	c.	c.	PROPN
easat-10905	416	14	bosco	bosco	PROPN
easat-10905	416	15	,	,	PUNCT
easat-10905	416	16	and	and	CCONJ
easat-10905	417	1	v.	v.	ADP
easat-10905	417	2	patti	patti	PROPN
easat-10905	417	3	,	,	PUNCT
easat-10905	417	4	"	"	PUNCT
easat-10905	417	5	resources	resource	NOUN
easat-10905	417	6	and	and	CCONJ
easat-10905	417	7	benchmark	benchmark	NOUN
easat-10905	417	8	corpora	corpora	NOUN
easat-10905	417	9	for	for	ADP
easat-10905	417	10	hate	hate	NOUN
easat-10905	417	11	speech	speech	NOUN
easat-10905	417	12	detection	detection	NOUN
easat-10905	417	13	:	:	PUNCT
easat-10905	417	14	a	a	DET
easat-10905	417	15	systematic	systematic	ADJ
easat-10905	417	16	review	review	NOUN
easat-10905	417	17	,	,	PUNCT
easat-10905	417	18	"	"	PUNCT
easat-10905	417	19	language	language	NOUN
easat-10905	417	20	resources	resource	NOUN
easat-10905	417	21	and	and	CCONJ
easat-10905	417	22	evaluation	evaluation	NOUN
easat-10905	417	23	,	,	PUNCT
easat-10905	417	24	vol	vol	NOUN
easat-10905	417	25	.	.	PROPN
easat-10905	418	1	55	55	NUM
easat-10905	418	2	,	,	PUNCT
easat-10905	418	3	pp	pp	ADJ
easat-10905	418	4	.	.	PUNCT
easat-10905	419	1	477	477	NUM
easat-10905	419	2	-	-	SYM
easat-10905	419	3	523	523	NUM
easat-10905	419	4	,	,	PUNCT
easat-10905	419	5	2021	2021	NUM
easat-10905	419	6	.	.	PUNCT
easat-10905	420	1	https://doi.org/10.1007/s10579-020-09502-8	https://doi.org/10.1007/s10579-020-09502-8	PRON
easat-10905	421	1	[	[	X
easat-10905	421	2	6	6	NUM
easat-10905	421	3	]	]	X
easat-10905	421	4	n.	n.	PROPN
easat-10905	421	5	d.	d.	PROPN
easat-10905	421	6	gitari	gitari	PROPN
easat-10905	421	7	,	,	PUNCT
easat-10905	421	8	z.	z.	PROPN
easat-10905	421	9	zuping	zuping	PROPN
easat-10905	421	10	,	,	PUNCT
easat-10905	421	11	h.	h.	PROPN
easat-10905	421	12	damien	damien	PROPN
easat-10905	421	13	,	,	PUNCT
easat-10905	421	14	and	and	CCONJ
easat-10905	421	15	j.	j.	PROPN
easat-10905	421	16	long	long	PROPN
easat-10905	421	17	,	,	PUNCT
easat-10905	421	18	"	"	PUNCT
easat-10905	421	19	a	a	DET
easat-10905	421	20	lexicon	lexicon	NOUN
easat-10905	421	21	-	-	PUNCT
easat-10905	421	22	based	base	VERB
easat-10905	421	23	approach	approach	NOUN
easat-10905	421	24	for	for	ADP
easat-10905	421	25	hate	hate	NOUN
easat-10905	421	26	speech	speech	NOUN
easat-10905	421	27	detection	detection	NOUN
easat-10905	421	28	,	,	PUNCT
easat-10905	421	29	"	"	PUNCT
easat-10905	421	30	international	international	ADJ
easat-10905	421	31	journal	journal	NOUN
easat-10905	421	32	of	of	ADP
easat-10905	421	33	multimedia	multimedia	NOUN
easat-10905	421	34	and	and	CCONJ
easat-10905	421	35	ubiquitous	ubiquitous	ADJ
easat-10905	421	36	engineering	engineering	NOUN
easat-10905	421	37	,	,	PUNCT
easat-10905	421	38	vol	vol	NOUN
easat-10905	421	39	.	.	PROPN
easat-10905	421	40	10	10	NUM
easat-10905	421	41	,	,	PUNCT
easat-10905	421	42	no	no	INTJ
easat-10905	421	43	.	.	NOUN
easat-10905	421	44	4	4	NUM
easat-10905	421	45	,	,	PUNCT
easat-10905	421	46	pp	pp	ADJ
easat-10905	421	47	.	.	PUNCT
easat-10905	422	1	215	215	NUM
easat-10905	422	2	-	-	SYM
easat-10905	422	3	230	230	NUM
easat-10905	422	4	,	,	PUNCT
easat-10905	422	5	2015	2015	NUM
easat-10905	422	6	.	.	PUNCT
easat-10905	423	1	[	[	X
easat-10905	423	2	7	7	X
easat-10905	423	3	]	]	X
easat-10905	423	4	g.	g.	PROPN
easat-10905	423	5	i.	i.	PROPN
easat-10905	423	6	sigurbergsson	sigurbergsson	PROPN
easat-10905	423	7	and	and	CCONJ
easat-10905	423	8	l.	l.	PROPN
easat-10905	423	9	derczynski	derczynski	PROPN
easat-10905	423	10	,	,	PUNCT
easat-10905	423	11	"	"	PUNCT
easat-10905	423	12	offensive	offensive	ADJ
easat-10905	423	13	language	language	NOUN
easat-10905	423	14	and	and	CCONJ
easat-10905	423	15	hate	hate	VERB
easat-10905	423	16	speech	speech	NOUN
easat-10905	423	17	detection	detection	NOUN
easat-10905	423	18	for	for	ADP
easat-10905	423	19	danish	danish	ADJ
easat-10905	423	20	,	,	PUNCT
easat-10905	423	21	"	"	PUNCT
easat-10905	423	22	arxiv	arxiv	PROPN
easat-10905	423	23	preprint	preprint	NOUN
easat-10905	423	24	arxiv:1908.04531	arxiv:1908.04531	NOUN
easat-10905	423	25	,	,	PUNCT
easat-10905	423	26	2019	2019	NUM
easat-10905	423	27	.	.	PUNCT
easat-10905	424	1	https://doi.org/10.48550/arxiv.1908.04531	https://doi.org/10.48550/arxiv.1908.04531	PROPN
easat-10905	425	1	[	[	X
easat-10905	425	2	8	8	NUM
easat-10905	425	3	]	]	PUNCT
easat-10905	425	4	j.	j.	PROPN
easat-10905	425	5	h.	h.	PROPN
easat-10905	425	6	park	park	PROPN
easat-10905	425	7	and	and	CCONJ
easat-10905	425	8	p.	p.	PROPN
easat-10905	425	9	fung	fung	PROPN
easat-10905	425	10	,	,	PUNCT
easat-10905	425	11	"	"	PUNCT
easat-10905	425	12	one	one	NUM
easat-10905	425	13	-	-	PUNCT
easat-10905	425	14	step	step	NOUN
easat-10905	425	15	and	and	CCONJ
easat-10905	425	16	two	two	NUM
easat-10905	425	17	-	-	PUNCT
easat-10905	425	18	step	step	NOUN
easat-10905	425	19	classification	classification	NOUN
easat-10905	425	20	for	for	ADP
easat-10905	425	21	abusive	abusive	ADJ
easat-10905	425	22	language	language	NOUN
easat-10905	425	23	detection	detection	NOUN
easat-10905	425	24	on	on	ADP
easat-10905	425	25	twitter	twitter	NOUN
easat-10905	425	26	,	,	PUNCT
easat-10905	425	27	"	"	PUNCT
easat-10905	425	28	arxiv	arxiv	PROPN
easat-10905	425	29	preprint	preprint	VERB
easat-10905	425	30	arxiv:1706.01206	arxiv:1706.01206	NOUN
easat-10905	425	31	,	,	PUNCT
easat-10905	425	32	2017	2017	NUM
easat-10905	425	33	.	.	PUNCT
easat-10905	426	1	https://doi.org/10.48550/arxiv.1706.01206	https://doi.org/10.48550/arxiv.1706.01206	PROPN
easat-10905	427	1	[	[	X
easat-10905	427	2	9	9	NUM
easat-10905	427	3	]	]	PUNCT
easat-10905	427	4	a.	a.	NOUN
easat-10905	427	5	khan	khan	PROPN
easat-10905	427	6	,	,	PUNCT
easat-10905	427	7	a.	a.	PROPN
easat-10905	427	8	ahmed	ahmed	PROPN
easat-10905	427	9	,	,	PUNCT
easat-10905	427	10	s.	s.	PROPN
easat-10905	427	11	jan	jan	PROPN
easat-10905	427	12	,	,	PUNCT
easat-10905	427	13	m.	m.	NOUN
easat-10905	427	14	bilal	bilal	PROPN
easat-10905	427	15	,	,	PUNCT
easat-10905	427	16	and	and	CCONJ
easat-10905	427	17	m.	m.	PROPN
easat-10905	427	18	f.	f.	PROPN
easat-10905	427	19	zuhairi	zuhairi	PROPN
easat-10905	427	20	,	,	PUNCT
easat-10905	427	21	"	"	PUNCT
easat-10905	427	22	abusive	abusive	ADJ
easat-10905	427	23	language	language	NOUN
easat-10905	427	24	detection	detection	NOUN
easat-10905	427	25	in	in	ADP
easat-10905	427	26	urdu	urdu	PROPN
easat-10905	427	27	text	text	NOUN
easat-10905	427	28	:	:	PUNCT
easat-10905	427	29	leveraging	leverage	VERB
easat-10905	427	30	deep	deep	ADJ
easat-10905	427	31	learning	learning	NOUN
easat-10905	427	32	and	and	CCONJ
easat-10905	427	33	attention	attention	NOUN
easat-10905	427	34	mechanism	mechanism	NOUN
easat-10905	427	35	,	,	PUNCT
easat-10905	427	36	"	"	PUNCT
easat-10905	427	37	ieee	ieee	NOUN
easat-10905	427	38	access	access	NOUN
easat-10905	427	39	,	,	PUNCT
easat-10905	427	40	vol	vol	NOUN
easat-10905	427	41	.	.	PROPN
easat-10905	427	42	12	12	NUM
easat-10905	427	43	,	,	PUNCT
easat-10905	427	44	pp	pp	ADJ
easat-10905	427	45	.	.	PUNCT
easat-10905	427	46	37418	37418	NUM
easat-10905	427	47	-	-	SYM
easat-10905	427	48	37431	37431	NUM
easat-10905	427	49	,	,	PUNCT
easat-10905	427	50	2024	2024	NUM
easat-10905	427	51	.	.	PUNCT
easat-10905	428	1	https://doi.org/10.1109/access.2024.3370232	https://doi.org/10.1109/access.2024.3370232	PROPN
easat-10905	429	1	[	[	X
easat-10905	429	2	10	10	NUM
easat-10905	429	3	]	]	X
easat-10905	429	4	r.	r.	PROPN
easat-10905	429	5	pelle	pelle	PROPN
easat-10905	429	6	,	,	PUNCT
easat-10905	429	7	c.	c.	PROPN
easat-10905	429	8	alcântara	alcântara	PROPN
easat-10905	429	9	,	,	PUNCT
easat-10905	429	10	and	and	CCONJ
easat-10905	429	11	v.	v.	ADP
easat-10905	429	12	p.	p.	PROPN
easat-10905	429	13	moreira	moreira	PROPN
easat-10905	429	14	,	,	PUNCT
easat-10905	429	15	"	"	PUNCT
easat-10905	429	16	a	a	DET
easat-10905	429	17	classifier	classifier	NOUN
easat-10905	429	18	ensemble	ensemble	ADJ
easat-10905	429	19	for	for	ADP
easat-10905	429	20	offensive	offensive	ADJ
easat-10905	429	21	text	text	NOUN
easat-10905	429	22	detection	detection	NOUN
easat-10905	429	23	,	,	PUNCT
easat-10905	429	24	"	"	PUNCT
easat-10905	429	25	in	in	ADP
easat-10905	429	26	proceedings	proceeding	NOUN
easat-10905	429	27	of	of	ADP
easat-10905	429	28	the	the	DET
easat-10905	429	29	24th	24th	ADJ
easat-10905	429	30	brazilian	brazilian	ADJ
easat-10905	429	31	symposium	symposium	NOUN
easat-10905	429	32	on	on	ADP
easat-10905	429	33	multimedia	multimedia	NOUN
easat-10905	429	34	and	and	CCONJ
easat-10905	429	35	the	the	DET
easat-10905	429	36	web	web	NOUN
easat-10905	429	37	,	,	PUNCT
easat-10905	429	38	2018	2018	NUM
easat-10905	429	39	.	.	PUNCT
easat-10905	430	1	[	[	X
easat-10905	430	2	11	11	NUM
easat-10905	430	3	]	]	PUNCT
easat-10905	430	4	r.	r.	PROPN
easat-10905	430	5	ahuja	ahuja	PROPN
easat-10905	430	6	,	,	PUNCT
easat-10905	430	7	a.	a.	PROPN
easat-10905	430	8	banga	banga	PROPN
easat-10905	430	9	,	,	PUNCT
easat-10905	430	10	and	and	CCONJ
easat-10905	430	11	s.	s.	PROPN
easat-10905	430	12	sharma	sharma	PROPN
easat-10905	430	13	,	,	PUNCT
easat-10905	430	14	detecting	detect	VERB
easat-10905	430	15	abusive	abusive	ADJ
easat-10905	430	16	comments	comment	NOUN
easat-10905	430	17	using	use	VERB
easat-10905	430	18	ensemble	ensemble	ADJ
easat-10905	430	19	deep	deep	ADJ
easat-10905	430	20	learning	learning	NOUN
easat-10905	430	21	algorithms	algorithm	NOUN
easat-10905	430	22	.	.	PUNCT
easat-10905	431	1	in	in	ADP
easat-10905	431	2	malware	malware	NOUN
easat-10905	431	3	analysis	analysis	NOUN
easat-10905	431	4	using	use	VERB
easat-10905	431	5	artificial	artificial	ADJ
easat-10905	431	6	intelligence	intelligence	NOUN
easat-10905	431	7	and	and	CCONJ
easat-10905	431	8	deep	deep	ADJ
easat-10905	431	9	learning	learning	NOUN
easat-10905	431	10	.	.	PUNCT
easat-10905	432	1	cham	cham	PROPN
easat-10905	432	2	:	:	PUNCT
easat-10905	432	3	springer	springer	NOUN
easat-10905	432	4	international	international	ADJ
easat-10905	432	5	publishing	publishing	NOUN
easat-10905	432	6	,	,	PUNCT
easat-10905	432	7	2020	2020	NUM
easat-10905	432	8	.	.	PUNCT
easat-10905	433	1	[	[	X
easat-10905	433	2	12	12	NUM
easat-10905	433	3	]	]	PUNCT
easat-10905	433	4	a.	a.	NOUN
easat-10905	433	5	baruah	baruah	PROPN
easat-10905	433	6	,	,	PUNCT
easat-10905	433	7	l.	l.	PROPN
easat-10905	433	8	wahlang	wahlang	PROPN
easat-10905	433	9	,	,	PUNCT
easat-10905	433	10	f.	f.	PROPN
easat-10905	433	11	jyrwa	jyrwa	PROPN
easat-10905	433	12	,	,	PUNCT
easat-10905	433	13	f.	f.	PROPN
easat-10905	433	14	shadap	shadap	PROPN
easat-10905	433	15	,	,	PUNCT
easat-10905	433	16	f.	f.	PROPN
easat-10905	433	17	barbhuiya	barbhuiya	PROPN
easat-10905	433	18	,	,	PUNCT
easat-10905	433	19	and	and	CCONJ
easat-10905	433	20	k.	k.	PROPN
easat-10905	433	21	dey	dey	PROPN
easat-10905	433	22	,	,	PUNCT
easat-10905	433	23	"	"	PUNCT
easat-10905	433	24	abusive	abusive	ADJ
easat-10905	433	25	language	language	NOUN
easat-10905	433	26	detection	detection	NOUN
easat-10905	433	27	in	in	ADP
easat-10905	433	28	khasi	khasi	ADJ
easat-10905	433	29	social	social	ADJ
easat-10905	433	30	media	medium	NOUN
easat-10905	433	31	comments	comment	NOUN
easat-10905	433	32	,	,	PUNCT
easat-10905	433	33	"	"	PUNCT
easat-10905	433	34	acm	acm	NOUN
easat-10905	433	35	transactions	transaction	NOUN
easat-10905	433	36	on	on	ADP
easat-10905	433	37	asian	asian	ADJ
easat-10905	433	38	and	and	CCONJ
easat-10905	433	39	low	low	ADJ
easat-10905	433	40	-	-	PUNCT
easat-10905	433	41	resource	resource	NOUN
easat-10905	433	42	language	language	NOUN
easat-10905	433	43	information	information	NOUN
easat-10905	433	44	processing	processing	NOUN
easat-10905	433	45	,	,	PUNCT
easat-10905	433	46	2024	2024	NUM
easat-10905	433	47	.	.	PUNCT
easat-10905	434	1	[	[	X
easat-10905	434	2	13	13	NUM
easat-10905	434	3	]	]	PUNCT
easat-10905	434	4	k.	k.	PROPN
easat-10905	434	5	b.	b.	PROPN
easat-10905	434	6	nelatoori	nelatoori	PROPN
easat-10905	434	7	and	and	CCONJ
easat-10905	434	8	h.	h.	PROPN
easat-10905	434	9	b.	b.	PROPN
easat-10905	434	10	kommanti	kommanti	PROPN
easat-10905	434	11	,	,	PUNCT
easat-10905	434	12	"	"	PUNCT
easat-10905	434	13	attention	attention	NOUN
easat-10905	434	14	-	-	PUNCT
easat-10905	434	15	based	base	VERB
easat-10905	434	16	bi	bi	ADJ
easat-10905	434	17	-	-	ADJ
easat-10905	434	18	lstm	lstm	ADJ
easat-10905	434	19	network	network	NOUN
easat-10905	434	20	for	for	ADP
easat-10905	434	21	abusive	abusive	ADJ
easat-10905	434	22	language	language	NOUN
easat-10905	434	23	detection	detection	NOUN
easat-10905	434	24	,	,	PUNCT
easat-10905	434	25	"	"	PUNCT
easat-10905	434	26	iete	iete	ADJ
easat-10905	434	27	journal	journal	NOUN
easat-10905	434	28	of	of	ADP
easat-10905	434	29	research	research	NOUN
easat-10905	434	30	,	,	PUNCT
easat-10905	434	31	vol	vol	NOUN
easat-10905	434	32	.	.	PROPN
easat-10905	434	33	69	69	NUM
easat-10905	434	34	,	,	PUNCT
easat-10905	434	35	no	no	INTJ
easat-10905	434	36	.	.	NOUN
easat-10905	434	37	11	11	NUM
easat-10905	434	38	,	,	PUNCT
easat-10905	434	39	pp	pp	ADJ
easat-10905	434	40	.	.	PUNCT
easat-10905	435	1	7884	7884	NUM
easat-10905	435	2	-	-	SYM
easat-10905	435	3	7892	7892	NUM
easat-10905	435	4	,	,	PUNCT
easat-10905	435	5	2023	2023	NUM
easat-10905	435	6	.	.	PUNCT
easat-10905	435	7	https://doi.org/10.1080/03772063.2022.2034534	https://doi.org/10.1080/03772063.2022.2034534	PUNCT
easat-10905	436	1	[	[	X
easat-10905	436	2	14	14	NUM
easat-10905	436	3	]	]	PUNCT
easat-10905	436	4	s.	s.	PROPN
easat-10905	436	5	a.	a.	PROPN
easat-10905	436	6	agnes	agnes	PROPN
easat-10905	436	7	,	,	PUNCT
easat-10905	436	8	a.	a.	PROPN
easat-10905	436	9	a.	a.	PROPN
easat-10905	436	10	solomon	solomon	PROPN
easat-10905	436	11	,	,	PUNCT
easat-10905	436	12	and	and	CCONJ
easat-10905	436	13	d.	d.	PROPN
easat-10905	436	14	j.	j.	PROPN
easat-10905	436	15	c.	c.	PROPN
easat-10905	436	16	tamilmaran	tamilmaran	PROPN
easat-10905	436	17	,	,	PUNCT
easat-10905	436	18	"	"	PUNCT
easat-10905	436	19	abusive	abusive	ADJ
easat-10905	436	20	comment	comment	NOUN
easat-10905	436	21	detection	detection	NOUN
easat-10905	436	22	in	in	ADP
easat-10905	436	23	social	social	ADJ
easat-10905	436	24	media	medium	NOUN
easat-10905	436	25	with	with	ADP
easat-10905	436	26	bidirectional	bidirectional	ADJ
easat-10905	436	27	lstm	lstm	PROPN
easat-10905	436	28	model	model	NOUN
easat-10905	436	29	,	,	PUNCT
easat-10905	436	30	"	"	PUNCT
easat-10905	436	31	in	in	ADP
easat-10905	436	32	2023	2023	NUM
easat-10905	436	33	5th	5th	ADJ
easat-10905	436	34	international	international	ADJ
easat-10905	436	35	conference	conference	NOUN
easat-10905	436	36	on	on	ADP
easat-10905	436	37	smart	smart	ADJ
easat-10905	436	38	systems	system	NOUN
easat-10905	436	39	and	and	CCONJ
easat-10905	436	40	inventive	inventive	ADJ
easat-10905	436	41	technology	technology	NOUN
easat-10905	436	42	(	(	PUNCT
easat-10905	436	43	icssit	icssit	NOUN
easat-10905	436	44	)	)	PUNCT
easat-10905	436	45	(	(	PUNCT
easat-10905	436	46	pp	pp	ADJ
easat-10905	436	47	.	.	PUNCT
easat-10905	437	1	1368	1368	NUM
easat-10905	437	2	-	-	SYM
easat-10905	437	3	1373	1373	NUM
easat-10905	437	4	)	)	PUNCT
easat-10905	437	5	.	.	PUNCT
easat-10905	438	1	ieee	ieee	NOUN
easat-10905	438	2	,	,	PUNCT
easat-10905	438	3	2023	2023	NUM
easat-10905	438	4	.	.	PUNCT
easat-10905	439	1	[	[	X
easat-10905	439	2	15	15	NUM
easat-10905	439	3	]	]	X
easat-10905	439	4	z.	z.	PROPN
easat-10905	439	5	mai	mai	PROPN
easat-10905	439	6	,	,	PUNCT
easat-10905	439	7	r.	r.	PROPN
easat-10905	439	8	li	li	PROPN
easat-10905	439	9	,	,	PUNCT
easat-10905	439	10	j.	j.	PROPN
easat-10905	439	11	jeong	jeong	PROPN
easat-10905	439	12	,	,	PUNCT
easat-10905	439	13	d.	d.	PROPN
easat-10905	439	14	quispe	quispe	PROPN
easat-10905	439	15	,	,	PUNCT
easat-10905	439	16	h.	h.	PROPN
easat-10905	439	17	kim	kim	PROPN
easat-10905	439	18	,	,	PUNCT
easat-10905	439	19	and	and	CCONJ
easat-10905	439	20	s.	s.	PROPN
easat-10905	439	21	sanner	sanner	PROPN
easat-10905	439	22	,	,	PUNCT
easat-10905	439	23	"	"	PUNCT
easat-10905	439	24	online	online	ADJ
easat-10905	439	25	continual	continual	ADJ
easat-10905	439	26	learning	learning	NOUN
easat-10905	439	27	in	in	ADP
easat-10905	439	28	image	image	NOUN
easat-10905	439	29	classification	classification	NOUN
easat-10905	439	30	:	:	PUNCT
easat-10905	439	31	an	an	DET
easat-10905	439	32	empirical	empirical	ADJ
easat-10905	439	33	survey	survey	NOUN
easat-10905	439	34	,	,	PUNCT
easat-10905	439	35	"	"	PUNCT
easat-10905	439	36	neurocomputing	neurocomputing	NOUN
easat-10905	439	37	,	,	PUNCT
easat-10905	439	38	vol	vol	NOUN
easat-10905	439	39	.	.	PROPN
easat-10905	439	40	469	469	NUM
easat-10905	439	41	,	,	PUNCT
easat-10905	439	42	pp	pp	ADJ
easat-10905	439	43	.	.	PUNCT
easat-10905	440	1	28	28	NUM
easat-10905	440	2	-	-	SYM
easat-10905	440	3	51	51	NUM
easat-10905	440	4	,	,	PUNCT
easat-10905	440	5	2022	2022	NUM
easat-10905	440	6	.	.	PUNCT
easat-10905	441	1	https://doi.org/10.1016/j.neucom.2021.10.021	https://doi.org/10.1016/j.neucom.2021.10.021	NOUN
easat-10905	442	1	[	[	X
easat-10905	442	2	16	16	NUM
easat-10905	442	3	]	]	PUNCT
easat-10905	442	4	z.	z.	PROPN
easat-10905	442	5	ke	ke	PROPN
easat-10905	442	6	and	and	CCONJ
easat-10905	442	7	b.	b.	PROPN
easat-10905	442	8	liu	liu	PROPN
easat-10905	442	9	,	,	PUNCT
easat-10905	442	10	"	"	PUNCT
easat-10905	442	11	continual	continual	ADJ
easat-10905	442	12	learning	learning	NOUN
easat-10905	442	13	of	of	ADP
easat-10905	442	14	natural	natural	ADJ
easat-10905	442	15	language	language	NOUN
easat-10905	442	16	processing	processing	NOUN
easat-10905	442	17	tasks	task	NOUN
easat-10905	442	18	:	:	PUNCT
easat-10905	442	19	a	a	DET
easat-10905	442	20	survey	survey	NOUN
easat-10905	442	21	,	,	PUNCT
easat-10905	442	22	"	"	PUNCT
easat-10905	442	23	arxiv	arxiv	PROPN
easat-10905	442	24	preprint	preprint	VERB
easat-10905	442	25	arxiv:2211.12701	arxiv:2211.12701	NOUN
easat-10905	442	26	,	,	PUNCT
easat-10905	442	27	2022	2022	NUM
easat-10905	442	28	.	.	PUNCT
easat-10905	443	1	https://doi.org/10.48550/arxiv.2211.12701	https://doi.org/10.48550/arxiv.2211.12701	PROPN
easat-10905	443	2	[	[	X
easat-10905	443	3	17	17	NUM
easat-10905	443	4	]	]	X
easat-10905	443	5	g.	g.	PROPN
easat-10905	443	6	i.	i.	PROPN
easat-10905	443	7	parisi	parisi	PROPN
easat-10905	443	8	,	,	PUNCT
easat-10905	443	9	r.	r.	PROPN
easat-10905	443	10	kemker	kemker	PROPN
easat-10905	443	11	,	,	PUNCT
easat-10905	443	12	j.	j.	PROPN
easat-10905	443	13	l.	l.	PROPN
easat-10905	443	14	part	part	PROPN
easat-10905	443	15	,	,	PUNCT
easat-10905	443	16	c.	c.	PROPN
easat-10905	443	17	kanan	kanan	PROPN
easat-10905	443	18	,	,	PUNCT
easat-10905	443	19	and	and	CCONJ
easat-10905	443	20	s.	s.	PROPN
easat-10905	443	21	wermter	wermter	PROPN
easat-10905	443	22	,	,	PUNCT
easat-10905	443	23	"	"	PUNCT
easat-10905	443	24	continual	continual	ADJ
easat-10905	443	25	lifelong	lifelong	ADJ
easat-10905	443	26	learning	learning	NOUN
easat-10905	443	27	with	with	ADP
easat-10905	443	28	neural	neural	ADJ
easat-10905	443	29	networks	network	NOUN
easat-10905	443	30	:	:	PUNCT
easat-10905	443	31	a	a	DET
easat-10905	443	32	review	review	NOUN
easat-10905	443	33	,	,	PUNCT
easat-10905	443	34	"	"	PUNCT
easat-10905	443	35	neural	neural	ADJ
easat-10905	443	36	networks	network	NOUN
easat-10905	443	37	,	,	PUNCT
easat-10905	443	38	vol	vol	NOUN
easat-10905	443	39	.	.	PROPN
easat-10905	443	40	113	113	NUM
easat-10905	443	41	,	,	PUNCT
easat-10905	443	42	pp	pp	ADJ
easat-10905	443	43	.	.	PUNCT
easat-10905	444	1	54	54	NUM
easat-10905	444	2	-	-	SYM
easat-10905	444	3	71	71	NUM
easat-10905	444	4	,	,	PUNCT
easat-10905	444	5	2019	2019	NUM
easat-10905	444	6	.	.	PUNCT
easat-10905	445	1	https://doi.org/10.1016/j.neunet.2019.01.012	https://doi.org/10.1016/j.neunet.2019.01.012	NOUN
easat-10905	445	2	[	[	X
easat-10905	445	3	18	18	NUM
easat-10905	445	4	]	]	PUNCT
easat-10905	445	5	l.	l.	PROPN
easat-10905	445	6	wang	wang	PROPN
easat-10905	445	7	,	,	PUNCT
easat-10905	445	8	x.	x.	PROPN
easat-10905	445	9	zhang	zhang	PROPN
easat-10905	445	10	,	,	PUNCT
easat-10905	445	11	h.	h.	PROPN
easat-10905	445	12	su	su	PROPN
easat-10905	445	13	,	,	PUNCT
easat-10905	445	14	and	and	CCONJ
easat-10905	445	15	j.	j.	PROPN
easat-10905	445	16	zhu	zhu	PROPN
easat-10905	445	17	,	,	PUNCT
easat-10905	445	18	"	"	PUNCT
easat-10905	445	19	a	a	DET
easat-10905	445	20	comprehensive	comprehensive	ADJ
easat-10905	445	21	survey	survey	NOUN
easat-10905	445	22	of	of	ADP
easat-10905	445	23	continual	continual	ADJ
easat-10905	445	24	learning	learning	NOUN
easat-10905	445	25	:	:	PUNCT
easat-10905	446	1	theory	theory	NOUN
easat-10905	446	2	,	,	PUNCT
easat-10905	446	3	method	method	NOUN
easat-10905	446	4	and	and	CCONJ
easat-10905	446	5	application	application	NOUN
easat-10905	446	6	,	,	PUNCT
easat-10905	446	7	"	"	PUNCT
easat-10905	446	8	ieee	ieee	NOUN
easat-10905	446	9	transactions	transaction	NOUN
easat-10905	446	10	on	on	ADP
easat-10905	446	11	pattern	pattern	NOUN
easat-10905	446	12	analysis	analysis	NOUN
easat-10905	446	13	and	and	CCONJ
easat-10905	446	14	machine	machine	NOUN
easat-10905	446	15	intelligence	intelligence	NOUN
easat-10905	446	16	,	,	PUNCT
easat-10905	446	17	vol	vol	NOUN
easat-10905	446	18	.	.	PROPN
easat-10905	446	19	46	46	NUM
easat-10905	446	20	,	,	PUNCT
easat-10905	446	21	no	no	INTJ
easat-10905	446	22	.	.	NOUN
easat-10905	446	23	8	8	NUM
easat-10905	446	24	,	,	PUNCT
easat-10905	446	25	pp	pp	ADJ
easat-10905	446	26	.	.	PUNCT
easat-10905	447	1	5362	5362	NUM
easat-10905	447	2	-	-	SYM
easat-10905	447	3	5383	5383	NUM
easat-10905	447	4	,	,	PUNCT
easat-10905	447	5	2024	2024	NUM
easat-10905	447	6	.	.	PUNCT
easat-10905	448	1	https://doi.org/10.1109/tpami.2024.3367329	https://doi.org/10.1109/tpami.2024.3367329	PROPN
easat-10905	448	2	[	[	X
easat-10905	448	3	19	19	NUM
easat-10905	448	4	]	]	PUNCT
easat-10905	448	5	a.	a.	NOUN
easat-10905	448	6	gepperth	gepperth	NOUN
easat-10905	448	7	and	and	CCONJ
easat-10905	448	8	b.	b.	PROPN
easat-10905	448	9	hammer	hammer	NOUN
easat-10905	448	10	,	,	PUNCT
easat-10905	448	11	"	"	PUNCT
easat-10905	448	12	incremental	incremental	ADJ
easat-10905	448	13	learning	learning	NOUN
easat-10905	448	14	algorithms	algorithm	NOUN
easat-10905	448	15	and	and	CCONJ
easat-10905	448	16	applications	application	NOUN
easat-10905	448	17	,	,	PUNCT
easat-10905	448	18	"	"	PUNCT
easat-10905	448	19	in	in	ADP
easat-10905	448	20	european	european	ADJ
easat-10905	448	21	symposium	symposium	NOUN
easat-10905	448	22	on	on	ADP
easat-10905	448	23	artificial	artificial	ADJ
easat-10905	448	24	neural	neural	ADJ
easat-10905	448	25	networks	network	NOUN
easat-10905	448	26	(	(	PUNCT
easat-10905	448	27	esann	esann	PROPN
easat-10905	448	28	)	)	PUNCT
easat-10905	448	29	,	,	PUNCT
easat-10905	448	30	2016	2016	NUM
easat-10905	448	31	.	.	PUNCT
easat-10905	449	1	[	[	X
easat-10905	449	2	20	20	NUM
easat-10905	449	3	]	]	X
easat-10905	449	4	m.	m.	NOUN
easat-10905	449	5	de	de	X
easat-10905	449	6	lange	lange	PROPN
easat-10905	449	7	et	et	PROPN
easat-10905	449	8	al	al	PROPN
easat-10905	449	9	.	.	PROPN
easat-10905	449	10	,	,	PUNCT
easat-10905	449	11	"	"	PUNCT
easat-10905	449	12	a	a	DET
easat-10905	449	13	continual	continual	ADJ
easat-10905	449	14	learning	learning	NOUN
easat-10905	449	15	survey	survey	NOUN
easat-10905	449	16	:	:	PUNCT
easat-10905	449	17	defying	defy	VERB
easat-10905	449	18	forgetting	forget	VERB
easat-10905	449	19	in	in	ADP
easat-10905	449	20	classification	classification	NOUN
easat-10905	449	21	tasks	task	NOUN
easat-10905	449	22	,	,	PUNCT
easat-10905	449	23	"	"	PUNCT
easat-10905	449	24	ieee	ieee	NOUN
easat-10905	449	25	transactions	transaction	NOUN
easat-10905	449	26	on	on	ADP
easat-10905	449	27	pattern	pattern	NOUN
easat-10905	449	28	analysis	analysis	NOUN
easat-10905	449	29	and	and	CCONJ
easat-10905	449	30	machine	machine	NOUN
easat-10905	449	31	intelligence	intelligence	NOUN
easat-10905	449	32	,	,	PUNCT
easat-10905	449	33	vol	vol	NOUN
easat-10905	449	34	.	.	PROPN
easat-10905	450	1	44	44	NUM
easat-10905	450	2	,	,	PUNCT
easat-10905	450	3	no	no	INTJ
easat-10905	450	4	.	.	NOUN
easat-10905	450	5	7	7	NUM
easat-10905	450	6	,	,	PUNCT
easat-10905	450	7	pp	pp	ADJ
easat-10905	450	8	.	.	PUNCT
easat-10905	451	1	3366	3366	NUM
easat-10905	451	2	-	-	SYM
easat-10905	451	3	3385	3385	NUM
easat-10905	451	4	,	,	PUNCT
easat-10905	451	5	2021	2021	NUM
easat-10905	451	6	.	.	PUNCT
easat-10905	452	1	https://doi.org/10.1109/tpami.2021.3057446	https://doi.org/10.1109/tpami.2021.3057446	PROPN
easat-10905	452	2	https://creativecommons.org/licenses/by/4.0/	https://creativecommons.org/licenses/by/4.0/	PROPN
easat-10905	452	3	https://doi.org/10.1016/j.avb.2021.101608	https://doi.org/10.1016/j.avb.2021.101608	PROPN
easat-10905	452	4	https://doi.org/10.1007/s00530-021-00784-8	https://doi.org/10.1007/s00530-021-00784-8	NUM
easat-10905	452	5	https://doi.org/10.1016/j.nlp.2023.100006	https://doi.org/10.1016/j.nlp.2023.100006	PROPN
easat-10905	452	6	https://doi.org/10.1007/s10579-020-09502-8	https://doi.org/10.1007/s10579-020-09502-8	PRON
easat-10905	452	7	https://doi.org/10.48550/arxiv.1908.04531	https://doi.org/10.48550/arxiv.1908.04531	PROPN
easat-10905	452	8	https://doi.org/10.48550/arxiv.1706.01206	https://doi.org/10.48550/arxiv.1706.01206	PROPN
easat-10905	452	9	https://doi.org/10.1109/access.2024.3370232	https://doi.org/10.1109/access.2024.3370232	PROPN
easat-10905	452	10	https://doi.org/10.1080/03772063.2022.2034534	https://doi.org/10.1080/03772063.2022.2034534	PROPN
easat-10905	452	11	https://doi.org/10.1016/j.neucom.2021.10.021	https://doi.org/10.1016/j.neucom.2021.10.021	NOUN
easat-10905	452	12	https://doi.org/10.48550/arxiv.2211.12701	https://doi.org/10.48550/arxiv.2211.12701	PROPN
easat-10905	452	13	https://doi.org/10.1016/j.neunet.2019.01.012	https://doi.org/10.1016/j.neunet.2019.01.012	NOUN
easat-10905	452	14	https://doi.org/10.1109/tpami.2024.3367329	https://doi.org/10.1109/tpami.2024.3367329	PROPN
easat-10905	452	15	https://doi.org/10.1109/tpami.2021.3057446	https://doi.org/10.1109/tpami.2021.3057446	PROPN
easat-10905	452	16	466	466	NUM
easat-10905	452	17	edelweiss	edelweiss	PROPN
easat-10905	452	18	applied	apply	VERB
easat-10905	452	19	science	science	NOUN
easat-10905	452	20	and	and	CCONJ
easat-10905	452	21	technology	technology	NOUN
easat-10905	452	22	issn	issn	PROPN
easat-10905	452	23	:	:	PUNCT
easat-10905	452	24	2576	2576	NUM
easat-10905	452	25	-	-	SYM
easat-10905	452	26	8484	8484	NUM
easat-10905	452	27	vol	vol	NOUN
easat-10905	452	28	.	.	PROPN
easat-10905	452	29	9	9	NUM
easat-10905	452	30	,	,	PUNCT
easat-10905	452	31	no	no	INTJ
easat-10905	452	32	.	.	NOUN
easat-10905	452	33	11	11	NUM
easat-10905	452	34	:	:	PUNCT
easat-10905	452	35	448	448	NUM
easat-10905	452	36	-	-	SYM
easat-10905	452	37	466	466	NUM
easat-10905	452	38	,	,	PUNCT
easat-10905	452	39	2025	2025	NUM
easat-10905	452	40	doi	doi	NOUN
easat-10905	452	41	:	:	PUNCT
easat-10905	452	42	10.55214/2576	10.55214/2576	NUM
easat-10905	452	43	-	-	PUNCT
easat-10905	452	44	8484.v9i11.10905	8484.v9i11.10905	NOUN
easat-10905	452	45	©	©	PROPN
easat-10905	452	46	2025	2025	NUM
easat-10905	452	47	by	by	ADP
easat-10905	452	48	the	the	DET
easat-10905	452	49	authors	author	NOUN
easat-10905	452	50	;	;	PUNCT
easat-10905	452	51	licensee	licensee	PROPN
easat-10905	452	52	learning	learning	NOUN
easat-10905	452	53	gate	gate	NOUN
easat-10905	452	54	[	[	X
easat-10905	452	55	21	21	NUM
easat-10905	452	56	]	]	PUNCT
easat-10905	452	57	m.	m.	NOUN
easat-10905	452	58	biesialska	biesialska	PROPN
easat-10905	452	59	,	,	PUNCT
easat-10905	452	60	k.	k.	PROPN
easat-10905	452	61	biesialska	biesialska	PROPN
easat-10905	452	62	,	,	PUNCT
easat-10905	452	63	and	and	CCONJ
easat-10905	452	64	m.	m.	PROPN
easat-10905	452	65	r.	r.	PROPN
easat-10905	452	66	costa	costa	PROPN
easat-10905	452	67	-	-	PUNCT
easat-10905	452	68	jussa	jussa	PROPN
easat-10905	452	69	,	,	PUNCT
easat-10905	452	70	"	"	PUNCT
easat-10905	452	71	continual	continual	ADJ
easat-10905	452	72	lifelong	lifelong	ADJ
easat-10905	452	73	learning	learning	NOUN
easat-10905	452	74	in	in	ADP
easat-10905	452	75	natural	natural	ADJ
easat-10905	452	76	language	language	NOUN
easat-10905	452	77	processing	processing	NOUN
easat-10905	452	78	:	:	PUNCT
easat-10905	452	79	a	a	DET
easat-10905	452	80	survey	survey	NOUN
easat-10905	452	81	,	,	PUNCT
easat-10905	452	82	"	"	PUNCT
easat-10905	452	83	arxiv	arxiv	PROPN
easat-10905	452	84	preprint	preprint	NOUN
easat-10905	452	85	arxiv:2012.09823	arxiv:2012.09823	NOUN
easat-10905	452	86	,	,	PUNCT
easat-10905	452	87	2020	2020	NUM
easat-10905	452	88	.	.	PUNCT
easat-10905	453	1	https://doi.org/10.18653/v1/2020.coling-main.574	https://doi.org/10.18653/v1/2020.coling-main.574	PRON
easat-10905	454	1	[	[	X
easat-10905	454	2	22	22	NUM
easat-10905	454	3	]	]	PUNCT
easat-10905	454	4	p.	p.	PROPN
easat-10905	454	5	yang	yang	PROPN
easat-10905	454	6	,	,	PUNCT
easat-10905	454	7	d.	d.	PROPN
easat-10905	454	8	li	li	PROPN
easat-10905	454	9	,	,	PUNCT
easat-10905	454	10	and	and	CCONJ
easat-10905	454	11	p.	p.	PROPN
easat-10905	454	12	li	li	PROPN
easat-10905	454	13	,	,	PUNCT
easat-10905	454	14	"	"	PUNCT
easat-10905	454	15	continual	continual	ADJ
easat-10905	454	16	learning	learning	NOUN
easat-10905	454	17	for	for	ADP
easat-10905	454	18	natural	natural	ADJ
easat-10905	454	19	language	language	NOUN
easat-10905	454	20	generations	generation	NOUN
easat-10905	454	21	with	with	ADP
easat-10905	454	22	transformer	transformer	NOUN
easat-10905	454	23	calibration	calibration	NOUN
easat-10905	454	24	,	,	PUNCT
easat-10905	454	25	"	"	PUNCT
easat-10905	454	26	in	in	ADP
easat-10905	454	27	proceedings	proceeding	NOUN
easat-10905	454	28	of	of	ADP
easat-10905	454	29	the	the	DET
easat-10905	454	30	26th	26th	ADJ
easat-10905	454	31	conference	conference	NOUN
easat-10905	454	32	on	on	ADP
easat-10905	454	33	computational	computational	ADJ
easat-10905	454	34	natural	natural	ADJ
easat-10905	454	35	language	language	NOUN
easat-10905	454	36	learning	learning	NOUN
easat-10905	454	37	(	(	PUNCT
easat-10905	454	38	conll	conll	PROPN
easat-10905	454	39	)	)	PUNCT
easat-10905	454	40	,	,	PUNCT
easat-10905	454	41	2022	2022	NUM
easat-10905	454	42	.	.	PUNCT
easat-10905	455	1	[	[	X
easat-10905	455	2	23	23	NUM
easat-10905	455	3	]	]	PUNCT
easat-10905	455	4	v.	v.	CCONJ
easat-10905	455	5	varshney	varshney	NOUN
easat-10905	455	6	,	,	PUNCT
easat-10905	455	7	m.	m.	NOUN
easat-10905	455	8	patidar	patidar	PROPN
easat-10905	455	9	,	,	PUNCT
easat-10905	455	10	r.	r.	PROPN
easat-10905	455	11	kumar	kumar	PROPN
easat-10905	455	12	,	,	PUNCT
easat-10905	455	13	g.	g.	PROPN
easat-10905	455	14	shroff	shroff	PROPN
easat-10905	455	15	,	,	PUNCT
easat-10905	455	16	and	and	CCONJ
easat-10905	455	17	l.	l.	PROPN
easat-10905	455	18	vig	vig	PROPN
easat-10905	455	19	,	,	PUNCT
easat-10905	455	20	"	"	PUNCT
easat-10905	455	21	prompt	prompt	ADJ
easat-10905	455	22	augmented	augment	VERB
easat-10905	455	23	generative	generative	ADJ
easat-10905	455	24	replay	replay	NOUN
easat-10905	455	25	via	via	ADP
easat-10905	455	26	supervised	supervised	ADJ
easat-10905	455	27	contrastive	contrastive	ADJ
easat-10905	455	28	training	training	NOUN
easat-10905	455	29	for	for	ADP
easat-10905	455	30	lifelong	lifelong	ADJ
easat-10905	455	31	intent	intent	NOUN
easat-10905	455	32	detection	detection	NOUN
easat-10905	455	33	,	,	PUNCT
easat-10905	455	34	"	"	PUNCT
easat-10905	455	35	presented	present	VERB
easat-10905	455	36	at	at	ADP
easat-10905	455	37	the	the	DET
easat-10905	455	38	findings	finding	NOUN
easat-10905	455	39	of	of	ADP
easat-10905	455	40	the	the	DET
easat-10905	455	41	2022	2022	NUM
easat-10905	455	42	conference	conference	NOUN
easat-10905	455	43	of	of	ADP
easat-10905	455	44	the	the	DET
easat-10905	455	45	north	north	ADJ
easat-10905	455	46	american	american	ADJ
easat-10905	455	47	chapter	chapter	NOUN
easat-10905	455	48	of	of	ADP
easat-10905	455	49	the	the	DET
easat-10905	455	50	association	association	NOUN
easat-10905	455	51	for	for	ADP
easat-10905	455	52	computational	computational	ADJ
easat-10905	455	53	linguistics	linguistic	NOUN
easat-10905	455	54	:	:	PUNCT
easat-10905	455	55	human	human	ADJ
easat-10905	455	56	language	language	NOUN
easat-10905	455	57	technologies	technology	NOUN
easat-10905	455	58	(	(	PUNCT
easat-10905	455	59	naacl‑hlt	naacl‑hlt	PROPN
easat-10905	455	60	2022	2022	NUM
easat-10905	455	61	)	)	PUNCT
easat-10905	455	62	(	(	PUNCT
easat-10905	455	63	pp	pp	ADP
easat-10905	455	64	.	.	PUNCT
easat-10905	455	65	1113–1127	1113–1127	NUM
easat-10905	455	66	)	)	PUNCT
easat-10905	455	67	.	.	PUNCT
easat-10905	456	1	seattle	seattle	PROPN
easat-10905	456	2	,	,	PUNCT
easat-10905	456	3	wa	wa	PROPN
easat-10905	456	4	:	:	PUNCT
easat-10905	456	5	association	association	NOUN
easat-10905	456	6	for	for	ADP
easat-10905	456	7	computational	computational	ADJ
easat-10905	456	8	linguistics	linguistic	NOUN
easat-10905	456	9	,	,	PUNCT
easat-10905	456	10	2022	2022	NUM
easat-10905	456	11	.	.	PUNCT
easat-10905	457	1	[	[	X
easat-10905	457	2	24	24	NUM
easat-10905	457	3	]	]	PUNCT
easat-10905	457	4	c.	c.	PROPN
easat-10905	457	5	wang	wang	PROPN
easat-10905	457	6	et	et	PROPN
easat-10905	457	7	al	al	PROPN
easat-10905	457	8	.	.	PROPN
easat-10905	457	9	,	,	PUNCT
easat-10905	457	10	"	"	PUNCT
easat-10905	457	11	mell	mell	NOUN
easat-10905	457	12	:	:	PUNCT
easat-10905	457	13	large	large	ADJ
easat-10905	457	14	-	-	PUNCT
easat-10905	457	15	scale	scale	NOUN
easat-10905	457	16	extensible	extensible	ADJ
easat-10905	457	17	user	user	NOUN
easat-10905	457	18	intent	intent	VERB
easat-10905	457	19	classification	classification	NOUN
easat-10905	457	20	for	for	ADP
easat-10905	457	21	dialogue	dialogue	NOUN
easat-10905	457	22	systems	system	NOUN
easat-10905	457	23	with	with	ADP
easat-10905	457	24	meta	meta	PRON
easat-10905	457	25	lifelong	lifelong	ADJ
easat-10905	457	26	learning	learning	NOUN
easat-10905	457	27	,	,	PUNCT
easat-10905	457	28	"	"	PUNCT
easat-10905	457	29	in	in	ADP
easat-10905	457	30	proceedings	proceeding	NOUN
easat-10905	457	31	of	of	ADP
easat-10905	457	32	the	the	DET
easat-10905	457	33	27th	27th	ADJ
easat-10905	457	34	acm	acm	PROPN
easat-10905	457	35	sigkdd	sigkdd	NOUN
easat-10905	457	36	conference	conference	NOUN
easat-10905	457	37	on	on	ADP
easat-10905	457	38	knowledge	knowledge	PROPN
easat-10905	457	39	discovery	discovery	PROPN
easat-10905	457	40	&	&	CCONJ
easat-10905	457	41	data	data	PROPN
easat-10905	457	42	mining	mining	NOUN
easat-10905	457	43	,	,	PUNCT
easat-10905	457	44	2021	2021	NUM
easat-10905	457	45	.	.	PUNCT
easat-10905	458	1	[	[	X
easat-10905	458	2	25	25	NUM
easat-10905	458	3	]	]	PUNCT
easat-10905	458	4	j.	j.	PROPN
easat-10905	458	5	qian	qian	PROPN
easat-10905	458	6	,	,	PUNCT
easat-10905	458	7	h.	h.	PROPN
easat-10905	458	8	wang	wang	PROPN
easat-10905	458	9	,	,	PUNCT
easat-10905	458	10	m.	m.	NOUN
easat-10905	458	11	elsherief	elsherief	PROPN
easat-10905	458	12	,	,	PUNCT
easat-10905	458	13	and	and	CCONJ
easat-10905	458	14	x.	x.	PROPN
easat-10905	458	15	yan	yan	PROPN
easat-10905	458	16	,	,	PUNCT
easat-10905	458	17	"	"	PUNCT
easat-10905	458	18	lifelong	lifelong	ADJ
easat-10905	458	19	learning	learning	NOUN
easat-10905	458	20	of	of	ADP
easat-10905	458	21	hate	hate	NOUN
easat-10905	458	22	speech	speech	NOUN
easat-10905	458	23	classification	classification	NOUN
easat-10905	458	24	on	on	ADP
easat-10905	458	25	social	social	ADJ
easat-10905	458	26	media	medium	NOUN
easat-10905	458	27	,	,	PUNCT
easat-10905	458	28	"	"	PUNCT
easat-10905	458	29	arxiv	arxiv	PROPN
easat-10905	458	30	preprint	preprint	NOUN
easat-10905	458	31	arxiv:2106.02821	arxiv:2106.02821	PROPN
easat-10905	458	32	,	,	PUNCT
easat-10905	458	33	2021	2021	NUM
easat-10905	458	34	.	.	PUNCT
easat-10905	459	1	https://doi.org/10.48550/arxiv.2106.02821	https://doi.org/10.48550/arxiv.2106.02821	PRON
easat-10905	460	1	[	[	X
easat-10905	460	2	26	26	NUM
easat-10905	460	3	]	]	PUNCT
easat-10905	460	4	z.	z.	PROPN
easat-10905	460	5	chen	chen	PROPN
easat-10905	460	6	,	,	PUNCT
easat-10905	460	7	n.	n.	PROPN
easat-10905	460	8	ma	ma	PROPN
easat-10905	460	9	,	,	PUNCT
easat-10905	460	10	and	and	CCONJ
easat-10905	460	11	b.	b.	PROPN
easat-10905	460	12	liu	liu	PROPN
easat-10905	460	13	,	,	PUNCT
easat-10905	460	14	"	"	PUNCT
easat-10905	460	15	lifelong	lifelong	ADJ
easat-10905	460	16	learning	learning	NOUN
easat-10905	460	17	for	for	ADP
easat-10905	460	18	sentiment	sentiment	NOUN
easat-10905	460	19	classification	classification	NOUN
easat-10905	460	20	,	,	PUNCT
easat-10905	460	21	"	"	PUNCT
easat-10905	460	22	arxiv	arxiv	PROPN
easat-10905	460	23	preprint	preprint	NOUN
easat-10905	460	24	arxiv:1801.02808	arxiv:1801.02808	NOUN
easat-10905	460	25	,	,	PUNCT
easat-10905	460	26	2018	2018	NUM
easat-10905	460	27	.	.	PUNCT
easat-10905	461	1	https://doi.org/10.48550/arxiv.1801.02808	https://doi.org/10.48550/arxiv.1801.02808	X
easat-10905	462	1	[	[	X
easat-10905	462	2	27	27	NUM
easat-10905	462	3	]	]	X
easat-10905	462	4	c.	c.	PROPN
easat-10905	462	5	de	de	PROPN
easat-10905	462	6	 	 	SPACE
easat-10905	462	7	masson	masson	PROPN
easat-10905	462	8	 	 	SPACE
easat-10905	462	9	d’autume	d’autume	PROPN
easat-10905	462	10	,	,	PUNCT
easat-10905	462	11	s.	s.	PROPN
easat-10905	462	12	ruder	ruder	PROPN
easat-10905	462	13	,	,	PUNCT
easat-10905	462	14	l.	l.	PROPN
easat-10905	462	15	kong	kong	PROPN
easat-10905	462	16	,	,	PUNCT
easat-10905	462	17	and	and	CCONJ
easat-10905	462	18	d.	d.	PROPN
easat-10905	462	19	yogatama	yogatama	PROPN
easat-10905	462	20	,	,	PUNCT
easat-10905	462	21	"	"	PUNCT
easat-10905	462	22	episodic	episodic	ADJ
easat-10905	462	23	memory	memory	NOUN
easat-10905	462	24	in	in	ADP
easat-10905	462	25	lifelong	lifelong	ADJ
easat-10905	462	26	language	language	NOUN
easat-10905	462	27	learning	learning	NOUN
easat-10905	462	28	,	,	PUNCT
easat-10905	462	29	"	"	PUNCT
easat-10905	462	30	presented	present	VERB
easat-10905	462	31	at	at	ADP
easat-10905	462	32	the	the	DET
easat-10905	462	33	advances	advance	NOUN
easat-10905	462	34	in	in	ADP
easat-10905	462	35	neural	neural	ADJ
easat-10905	462	36	information	information	NOUN
easat-10905	462	37	processing	processing	NOUN
easat-10905	462	38	systems	system	NOUN
easat-10905	462	39	32	32	NUM
easat-10905	462	40	:	:	PUNCT
easat-10905	462	41	annual	annual	ADJ
easat-10905	462	42	conference	conference	NOUN
easat-10905	462	43	on	on	ADP
easat-10905	462	44	neural	neural	ADJ
easat-10905	462	45	information	information	NOUN
easat-10905	462	46	processing	processing	NOUN
easat-10905	462	47	systems	system	NOUN
easat-10905	462	48	(	(	PUNCT
easat-10905	462	49	neurips	neurip	NOUN
easat-10905	462	50	2019	2019	NUM
easat-10905	462	51	)	)	PUNCT
easat-10905	463	1	(	(	PUNCT
easat-10905	463	2	pp	pp	ADV
easat-10905	463	3	.	.	PUNCT
easat-10905	463	4	 	 	SPACE
easat-10905	463	5	13122–13131	13122–13131	NUM
easat-10905	463	6	)	)	PUNCT
easat-10905	463	7	.	.	PUNCT
easat-10905	464	1	vancouver	vancouver	PROPN
easat-10905	464	2	,	,	PUNCT
easat-10905	464	3	bc	bc	PROPN
easat-10905	464	4	,	,	PUNCT
easat-10905	464	5	canada	canada	PROPN
easat-10905	464	6	,	,	PUNCT
easat-10905	464	7	2019	2019	NUM
easat-10905	464	8	.	.	PUNCT
easat-10905	465	1	[	[	X
easat-10905	465	2	28	28	NUM
easat-10905	465	3	]	]	X
easat-10905	465	4	z.	z.	PROPN
easat-10905	465	5	li	li	PROPN
easat-10905	465	6	,	,	PUNCT
easat-10905	465	7	d.	d.	PROPN
easat-10905	465	8	hoiem	hoiem	PROPN
easat-10905	465	9	,	,	PUNCT
easat-10905	465	10	and	and	CCONJ
easat-10905	465	11	d.	d.	PROPN
easat-10905	465	12	forsyth	forsyth	PROPN
easat-10905	465	13	,	,	PUNCT
easat-10905	465	14	"	"	PUNCT
easat-10905	465	15	learning	learn	VERB
easat-10905	465	16	without	without	ADP
easat-10905	465	17	forgetting	forget	VERB
easat-10905	465	18	,	,	PUNCT
easat-10905	465	19	"	"	PUNCT
easat-10905	465	20	ieee	ieee	NOUN
easat-10905	465	21	transactions	transaction	NOUN
easat-10905	465	22	on	on	ADP
easat-10905	465	23	pattern	pattern	NOUN
easat-10905	465	24	analysis	analysis	NOUN
easat-10905	465	25	and	and	CCONJ
easat-10905	465	26	machine	machine	NOUN
easat-10905	465	27	intelligence	intelligence	NOUN
easat-10905	465	28	,	,	PUNCT
easat-10905	465	29	vol	vol	NOUN
easat-10905	465	30	.	.	PROPN
easat-10905	465	31	40	40	NUM
easat-10905	465	32	,	,	PUNCT
easat-10905	465	33	no	no	INTJ
easat-10905	465	34	.	.	NOUN
easat-10905	465	35	12	12	NUM
easat-10905	465	36	,	,	PUNCT
easat-10905	465	37	pp	pp	ADJ
easat-10905	465	38	.	.	PUNCT
easat-10905	465	39	2935–2947	2935–2947	NUM
easat-10905	465	40	,	,	PUNCT
easat-10905	465	41	2017	2017	NUM
easat-10905	465	42	.	.	PUNCT
easat-10905	466	1	[	[	X
easat-10905	466	2	29	29	NUM
easat-10905	466	3	]	]	PUNCT
easat-10905	466	4	h.	h.	PROPN
easat-10905	466	5	huang	huang	PROPN
easat-10905	466	6	and	and	CCONJ
easat-10905	466	7	a.	a.	NOUN
easat-10905	466	8	asemi	asemi	PROPN
easat-10905	466	9	,	,	PUNCT
easat-10905	466	10	"	"	PUNCT
easat-10905	466	11	an	an	DET
easat-10905	466	12	experimental	experimental	ADJ
easat-10905	466	13	study	study	NOUN
easat-10905	466	14	on	on	ADP
easat-10905	466	15	dynamic	dynamic	ADJ
easat-10905	466	16	lifelong	lifelong	ADJ
easat-10905	466	17	learning	learning	NOUN
easat-10905	466	18	with	with	ADP
easat-10905	466	19	gpt	gpt	NOUN
easat-10905	466	20	for	for	ADP
easat-10905	466	21	mitigating	mitigate	VERB
easat-10905	466	22	catastrophic	catastrophic	ADJ
easat-10905	466	23	forgetting	forgetting	NOUN
easat-10905	466	24	in	in	ADP
easat-10905	466	25	aspect‑based	aspect‑base	VERB
easat-10905	466	26	sentiment	sentiment	NOUN
easat-10905	466	27	analysis	analysis	NOUN
easat-10905	466	28	,	,	PUNCT
easat-10905	466	29	"	"	PUNCT
easat-10905	466	30	ieee	ieee	NOUN
easat-10905	466	31	access	access	NOUN
easat-10905	466	32	,	,	PUNCT
easat-10905	466	33	vol	vol	NOUN
easat-10905	466	34	.	.	PROPN
easat-10905	466	35	13	13	NUM
easat-10905	466	36	,	,	PUNCT
easat-10905	466	37	pp	pp	ADJ
easat-10905	466	38	.	.	PUNCT
easat-10905	467	1	90	90	NUM
easat-10905	467	2	 	 	SPACE
easat-10905	467	3	316–90	316–90	NOUN
easat-10905	467	4	 	 	SPACE
easat-10905	467	5	332	332	NUM
easat-10905	467	6	,	,	PUNCT
easat-10905	467	7	2025	2025	NUM
easat-10905	467	8	.	.	PUNCT
easat-10905	468	1	[	[	X
easat-10905	468	2	30	30	NUM
easat-10905	468	3	]	]	PUNCT
easat-10905	468	4	m.	m.	NOUN
easat-10905	468	5	jain	jain	PROPN
easat-10905	468	6	,	,	PUNCT
easat-10905	468	7	h.	h.	PROPN
easat-10905	468	8	kaur	kaur	PROPN
easat-10905	468	9	,	,	PUNCT
easat-10905	468	10	b.	b.	PROPN
easat-10905	468	11	gupta	gupta	PROPN
easat-10905	468	12	,	,	PUNCT
easat-10905	468	13	j.	j.	PROPN
easat-10905	468	14	gera	gera	PROPN
easat-10905	468	15	,	,	PUNCT
easat-10905	468	16	and	and	CCONJ
easat-10905	468	17	v.	v.	ADP
easat-10905	468	18	kalra	kalra	NOUN
easat-10905	468	19	,	,	PUNCT
easat-10905	468	20	"	"	PUNCT
easat-10905	468	21	incremental	incremental	ADJ
easat-10905	468	22	learning	learning	NOUN
easat-10905	468	23	algorithm	algorithm	NOUN
easat-10905	468	24	for	for	ADP
easat-10905	468	25	dynamic	dynamic	ADJ
easat-10905	468	26	evolution	evolution	NOUN
easat-10905	468	27	of	of	ADP
easat-10905	468	28	domain	domain	NOUN
easat-10905	468	29	specific	specific	ADJ
easat-10905	468	30	vocabulary	vocabulary	NOUN
easat-10905	468	31	with	with	ADP
easat-10905	468	32	its	its	PRON
easat-10905	468	33	stability	stability	NOUN
easat-10905	468	34	and	and	CCONJ
easat-10905	468	35	plasticity	plasticity	NOUN
easat-10905	468	36	analysis	analysis	NOUN
easat-10905	468	37	,	,	PUNCT
easat-10905	468	38	"	"	PUNCT
easat-10905	468	39	scientific	scientific	ADJ
easat-10905	468	40	reports	report	NOUN
easat-10905	468	41	,	,	PUNCT
easat-10905	468	42	vol	vol	NOUN
easat-10905	468	43	.	.	PROPN
easat-10905	468	44	15	15	NUM
easat-10905	468	45	,	,	PUNCT
easat-10905	468	46	p.	p.	NOUN
easat-10905	468	47	272	272	NUM
easat-10905	468	48	,	,	PUNCT
easat-10905	468	49	2025	2025	NUM
easat-10905	468	50	.	.	PUNCT
easat-10905	468	51	https://doi.org/10.1038/s41598-024-78785-6	https://doi.org/10.1038/s41598-024-78785-6	X
easat-10905	469	1	[	[	X
easat-10905	469	2	31	31	NUM
easat-10905	469	3	]	]	PUNCT
easat-10905	469	4	j.	j.	PROPN
easat-10905	469	5	d.	d.	PROPN
easat-10905	469	6	m.-w	m.-w	PROPN
easat-10905	469	7	.	.	PUNCT
easat-10905	470	1	c.	c.	PROPN
easat-10905	470	2	kenton	kenton	PROPN
easat-10905	470	3	and	and	CCONJ
easat-10905	470	4	l.	l.	PROPN
easat-10905	470	5	k.	k.	PROPN
easat-10905	470	6	toutanova	toutanova	PROPN
easat-10905	470	7	,	,	PUNCT
easat-10905	470	8	"	"	PUNCT
easat-10905	470	9	bert	bert	NOUN
easat-10905	470	10	:	:	PUNCT
easat-10905	470	11	pre	pre	ADJ
easat-10905	470	12	-	-	NOUN
easat-10905	470	13	training	training	NOUN
easat-10905	470	14	of	of	ADP
easat-10905	470	15	deep	deep	ADJ
easat-10905	470	16	bidirectional	bidirectional	ADJ
easat-10905	470	17	transformers	transformer	NOUN
easat-10905	470	18	for	for	ADP
easat-10905	470	19	language	language	NOUN
easat-10905	470	20	understanding	understanding	NOUN
easat-10905	470	21	,	,	PUNCT
easat-10905	470	22	"	"	PUNCT
easat-10905	470	23	in	in	ADP
easat-10905	470	24	proceedings	proceeding	NOUN
easat-10905	470	25	of	of	ADP
easat-10905	470	26	naacl	naacl	PROPN
easat-10905	470	27	-	-	PUNCT
easat-10905	470	28	hlt	hlt	NOUN
easat-10905	470	29	,	,	PUNCT
easat-10905	470	30	2019	2019	NUM
easat-10905	470	31	:	:	PUNCT
easat-10905	470	32	minneapolis	minneapolis	PROPN
easat-10905	470	33	,	,	PUNCT
easat-10905	470	34	minnesota	minnesota	PROPN
easat-10905	470	35	.	.	PUNCT
easat-10905	471	1	[	[	X
easat-10905	471	2	32	32	NUM
easat-10905	471	3	]	]	PUNCT
easat-10905	471	4	a.	a.	NOUN
easat-10905	471	5	vaswani	vaswani	PROPN
easat-10905	471	6	et	et	PROPN
easat-10905	471	7	al	al	PROPN
easat-10905	471	8	.	.	PROPN
easat-10905	471	9	,	,	PUNCT
easat-10905	471	10	"	"	PUNCT
easat-10905	471	11	attention	attention	NOUN
easat-10905	471	12	is	be	AUX
easat-10905	471	13	all	all	PRON
easat-10905	471	14	you	you	PRON
easat-10905	471	15	need	need	VERB
easat-10905	471	16	,	,	PUNCT
easat-10905	471	17	"	"	PUNCT
easat-10905	471	18	presented	present	VERB
easat-10905	471	19	at	at	ADP
easat-10905	471	20	the	the	DET
easat-10905	471	21	advances	advance	NOUN
easat-10905	471	22	in	in	ADP
easat-10905	471	23	neural	neural	ADJ
easat-10905	471	24	information	information	NOUN
easat-10905	471	25	processing	processing	NOUN
easat-10905	471	26	systems	system	NOUN
easat-10905	471	27	30	30	NUM
easat-10905	471	28	:	:	PUNCT
easat-10905	471	29	annual	annual	ADJ
easat-10905	471	30	conference	conference	NOUN
easat-10905	471	31	on	on	ADP
easat-10905	471	32	neural	neural	ADJ
easat-10905	471	33	information	information	NOUN
easat-10905	471	34	processing	processing	NOUN
easat-10905	471	35	systems	system	NOUN
easat-10905	471	36	(	(	PUNCT
easat-10905	471	37	neurips	neurip	NOUN
easat-10905	471	38	2017	2017	NUM
easat-10905	471	39	)	)	PUNCT
easat-10905	471	40	(	(	PUNCT
easat-10905	471	41	pp	pp	ADJ
easat-10905	471	42	.	.	PUNCT
easat-10905	472	1	5998–6008	5998–6008	NUM
easat-10905	472	2	)	)	PUNCT
easat-10905	472	3	.	.	PUNCT
easat-10905	473	1	long	long	ADJ
easat-10905	473	2	beach	beach	NOUN
easat-10905	473	3	,	,	PUNCT
easat-10905	473	4	ca	ca	PROPN
easat-10905	473	5	,	,	PUNCT
easat-10905	473	6	usa	usa	PROPN
easat-10905	473	7	,	,	PUNCT
easat-10905	473	8	2017	2017	NUM
easat-10905	473	9	.	.	PUNCT
easat-10905	474	1	[	[	X
easat-10905	474	2	33	33	NUM
easat-10905	474	3	]	]	X
easat-10905	474	4	v.	v.	X
easat-10905	474	5	sanh	sanh	PROPN
easat-10905	474	6	,	,	PUNCT
easat-10905	474	7	l.	l.	PROPN
easat-10905	474	8	debut	debut	PROPN
easat-10905	474	9	,	,	PUNCT
easat-10905	474	10	j.	j.	PROPN
easat-10905	474	11	chaumond	chaumond	PROPN
easat-10905	474	12	,	,	PUNCT
easat-10905	474	13	and	and	CCONJ
easat-10905	474	14	t.	t.	PROPN
easat-10905	474	15	wolf	wolf	PROPN
easat-10905	474	16	,	,	PUNCT
easat-10905	474	17	"	"	PUNCT
easat-10905	474	18	distilbert	distilbert	NOUN
easat-10905	474	19	,	,	PUNCT
easat-10905	474	20	a	a	DET
easat-10905	474	21	distilled	distil	VERB
easat-10905	474	22	version	version	NOUN
easat-10905	474	23	of	of	ADP
easat-10905	474	24	bert	bert	NOUN
easat-10905	474	25	:	:	PUNCT
easat-10905	474	26	smaller	small	ADJ
easat-10905	474	27	,	,	PUNCT
easat-10905	474	28	faster	fast	ADJ
easat-10905	474	29	,	,	PUNCT
easat-10905	474	30	cheaper	cheap	ADJ
easat-10905	474	31	and	and	CCONJ
easat-10905	474	32	lighter	light	ADJ
easat-10905	474	33	,	,	PUNCT
easat-10905	474	34	"	"	PUNCT
easat-10905	474	35	arxiv	arxiv	PROPN
easat-10905	474	36	preprint	preprint	NOUN
easat-10905	474	37	arxiv:1910.01108	arxiv:1910.01108	NOUN
easat-10905	474	38	,	,	PUNCT
easat-10905	474	39	2019	2019	NUM
easat-10905	474	40	.	.	PUNCT
easat-10905	475	1	https://doi.org/10.48550/arxiv.1910.01108	https://doi.org/10.48550/arxiv.1910.01108	PRON
easat-10905	476	1	[	[	X
easat-10905	476	2	34	34	NUM
easat-10905	476	3	]	]	PUNCT
easat-10905	476	4	z.	z.	PROPN
easat-10905	476	5	lan	lan	PROPN
easat-10905	476	6	,	,	PUNCT
easat-10905	476	7	m.	m.	PROPN
easat-10905	476	8	chen	chen	PROPN
easat-10905	476	9	,	,	PUNCT
easat-10905	476	10	s.	s.	PROPN
easat-10905	476	11	goodman	goodman	PROPN
easat-10905	476	12	,	,	PUNCT
easat-10905	476	13	k.	k.	PROPN
easat-10905	476	14	gimpel	gimpel	PROPN
easat-10905	476	15	,	,	PUNCT
easat-10905	476	16	p.	p.	PROPN
easat-10905	476	17	sharma	sharma	PROPN
easat-10905	476	18	,	,	PUNCT
easat-10905	476	19	and	and	CCONJ
easat-10905	476	20	r.	r.	PROPN
easat-10905	476	21	soricut	soricut	PROPN
easat-10905	476	22	,	,	PUNCT
easat-10905	476	23	"	"	PUNCT
easat-10905	476	24	albert	albert	NOUN
easat-10905	476	25	:	:	PUNCT
easat-10905	476	26	a	a	DET
easat-10905	476	27	lite	lite	ADJ
easat-10905	476	28	bert	bert	NOUN
easat-10905	476	29	for	for	ADP
easat-10905	476	30	self	self	NOUN
easat-10905	476	31	-	-	PUNCT
easat-10905	476	32	supervised	supervise	VERB
easat-10905	476	33	learning	learning	NOUN
easat-10905	476	34	of	of	ADP
easat-10905	476	35	language	language	NOUN
easat-10905	476	36	representations	representation	NOUN
easat-10905	476	37	,	,	PUNCT
easat-10905	476	38	"	"	PUNCT
easat-10905	476	39	arxiv	arxiv	PROPN
easat-10905	476	40	preprint	preprint	NOUN
easat-10905	476	41	arxiv:1909.11942	arxiv:1909.11942	NOUN
easat-10905	476	42	,	,	PUNCT
easat-10905	476	43	2019	2019	NUM
easat-10905	476	44	.	.	PUNCT
easat-10905	477	1	https://doi.org/10.48550/arxiv.1909.11942	https://doi.org/10.48550/arxiv.1909.11942	PROPN
easat-10905	478	1	[	[	X
easat-10905	478	2	35	35	NUM
easat-10905	478	3	]	]	X
easat-10905	478	4	y.	y.	PROPN
easat-10905	478	5	liu	liu	PROPN
easat-10905	478	6	et	et	PROPN
easat-10905	478	7	al	al	PROPN
easat-10905	478	8	.	.	PROPN
easat-10905	478	9	,	,	PUNCT
easat-10905	478	10	"	"	PUNCT
easat-10905	478	11	roberta	roberta	NOUN
easat-10905	478	12	:	:	PUNCT
easat-10905	478	13	a	a	DET
easat-10905	478	14	robustly	robustly	ADV
easat-10905	478	15	optimized	optimize	VERB
easat-10905	478	16	bert	bert	NOUN
easat-10905	478	17	pretraining	pretraine	VERB
easat-10905	478	18	approach	approach	NOUN
easat-10905	478	19	,	,	PUNCT
easat-10905	478	20	"	"	PUNCT
easat-10905	478	21	arxiv	arxiv	PROPN
easat-10905	478	22	preprint	preprint	NOUN
easat-10905	478	23	arxiv:1907.11692	arxiv:1907.11692	NOUN
easat-10905	478	24	,	,	PUNCT
easat-10905	478	25	2019	2019	NUM
easat-10905	478	26	.	.	PUNCT
easat-10905	479	1	https://doi.org/10.48550/arxiv.1907.11692	https://doi.org/10.48550/arxiv.1907.11692	PROPN
easat-10905	479	2	[	[	X
easat-10905	479	3	36	36	NUM
easat-10905	479	4	]	]	PUNCT
easat-10905	479	5	s.	s.	PROPN
easat-10905	479	6	liu	liu	PROPN
easat-10905	479	7	et	et	PROPN
easat-10905	479	8	al	al	PROPN
easat-10905	479	9	.	.	PROPN
easat-10905	479	10	,	,	PUNCT
easat-10905	479	11	"	"	PUNCT
easat-10905	479	12	a	a	DET
easat-10905	479	13	comprehensive	comprehensive	ADJ
easat-10905	479	14	review	review	NOUN
easat-10905	479	15	of	of	ADP
easat-10905	479	16	continual	continual	ADJ
easat-10905	479	17	learning	learning	NOUN
easat-10905	479	18	with	with	ADP
easat-10905	479	19	machine	machine	NOUN
easat-10905	479	20	learning	learning	NOUN
easat-10905	479	21	models	model	NOUN
easat-10905	479	22	,	,	PUNCT
easat-10905	479	23	"	"	PUNCT
easat-10905	479	24	in	in	ADP
easat-10905	479	25	international	international	ADJ
easat-10905	479	26	conference	conference	NOUN
easat-10905	479	27	on	on	ADP
easat-10905	479	28	image	image	NOUN
easat-10905	479	29	,	,	PUNCT
easat-10905	479	30	vision	vision	NOUN
easat-10905	479	31	and	and	CCONJ
easat-10905	479	32	intelligent	intelligent	ADJ
easat-10905	479	33	systems	system	NOUN
easat-10905	479	34	,	,	PUNCT
easat-10905	479	35	pp	pp	ADJ
easat-10905	479	36	.	.	PUNCT
easat-10905	480	1	504	504	NUM
easat-10905	480	2	-	-	SYM
easat-10905	480	3	512	512	NUM
easat-10905	480	4	.	.	PUNCT
easat-10905	481	1	singapore	singapore	PROPN
easat-10905	481	2	:	:	PUNCT
easat-10905	481	3	springer	springer	NOUN
easat-10905	481	4	nature	nature	PROPN
easat-10905	481	5	singapore	singapore	PROPN
easat-10905	481	6	,	,	PUNCT
easat-10905	481	7	2023	2023	NUM
easat-10905	481	8	.	.	PUNCT
easat-10905	482	1	[	[	X
easat-10905	482	2	37	37	NUM
easat-10905	482	3	]	]	PUNCT
easat-10905	482	4	h.-g	h.-g	PROPN
easat-10905	482	5	.	.	PUNCT
easat-10905	483	1	doan	doan	PROPN
easat-10905	483	2	,	,	PUNCT
easat-10905	483	3	h.-q	h.-q	PROPN
easat-10905	483	4	.	.	PUNCT
easat-10905	484	1	luong	luong	PROPN
easat-10905	484	2	,	,	PUNCT
easat-10905	484	3	t.-o	t.-o	NOUN
easat-10905	484	4	.	.	PUNCT
easat-10905	485	1	ha	ha	INTJ
easat-10905	485	2	,	,	PUNCT
easat-10905	485	3	and	and	CCONJ
easat-10905	485	4	t.	t.	PROPN
easat-10905	485	5	t.	t.	PROPN
easat-10905	485	6	t.	t.	PROPN
easat-10905	485	7	pham	pham	PROPN
easat-10905	485	8	,	,	PUNCT
easat-10905	485	9	"	"	PUNCT
easat-10905	485	10	an	an	DET
easat-10905	485	11	efficient	efficient	ADJ
easat-10905	485	12	strategy	strategy	NOUN
easat-10905	485	13	for	for	ADP
easat-10905	485	14	catastrophic	catastrophic	ADJ
easat-10905	485	15	forgetting	forgetting	NOUN
easat-10905	485	16	reduction	reduction	NOUN
easat-10905	485	17	in	in	ADP
easat-10905	485	18	incremental	incremental	ADJ
easat-10905	485	19	learning	learning	NOUN
easat-10905	485	20	,	,	PUNCT
easat-10905	485	21	"	"	PUNCT
easat-10905	485	22	electronics	electronic	NOUN
easat-10905	485	23	,	,	PUNCT
easat-10905	485	24	vol	vol	NOUN
easat-10905	485	25	.	.	PROPN
easat-10905	486	1	12	12	NUM
easat-10905	486	2	,	,	PUNCT
easat-10905	486	3	no	no	INTJ
easat-10905	486	4	.	.	NOUN
easat-10905	486	5	10	10	NUM
easat-10905	486	6	,	,	PUNCT
easat-10905	486	7	p.	p.	NOUN
easat-10905	486	8	2265	2265	NUM
easat-10905	486	9	,	,	PUNCT
easat-10905	486	10	2023	2023	NUM
easat-10905	486	11	.	.	PUNCT
easat-10905	487	1	https://doi.org/10.3390/electronics12102265	https://doi.org/10.3390/electronics12102265	NOUN
easat-10905	488	1	[	[	X
easat-10905	488	2	38	38	NUM
easat-10905	488	3	]	]	PUNCT
easat-10905	488	4	g.	g.	PROPN
easat-10905	488	5	m.	m.	PROPN
easat-10905	488	6	van	van	PROPN
easat-10905	488	7	de	de	X
easat-10905	488	8	ven	ven	PROPN
easat-10905	488	9	,	,	PUNCT
easat-10905	488	10	t.	t.	NOUN
easat-10905	488	11	tuytelaars	tuytelaar	NOUN
easat-10905	488	12	,	,	PUNCT
easat-10905	488	13	and	and	CCONJ
easat-10905	488	14	a.	a.	PROPN
easat-10905	488	15	s.	s.	PROPN
easat-10905	488	16	tolias	tolias	PROPN
easat-10905	488	17	,	,	PUNCT
easat-10905	488	18	"	"	PUNCT
easat-10905	488	19	three	three	NUM
easat-10905	488	20	types	type	NOUN
easat-10905	488	21	of	of	ADP
easat-10905	488	22	incremental	incremental	ADJ
easat-10905	488	23	learning	learning	NOUN
easat-10905	488	24	,	,	PUNCT
easat-10905	488	25	"	"	PUNCT
easat-10905	488	26	nature	nature	NOUN
easat-10905	488	27	machine	machine	NOUN
easat-10905	488	28	intelligence	intelligence	NOUN
easat-10905	488	29	,	,	PUNCT
easat-10905	488	30	vol	vol	NOUN
easat-10905	488	31	.	.	PROPN
easat-10905	488	32	4	4	NUM
easat-10905	488	33	,	,	PUNCT
easat-10905	488	34	pp	pp	ADJ
easat-10905	488	35	.	.	PUNCT
easat-10905	489	1	1185	1185	NUM
easat-10905	489	2	-	-	SYM
easat-10905	489	3	1197	1197	NUM
easat-10905	489	4	,	,	PUNCT
easat-10905	489	5	2022	2022	NUM
easat-10905	489	6	.	.	PUNCT
easat-10905	490	1	https://doi.org/10.1038/s42256-022-00568-3	https://doi.org/10.1038/s42256-022-00568-3	NUM
easat-10905	491	1	[	[	X
easat-10905	491	2	39	39	NUM
easat-10905	491	3	]	]	PUNCT
easat-10905	491	4	q.	q.	PROPN
easat-10905	491	5	gao	gao	PROPN
easat-10905	491	6	,	,	PUNCT
easat-10905	491	7	x.	x.	PROPN
easat-10905	491	8	shan	shan	PROPN
easat-10905	491	9	,	,	PUNCT
easat-10905	491	10	y.	y.	PROPN
easat-10905	491	11	zhang	zhang	PROPN
easat-10905	491	12	,	,	PUNCT
easat-10905	491	13	and	and	CCONJ
easat-10905	491	14	f.	f.	PROPN
easat-10905	491	15	zhou	zhou	PROPN
easat-10905	491	16	,	,	PUNCT
easat-10905	491	17	"	"	PUNCT
easat-10905	491	18	enhancing	enhance	VERB
easat-10905	491	19	knowledge	knowledge	NOUN
easat-10905	491	20	transfer	transfer	NOUN
easat-10905	491	21	for	for	ADP
easat-10905	491	22	task	task	NOUN
easat-10905	491	23	incremental	incremental	ADJ
easat-10905	491	24	learning	learning	NOUN
easat-10905	491	25	with	with	ADP
easat-10905	491	26	data	data	NOUN
easat-10905	491	27	-	-	PUNCT
easat-10905	491	28	free	free	ADJ
easat-10905	491	29	subnetwork	subnetwork	NOUN
easat-10905	491	30	,	,	PUNCT
easat-10905	491	31	"	"	PUNCT
easat-10905	491	32	advances	advance	NOUN
easat-10905	491	33	in	in	ADP
easat-10905	491	34	neural	neural	ADJ
easat-10905	491	35	information	information	NOUN
easat-10905	491	36	processing	processing	NOUN
easat-10905	491	37	systems	system	NOUN
easat-10905	491	38	,	,	PUNCT
easat-10905	491	39	vol	vol	NOUN
easat-10905	491	40	.	.	PROPN
easat-10905	491	41	36	36	NUM
easat-10905	491	42	,	,	PUNCT
easat-10905	491	43	pp	pp	ADJ
easat-10905	491	44	.	.	PUNCT
easat-10905	491	45	68471	68471	NUM
easat-10905	491	46	-	-	SYM
easat-10905	491	47	68484	68484	NUM
easat-10905	491	48	,	,	PUNCT
easat-10905	491	49	2023	2023	NUM
easat-10905	491	50	.	.	PUNCT
easat-10905	492	1	[	[	X
easat-10905	492	2	40	40	NUM
easat-10905	492	3	]	]	PUNCT
easat-10905	492	4	s.	s.	PROPN
easat-10905	492	5	tian	tian	PROPN
easat-10905	492	6	,	,	PUNCT
easat-10905	492	7	l.	l.	PROPN
easat-10905	492	8	li	li	PROPN
easat-10905	492	9	,	,	PUNCT
easat-10905	492	10	w.	w.	PROPN
easat-10905	492	11	li	li	PROPN
easat-10905	492	12	,	,	PUNCT
easat-10905	492	13	h.	h.	PROPN
easat-10905	492	14	ran	run	VERB
easat-10905	492	15	,	,	PUNCT
easat-10905	492	16	x.	x.	NOUN
easat-10905	492	17	ning	ning	PROPN
easat-10905	492	18	,	,	PUNCT
easat-10905	492	19	and	and	CCONJ
easat-10905	492	20	p.	p.	PROPN
easat-10905	492	21	tiwari	tiwari	PROPN
easat-10905	492	22	,	,	PUNCT
easat-10905	492	23	"	"	PUNCT
easat-10905	492	24	a	a	DET
easat-10905	492	25	survey	survey	NOUN
easat-10905	492	26	on	on	ADP
easat-10905	492	27	few	few	ADJ
easat-10905	492	28	-	-	PUNCT
easat-10905	492	29	shot	shot	NOUN
easat-10905	492	30	class	class	NOUN
easat-10905	492	31	-	-	PUNCT
easat-10905	492	32	incremental	incremental	NOUN
easat-10905	492	33	learning	learning	NOUN
easat-10905	492	34	,	,	PUNCT
easat-10905	492	35	"	"	PUNCT
easat-10905	492	36	neural	neural	ADJ
easat-10905	492	37	networks	network	NOUN
easat-10905	492	38	,	,	PUNCT
easat-10905	492	39	vol	vol	NOUN
easat-10905	492	40	.	.	PROPN
easat-10905	492	41	169	169	NUM
easat-10905	492	42	,	,	PUNCT
easat-10905	492	43	pp	pp	ADJ
easat-10905	492	44	.	.	PUNCT
easat-10905	493	1	307	307	NUM
easat-10905	493	2	-	-	SYM
easat-10905	493	3	324	324	NUM
easat-10905	493	4	,	,	PUNCT
easat-10905	493	5	2024	2024	NUM
easat-10905	493	6	.	.	PUNCT
easat-10905	494	1	https://doi.org/10.1016/j.neunet.2023.10.039	https://doi.org/10.1016/j.neunet.2023.10.039	NOUN
easat-10905	494	2	[	[	X
easat-10905	494	3	41	41	NUM
easat-10905	494	4	]	]	X
easat-10905	494	5	h.	h.	PROPN
easat-10905	494	6	shi	shi	PROPN
easat-10905	494	7	and	and	CCONJ
easat-10905	494	8	h.	h.	PROPN
easat-10905	494	9	wang	wang	PROPN
easat-10905	494	10	,	,	PUNCT
easat-10905	494	11	"	"	PUNCT
easat-10905	494	12	a	a	DET
easat-10905	494	13	unified	unified	ADJ
easat-10905	494	14	approach	approach	NOUN
easat-10905	494	15	to	to	AUX
easat-10905	494	16	domain	domain	VERB
easat-10905	494	17	incremental	incremental	ADJ
easat-10905	494	18	learning	learning	NOUN
easat-10905	494	19	with	with	ADP
easat-10905	494	20	memory	memory	NOUN
easat-10905	494	21	:	:	PUNCT
easat-10905	494	22	theory	theory	NOUN
easat-10905	494	23	and	and	CCONJ
easat-10905	494	24	algorithm	algorithm	NOUN
easat-10905	494	25	,	,	PUNCT
easat-10905	494	26	"	"	PUNCT
easat-10905	494	27	advances	advance	NOUN
easat-10905	494	28	in	in	ADP
easat-10905	494	29	neural	neural	ADJ
easat-10905	494	30	information	information	NOUN
easat-10905	494	31	processing	processing	NOUN
easat-10905	494	32	systems	system	NOUN
easat-10905	494	33	,	,	PUNCT
easat-10905	494	34	vol	vol	NOUN
easat-10905	494	35	.	.	PROPN
easat-10905	495	1	36	36	NUM
easat-10905	495	2	,	,	PUNCT
easat-10905	495	3	pp	pp	ADJ
easat-10905	495	4	.	.	PUNCT
easat-10905	496	1	15027	15027	NUM
easat-10905	496	2	-	-	SYM
easat-10905	496	3	15059	15059	NUM
easat-10905	496	4	,	,	PUNCT
easat-10905	496	5	2023	2023	NUM
easat-10905	496	6	.	.	PUNCT
easat-10905	497	1	[	[	X
easat-10905	497	2	42	42	NUM
easat-10905	497	3	]	]	X
easat-10905	497	4	e.	e.	PROPN
easat-10905	497	5	l.	l.	PROPN
easat-10905	497	6	aleixo	aleixo	PROPN
easat-10905	497	7	,	,	PUNCT
easat-10905	497	8	j.	j.	PROPN
easat-10905	497	9	g.	g.	PROPN
easat-10905	497	10	colonna	colonna	PROPN
easat-10905	497	11	,	,	PUNCT
easat-10905	497	12	m.	m.	NOUN
easat-10905	497	13	cristo	cristo	PROPN
easat-10905	497	14	,	,	PUNCT
easat-10905	497	15	and	and	CCONJ
easat-10905	497	16	e.	e.	PROPN
easat-10905	497	17	fernandes	fernandes	PROPN
easat-10905	497	18	,	,	PUNCT
easat-10905	497	19	"	"	PUNCT
easat-10905	497	20	catastrophic	catastrophic	ADJ
easat-10905	497	21	forgetting	forgetting	NOUN
easat-10905	497	22	in	in	ADP
easat-10905	497	23	deep	deep	ADJ
easat-10905	497	24	learning	learning	NOUN
easat-10905	497	25	:	:	PUNCT
easat-10905	497	26	a	a	DET
easat-10905	497	27	comprehensive	comprehensive	ADJ
easat-10905	497	28	taxonomy	taxonomy	NOUN
easat-10905	497	29	,	,	PUNCT
easat-10905	497	30	"	"	PUNCT
easat-10905	497	31	arxiv	arxiv	PROPN
easat-10905	497	32	preprint	preprint	VERB
easat-10905	497	33	arxiv:2312.10549	arxiv:2312.10549	NOUN
easat-10905	497	34	,	,	PUNCT
easat-10905	497	35	2023	2023	NUM
easat-10905	497	36	.	.	PUNCT
easat-10905	498	1	https://doi.org/10.48550/arxiv.2312.10549	https://doi.org/10.48550/arxiv.2312.10549	PROPN
easat-10905	498	2	[	[	X
easat-10905	498	3	43	43	NUM
easat-10905	498	4	]	]	PUNCT
easat-10905	498	5	j.	j.	PROPN
easat-10905	498	6	huang	huang	PROPN
easat-10905	498	7	et	et	PROPN
easat-10905	498	8	al	al	PROPN
easat-10905	498	9	.	.	PROPN
easat-10905	498	10	,	,	PUNCT
easat-10905	498	11	"	"	PUNCT
easat-10905	498	12	mitigating	mitigate	VERB
easat-10905	498	13	catastrophic	catastrophic	ADJ
easat-10905	498	14	forgetting	forgetting	NOUN
easat-10905	498	15	in	in	ADP
easat-10905	498	16	large	large	ADJ
easat-10905	498	17	language	language	NOUN
easat-10905	498	18	models	model	NOUN
easat-10905	498	19	with	with	ADP
easat-10905	498	20	self	self	NOUN
easat-10905	498	21	-	-	PUNCT
easat-10905	498	22	synthesized	synthesize	VERB
easat-10905	498	23	rehearsal	rehearsal	NOUN
easat-10905	498	24	,	,	PUNCT
easat-10905	498	25	"	"	PUNCT
easat-10905	498	26	arxiv	arxiv	PROPN
easat-10905	498	27	preprint	preprint	NOUN
easat-10905	498	28	arxiv:2403.01244	arxiv:2403.01244	NOUN
easat-10905	498	29	,	,	PUNCT
easat-10905	498	30	2024	2024	NUM
easat-10905	498	31	.	.	PUNCT
easat-10905	499	1	https://doi.org/10.48550/arxiv.2403.01244	https://doi.org/10.48550/arxiv.2403.01244	PROPN
easat-10905	499	2	https://doi.org/10.18653/v1/2020.coling-main.574	https://doi.org/10.18653/v1/2020.coling-main.574	VERB
easat-10905	500	1	https://doi.org/10.48550/arxiv.2106.02821	https://doi.org/10.48550/arxiv.2106.02821	PROPN
easat-10905	500	2	https://doi.org/10.48550/arxiv.1801.02808	https://doi.org/10.48550/arxiv.1801.02808	PROPN
easat-10905	500	3	https://doi.org/10.1038/s41598-024-78785-6	https://doi.org/10.1038/s41598-024-78785-6	NOUN
easat-10905	501	1	https://doi.org/10.48550/arxiv.1910.01108	https://doi.org/10.48550/arxiv.1910.01108	PUNCT
easat-10905	501	2	https://doi.org/10.48550/arxiv.1909.11942	https://doi.org/10.48550/arxiv.1909.11942	PROPN
easat-10905	501	3	https://doi.org/10.48550/arxiv.1907.11692	https://doi.org/10.48550/arxiv.1907.11692	PROPN
easat-10905	501	4	https://doi.org/10.3390/electronics12102265	https://doi.org/10.3390/electronics12102265	VERB
easat-10905	501	5	https://doi.org/10.1038/s42256-022-00568-3	https://doi.org/10.1038/s42256-022-00568-3	NUM
easat-10905	501	6	https://doi.org/10.1016/j.neunet.2023.10.039	https://doi.org/10.1016/j.neunet.2023.10.039	NOUN
easat-10905	501	7	https://doi.org/10.48550/arxiv.2312.10549	https://doi.org/10.48550/arxiv.2312.10549	PROPN
easat-10905	501	8	https://doi.org/10.48550/arxiv.2403.01244	https://doi.org/10.48550/arxiv.2403.01244	PROPN