id	sid	tid	token	lemma	pos
fcis-5214	1	1	frontiers	frontier	NOUN
fcis-5214	1	2	in	in	ADP
fcis-5214	1	3	computing	computing	NOUN
fcis-5214	1	4	and	and	CCONJ
fcis-5214	1	5	intelligent	intelligent	ADJ
fcis-5214	1	6	systems	system	NOUN
fcis-5214	1	7	issn	issn	VERB
fcis-5214	1	8	:	:	PUNCT
fcis-5214	1	9	2832	2832	NUM
fcis-5214	1	10	-	-	SYM
fcis-5214	1	11	6024	6024	NUM
fcis-5214	1	12	|	|	NOUN
fcis-5214	1	13	vol	vol	NOUN
fcis-5214	1	14	.	.	PROPN
fcis-5214	2	1	2	2	NUM
fcis-5214	2	2	,	,	PUNCT
fcis-5214	2	3	no	no	INTJ
fcis-5214	2	4	.	.	NOUN
fcis-5214	2	5	3	3	NUM
fcis-5214	2	6	,	,	PUNCT
fcis-5214	2	7	2022	2022	NUM
fcis-5214	2	8	63	63	NUM
fcis-5214	2	9	communication	communication	NOUN
fcis-5214	2	10	efficient	efficient	ADJ
fcis-5214	2	11	federated	federated	ADJ
fcis-5214	2	12	personalized	personalize	VERB
fcis-5214	2	13	recommendation	recommendation	NOUN
fcis-5214	2	14	lingtao	lingtao	PROPN
fcis-5214	2	15	wei	wei	PROPN
fcis-5214	2	16	*	*	PUNCT
fcis-5214	2	17	college	college	PROPN
fcis-5214	2	18	of	of	ADP
fcis-5214	2	19	computer	computer	NOUN
fcis-5214	2	20	science	science	PROPN
fcis-5214	2	21	&	&	CCONJ
fcis-5214	2	22	technology	technology	PROPN
fcis-5214	2	23	,	,	PUNCT
fcis-5214	2	24	qingdao	qingdao	PROPN
fcis-5214	2	25	university	university	PROPN
fcis-5214	2	26	,	,	PUNCT
fcis-5214	2	27	qingdao	qingdao	PROPN
fcis-5214	2	28	shandong	shandong	PROPN
fcis-5214	2	29	,	,	PUNCT
fcis-5214	2	30	china	china	PROPN
fcis-5214	2	31	*	*	PUNCT
fcis-5214	2	32	corresponding	correspond	VERB
fcis-5214	2	33	author	author	NOUN
fcis-5214	2	34	:	:	PUNCT
fcis-5214	2	35	email	email	NOUN
fcis-5214	2	36	:	:	PUNCT
fcis-5214	2	37	qduwlt@163.com	qduwlt@163.com	PROPN
fcis-5214	2	38	abstract	abstract	NOUN
fcis-5214	2	39	:	:	PUNCT
fcis-5214	2	40	recommendation	recommendation	NOUN
fcis-5214	2	41	systems	system	NOUN
fcis-5214	2	42	that	that	PRON
fcis-5214	2	43	can	can	AUX
fcis-5214	2	44	correctly	correctly	ADV
fcis-5214	2	45	predict	predict	VERB
fcis-5214	2	46	user	user	NOUN
fcis-5214	2	47	preferences	preference	NOUN
fcis-5214	2	48	in	in	ADP
fcis-5214	2	49	the	the	DET
fcis-5214	2	50	information	information	NOUN
fcis-5214	2	51	age	age	NOUN
fcis-5214	2	52	have	have	AUX
fcis-5214	2	53	become	become	VERB
fcis-5214	2	54	an	an	DET
fcis-5214	2	55	important	important	ADJ
fcis-5214	2	56	factor	factor	NOUN
fcis-5214	2	57	for	for	ADP
fcis-5214	2	58	business	business	NOUN
fcis-5214	2	59	success	success	NOUN
fcis-5214	2	60	.	.	PUNCT
fcis-5214	3	1	however	however	ADV
fcis-5214	3	2	,	,	PUNCT
fcis-5214	3	3	recommendation	recommendation	NOUN
fcis-5214	3	4	systems	system	NOUN
fcis-5214	3	5	require	require	VERB
fcis-5214	3	6	users	user	NOUN
fcis-5214	3	7	'	'	PART
fcis-5214	3	8	personal	personal	ADJ
fcis-5214	3	9	information	information	NOUN
fcis-5214	3	10	,	,	PUNCT
fcis-5214	3	11	and	and	CCONJ
fcis-5214	3	12	centralized	centralized	ADJ
fcis-5214	3	13	collection	collection	NOUN
fcis-5214	3	14	and	and	CCONJ
fcis-5214	3	15	processing	processing	NOUN
fcis-5214	3	16	of	of	ADP
fcis-5214	3	17	user	user	NOUN
fcis-5214	3	18	data	datum	NOUN
fcis-5214	3	19	may	may	AUX
fcis-5214	3	20	lead	lead	VERB
fcis-5214	3	21	to	to	ADP
fcis-5214	3	22	serious	serious	ADJ
fcis-5214	3	23	privacy	privacy	NOUN
fcis-5214	3	24	risks	risk	NOUN
fcis-5214	3	25	.	.	PUNCT
fcis-5214	4	1	good	good	ADJ
fcis-5214	4	2	progress	progress	NOUN
fcis-5214	4	3	has	have	AUX
fcis-5214	4	4	been	be	AUX
fcis-5214	4	5	made	make	VERB
fcis-5214	4	6	in	in	ADP
fcis-5214	4	7	recent	recent	ADJ
fcis-5214	4	8	years	year	NOUN
fcis-5214	4	9	using	use	VERB
fcis-5214	4	10	federated	federated	ADJ
fcis-5214	4	11	learning	learning	NOUN
fcis-5214	4	12	techniques	technique	NOUN
fcis-5214	4	13	for	for	ADP
fcis-5214	4	14	privacy	privacy	NOUN
fcis-5214	4	15	-	-	PUNCT
fcis-5214	4	16	preserving	preserve	VERB
fcis-5214	4	17	recommendations	recommendation	NOUN
fcis-5214	4	18	,	,	PUNCT
fcis-5214	4	19	but	but	CCONJ
fcis-5214	4	20	several	several	ADJ
fcis-5214	4	21	key	key	ADJ
fcis-5214	4	22	challenges	challenge	NOUN
fcis-5214	4	23	remain	remain	VERB
fcis-5214	4	24	to	to	PART
fcis-5214	4	25	be	be	AUX
fcis-5214	4	26	addressed	address	VERB
fcis-5214	4	27	:	:	PUNCT
fcis-5214	4	28	most	most	ADV
fcis-5214	4	29	federated	federated	ADJ
fcis-5214	4	30	recommender	recommender	NOUN
fcis-5214	4	31	systems	system	NOUN
fcis-5214	4	32	ignore	ignore	VERB
fcis-5214	4	33	communication	communication	NOUN
fcis-5214	4	34	process	process	NOUN
fcis-5214	4	35	optimization	optimization	NOUN
fcis-5214	4	36	,	,	PUNCT
fcis-5214	4	37	inequities	inequity	NOUN
fcis-5214	4	38	in	in	ADP
fcis-5214	4	39	aggregation	aggregation	NOUN
fcis-5214	4	40	of	of	ADP
fcis-5214	4	41	federated	federated	ADJ
fcis-5214	4	42	models	model	NOUN
fcis-5214	4	43	,	,	PUNCT
fcis-5214	4	44	and	and	CCONJ
fcis-5214	4	45	lack	lack	NOUN
fcis-5214	4	46	of	of	ADP
fcis-5214	4	47	personalization	personalization	NOUN
fcis-5214	4	48	to	to	ADP
fcis-5214	4	49	users	user	NOUN
fcis-5214	4	50	.	.	PUNCT
fcis-5214	5	1	in	in	ADP
fcis-5214	5	2	this	this	DET
fcis-5214	5	3	paper	paper	NOUN
fcis-5214	5	4	,	,	PUNCT
fcis-5214	5	5	we	we	PRON
fcis-5214	5	6	propose	propose	VERB
fcis-5214	5	7	a	a	DET
fcis-5214	5	8	communication	communication	NOUN
fcis-5214	5	9	efficient	efficient	ADJ
fcis-5214	5	10	and	and	CCONJ
fcis-5214	5	11	fair	fair	ADJ
fcis-5214	5	12	personalized	personalized	ADJ
fcis-5214	5	13	federated	federated	ADJ
fcis-5214	5	14	recommendation	recommendation	NOUN
fcis-5214	5	15	approach	approach	NOUN
fcis-5214	5	16	(	(	PUNCT
fcis-5214	5	17	cffr	cffr	NOUN
fcis-5214	5	18	)	)	PUNCT
fcis-5214	5	19	to	to	PART
fcis-5214	5	20	address	address	VERB
fcis-5214	5	21	these	these	DET
fcis-5214	5	22	challenges	challenge	NOUN
fcis-5214	5	23	.	.	PUNCT
fcis-5214	6	1	cffr	cffr	PROPN
fcis-5214	6	2	uses	use	VERB
fcis-5214	6	3	adaptive	adaptive	ADJ
fcis-5214	6	4	client	client	NOUN
fcis-5214	6	5	group	group	NOUN
fcis-5214	6	6	selection	selection	NOUN
fcis-5214	6	7	to	to	PART
fcis-5214	6	8	personalize	personalize	VERB
fcis-5214	6	9	models	model	NOUN
fcis-5214	6	10	while	while	SCONJ
fcis-5214	6	11	accelerating	accelerate	VERB
fcis-5214	6	12	the	the	DET
fcis-5214	6	13	training	training	NOUN
fcis-5214	6	14	process	process	NOUN
fcis-5214	6	15	.	.	PUNCT
fcis-5214	7	1	a	a	DET
fcis-5214	7	2	fair	fair	ADJ
fcis-5214	7	3	-	-	PUNCT
fcis-5214	7	4	aware	aware	ADJ
fcis-5214	7	5	model	model	NOUN
fcis-5214	7	6	aggregation	aggregation	NOUN
fcis-5214	7	7	algorithm	algorithm	NOUN
fcis-5214	7	8	is	be	AUX
fcis-5214	7	9	proposed	propose	VERB
fcis-5214	7	10	that	that	SCONJ
fcis-5214	7	11	adaptively	adaptively	ADV
fcis-5214	7	12	captures	capture	VERB
fcis-5214	7	13	the	the	DET
fcis-5214	7	14	performance	performance	NOUN
fcis-5214	7	15	and	and	CCONJ
fcis-5214	7	16	data	datum	NOUN
fcis-5214	7	17	imbalance	imbalance	NOUN
fcis-5214	7	18	among	among	ADP
fcis-5214	7	19	different	different	ADJ
fcis-5214	7	20	clients	client	NOUN
fcis-5214	7	21	to	to	PART
fcis-5214	7	22	address	address	VERB
fcis-5214	7	23	the	the	DET
fcis-5214	7	24	unfairness	unfairness	ADJ
fcis-5214	7	25	problem	problem	NOUN
fcis-5214	7	26	.	.	PUNCT
fcis-5214	8	1	extensive	extensive	ADJ
fcis-5214	8	2	experimental	experimental	ADJ
fcis-5214	8	3	results	result	NOUN
fcis-5214	8	4	demonstrate	demonstrate	VERB
fcis-5214	8	5	the	the	DET
fcis-5214	8	6	effectiveness	effectiveness	NOUN
fcis-5214	8	7	and	and	CCONJ
fcis-5214	8	8	efficiency	efficiency	NOUN
fcis-5214	8	9	of	of	ADP
fcis-5214	8	10	our	our	PRON
fcis-5214	8	11	proposed	propose	VERB
fcis-5214	8	12	method	method	NOUN
fcis-5214	8	13	.	.	PUNCT
fcis-5214	9	1	keywords	keyword	NOUN
fcis-5214	9	2	:	:	PUNCT
fcis-5214	9	3	recommendation	recommendation	NOUN
fcis-5214	9	4	;	;	PUNCT
fcis-5214	9	5	federated	federated	ADJ
fcis-5214	9	6	learning	learning	NOUN
fcis-5214	9	7	;	;	PUNCT
fcis-5214	9	8	grouping	group	VERB
fcis-5214	9	9	.	.	PUNCT
fcis-5214	10	1	1	1	X
fcis-5214	10	2	.	.	X
fcis-5214	10	3	introduction	introduction	NOUN
fcis-5214	10	4	users	user	NOUN
fcis-5214	10	5	in	in	ADP
fcis-5214	10	6	the	the	DET
fcis-5214	10	7	digital	digital	ADJ
fcis-5214	10	8	age	age	NOUN
fcis-5214	10	9	are	be	AUX
fcis-5214	10	10	overwhelmed	overwhelm	VERB
fcis-5214	10	11	by	by	ADP
fcis-5214	10	12	the	the	DET
fcis-5214	10	13	vast	vast	ADJ
fcis-5214	10	14	amount	amount	NOUN
fcis-5214	10	15	of	of	ADP
fcis-5214	10	16	information	information	NOUN
fcis-5214	10	17	on	on	ADP
fcis-5214	10	18	the	the	DET
fcis-5214	10	19	internet	internet	NOUN
fcis-5214	10	20	and	and	CCONJ
fcis-5214	10	21	have	have	VERB
fcis-5214	10	22	difficulty	difficulty	NOUN
fcis-5214	10	23	getting	get	VERB
fcis-5214	10	24	the	the	DET
fcis-5214	10	25	information	information	NOUN
fcis-5214	10	26	they	they	PRON
fcis-5214	10	27	need	need	VERB
fcis-5214	10	28	.	.	PUNCT
fcis-5214	11	1	recommender	recommender	NOUN
fcis-5214	11	2	systems	system	NOUN
fcis-5214	11	3	(	(	PUNCT
fcis-5214	11	4	rs	rs	NOUN
fcis-5214	11	5	)	)	PUNCT
fcis-5214	11	6	help	help	VERB
fcis-5214	11	7	users	user	NOUN
fcis-5214	11	8	discover	discover	VERB
fcis-5214	11	9	the	the	DET
fcis-5214	11	10	most	most	ADV
fcis-5214	11	11	useful	useful	ADJ
fcis-5214	11	12	information	information	NOUN
fcis-5214	11	13	or	or	CCONJ
fcis-5214	11	14	services	service	NOUN
fcis-5214	11	15	at	at	ADP
fcis-5214	11	16	the	the	DET
fcis-5214	11	17	right	right	ADJ
fcis-5214	11	18	time	time	NOUN
fcis-5214	11	19	and	and	CCONJ
fcis-5214	11	20	in	in	ADP
fcis-5214	11	21	the	the	DET
fcis-5214	11	22	most	most	ADV
fcis-5214	11	23	appropriate	appropriate	ADJ
fcis-5214	11	24	way	way	NOUN
fcis-5214	11	25	.	.	PUNCT
fcis-5214	12	1	a	a	DET
fcis-5214	12	2	recommendation	recommendation	NOUN
fcis-5214	12	3	system	system	NOUN
fcis-5214	12	4	that	that	PRON
fcis-5214	12	5	can	can	AUX
fcis-5214	12	6	correctly	correctly	ADV
fcis-5214	12	7	predict	predict	VERB
fcis-5214	12	8	user	user	NOUN
fcis-5214	12	9	preferences	preference	NOUN
fcis-5214	12	10	becomes	become	VERB
fcis-5214	12	11	a	a	DET
fcis-5214	12	12	key	key	ADJ
fcis-5214	12	13	element	element	NOUN
fcis-5214	12	14	of	of	ADP
fcis-5214	12	15	business	business	NOUN
fcis-5214	12	16	success	success	NOUN
fcis-5214	12	17	.	.	PUNCT
fcis-5214	13	1	recommender	recommender	NOUN
fcis-5214	13	2	systems	system	NOUN
fcis-5214	13	3	need	need	VERB
fcis-5214	13	4	to	to	PART
fcis-5214	13	5	collect	collect	VERB
fcis-5214	13	6	user	user	NOUN
fcis-5214	13	7	data	datum	NOUN
fcis-5214	13	8	to	to	PART
fcis-5214	13	9	perform	perform	VERB
fcis-5214	13	10	analytical	analytical	ADJ
fcis-5214	13	11	modeling	modeling	NOUN
fcis-5214	13	12	to	to	PART
fcis-5214	13	13	analyze	analyze	VERB
fcis-5214	13	14	user	user	NOUN
fcis-5214	13	15	preferences	preference	NOUN
fcis-5214	13	16	in	in	ADP
fcis-5214	13	17	order	order	NOUN
fcis-5214	13	18	to	to	PART
fcis-5214	13	19	provide	provide	VERB
fcis-5214	13	20	services	service	NOUN
fcis-5214	13	21	to	to	ADP
fcis-5214	13	22	users	user	NOUN
fcis-5214	13	23	.	.	PUNCT
fcis-5214	14	1	however	however	ADV
fcis-5214	14	2	,	,	PUNCT
fcis-5214	14	3	with	with	ADP
fcis-5214	14	4	introduction	introduction	NOUN
fcis-5214	14	5	of	of	ADP
fcis-5214	14	6	data	datum	NOUN
fcis-5214	14	7	privacy	privacy	NOUN
fcis-5214	14	8	laws	law	NOUN
fcis-5214	14	9	,	,	PUNCT
fcis-5214	14	10	increased	increase	VERB
fcis-5214	14	11	regulation	regulation	NOUN
fcis-5214	14	12	,	,	PUNCT
fcis-5214	14	13	and	and	CCONJ
fcis-5214	14	14	growing	grow	VERB
fcis-5214	14	15	public	public	ADJ
fcis-5214	14	16	unease	unease	NOUN
fcis-5214	14	17	about	about	ADP
fcis-5214	14	18	how	how	SCONJ
fcis-5214	14	19	personal	personal	ADJ
fcis-5214	14	20	data	data	NOUN
fcis-5214	14	21	is	be	AUX
fcis-5214	14	22	collected	collect	VERB
fcis-5214	14	23	and	and	CCONJ
fcis-5214	14	24	used	use	VERB
fcis-5214	14	25	,	,	PUNCT
fcis-5214	14	26	it	it	PRON
fcis-5214	14	27	has	have	AUX
fcis-5214	14	28	become	become	VERB
fcis-5214	14	29	increasingly	increasingly	ADV
fcis-5214	14	30	important	important	ADJ
fcis-5214	14	31	for	for	SCONJ
fcis-5214	14	32	recommendation	recommendation	NOUN
fcis-5214	14	33	systems	system	NOUN
fcis-5214	14	34	to	to	PART
fcis-5214	14	35	find	find	VERB
fcis-5214	14	36	a	a	DET
fcis-5214	14	37	better	well	ADJ
fcis-5214	14	38	balance	balance	NOUN
fcis-5214	14	39	between	between	ADP
fcis-5214	14	40	personalization	personalization	NOUN
fcis-5214	14	41	and	and	CCONJ
fcis-5214	14	42	privacy	privacy	NOUN
fcis-5214	14	43	.	.	PUNCT
fcis-5214	15	1	to	to	PART
fcis-5214	15	2	address	address	VERB
fcis-5214	15	3	this	this	DET
fcis-5214	15	4	issue	issue	NOUN
fcis-5214	15	5	,	,	PUNCT
fcis-5214	15	6	many	many	ADJ
fcis-5214	15	7	studies	study	NOUN
fcis-5214	15	8	on	on	ADP
fcis-5214	15	9	the	the	DET
fcis-5214	15	10	protection	protection	NOUN
fcis-5214	15	11	of	of	ADP
fcis-5214	15	12	personal	personal	ADJ
fcis-5214	15	13	information	information	NOUN
fcis-5214	15	14	in	in	ADP
fcis-5214	15	15	recommender	recommender	NOUN
fcis-5214	15	16	systems	system	NOUN
fcis-5214	15	17	have	have	AUX
fcis-5214	15	18	been	be	AUX
fcis-5214	15	19	conducted	conduct	VERB
fcis-5214	15	20	,	,	PUNCT
fcis-5214	15	21	and	and	CCONJ
fcis-5214	15	22	good	good	ADJ
fcis-5214	15	23	progress	progress	NOUN
fcis-5214	15	24	has	have	AUX
fcis-5214	15	25	been	be	AUX
fcis-5214	15	26	made	make	VERB
fcis-5214	15	27	in	in	ADP
fcis-5214	15	28	recent	recent	ADJ
fcis-5214	15	29	years	year	NOUN
fcis-5214	15	30	on	on	ADP
fcis-5214	15	31	privacy	privacy	NOUN
fcis-5214	15	32	-	-	PUNCT
fcis-5214	15	33	preserving	preserve	VERB
fcis-5214	15	34	recommendations	recommendation	NOUN
fcis-5214	15	35	using	use	VERB
fcis-5214	15	36	federated	federated	ADJ
fcis-5214	15	37	learning	learning	NOUN
fcis-5214	15	38	techniques	technique	NOUN
fcis-5214	15	39	,	,	PUNCT
fcis-5214	15	40	but	but	CCONJ
fcis-5214	15	41	several	several	ADJ
fcis-5214	15	42	key	key	ADJ
fcis-5214	15	43	challenges	challenge	NOUN
fcis-5214	15	44	remain	remain	VERB
fcis-5214	15	45	to	to	PART
fcis-5214	15	46	be	be	AUX
fcis-5214	15	47	addressed	address	VERB
fcis-5214	15	48	:	:	PUNCT
fcis-5214	15	49	most	most	ADV
fcis-5214	15	50	federated	federated	ADJ
fcis-5214	15	51	recommender	recommender	NOUN
fcis-5214	15	52	systems	system	NOUN
fcis-5214	15	53	only	only	ADV
fcis-5214	15	54	consider	consider	VERB
fcis-5214	15	55	model	model	NOUN
fcis-5214	15	56	performance	performance	NOUN
fcis-5214	15	57	and	and	CCONJ
fcis-5214	15	58	privacy	privacy	NOUN
fcis-5214	15	59	-	-	PUNCT
fcis-5214	15	60	preserving	preserve	VERB
fcis-5214	15	61	capabilities	capability	NOUN
fcis-5214	15	62	and	and	CCONJ
fcis-5214	15	63	ignore	ignore	VERB
fcis-5214	15	64	communication	communication	NOUN
fcis-5214	15	65	process	process	NOUN
fcis-5214	15	66	optimization	optimization	NOUN
fcis-5214	15	67	,	,	PUNCT
fcis-5214	15	68	inequity	inequity	NOUN
fcis-5214	15	69	problems	problem	NOUN
fcis-5214	15	70	in	in	ADP
fcis-5214	15	71	aggregation	aggregation	NOUN
fcis-5214	15	72	of	of	ADP
fcis-5214	15	73	federated	federated	ADJ
fcis-5214	15	74	models	model	NOUN
fcis-5214	15	75	,	,	PUNCT
fcis-5214	15	76	and	and	CCONJ
fcis-5214	15	77	the	the	DET
fcis-5214	15	78	systems	system	NOUN
fcis-5214	15	79	,	,	PUNCT
fcis-5214	15	80	and	and	CCONJ
fcis-5214	15	81	insufficient	insufficient	ADJ
fcis-5214	15	82	research	research	NOUN
fcis-5214	15	83	on	on	ADP
fcis-5214	15	84	personalization	personalization	NOUN
fcis-5214	15	85	of	of	ADP
fcis-5214	15	86	federated	federated	ADJ
fcis-5214	15	87	recommender	recommender	NOUN
fcis-5214	15	88	systems	system	NOUN
fcis-5214	15	89	.	.	PUNCT
fcis-5214	16	1	we	we	PRON
fcis-5214	16	2	aim	aim	VERB
fcis-5214	16	3	to	to	PART
fcis-5214	16	4	develop	develop	VERB
fcis-5214	16	5	a	a	DET
fcis-5214	16	6	more	more	ADV
fcis-5214	16	7	efficient	efficient	ADJ
fcis-5214	16	8	way	way	NOUN
fcis-5214	16	9	to	to	PART
fcis-5214	16	10	train	train	VERB
fcis-5214	16	11	recommendation	recommendation	NOUN
fcis-5214	16	12	models	model	NOUN
fcis-5214	16	13	faster	fast	ADV
fcis-5214	16	14	in	in	ADP
fcis-5214	16	15	a	a	DET
fcis-5214	16	16	federated	federated	ADJ
fcis-5214	16	17	environment	environment	NOUN
fcis-5214	16	18	so	so	SCONJ
fcis-5214	16	19	that	that	SCONJ
fcis-5214	16	20	users	user	NOUN
fcis-5214	16	21	can	can	AUX
fcis-5214	16	22	enjoy	enjoy	VERB
fcis-5214	16	23	accurate	accurate	ADJ
fcis-5214	16	24	recommendations	recommendation	NOUN
fcis-5214	16	25	without	without	ADP
fcis-5214	16	26	spending	spend	VERB
fcis-5214	16	27	unnecessary	unnecessary	ADJ
fcis-5214	16	28	effort	effort	NOUN
fcis-5214	16	29	and	and	CCONJ
fcis-5214	16	30	suffering	suffer	VERB
fcis-5214	16	31	from	from	ADP
fcis-5214	16	32	high	high	ADJ
fcis-5214	16	33	communication	communication	NOUN
fcis-5214	16	34	costs	cost	NOUN
fcis-5214	16	35	.	.	PUNCT
fcis-5214	17	1	the	the	DET
fcis-5214	17	2	user	user	NOUN
fcis-5214	17	3	in	in	ADP
fcis-5214	17	4	a	a	DET
fcis-5214	17	5	federation	federation	NOUN
fcis-5214	17	6	learning	learn	VERB
fcis-5214	17	7	scenario	scenario	NOUN
fcis-5214	17	8	bears	bear	VERB
fcis-5214	17	9	most	most	ADJ
fcis-5214	17	10	of	of	ADP
fcis-5214	17	11	the	the	DET
fcis-5214	17	12	costs	cost	NOUN
fcis-5214	17	13	involved	involve	VERB
fcis-5214	17	14	in	in	ADP
fcis-5214	17	15	training	train	VERB
fcis-5214	17	16	the	the	DET
fcis-5214	17	17	model	model	NOUN
fcis-5214	17	18	.	.	PUNCT
fcis-5214	18	1	specifically	specifically	ADV
fcis-5214	18	2	,	,	PUNCT
fcis-5214	18	3	users	user	NOUN
fcis-5214	18	4	may	may	AUX
fcis-5214	18	5	experience	experience	VERB
fcis-5214	18	6	performance	performance	NOUN
fcis-5214	18	7	degradation	degradation	NOUN
fcis-5214	18	8	and	and	CCONJ
fcis-5214	18	9	increased	increase	VERB
fcis-5214	18	10	communication	communication	NOUN
fcis-5214	18	11	load	load	NOUN
fcis-5214	18	12	during	during	ADP
fcis-5214	18	13	training	training	NOUN
fcis-5214	18	14	,	,	PUNCT
fcis-5214	18	15	especially	especially	ADV
fcis-5214	18	16	if	if	SCONJ
fcis-5214	18	17	they	they	PRON
fcis-5214	18	18	have	have	VERB
fcis-5214	18	19	to	to	PART
fcis-5214	18	20	wait	wait	VERB
fcis-5214	18	21	for	for	ADP
fcis-5214	18	22	a	a	DET
fcis-5214	18	23	long	long	ADJ
fcis-5214	18	24	time	time	NOUN
fcis-5214	18	25	to	to	PART
fcis-5214	18	26	enjoy	enjoy	VERB
fcis-5214	18	27	quality	quality	NOUN
fcis-5214	18	28	personalized	personalized	ADJ
fcis-5214	18	29	recommendations	recommendation	NOUN
fcis-5214	18	30	,	,	PUNCT
fcis-5214	18	31	which	which	PRON
fcis-5214	18	32	can	can	AUX
fcis-5214	18	33	reduce	reduce	VERB
fcis-5214	18	34	their	their	PRON
fcis-5214	18	35	acceptance	acceptance	NOUN
fcis-5214	18	36	of	of	ADP
fcis-5214	18	37	the	the	DET
fcis-5214	18	38	service	service	NOUN
fcis-5214	18	39	.	.	PUNCT
fcis-5214	19	1	in	in	ADP
fcis-5214	19	2	this	this	DET
fcis-5214	19	3	paper	paper	NOUN
fcis-5214	19	4	,	,	PUNCT
fcis-5214	19	5	we	we	PRON
fcis-5214	19	6	propose	propose	VERB
fcis-5214	19	7	a	a	DET
fcis-5214	19	8	communication	communication	NOUN
fcis-5214	19	9	-	-	PUNCT
fcis-5214	19	10	efficient	efficient	ADJ
fcis-5214	19	11	and	and	CCONJ
fcis-5214	19	12	fair	fair	ADJ
fcis-5214	19	13	personalized	personalized	ADJ
fcis-5214	19	14	federated	federated	ADJ
fcis-5214	19	15	recommendation	recommendation	NOUN
fcis-5214	19	16	method	method	NOUN
fcis-5214	19	17	(	(	PUNCT
fcis-5214	19	18	cffr	cffr	NOUN
fcis-5214	19	19	)	)	PUNCT
fcis-5214	19	20	to	to	PART
fcis-5214	19	21	address	address	VERB
fcis-5214	19	22	these	these	DET
fcis-5214	19	23	challenges	challenge	NOUN
fcis-5214	19	24	.	.	PUNCT
fcis-5214	20	1	cffr	cffr	NOUN
fcis-5214	20	2	introduces	introduce	VERB
fcis-5214	20	3	a	a	DET
fcis-5214	20	4	communication	communication	NOUN
fcis-5214	20	5	-	-	PUNCT
fcis-5214	20	6	efficient	efficient	ADJ
fcis-5214	20	7	scheme	scheme	NOUN
fcis-5214	20	8	that	that	PRON
fcis-5214	20	9	uses	use	VERB
fcis-5214	20	10	adaptive	adaptive	ADJ
fcis-5214	20	11	client	client	NOUN
fcis-5214	20	12	group	group	NOUN
fcis-5214	20	13	selection	selection	NOUN
fcis-5214	20	14	to	to	PART
fcis-5214	20	15	achieve	achieve	VERB
fcis-5214	20	16	model	model	NOUN
fcis-5214	20	17	personalization	personalization	NOUN
fcis-5214	20	18	while	while	SCONJ
fcis-5214	20	19	accelerating	accelerate	VERB
fcis-5214	20	20	the	the	DET
fcis-5214	20	21	training	training	NOUN
fcis-5214	20	22	process	process	NOUN
fcis-5214	20	23	.	.	PUNCT
fcis-5214	21	1	a	a	DET
fcis-5214	21	2	fair	fair	ADJ
fcis-5214	21	3	-	-	PUNCT
fcis-5214	21	4	aware	aware	ADJ
fcis-5214	21	5	model	model	NOUN
fcis-5214	21	6	aggregation	aggregation	NOUN
fcis-5214	21	7	algorithm	algorithm	NOUN
fcis-5214	21	8	is	be	AUX
fcis-5214	21	9	proposed	propose	VERB
fcis-5214	21	10	,	,	PUNCT
fcis-5214	21	11	which	which	PRON
fcis-5214	21	12	can	can	AUX
fcis-5214	21	13	adjust	adjust	VERB
fcis-5214	21	14	the	the	DET
fcis-5214	21	15	parameter	parameter	NOUN
fcis-5214	21	16	aggregation	aggregation	NOUN
fcis-5214	21	17	weights	weight	NOUN
fcis-5214	21	18	according	accord	VERB
fcis-5214	21	19	to	to	ADP
fcis-5214	21	20	the	the	DET
fcis-5214	21	21	imbalance	imbalance	NOUN
fcis-5214	21	22	state	state	NOUN
fcis-5214	21	23	of	of	ADP
fcis-5214	21	24	performance	performance	NOUN
fcis-5214	21	25	and	and	CCONJ
fcis-5214	21	26	data	datum	NOUN
fcis-5214	21	27	among	among	ADP
fcis-5214	21	28	different	different	ADJ
fcis-5214	21	29	clients	client	NOUN
fcis-5214	21	30	to	to	PART
fcis-5214	21	31	solve	solve	VERB
fcis-5214	21	32	the	the	DET
fcis-5214	21	33	unfairness	unfairness	NOUN
fcis-5214	21	34	problem	problem	NOUN
fcis-5214	21	35	.	.	PUNCT
fcis-5214	22	1	extensive	extensive	ADJ
fcis-5214	22	2	experimental	experimental	ADJ
fcis-5214	22	3	results	result	NOUN
fcis-5214	22	4	demonstrate	demonstrate	VERB
fcis-5214	22	5	the	the	DET
fcis-5214	22	6	effectiveness	effectiveness	NOUN
fcis-5214	22	7	and	and	CCONJ
fcis-5214	22	8	efficiency	efficiency	NOUN
fcis-5214	22	9	of	of	ADP
fcis-5214	22	10	our	our	PRON
fcis-5214	22	11	proposed	propose	VERB
fcis-5214	22	12	method	method	NOUN
fcis-5214	22	13	.	.	PUNCT
fcis-5214	23	1	the	the	DET
fcis-5214	23	2	remainder	remainder	NOUN
fcis-5214	23	3	of	of	ADP
fcis-5214	23	4	this	this	DET
fcis-5214	23	5	paper	paper	NOUN
fcis-5214	23	6	is	be	AUX
fcis-5214	23	7	organized	organize	VERB
fcis-5214	23	8	as	as	SCONJ
fcis-5214	23	9	follows	follow	VERB
fcis-5214	23	10	.	.	PUNCT
fcis-5214	24	1	section	section	NOUN
fcis-5214	24	2	2	2	NUM
fcis-5214	24	3	discusses	discuss	VERB
fcis-5214	24	4	related	related	ADJ
fcis-5214	24	5	studies	study	NOUN
fcis-5214	24	6	.	.	PUNCT
fcis-5214	25	1	section	section	NOUN
fcis-5214	25	2	3	3	NUM
fcis-5214	25	3	elaborates	elaborate	VERB
fcis-5214	25	4	on	on	ADP
fcis-5214	25	5	our	our	PRON
fcis-5214	25	6	proposed	propose	VERB
fcis-5214	25	7	cffr	cffr	NOUN
fcis-5214	25	8	method	method	PROPN
fcis-5214	25	9	.	.	PUNCT
fcis-5214	26	1	section	section	NOUN
fcis-5214	26	2	4	4	NUM
fcis-5214	26	3	describes	describe	VERB
fcis-5214	26	4	extensive	extensive	ADJ
fcis-5214	26	5	experiments	experiment	NOUN
fcis-5214	26	6	.	.	PUNCT
fcis-5214	27	1	finally	finally	ADV
fcis-5214	27	2	,	,	PUNCT
fcis-5214	27	3	we	we	PRON
fcis-5214	27	4	summarize	summarize	VERB
fcis-5214	27	5	our	our	PRON
fcis-5214	27	6	work	work	NOUN
fcis-5214	27	7	in	in	ADP
fcis-5214	27	8	section	section	NOUN
fcis-5214	27	9	5	5	NUM
fcis-5214	27	10	.	.	SYM
fcis-5214	27	11	2	2	NUM
fcis-5214	27	12	.	.	X
fcis-5214	27	13	related	relate	VERB
fcis-5214	27	14	works	work	NOUN
fcis-5214	27	15	2.1	2.1	NUM
fcis-5214	27	16	.	.	PUNCT
fcis-5214	27	17	federated	federated	ADJ
fcis-5214	27	18	learning	learning	NOUN
fcis-5214	27	19	for	for	ADP
fcis-5214	27	20	recommender	recommender	NOUN
fcis-5214	27	21	system	system	NOUN
fcis-5214	27	22	federated	federate	VERB
fcis-5214	27	23	learning	learn	VERB
fcis-5214	27	24	[	[	X
fcis-5214	27	25	1	1	NUM
fcis-5214	27	26	]	]	PUNCT
fcis-5214	27	27	is	be	AUX
fcis-5214	27	28	a	a	DET
fcis-5214	27	29	machine	machine	NOUN
fcis-5214	27	30	learning	learning	NOUN
fcis-5214	27	31	technique	technique	NOUN
fcis-5214	27	32	that	that	PRON
fcis-5214	27	33	allows	allow	VERB
fcis-5214	27	34	multiple	multiple	ADJ
fcis-5214	27	35	clients	client	NOUN
fcis-5214	27	36	to	to	PART
fcis-5214	27	37	directly	directly	ADV
fcis-5214	27	38	participate	participate	VERB
fcis-5214	27	39	in	in	ADP
fcis-5214	27	40	model	model	NOUN
fcis-5214	27	41	training	training	NOUN
fcis-5214	27	42	without	without	ADP
fcis-5214	27	43	directly	directly	ADV
fcis-5214	27	44	sharing	share	VERB
fcis-5214	27	45	data	datum	NOUN
fcis-5214	27	46	.	.	PUNCT
fcis-5214	28	1	recommendation	recommendation	NOUN
fcis-5214	28	2	algorithms	algorithm	NOUN
fcis-5214	28	3	can	can	AUX
fcis-5214	28	4	be	be	AUX
fcis-5214	28	5	migrated	migrate	VERB
fcis-5214	28	6	to	to	ADP
fcis-5214	28	7	a	a	DET
fcis-5214	28	8	federation	federation	NOUN
fcis-5214	28	9	learning	learn	VERB
fcis-5214	28	10	scenario	scenario	NOUN
fcis-5214	28	11	in	in	ADP
fcis-5214	28	12	order	order	NOUN
fcis-5214	28	13	to	to	PART
fcis-5214	28	14	achieve	achieve	VERB
fcis-5214	28	15	protection	protection	NOUN
fcis-5214	28	16	of	of	ADP
fcis-5214	28	17	users	user	NOUN
fcis-5214	28	18	'	'	PART
fcis-5214	28	19	private	private	ADJ
fcis-5214	28	20	information	information	NOUN
fcis-5214	28	21	,	,	PUNCT
fcis-5214	28	22	using	use	VERB
fcis-5214	28	23	users	user	NOUN
fcis-5214	28	24	'	'	PART
fcis-5214	28	25	local	local	ADJ
fcis-5214	28	26	data	datum	NOUN
fcis-5214	28	27	for	for	ADP
fcis-5214	28	28	model	model	NOUN
fcis-5214	28	29	training	training	NOUN
fcis-5214	28	30	and	and	CCONJ
fcis-5214	28	31	implementing	implement	VERB
fcis-5214	28	32	predictive	predictive	ADJ
fcis-5214	28	33	inference	inference	NOUN
fcis-5214	28	34	.	.	PUNCT
fcis-5214	29	1	typical	typical	ADJ
fcis-5214	29	2	federated	federated	ADJ
fcis-5214	29	3	learning	learning	NOUN
fcis-5214	29	4	-	-	PUNCT
fcis-5214	29	5	based	base	VERB
fcis-5214	29	6	recommender	recommender	NOUN
fcis-5214	29	7	systems	system	NOUN
fcis-5214	29	8	are	be	AUX
fcis-5214	29	9	fcf	fcf	NOUN
fcis-5214	29	10	[	[	X
fcis-5214	29	11	2	2	NUM
fcis-5214	29	12	]	]	PUNCT
fcis-5214	29	13	and	and	CCONJ
fcis-5214	29	14	fed	feed	VERB
fcis-5214	29	15	-	-	PUNCT
fcis-5214	29	16	mvmf	mvmf	PROPN
fcis-5214	29	17	[	[	X
fcis-5214	29	18	3	3	NUM
fcis-5214	29	19	]	]	PUNCT
fcis-5214	29	20	,	,	PUNCT
fcis-5214	29	21	but	but	CCONJ
fcis-5214	29	22	the	the	DET
fcis-5214	29	23	protection	protection	NOUN
fcis-5214	29	24	of	of	ADP
fcis-5214	29	25	the	the	DET
fcis-5214	29	26	gradient	gradient	NOUN
fcis-5214	29	27	data	datum	NOUN
fcis-5214	29	28	uploaded	upload	VERB
fcis-5214	29	29	by	by	ADP
fcis-5214	29	30	each	each	DET
fcis-5214	29	31	client	client	NOUN
fcis-5214	29	32	is	be	AUX
fcis-5214	29	33	not	not	PART
fcis-5214	29	34	implemented	implement	VERB
fcis-5214	29	35	in	in	ADP
fcis-5214	29	36	the	the	DET
fcis-5214	29	37	two	two	NUM
fcis-5214	29	38	methods	method	NOUN
fcis-5214	29	39	,	,	PUNCT
fcis-5214	29	40	since	since	SCONJ
fcis-5214	29	41	the	the	DET
fcis-5214	29	42	user	user	NOUN
fcis-5214	29	43	's	's	PART
fcis-5214	29	44	rating	rating	NOUN
fcis-5214	29	45	information	information	NOUN
fcis-5214	29	46	can	can	AUX
fcis-5214	29	47	be	be	AUX
fcis-5214	29	48	inferred	infer	VERB
fcis-5214	29	49	if	if	SCONJ
fcis-5214	29	50	given	give	VERB
fcis-5214	29	51	the	the	DET
fcis-5214	29	52	gradients	gradient	NOUN
fcis-5214	29	53	of	of	ADP
fcis-5214	29	54	a	a	DET
fcis-5214	29	55	user	user	NOUN
fcis-5214	29	56	uploaded	upload	VERB
fcis-5214	29	57	in	in	ADP
fcis-5214	29	58	two	two	NUM
fcis-5214	29	59	continuous	continuous	ADJ
fcis-5214	29	60	steps	step	NOUN
fcis-5214	29	61	[	[	X
fcis-5214	29	62	4	4	NUM
fcis-5214	29	63	]	]	PUNCT
fcis-5214	29	64	.	.	PUNCT
fcis-5214	30	1	the	the	DET
fcis-5214	30	2	model	model	NOUN
fcis-5214	30	3	updates	update	NOUN
fcis-5214	30	4	sent	send	VERB
fcis-5214	30	5	to	to	ADP
fcis-5214	30	6	the	the	DET
fcis-5214	30	7	server	server	NOUN
fcis-5214	30	8	may	may	AUX
fcis-5214	30	9	contain	contain	VERB
fcis-5214	30	10	enough	enough	ADJ
fcis-5214	30	11	information	information	NOUN
fcis-5214	30	12	to	to	PART
fcis-5214	30	13	discover	discover	VERB
fcis-5214	30	14	the	the	DET
fcis-5214	30	15	original	original	ADJ
fcis-5214	30	16	data	datum	NOUN
fcis-5214	30	17	,	,	PUNCT
fcis-5214	30	18	fedrec	fedrec	NOUN
fcis-5214	30	19	[	[	X
fcis-5214	30	20	5	5	X
fcis-5214	30	21	]	]	X
fcis-5214	30	22	propose	propose	VERB
fcis-5214	30	23	to	to	PART
fcis-5214	30	24	give	give	VERB
fcis-5214	30	25	some	some	DET
fcis-5214	30	26	(	(	PUNCT
fcis-5214	30	27	instead	instead	ADV
fcis-5214	30	28	of	of	ADP
fcis-5214	30	29	all	all	PRON
fcis-5214	30	30	)	)	PUNCT
fcis-5214	30	31	unrated	unrated	ADJ
fcis-5214	30	32	item	item	NOUN
fcis-5214	30	33	to	to	PART
fcis-5214	30	34	be	be	AUX
fcis-5214	30	35	randomly	randomly	ADV
fcis-5214	30	36	sampled	sample	VERB
fcis-5214	30	37	and	and	CCONJ
fcis-5214	30	38	to	to	PART
fcis-5214	30	39	assign	assign	VERB
fcis-5214	30	40	with	with	ADP
fcis-5214	30	41	some	some	DET
fcis-5214	30	42	virtual	virtual	ADJ
fcis-5214	30	43	ratings	rating	NOUN
fcis-5214	30	44	.	.	PUNCT
fcis-5214	31	1	fedrec++	fedrec++	NOUN
fcis-5214	32	1	[	[	X
fcis-5214	32	2	6	6	NUM
fcis-5214	32	3	]	]	PUNCT
fcis-5214	32	4	extends	extend	VERB
fcis-5214	32	5	the	the	DET
fcis-5214	32	6	fedrec	fedrec	NOUN
fcis-5214	33	1	[	[	X
fcis-5214	33	2	5	5	NUM
fcis-5214	33	3	]	]	PUNCT
fcis-5214	33	4	to	to	PART
fcis-5214	33	5	eliminate	eliminate	VERB
fcis-5214	33	6	the	the	DET
fcis-5214	33	7	nose	nose	NOUN
fcis-5214	33	8	in	in	ADP
fcis-5214	33	9	a	a	DET
fcis-5214	33	10	privacy	privacy	NOUN
fcis-5214	33	11	-	-	PUNCT
fcis-5214	33	12	aware	aware	ADJ
fcis-5214	33	13	manner	manner	NOUN
fcis-5214	33	14	by	by	ADP
fcis-5214	33	15	allocating	allocate	VERB
fcis-5214	33	16	64	64	NUM
fcis-5214	33	17	some	some	DET
fcis-5214	33	18	denoising	denoise	VERB
fcis-5214	33	19	clients	client	NOUN
fcis-5214	33	20	.	.	PUNCT
fcis-5214	34	1	2.2	2.2	NUM
fcis-5214	34	2	.	.	PUNCT
fcis-5214	35	1	local	local	ADJ
fcis-5214	35	2	differential	differential	PROPN
fcis-5214	35	3	privacy	privacy	NOUN
fcis-5214	35	4	(	(	PUNCT
fcis-5214	35	5	ε	ε	PROPN
fcis-5214	35	6	-local	-local	PROPN
fcis-5214	35	7	differential	differential	PROPN
fcis-5214	35	8	privacy	privacy	NOUN
fcis-5214	35	9	,	,	PUNCT
fcis-5214	35	10	ε	ε	PROPN
fcis-5214	35	11	-ldp	-ldp	PUNCT
fcis-5214	36	1	[	[	X
fcis-5214	36	2	7	7	NUM
fcis-5214	36	3	]	]	SYM
fcis-5214	36	4	)	)	PUNCT
fcis-5214	36	5	a	a	DET
fcis-5214	36	6	randomized	randomized	ADJ
fcis-5214	36	7	algorithm	algorithm	NOUN
fcis-5214	36	8	f	f	NOUN
fcis-5214	36	9	:	:	PUNCT
fcis-5214	36	10	𝕋	𝕋	NOUN
fcis-5214	36	11	→	→	SYM
fcis-5214	36	12	𝕐	𝕐	NOUN
fcis-5214	36	13	with	with	ADP
fcis-5214	36	14	domain	domain	NOUN
fcis-5214	36	15	𝕋	𝕋	NOUN
fcis-5214	36	16	and	and	CCONJ
fcis-5214	36	17	range	range	NOUN
fcis-5214	36	18	𝕐	𝕐	PROPN
fcis-5214	36	19	satisfies	satisfy	VERB
fcis-5214	36	20	ϵ-ldp	ϵ-ldp	NOUN
fcis-5214	37	1	if	if	SCONJ
fcis-5214	37	2	and	and	CCONJ
fcis-5214	37	3	only	only	ADV
fcis-5214	37	4	if	if	SCONJ
fcis-5214	37	5	,	,	PUNCT
fcis-5214	37	6	for	for	ADP
fcis-5214	37	7	any	any	DET
fcis-5214	37	8	inputs	input	NOUN
fcis-5214	37	9	t	t	PROPN
fcis-5214	37	10	,	,	PUNCT
fcis-5214	37	11	t′	t′	NUM
fcis-5214	37	12	∈	∈	PROPN
fcis-5214	37	13	𝕋	𝕋	NOUN
fcis-5214	37	14	and	and	CCONJ
fcis-5214	37	15	output	output	NOUN
fcis-5214	37	16	y	y	PROPN
fcis-5214	37	17	∈	∈	PROPN
fcis-5214	37	18	𝕐	𝕐	NOUN
fcis-5214	37	19	:	:	PUNCT
fcis-5214	37	20	pr[f(t	pr[f(t	PROPN
fcis-5214	37	21	)	)	PUNCT
fcis-5214	37	22	=	=	SYM
fcis-5214	37	23	y	y	PROPN
fcis-5214	37	24	]	]	PUNCT
fcis-5214	37	25	≤	≤	NUM
fcis-5214	37	26	eε	eε	ADP
fcis-5214	37	27	pr[f(t′	pr[f(t′	ADJ
fcis-5214	37	28	)	)	PUNCT
fcis-5214	37	29	=	=	SYM
fcis-5214	38	1	y	y	NOUN
fcis-5214	38	2	]	]	X
fcis-5214	38	3	(	(	PUNCT
fcis-5214	38	4	1	1	X
fcis-5214	38	5	)	)	PUNCT
fcis-5214	38	6	where	where	SCONJ
fcis-5214	38	7	ε	ε	PROPN
fcis-5214	38	8	is	be	AUX
fcis-5214	38	9	the	the	DET
fcis-5214	38	10	privacy	privacy	NOUN
fcis-5214	38	11	budget	budget	NOUN
fcis-5214	38	12	that	that	PRON
fcis-5214	38	13	controls	control	VERB
fcis-5214	38	14	the	the	DET
fcis-5214	38	15	balance	balance	NOUN
fcis-5214	38	16	between	between	ADP
fcis-5214	38	17	utility	utility	NOUN
fcis-5214	38	18	and	and	CCONJ
fcis-5214	38	19	privacy	privacy	NOUN
fcis-5214	38	20	protection	protection	NOUN
fcis-5214	38	21	?	?	PUNCT
fcis-5214	39	1	we	we	PRON
fcis-5214	39	2	implement	implement	VERB
fcis-5214	39	3	𝑓	𝑓	PRON
fcis-5214	39	4	with	with	ADP
fcis-5214	39	5	a	a	DET
fcis-5214	39	6	laplace	laplace	NOUN
fcis-5214	39	7	mechanism	mechanism	NOUN
fcis-5214	39	8	by	by	ADP
fcis-5214	39	9	adding	add	VERB
fcis-5214	39	10	well	well	ADV
fcis-5214	39	11	designed	design	VERB
fcis-5214	39	12	noise	noise	NOUN
fcis-5214	39	13	to	to	ADP
fcis-5214	39	14	the	the	DET
fcis-5214	39	15	data	datum	NOUN
fcis-5214	39	16	.	.	PUNCT
fcis-5214	40	1	3	3	X
fcis-5214	40	2	.	.	NUM
fcis-5214	40	3	methodologies	methodology	NOUN
fcis-5214	40	4	3.1	3.1	NUM
fcis-5214	40	5	.	.	PUNCT
fcis-5214	41	1	system	system	NOUN
fcis-5214	41	2	overview	overview	VERB
fcis-5214	41	3	some	some	DET
fcis-5214	41	4	commercial	commercial	ADJ
fcis-5214	41	5	behaviors	behavior	NOUN
fcis-5214	41	6	of	of	ADP
fcis-5214	41	7	users	user	NOUN
fcis-5214	41	8	,	,	PUNCT
fcis-5214	41	9	such	such	ADJ
fcis-5214	41	10	as	as	ADP
fcis-5214	41	11	paid	pay	VERB
fcis-5214	41	12	purchase	purchase	NOUN
fcis-5214	41	13	records	record	NOUN
fcis-5214	41	14	on	on	ADP
fcis-5214	41	15	movie	movie	NOUN
fcis-5214	41	16	and	and	CCONJ
fcis-5214	41	17	video	video	NOUN
fcis-5214	41	18	websites	website	NOUN
fcis-5214	41	19	are	be	AUX
fcis-5214	41	20	required	require	VERB
fcis-5214	41	21	to	to	PART
fcis-5214	41	22	be	be	AUX
fcis-5214	41	23	kept	keep	VERB
fcis-5214	41	24	by	by	ADP
fcis-5214	41	25	the	the	DET
fcis-5214	41	26	server	server	NOUN
fcis-5214	41	27	,	,	PUNCT
fcis-5214	41	28	i.e.	i.e.	X
fcis-5214	41	29	,	,	PUNCT
fcis-5214	41	30	user	user	NOUN
fcis-5214	41	31	item	item	NOUN
fcis-5214	41	32	interaction	interaction	NOUN
fcis-5214	41	33	records	record	NOUN
fcis-5214	41	34	can	can	AUX
fcis-5214	41	35	not	not	PART
fcis-5214	41	36	be	be	AUX
fcis-5214	41	37	hidden	hide	VERB
fcis-5214	41	38	from	from	ADP
fcis-5214	41	39	the	the	DET
fcis-5214	41	40	server	server	NOUN
fcis-5214	41	41	,	,	PUNCT
fcis-5214	41	42	but	but	CCONJ
fcis-5214	41	43	the	the	DET
fcis-5214	41	44	specific	specific	ADJ
fcis-5214	41	45	preferences	preference	NOUN
fcis-5214	41	46	of	of	ADP
fcis-5214	41	47	users	user	NOUN
fcis-5214	41	48	for	for	ADP
fcis-5214	41	49	items	item	NOUN
fcis-5214	41	50	,	,	PUNCT
fcis-5214	41	51	i.e.	i.e.	X
fcis-5214	41	52	,	,	PUNCT
fcis-5214	41	53	user	user	NOUN
fcis-5214	41	54	item	item	NOUN
fcis-5214	41	55	rating	rating	NOUN
fcis-5214	41	56	data	datum	NOUN
fcis-5214	41	57	,	,	PUNCT
fcis-5214	41	58	can	can	AUX
fcis-5214	41	59	still	still	ADV
fcis-5214	41	60	be	be	AUX
fcis-5214	41	61	stored	store	VERB
fcis-5214	41	62	locally	locally	ADV
fcis-5214	41	63	on	on	ADP
fcis-5214	41	64	the	the	DET
fcis-5214	41	65	user	user	NOUN
fcis-5214	41	66	's	's	PART
fcis-5214	41	67	device	device	NOUN
fcis-5214	41	68	rather	rather	ADV
fcis-5214	41	69	than	than	ADP
fcis-5214	41	70	uploaded	upload	VERB
fcis-5214	41	71	to	to	ADP
fcis-5214	41	72	the	the	DET
fcis-5214	41	73	server	server	NOUN
fcis-5214	41	74	,	,	PUNCT
fcis-5214	41	75	thus	thus	ADV
fcis-5214	41	76	reducing	reduce	VERB
fcis-5214	41	77	the	the	DET
fcis-5214	41	78	privacy	privacy	NOUN
fcis-5214	41	79	risk	risk	NOUN
fcis-5214	41	80	of	of	ADP
fcis-5214	41	81	users	user	NOUN
fcis-5214	41	82	to	to	ADP
fcis-5214	41	83	some	some	DET
fcis-5214	41	84	extent	extent	NOUN
fcis-5214	41	85	.	.	PUNCT
fcis-5214	42	1	we	we	PRON
fcis-5214	42	2	assume	assume	VERB
fcis-5214	42	3	that	that	SCONJ
fcis-5214	42	4	there	there	PRON
fcis-5214	42	5	are	be	VERB
fcis-5214	42	6	n	n	PRON
fcis-5214	42	7	users	user	NOUN
fcis-5214	42	8	u1	u1	NOUN
fcis-5214	42	9	,	,	PUNCT
fcis-5214	42	10	…	…	PUNCT
fcis-5214	42	11	,	,	PUNCT
fcis-5214	42	12	un	un	PROPN
fcis-5214	42	13	and	and	CCONJ
fcis-5214	42	14	m	m	PROPN
fcis-5214	42	15	items	item	NOUN
fcis-5214	42	16	i1	i1	PROPN
fcis-5214	42	17	,	,	PUNCT
fcis-5214	42	18	…	…	PUNCT
fcis-5214	42	19	,	,	PUNCT
fcis-5214	42	20	i	i	PRON
fcis-5214	42	21	m	m	VERB
fcis-5214	42	22	.	.	PUNCT
fcis-5214	43	1	firstly	firstly	ADV
fcis-5214	43	2	,	,	PUNCT
fcis-5214	43	3	we	we	PRON
fcis-5214	43	4	calculate	calculate	VERB
fcis-5214	43	5	user	user	NOUN
fcis-5214	43	6	similarity	similarity	NOUN
fcis-5214	43	7	based	base	VERB
fcis-5214	43	8	on	on	ADP
fcis-5214	43	9	the	the	DET
fcis-5214	43	10	user	user	NOUN
fcis-5214	43	11	item	item	NOUN
fcis-5214	43	12	interaction	interaction	NOUN
fcis-5214	43	13	records	record	NOUN
fcis-5214	43	14	stored	store	VERB
fcis-5214	43	15	on	on	ADP
fcis-5214	43	16	the	the	DET
fcis-5214	43	17	server	server	NOUN
fcis-5214	43	18	,	,	PUNCT
fcis-5214	43	19	and	and	CCONJ
fcis-5214	43	20	generate	generate	VERB
fcis-5214	43	21	a	a	DET
fcis-5214	43	22	grouping	grouping	NOUN
fcis-5214	43	23	of	of	ADP
fcis-5214	43	24	target	target	NOUN
fcis-5214	43	25	user	user	NOUN
fcis-5214	43	26	i	i	PRON
fcis-5214	43	27	based	base	VERB
fcis-5214	43	28	on	on	ADP
fcis-5214	43	29	the	the	DET
fcis-5214	43	30	user	user	NOUN
fcis-5214	43	31	similarity	similarity	NOUN
fcis-5214	43	32	,	,	PUNCT
fcis-5214	43	33	and	and	CCONJ
fcis-5214	43	34	the	the	DET
fcis-5214	43	35	users	user	NOUN
fcis-5214	43	36	in	in	ADP
fcis-5214	43	37	the	the	DET
fcis-5214	43	38	group	group	NOUN
fcis-5214	43	39	collaborate	collaborate	VERB
fcis-5214	43	40	to	to	PART
fcis-5214	43	41	train	train	VERB
fcis-5214	43	42	the	the	DET
fcis-5214	43	43	recommendation	recommendation	NOUN
fcis-5214	43	44	model	model	NOUN
fcis-5214	43	45	,	,	PUNCT
fcis-5214	43	46	and	and	CCONJ
fcis-5214	43	47	we	we	PRON
fcis-5214	43	48	also	also	ADV
fcis-5214	43	49	propose	propose	VERB
fcis-5214	43	50	a	a	DET
fcis-5214	43	51	new	new	ADJ
fcis-5214	43	52	model	model	NOUN
fcis-5214	43	53	aggregation	aggregation	NOUN
fcis-5214	43	54	method	method	NOUN
fcis-5214	43	55	to	to	PART
fcis-5214	43	56	speed	speed	VERB
fcis-5214	43	57	up	up	ADP
fcis-5214	43	58	the	the	DET
fcis-5214	43	59	model	model	NOUN
fcis-5214	43	60	training	training	NOUN
fcis-5214	43	61	in	in	ADP
fcis-5214	43	62	the	the	DET
fcis-5214	43	63	model	model	NOUN
fcis-5214	43	64	training	training	NOUN
fcis-5214	43	65	.	.	PUNCT
fcis-5214	44	1	our	our	PRON
fcis-5214	44	2	method	method	NOUN
fcis-5214	44	3	can	can	AUX
fcis-5214	44	4	serve	serve	VERB
fcis-5214	44	5	small	small	ADJ
fcis-5214	44	6	batches	batch	NOUN
fcis-5214	44	7	of	of	ADP
fcis-5214	44	8	users	user	NOUN
fcis-5214	44	9	with	with	ADP
fcis-5214	44	10	similar	similar	ADJ
fcis-5214	44	11	preferences	preference	NOUN
fcis-5214	44	12	quickly	quickly	ADV
fcis-5214	44	13	and	and	CCONJ
fcis-5214	44	14	accurately	accurately	ADV
fcis-5214	44	15	.	.	PUNCT
fcis-5214	45	1	3.2	3.2	NUM
fcis-5214	45	2	.	.	PUNCT
fcis-5214	46	1	the	the	DET
fcis-5214	46	2	recommendation	recommendation	NOUN
fcis-5214	46	3	model	model	NOUN
fcis-5214	46	4	any	any	DET
fcis-5214	46	5	representation	representation	NOUN
fcis-5214	46	6	learning	learning	NOUN
fcis-5214	46	7	model	model	NOUN
fcis-5214	46	8	can	can	AUX
fcis-5214	46	9	be	be	AUX
fcis-5214	46	10	used	use	VERB
fcis-5214	46	11	in	in	ADP
fcis-5214	46	12	our	our	PRON
fcis-5214	46	13	framework	framework	NOUN
fcis-5214	46	14	to	to	PART
fcis-5214	46	15	obtain	obtain	VERB
fcis-5214	46	16	embeddings	embedding	NOUN
fcis-5214	46	17	of	of	ADP
fcis-5214	46	18	users	user	NOUN
fcis-5214	46	19	and	and	CCONJ
fcis-5214	46	20	items	item	NOUN
fcis-5214	46	21	(	(	PUNCT
fcis-5214	46	22	e.g.	e.g.	ADV
fcis-5214	46	23	,	,	PUNCT
fcis-5214	46	24	mf	mf	X
fcis-5214	46	25	[	[	X
fcis-5214	46	26	8	8	NUM
fcis-5214	46	27	]	]	PUNCT
fcis-5214	46	28	,	,	PUNCT
fcis-5214	46	29	bpr	bpr	ADV
fcis-5214	47	1	[	[	X
fcis-5214	47	2	9	9	NUM
fcis-5214	47	3	]	]	PUNCT
fcis-5214	47	4	,	,	PUNCT
fcis-5214	47	5	and	and	CCONJ
fcis-5214	47	6	gmf	gmf	NOUN
fcis-5214	47	7	[	[	X
fcis-5214	47	8	10	10	NUM
fcis-5214	47	9	]	]	NUM
fcis-5214	47	10	)	)	PUNCT
fcis-5214	47	11	.	.	PUNCT
fcis-5214	48	1	here	here	ADV
fcis-5214	48	2	,	,	PUNCT
fcis-5214	48	3	we	we	PRON
fcis-5214	48	4	employ	employ	VERB
fcis-5214	48	5	generalized	generalized	ADJ
fcis-5214	48	6	matrix	matrix	NOUN
fcis-5214	48	7	factorization	factorization	NOUN
fcis-5214	48	8	(	(	PUNCT
fcis-5214	48	9	gmf	gmf	NOUN
fcis-5214	48	10	)	)	PUNCT
fcis-5214	48	11	to	to	PART
fcis-5214	48	12	learn	learn	VERB
fcis-5214	48	13	the	the	DET
fcis-5214	48	14	representation	representation	NOUN
fcis-5214	48	15	vector	vector	NOUN
fcis-5214	48	16	of	of	ADP
fcis-5214	48	17	users	user	NOUN
fcis-5214	48	18	and	and	CCONJ
fcis-5214	48	19	items	item	NOUN
fcis-5214	48	20	.	.	PUNCT
fcis-5214	49	1	we	we	PRON
fcis-5214	49	2	define	define	VERB
fcis-5214	49	3	the	the	DET
fcis-5214	49	4	user	user	NOUN
fcis-5214	49	5	embedding	embed	VERB
fcis-5214	49	6	as	as	ADP
fcis-5214	49	7	eu	eu	PROPN
fcis-5214	49	8	∈	∈	PROPN
fcis-5214	49	9	rk	rk	NOUN
fcis-5214	49	10	and	and	CCONJ
fcis-5214	49	11	the	the	DET
fcis-5214	49	12	item	item	NOUN
fcis-5214	49	13	embedding	embed	VERB
fcis-5214	49	14	as	as	ADP
fcis-5214	49	15	ev	ev	PRON
fcis-5214	49	16	∈	∈	PROPN
fcis-5214	49	17	rk	rk	NOUN
fcis-5214	49	18	,	,	PUNCT
fcis-5214	49	19	with	with	ADP
fcis-5214	49	20	k	k	PROPN
fcis-5214	49	21	denoting	denote	VERB
fcis-5214	49	22	the	the	DET
fcis-5214	49	23	embedding	embed	VERB
fcis-5214	49	24	dimension	dimension	NOUN
fcis-5214	49	25	.	.	PUNCT
fcis-5214	50	1	the	the	DET
fcis-5214	50	2	user	user	NOUN
fcis-5214	50	3	embedding	embed	VERB
fcis-5214	50	4	and	and	CCONJ
fcis-5214	50	5	item	item	NOUN
fcis-5214	50	6	embedding	embed	VERB
fcis-5214	50	7	are	be	AUX
fcis-5214	50	8	fed	feed	VERB
fcis-5214	50	9	into	into	ADP
fcis-5214	50	10	the	the	DET
fcis-5214	50	11	fully	fully	ADV
fcis-5214	50	12	connected	connected	ADJ
fcis-5214	50	13	layer	layer	NOUN
fcis-5214	50	14	to	to	PART
fcis-5214	50	15	obtain	obtain	VERB
fcis-5214	50	16	the	the	DET
fcis-5214	50	17	prediction	prediction	NOUN
fcis-5214	50	18	scores	score	NOUN
fcis-5214	50	19	r̂uv	r̂uv	VERB
fcis-5214	50	20	.	.	PUNCT
fcis-5214	51	1	�	�	PROPN
fcis-5214	51	2	̂	̂	VERB
fcis-5214	51	3	�	�	NOUN
fcis-5214	51	4	𝑢𝑣	𝑢𝑣	NOUN
fcis-5214	51	5	=	=	SYM
fcis-5214	51	6	𝑤𝑇(𝑒𝑢	𝑤𝑇(𝑒𝑢	PROPN
fcis-5214	51	7	⊙	⊙	PROPN
fcis-5214	51	8	𝑒𝑣	𝑒𝑣	ADP
fcis-5214	51	9	)	)	PUNCT
fcis-5214	51	10	(	(	PUNCT
fcis-5214	51	11	2	2	X
fcis-5214	51	12	)	)	PUNCT
fcis-5214	51	13	where	where	SCONJ
fcis-5214	51	14	w	w	PROPN
fcis-5214	51	15	∈	∈	PROPN
fcis-5214	51	16	r1×k	r1×k	NOUN
fcis-5214	51	17	are	be	AUX
fcis-5214	51	18	the	the	DET
fcis-5214	51	19	parameters	parameter	NOUN
fcis-5214	51	20	of	of	ADP
fcis-5214	51	21	the	the	DET
fcis-5214	51	22	fully	fully	ADV
fcis-5214	51	23	connected	connected	ADJ
fcis-5214	51	24	layer	layer	NOUN
fcis-5214	51	25	.	.	PUNCT
fcis-5214	52	1	gmf	gmf	NOUN
fcis-5214	52	2	is	be	AUX
fcis-5214	52	3	used	use	VERB
fcis-5214	52	4	to	to	PART
fcis-5214	52	5	train	train	VERB
fcis-5214	52	6	the	the	DET
fcis-5214	52	7	model	model	NOUN
fcis-5214	52	8	by	by	ADP
fcis-5214	52	9	minimizing	minimize	VERB
fcis-5214	52	10	the	the	DET
fcis-5214	52	11	loss	loss	NOUN
fcis-5214	52	12	function	function	NOUN
fcis-5214	52	13	between	between	ADP
fcis-5214	52	14	the	the	DET
fcis-5214	52	15	predicted	predict	VERB
fcis-5214	52	16	scores	score	NOUN
fcis-5214	52	17	and	and	CCONJ
fcis-5214	52	18	their	their	PRON
fcis-5214	52	19	true	true	ADJ
fcis-5214	52	20	values	value	NOUN
fcis-5214	52	21	,	,	PUNCT
fcis-5214	52	22	and	and	CCONJ
fcis-5214	52	23	here	here	ADV
fcis-5214	52	24	we	we	PRON
fcis-5214	52	25	use	use	VERB
fcis-5214	52	26	the	the	DET
fcis-5214	52	27	mean	mean	ADJ
fcis-5214	52	28	squared	square	VERB
fcis-5214	52	29	error	error	NOUN
fcis-5214	52	30	for	for	ADP
fcis-5214	52	31	explicit	explicit	ADJ
fcis-5214	52	32	recommendation	recommendation	NOUN
fcis-5214	52	33	.	.	PUNCT
fcis-5214	53	1	the	the	DET
fcis-5214	53	2	optimization	optimization	NOUN
fcis-5214	53	3	problem	problem	NOUN
fcis-5214	53	4	that	that	PRON
fcis-5214	53	5	minimizes	minimize	VERB
fcis-5214	53	6	the	the	DET
fcis-5214	53	7	loss	loss	NOUN
fcis-5214	53	8	function	function	NOUN
fcis-5214	53	9	can	can	AUX
fcis-5214	53	10	be	be	AUX
fcis-5214	53	11	expressed	express	VERB
fcis-5214	53	12	as	as	SCONJ
fcis-5214	53	13	follows	follow	VERB
fcis-5214	53	14	:	:	PUNCT
fcis-5214	53	15	min	min	NOUN
fcis-5214	53	16	∑	∑	PROPN
fcis-5214	53	17	(	(	PUNCT
fcis-5214	53	18	ruv	ruv	PROPN
fcis-5214	53	19	−	−	PROPN
fcis-5214	53	20	r̂uv)2	r̂uv)2	NOUN
fcis-5214	53	21	(	(	PUNCT
fcis-5214	53	22	u，v)∈rm×n	u，v)∈rm×n	PROPN
fcis-5214	53	23	+	+	NUM
fcis-5214	53	24	λ(||eu||2	λ(||eu||2	ADJ
fcis-5214	53	25	+	+	CCONJ
fcis-5214	53	26	||ev||2	||ev||2	ADJ
fcis-5214	53	27	)	)	PUNCT
fcis-5214	53	28	(	(	PUNCT
fcis-5214	53	29	3	3	X
fcis-5214	53	30	)	)	PUNCT
fcis-5214	53	31	during	during	ADP
fcis-5214	53	32	model	model	NOUN
fcis-5214	53	33	training	training	NOUN
fcis-5214	53	34	,	,	PUNCT
fcis-5214	53	35	we	we	PRON
fcis-5214	53	36	use	use	VERB
fcis-5214	53	37	ldp	ldp	PROPN
fcis-5214	53	38	to	to	PART
fcis-5214	53	39	add	add	VERB
fcis-5214	53	40	perturbations	perturbation	NOUN
fcis-5214	53	41	to	to	ADP
fcis-5214	53	42	the	the	DET
fcis-5214	53	43	gradient	gradient	NOUN
fcis-5214	53	44	to	to	PART
fcis-5214	53	45	better	well	ADV
fcis-5214	53	46	protect	protect	VERB
fcis-5214	53	47	user	user	NOUN
fcis-5214	53	48	privacy	privacy	NOUN
fcis-5214	53	49	since	since	SCONJ
fcis-5214	53	50	the	the	DET
fcis-5214	53	51	gradient	gradient	NOUN
fcis-5214	53	52	can	can	AUX
fcis-5214	53	53	be	be	AUX
fcis-5214	53	54	used	use	VERB
fcis-5214	53	55	to	to	PART
fcis-5214	53	56	infer	infer	VERB
fcis-5214	53	57	the	the	DET
fcis-5214	53	58	real	real	ADJ
fcis-5214	53	59	information	information	NOUN
fcis-5214	53	60	of	of	ADP
fcis-5214	53	61	the	the	DET
fcis-5214	53	62	user	user	NOUN
fcis-5214	53	63	.	.	PUNCT
fcis-5214	54	1	the	the	DET
fcis-5214	54	2	local	local	ADJ
fcis-5214	54	3	model	model	NOUN
fcis-5214	54	4	gradient	gradient	NOUN
fcis-5214	54	5	is	be	AUX
fcis-5214	54	6	defined	define	VERB
fcis-5214	54	7	as	as	ADP
fcis-5214	54	8	g	g	PROPN
fcis-5214	54	9	,	,	PUNCT
fcis-5214	54	10	which	which	PRON
fcis-5214	54	11	is	be	AUX
fcis-5214	54	12	perturbed	perturb	VERB
fcis-5214	54	13	by	by	ADP
fcis-5214	54	14	the	the	DET
fcis-5214	54	15	stochastic	stochastic	ADJ
fcis-5214	54	16	algorithm	algorithm	NOUN
fcis-5214	54	17	m	m	VERB
fcis-5214	54	18	as	as	SCONJ
fcis-5214	54	19	follows	follow	VERB
fcis-5214	54	20	:	:	PUNCT
fcis-5214	54	21	m(g	m(g	X
fcis-5214	54	22	)	)	PUNCT
fcis-5214	55	1	=	=	SYM
fcis-5214	55	2	clip(g，δ	clip(g，δ	NOUN
fcis-5214	55	3	)	)	PUNCT
fcis-5214	55	4	+	+	CCONJ
fcis-5214	55	5	la(0，λ	la(0，λ	NOUN
fcis-5214	55	6	)	)	PUNCT
fcis-5214	55	7	(	(	PUNCT
fcis-5214	55	8	4	4	X
fcis-5214	55	9	)	)	PUNCT
fcis-5214	55	10	clip(g，δ	clip(g，δ	NOUN
fcis-5214	55	11	)	)	PUNCT
fcis-5214	55	12	denotes	denote	VERB
fcis-5214	55	13	a	a	DET
fcis-5214	55	14	gradient	gradient	NOUN
fcis-5214	55	15	clipping	clip	VERB
fcis-5214	55	16	technique	technique	NOUN
fcis-5214	55	17	that	that	PRON
fcis-5214	55	18	limits	limit	VERB
fcis-5214	55	19	the	the	DET
fcis-5214	55	20	gradient	gradient	NOUN
fcis-5214	55	21	g	g	NOUN
fcis-5214	55	22	to	to	ADP
fcis-5214	55	23	the	the	DET
fcis-5214	55	24	δ	δ	PROPN
fcis-5214	55	25	scale	scale	NOUN
fcis-5214	55	26	to	to	PART
fcis-5214	55	27	prevent	prevent	VERB
fcis-5214	55	28	gradient	gradient	ADJ
fcis-5214	55	29	explosion	explosion	NOUN
fcis-5214	55	30	and	and	CCONJ
fcis-5214	55	31	better	well	ADJ
fcis-5214	55	32	train	train	VERB
fcis-5214	55	33	the	the	DET
fcis-5214	55	34	model	model	NOUN
fcis-5214	55	35	.	.	PUNCT
fcis-5214	56	1	la(0，λ	la(0，λ	PROPN
fcis-5214	56	2	)	)	PUNCT
fcis-5214	56	3	denotes	denote	VERB
fcis-5214	56	4	laplace	laplace	NOUN
fcis-5214	56	5	noise	noise	NOUN
fcis-5214	56	6	with	with	ADP
fcis-5214	56	7	mean	mean	NOUN
fcis-5214	56	8	0	0	NUM
fcis-5214	56	9	.	.	PUNCT
fcis-5214	57	1	λ	λ	NOUN
fcis-5214	57	2	controls	control	VERB
fcis-5214	57	3	the	the	DET
fcis-5214	57	4	intensity	intensity	NOUN
fcis-5214	57	5	of	of	ADP
fcis-5214	57	6	laplace	laplace	NOUN
fcis-5214	57	7	noise	noise	NOUN
fcis-5214	57	8	.	.	PUNCT
fcis-5214	58	1	the	the	DET
fcis-5214	58	2	larger	large	ADJ
fcis-5214	58	3	λ	λ	NOUN
fcis-5214	58	4	is	be	AUX
fcis-5214	58	5	,	,	PUNCT
fcis-5214	58	6	the	the	PRON
fcis-5214	58	7	higher	high	ADJ
fcis-5214	58	8	the	the	DET
fcis-5214	58	9	noise	noise	NOUN
fcis-5214	58	10	is	be	AUX
fcis-5214	58	11	and	and	CCONJ
fcis-5214	58	12	the	the	PRON
fcis-5214	58	13	higher	high	ADJ
fcis-5214	58	14	the	the	DET
fcis-5214	58	15	degree	degree	NOUN
fcis-5214	58	16	of	of	ADP
fcis-5214	58	17	privacy	privacy	NOUN
fcis-5214	58	18	protection	protection	NOUN
fcis-5214	58	19	is	be	AUX
fcis-5214	58	20	.	.	PUNCT
fcis-5214	59	1	after	after	ADP
fcis-5214	59	2	randomizing	randomize	VERB
fcis-5214	59	3	the	the	DET
fcis-5214	59	4	gradient	gradient	ADJ
fcis-5214	59	5	operation	operation	NOUN
fcis-5214	59	6	using	use	VERB
fcis-5214	59	7	ldp	ldp	PROPN
fcis-5214	59	8	technique	technique	NOUN
fcis-5214	59	9	,	,	PUNCT
fcis-5214	59	10	it	it	PRON
fcis-5214	59	11	is	be	AUX
fcis-5214	59	12	almost	almost	ADV
fcis-5214	59	13	impossible	impossible	ADJ
fcis-5214	59	14	to	to	PART
fcis-5214	59	15	infer	infer	VERB
fcis-5214	59	16	the	the	DET
fcis-5214	59	17	original	original	ADJ
fcis-5214	59	18	user	user	NOUN
fcis-5214	59	19	data	datum	NOUN
fcis-5214	59	20	from	from	ADP
fcis-5214	59	21	the	the	DET
fcis-5214	59	22	weights	weight	NOUN
fcis-5214	59	23	.	.	PUNCT
fcis-5214	60	1	the	the	DET
fcis-5214	60	2	weights	weight	NOUN
fcis-5214	60	3	on	on	ADP
fcis-5214	60	4	the	the	DET
fcis-5214	60	5	client	client	NOUN
fcis-5214	60	6	are	be	AUX
fcis-5214	60	7	then	then	ADV
fcis-5214	60	8	uploaded	upload	VERB
fcis-5214	60	9	to	to	ADP
fcis-5214	60	10	a	a	DET
fcis-5214	60	11	central	central	ADJ
fcis-5214	60	12	server	server	NOUN
fcis-5214	60	13	for	for	ADP
fcis-5214	60	14	aggregation	aggregation	NOUN
fcis-5214	60	15	.	.	PUNCT
fcis-5214	61	1	3.3	3.3	NUM
fcis-5214	61	2	.	.	PUNCT
fcis-5214	62	1	grouping	group	VERB
fcis-5214	62	2	method	method	NOUN
fcis-5214	62	3	the	the	DET
fcis-5214	62	4	ideal	ideal	ADJ
fcis-5214	62	5	user	user	NOUN
fcis-5214	62	6	grouping	grouping	NOUN
fcis-5214	62	7	should	should	AUX
fcis-5214	62	8	be	be	AUX
fcis-5214	62	9	a	a	DET
fcis-5214	62	10	good	good	ADJ
fcis-5214	62	11	representation	representation	NOUN
fcis-5214	62	12	of	of	ADP
fcis-5214	62	13	the	the	DET
fcis-5214	62	14	user	user	NOUN
fcis-5214	62	15	pattern	pattern	NOUN
fcis-5214	62	16	,	,	PUNCT
fcis-5214	62	17	and	and	CCONJ
fcis-5214	62	18	the	the	DET
fcis-5214	62	19	interaction	interaction	NOUN
fcis-5214	62	20	records	record	NOUN
fcis-5214	62	21	of	of	ADP
fcis-5214	62	22	user	user	NOUN
fcis-5214	62	23	items	item	NOUN
fcis-5214	62	24	within	within	ADP
fcis-5214	62	25	that	that	DET
fcis-5214	62	26	grouping	grouping	NOUN
fcis-5214	62	27	must	must	AUX
fcis-5214	62	28	have	have	VERB
fcis-5214	62	29	a	a	DET
fcis-5214	62	30	high	high	ADJ
fcis-5214	62	31	density	density	NOUN
fcis-5214	62	32	.	.	PUNCT
fcis-5214	63	1	therefore	therefore	ADV
fcis-5214	63	2	,	,	PUNCT
fcis-5214	63	3	user	user	NOUN
fcis-5214	63	4	grouping	grouping	NOUN
fcis-5214	63	5	should	should	AUX
fcis-5214	63	6	consider	consider	VERB
fcis-5214	63	7	the	the	DET
fcis-5214	63	8	following	follow	VERB
fcis-5214	63	9	two	two	NUM
fcis-5214	63	10	factors	factor	NOUN
fcis-5214	63	11	.	.	PUNCT
fcis-5214	64	1	one	one	NUM
fcis-5214	64	2	is	be	AUX
fcis-5214	64	3	to	to	PART
fcis-5214	64	4	estimate	estimate	VERB
fcis-5214	64	5	how	how	SCONJ
fcis-5214	64	6	similar	similar	ADJ
fcis-5214	64	7	the	the	DET
fcis-5214	64	8	interaction	interaction	NOUN
fcis-5214	64	9	records	record	NOUN
fcis-5214	64	10	of	of	ADP
fcis-5214	64	11	other	other	ADJ
fcis-5214	64	12	users	user	NOUN
fcis-5214	64	13	are	be	AUX
fcis-5214	64	14	to	to	ADP
fcis-5214	64	15	those	those	PRON
fcis-5214	64	16	of	of	ADP
fcis-5214	64	17	the	the	DET
fcis-5214	64	18	target	target	NOUN
fcis-5214	64	19	user	user	NOUN
fcis-5214	64	20	.	.	PUNCT
fcis-5214	65	1	recommendation	recommendation	NOUN
fcis-5214	65	2	performance	performance	NOUN
fcis-5214	65	3	will	will	AUX
fcis-5214	65	4	be	be	AUX
fcis-5214	65	5	enhanced	enhance	VERB
fcis-5214	65	6	if	if	SCONJ
fcis-5214	65	7	the	the	DET
fcis-5214	65	8	target	target	NOUN
fcis-5214	65	9	user	user	NOUN
fcis-5214	65	10	is	be	AUX
fcis-5214	65	11	grouped	group	VERB
fcis-5214	65	12	with	with	ADP
fcis-5214	65	13	other	other	ADJ
fcis-5214	65	14	users	user	NOUN
fcis-5214	65	15	having	have	VERB
fcis-5214	65	16	similar	similar	ADJ
fcis-5214	65	17	interaction	interaction	NOUN
fcis-5214	65	18	patterns	pattern	NOUN
fcis-5214	65	19	.	.	PUNCT
fcis-5214	66	1	the	the	DET
fcis-5214	66	2	second	second	NOUN
fcis-5214	66	3	is	be	AUX
fcis-5214	66	4	to	to	PART
fcis-5214	66	5	evaluate	evaluate	VERB
fcis-5214	66	6	the	the	DET
fcis-5214	66	7	density	density	NOUN
fcis-5214	66	8	of	of	ADP
fcis-5214	66	9	user	user	NOUN
fcis-5214	66	10	records	record	NOUN
fcis-5214	66	11	.	.	PUNCT
fcis-5214	67	1	the	the	DET
fcis-5214	67	2	more	more	ADJ
fcis-5214	67	3	user	user	NOUN
fcis-5214	67	4	rating	rating	NOUN
fcis-5214	67	5	data	datum	NOUN
fcis-5214	67	6	there	there	PRON
fcis-5214	67	7	is	be	VERB
fcis-5214	67	8	,	,	PUNCT
fcis-5214	67	9	the	the	DET
fcis-5214	67	10	denser	dense	ADJ
fcis-5214	67	11	the	the	DET
fcis-5214	67	12	data	datum	NOUN
fcis-5214	67	13	within	within	ADP
fcis-5214	67	14	the	the	DET
fcis-5214	67	15	grouping	grouping	NOUN
fcis-5214	67	16	and	and	CCONJ
fcis-5214	67	17	the	the	DET
fcis-5214	67	18	more	more	ADV
fcis-5214	67	19	useful	useful	ADJ
fcis-5214	67	20	information	information	NOUN
fcis-5214	67	21	can	can	AUX
fcis-5214	67	22	be	be	AUX
fcis-5214	67	23	captured	capture	VERB
fcis-5214	67	24	.	.	PUNCT
fcis-5214	68	1	therefore	therefore	ADV
fcis-5214	68	2	,	,	PUNCT
fcis-5214	68	3	we	we	PRON
fcis-5214	68	4	define	define	VERB
fcis-5214	68	5	a	a	DET
fcis-5214	68	6	group	group	NOUN
fcis-5214	68	7	scoring	scoring	NOUN
fcis-5214	68	8	function	function	NOUN
fcis-5214	68	9	as	as	ADP
fcis-5214	68	10	a	a	DET
fcis-5214	68	11	weighted	weighted	ADJ
fcis-5214	68	12	sum	sum	NOUN
fcis-5214	68	13	of	of	ADP
fcis-5214	68	14	user	user	NOUN
fcis-5214	68	15	rating	rating	NOUN
fcis-5214	68	16	frequency	frequency	NOUN
fcis-5214	68	17	scores	score	NOUN
fcis-5214	68	18	and	and	CCONJ
fcis-5214	68	19	similarity	similarity	NOUN
fcis-5214	68	20	scores	score	NOUN
fcis-5214	68	21	.	.	PUNCT
fcis-5214	69	1	for	for	ADP
fcis-5214	69	2	a	a	DET
fcis-5214	69	3	group	group	NOUN
fcis-5214	69	4	of	of	ADP
fcis-5214	69	5	size	size	NOUN
fcis-5214	69	6	k	k	PROPN
fcis-5214	69	7	,	,	PUNCT
fcis-5214	69	8	user	user	NOUN
fcis-5214	69	9	i	i	PROPN
fcis-5214	69	10	,	,	PUNCT
fcis-5214	69	11	other	other	ADJ
fcis-5214	69	12	user	user	PROPN
fcis-5214	69	13	j	j	PROPN
fcis-5214	69	14	,	,	PUNCT
fcis-5214	69	15	(	(	PUNCT
fcis-5214	69	16	1	1	NUM
fcis-5214	69	17	≤	≤	NUM
fcis-5214	69	18	j	j	PROPN
fcis-5214	69	19	≤	≤	NUM
fcis-5214	69	20	n	n	CCONJ
fcis-5214	69	21	)	)	PUNCT
fcis-5214	69	22	,	,	PUNCT
fcis-5214	69	23	the	the	DET
fcis-5214	69	24	group	group	NOUN
fcis-5214	69	25	score	score	NOUN
fcis-5214	69	26	function	function	NOUN
fcis-5214	69	27	is	be	AUX
fcis-5214	69	28	defined	define	VERB
fcis-5214	69	29	as	as	ADP
fcis-5214	69	30	follows	follow	VERB
fcis-5214	69	31	.	.	PUNCT
fcis-5214	70	1	score(i，j	score(i，j	NOUN
fcis-5214	70	2	)	)	PUNCT
fcis-5214	70	3	=	=	PUNCT
fcis-5214	70	4	{	{	PUNCT
fcis-5214	70	5	1，i	1，i	NUM
fcis-5214	70	6	=	=	SYM
fcis-5214	70	7	j	j	PROPN
fcis-5214	70	8	l	l	NOUN
fcis-5214	70	9	∗	∗	X
fcis-5214	70	10	freq(j	freq(j	PROPN
fcis-5214	70	11	)	)	PUNCT
fcis-5214	70	12	+	+	CCONJ
fcis-5214	70	13	(	(	PUNCT
fcis-5214	70	14	1	1	NUM
fcis-5214	70	15	−	−	PROPN
fcis-5214	70	16	l)sim(si	l)sim(si	PROPN
fcis-5214	70	17	,	,	PUNCT
fcis-5214	70	18	sj)，i	sj)，i	NOUN
fcis-5214	70	19	≠	≠	PROPN
fcis-5214	70	20	j	j	PROPN
fcis-5214	70	21	(	(	PUNCT
fcis-5214	70	22	5	5	NUM
fcis-5214	70	23	)	)	PUNCT
fcis-5214	70	24	where	where	SCONJ
fcis-5214	70	25	l	l	NOUN
fcis-5214	70	26	=	=	SYM
fcis-5214	70	27	k	k	PROPN
fcis-5214	70	28	n	n	PROPN
fcis-5214	70	29	.	.	PUNCT
fcis-5214	71	1	jaccard	jaccard	PROPN
fcis-5214	71	2	similarity	similarity	NOUN
fcis-5214	71	3	[	[	X
fcis-5214	71	4	11	11	NUM
fcis-5214	71	5	]	]	PUNCT
fcis-5214	71	6	is	be	AUX
fcis-5214	71	7	used	use	VERB
fcis-5214	71	8	to	to	PART
fcis-5214	71	9	evaluate	evaluate	VERB
fcis-5214	71	10	the	the	DET
fcis-5214	71	11	behavioral	behavioral	ADJ
fcis-5214	71	12	similarity	similarity	NOUN
fcis-5214	71	13	of	of	ADP
fcis-5214	71	14	users	user	NOUN
fcis-5214	71	15	,	,	PUNCT
fcis-5214	71	16	and	and	CCONJ
fcis-5214	71	17	si	si	PROPN
fcis-5214	71	18	and	and	CCONJ
fcis-5214	71	19	sj	sj	PROPN
fcis-5214	71	20	are	be	AUX
fcis-5214	71	21	the	the	DET
fcis-5214	71	22	interaction	interaction	NOUN
fcis-5214	71	23	records	record	NOUN
fcis-5214	71	24	of	of	ADP
fcis-5214	71	25	users	user	NOUN
fcis-5214	71	26	stored	store	VERB
fcis-5214	71	27	on	on	ADP
fcis-5214	71	28	the	the	DET
fcis-5214	71	29	server	server	NOUN
fcis-5214	71	30	as	as	ADP
fcis-5214	71	31	bit	bit	NOUN
fcis-5214	71	32	vectors	vector	NOUN
fcis-5214	71	33	.	.	PUNCT
fcis-5214	72	1	sim(si	sim(si	PROPN
fcis-5214	72	2	,	,	PUNCT
fcis-5214	72	3	sj	sj	X
fcis-5214	72	4	)	)	PUNCT
fcis-5214	72	5	=	=	SYM
fcis-5214	72	6	|si∩sj|	|si∩sj|	PROPN
fcis-5214	72	7	|si∪sj|	|si∪sj|	PROPN
fcis-5214	72	8	(	(	PUNCT
fcis-5214	72	9	6	6	NUM
fcis-5214	72	10	)	)	PUNCT
fcis-5214	72	11	freq(j	freq(j	NOUN
fcis-5214	72	12	)	)	PUNCT
fcis-5214	72	13	is	be	AUX
fcis-5214	72	14	the	the	DET
fcis-5214	72	15	rating	rating	NOUN
fcis-5214	72	16	frequency	frequency	NOUN
fcis-5214	72	17	function	function	NOUN
fcis-5214	72	18	of	of	ADP
fcis-5214	72	19	user	user	PROPN
fcis-5214	72	20	j	j	PROPN
fcis-5214	72	21	,	,	PUNCT
fcis-5214	72	22	defined	define	VERB
fcis-5214	72	23	as	as	ADP
fcis-5214	72	24	the	the	DET
fcis-5214	72	25	proportion	proportion	NOUN
fcis-5214	72	26	of	of	ADP
fcis-5214	72	27	the	the	DET
fcis-5214	72	28	number	number	NOUN
fcis-5214	72	29	of	of	ADP
fcis-5214	72	30	interaction	interaction	NOUN
fcis-5214	72	31	items	item	NOUN
fcis-5214	72	32	of	of	ADP
fcis-5214	72	33	user	user	NOUN
fcis-5214	72	34	j	j	PROPN
fcis-5214	72	35	to	to	ADP
fcis-5214	72	36	the	the	DET
fcis-5214	72	37	total	total	ADJ
fcis-5214	72	38	number	number	NOUN
fcis-5214	72	39	of	of	ADP
fcis-5214	72	40	items	item	NOUN
fcis-5214	72	41	?	?	PUNCT
fcis-5214	73	1	the	the	DET
fcis-5214	73	2	grouping	group	VERB
fcis-5214	73	3	process	process	NOUN
fcis-5214	73	4	is	be	AUX
fcis-5214	73	5	as	as	SCONJ
fcis-5214	73	6	follows	follow	VERB
fcis-5214	73	7	,	,	PUNCT
fcis-5214	73	8	the	the	DET
fcis-5214	73	9	server	server	NOUN
fcis-5214	73	10	calculates	calculate	VERB
fcis-5214	73	11	the	the	DET
fcis-5214	73	12	grouping	grouping	NOUN
fcis-5214	73	13	scores	score	NOUN
fcis-5214	73	14	with	with	ADP
fcis-5214	73	15	other	other	ADJ
fcis-5214	73	16	users	user	NOUN
fcis-5214	73	17	for	for	ADP
fcis-5214	73	18	the	the	DET
fcis-5214	73	19	target	target	NOUN
fcis-5214	73	20	user	user	NOUN
fcis-5214	73	21	i	i	PRON
fcis-5214	73	22	,	,	PUNCT
fcis-5214	73	23	sorts	sort	VERB
fcis-5214	73	24	the	the	DET
fcis-5214	73	25	users	user	NOUN
fcis-5214	73	26	in	in	ADP
fcis-5214	73	27	descending	descend	VERB
fcis-5214	73	28	order	order	NOUN
fcis-5214	73	29	of	of	ADP
fcis-5214	73	30	the	the	DET
fcis-5214	73	31	grouping	group	VERB
fcis-5214	73	32	scores	score	NOUN
fcis-5214	73	33	,	,	PUNCT
fcis-5214	73	34	and	and	CCONJ
fcis-5214	73	35	extracts	extract	VERB
fcis-5214	73	36	the	the	DET
fcis-5214	73	37	top	top	NOUN
fcis-5214	74	1	k	k	NOUN
fcis-5214	74	2	−	−	PROPN
fcis-5214	74	3	1	1	NUM
fcis-5214	74	4	users	user	NOUN
fcis-5214	74	5	to	to	PART
fcis-5214	74	6	form	form	VERB
fcis-5214	74	7	a	a	DET
fcis-5214	74	8	grouping	group	VERB
fcis-5214	74	9	ui	ui	NOUN
fcis-5214	74	10	(	(	PUNCT
fcis-5214	74	11	k	k	NOUN
fcis-5214	74	12	)	)	PUNCT
fcis-5214	74	13	with	with	ADP
fcis-5214	74	14	grouping	group	VERB
fcis-5214	74	15	size	size	NOUN
fcis-5214	74	16	k	k	PROPN
fcis-5214	74	17	with	with	ADP
fcis-5214	74	18	the	the	DET
fcis-5214	74	19	target	target	NOUN
fcis-5214	74	20	user	user	NOUN
fcis-5214	74	21	i.	i.	PROPN
fcis-5214	74	22	3.4	3.4	NUM
fcis-5214	74	23	.	.	PUNCT
fcis-5214	75	1	parameter	parameter	NOUN
fcis-5214	75	2	aggregation	aggregation	NOUN
fcis-5214	75	3	method	method	VERB
fcis-5214	75	4	the	the	DET
fcis-5214	75	5	data	datum	NOUN
fcis-5214	75	6	on	on	ADP
fcis-5214	75	7	the	the	DET
fcis-5214	75	8	client	client	NOUN
fcis-5214	75	9	side	side	NOUN
fcis-5214	75	10	is	be	AUX
fcis-5214	75	11	non	non	ADJ
fcis-5214	75	12	-	-	NOUN
fcis-5214	75	13	iid	iid	VERB
fcis-5214	75	14	,	,	PUNCT
fcis-5214	75	15	the	the	DET
fcis-5214	75	16	local	local	ADJ
fcis-5214	75	17	dataset	dataset	NOUN
fcis-5214	75	18	is	be	AUX
fcis-5214	75	19	usually	usually	ADV
fcis-5214	75	20	heterogeneous	heterogeneous	ADJ
fcis-5214	75	21	between	between	ADP
fcis-5214	75	22	clients	client	NOUN
fcis-5214	75	23	in	in	ADP
fcis-5214	75	24	terms	term	NOUN
fcis-5214	75	25	of	of	ADP
fcis-5214	75	26	size	size	NOUN
fcis-5214	75	27	and	and	CCONJ
fcis-5214	75	28	distribution	distribution	NOUN
fcis-5214	75	29	,	,	PUNCT
fcis-5214	75	30	and	and	CCONJ
fcis-5214	75	31	the	the	DET
fcis-5214	75	32	effect	effect	NOUN
fcis-5214	75	33	of	of	ADP
fcis-5214	75	34	model	model	NOUN
fcis-5214	75	35	initialization	initialization	NOUN
fcis-5214	75	36	,	,	PUNCT
fcis-5214	75	37	the	the	DET
fcis-5214	75	38	local	local	ADJ
fcis-5214	75	39	model	model	NOUN
fcis-5214	75	40	performance	performance	NOUN
fcis-5214	75	41	of	of	ADP
fcis-5214	75	42	each	each	DET
fcis-5214	75	43	client	client	NOUN
fcis-5214	75	44	during	during	ADP
fcis-5214	75	45	model	model	NOUN
fcis-5214	75	46	training	training	NOUN
fcis-5214	75	47	may	may	AUX
fcis-5214	75	48	be	be	AUX
fcis-5214	75	49	less	less	ADJ
fcis-5214	75	50	than	than	ADP
fcis-5214	75	51	optimal	optimal	ADJ
fcis-5214	75	52	.	.	PUNCT
fcis-5214	76	1	therefore	therefore	ADV
fcis-5214	76	2	,	,	PUNCT
fcis-5214	76	3	we	we	PRON
fcis-5214	76	4	propose	propose	VERB
fcis-5214	76	5	a	a	DET
fcis-5214	76	6	weighting	weighting	NOUN
fcis-5214	76	7	strategy	strategy	NOUN
fcis-5214	76	8	to	to	PART
fcis-5214	76	9	handle	handle	VERB
fcis-5214	76	10	the	the	DET
fcis-5214	76	11	differences	difference	NOUN
fcis-5214	76	12	in	in	ADP
fcis-5214	76	13	model	model	NOUN
fcis-5214	76	14	performance	performance	NOUN
fcis-5214	76	15	and	and	CCONJ
fcis-5214	76	16	data	data	NOUN
fcis-5214	76	17	size	size	NOUN
fcis-5214	76	18	between	between	ADP
fcis-5214	76	19	clients	client	NOUN
fcis-5214	76	20	.	.	PUNCT
fcis-5214	77	1	first	first	ADV
fcis-5214	77	2	,	,	PUNCT
fcis-5214	77	3	we	we	PRON
fcis-5214	77	4	denote	denote	VERB
fcis-5214	77	5	the	the	DET
fcis-5214	77	6	performance	performance	NOUN
fcis-5214	77	7	of	of	ADP
fcis-5214	77	8	client	client	NOUN
fcis-5214	77	9	i	i	PROPN
fcis-5214	77	10	as	as	ADP
fcis-5214	77	11	pi	pi	PROPN
fcis-5214	77	12	,	,	PUNCT
fcis-5214	77	13	denoted	denote	VERB
fcis-5214	77	14	as	as	ADP
fcis-5214	77	15	pi	pi	NOUN
fcis-5214	77	16	=	=	SYM
fcis-5214	77	17	1	1	NUM
fcis-5214	77	18	rmse(i	rmse(i	NOUN
fcis-5214	77	19	)	)	PUNCT
fcis-5214	77	20	(	(	PUNCT
fcis-5214	77	21	7	7	X
fcis-5214	77	22	)	)	PUNCT
fcis-5214	77	23	rmse(i	rmse(i	PROPN
fcis-5214	77	24	)	)	PUNCT
fcis-5214	77	25	is	be	AUX
fcis-5214	77	26	the	the	DET
fcis-5214	77	27	error	error	NOUN
fcis-5214	77	28	between	between	ADP
fcis-5214	77	29	the	the	DET
fcis-5214	77	30	predicted	predict	VERB
fcis-5214	77	31	score	score	NOUN
fcis-5214	77	32	of	of	ADP
fcis-5214	77	33	the	the	DET
fcis-5214	77	34	user	user	NOUN
fcis-5214	77	35	i	i	PRON
fcis-5214	77	36	's	be	AUX
fcis-5214	77	37	local	local	ADJ
fcis-5214	77	38	model	model	NOUN
fcis-5214	77	39	and	and	CCONJ
fcis-5214	77	40	the	the	DET
fcis-5214	77	41	true	true	ADJ
fcis-5214	77	42	value	value	NOUN
fcis-5214	77	43	,	,	PUNCT
fcis-5214	77	44	and	and	CCONJ
fcis-5214	77	45	a	a	DET
fcis-5214	77	46	larger	large	ADJ
fcis-5214	77	47	value	value	NOUN
fcis-5214	77	48	indicates	indicate	VERB
fcis-5214	77	49	a	a	DET
fcis-5214	77	50	larger	large	ADJ
fcis-5214	77	51	local	local	ADJ
fcis-5214	77	52	model	model	NOUN
fcis-5214	77	53	error	error	NOUN
fcis-5214	77	54	and	and	CCONJ
fcis-5214	77	55	a	a	DET
fcis-5214	77	56	worse	bad	ADJ
fcis-5214	77	57	performance	performance	NOUN
fcis-5214	77	58	of	of	ADP
fcis-5214	77	59	its	its	PRON
fcis-5214	77	60	local	local	ADJ
fcis-5214	77	61	model	model	NOUN
fcis-5214	77	62	.	.	PUNCT
fcis-5214	78	1	then	then	ADV
fcis-5214	78	2	we	we	PRON
fcis-5214	78	3	normalize	normalize	VERB
fcis-5214	78	4	it	it	PRON
fcis-5214	78	5	:	:	PUNCT
fcis-5214	78	6	pi	pi	NOUN
fcis-5214	78	7	′	′	NUM
fcis-5214	79	1	=	=	PUNCT
fcis-5214	79	2	pi	pi	ADJ
fcis-5214	79	3	∑	∑	PROPN
fcis-5214	79	4	pi	pi	PROPN
fcis-5214	79	5	i∈u	i∈u	PROPN
fcis-5214	79	6	i	i	PRON
fcis-5214	79	7	(	(	PUNCT
fcis-5214	79	8	k	k	NOUN
fcis-5214	79	9	)	)	PUNCT
fcis-5214	79	10	(	(	PUNCT
fcis-5214	79	11	8)	8)	NUM
fcis-5214	79	12	secondly	secondly	ADV
fcis-5214	79	13	,	,	PUNCT
fcis-5214	79	14	for	for	ADP
fcis-5214	79	15	the	the	DET
fcis-5214	79	16	effect	effect	NOUN
fcis-5214	79	17	of	of	ADP
fcis-5214	79	18	data	datum	NOUN
fcis-5214	79	19	sample	sample	NOUN
fcis-5214	79	20	size	size	NOUN
fcis-5214	79	21	imbalance	imbalance	NOUN
fcis-5214	79	22	,	,	PUNCT
fcis-5214	79	23	the	the	DET
fcis-5214	79	24	user	user	NOUN
fcis-5214	79	25	rating	rating	NOUN
fcis-5214	79	26	frequency	frequency	NOUN
fcis-5214	79	27	score	score	NOUN
fcis-5214	79	28	reflects	reflect	VERB
fcis-5214	79	29	the	the	DET
fcis-5214	79	30	sample	sample	NOUN
fcis-5214	79	31	data	data	NOUN
fcis-5214	79	32	size	size	NOUN
fcis-5214	79	33	on	on	ADP
fcis-5214	79	34	the	the	DET
fcis-5214	79	35	user	user	NOUN
fcis-5214	79	36	's	's	PART
fcis-5214	79	37	device	device	NOUN
fcis-5214	79	38	,	,	PUNCT
fcis-5214	79	39	thus	thus	ADV
fcis-5214	79	40	we	we	PRON
fcis-5214	79	41	define	define	VERB
fcis-5214	79	42	qi	qi	NOUN
fcis-5214	79	43	=	=	SYM
fcis-5214	79	44	freq(i	freq(i	PROPN
fcis-5214	79	45	)	)	PUNCT
fcis-5214	79	46	,	,	PUNCT
fcis-5214	79	47	which	which	PRON
fcis-5214	79	48	is	be	AUX
fcis-5214	79	49	normalized	normalize	VERB
fcis-5214	79	50	as	as	SCONJ
fcis-5214	79	51	follows	follow	VERB
fcis-5214	79	52	:	:	PUNCT
fcis-5214	79	53	65	65	NUM
fcis-5214	79	54	qi	qi	NOUN
fcis-5214	79	55	′	′	NUM
fcis-5214	79	56	=	=	SYM
fcis-5214	79	57	qi	qi	PROPN
fcis-5214	79	58	∑	∑	PROPN
fcis-5214	79	59	qi	qi	PROPN
fcis-5214	79	60	i∈u	i∈u	PROPN
fcis-5214	80	1	i	i	PRON
fcis-5214	80	2	(	(	PUNCT
fcis-5214	80	3	k	k	NOUN
fcis-5214	80	4	)	)	PUNCT
fcis-5214	80	5	(	(	PUNCT
fcis-5214	80	6	9	9	X
fcis-5214	80	7	)	)	PUNCT
fcis-5214	80	8	finally	finally	ADV
fcis-5214	80	9	,	,	PUNCT
fcis-5214	80	10	we	we	PRON
fcis-5214	80	11	get	get	VERB
fcis-5214	80	12	the	the	DET
fcis-5214	80	13	final	final	ADJ
fcis-5214	80	14	parameter	parameter	NOUN
fcis-5214	80	15	aggregation	aggregation	NOUN
fcis-5214	80	16	function	function	NOUN
fcis-5214	80	17	:	:	PUNCT
fcis-5214	80	18	aggi	aggi	PROPN
fcis-5214	80	19	=	=	PUNCT
fcis-5214	80	20	γαtpi	γαtpi	PROPN
fcis-5214	81	1	′	′	NUM
fcis-5214	82	1	+	+	CCONJ
fcis-5214	82	2	βqi	βqi	X
fcis-5214	82	3	′	′	NUM
fcis-5214	82	4	(	(	PUNCT
fcis-5214	82	5	10	10	NUM
fcis-5214	82	6	)	)	PUNCT
fcis-5214	82	7	wt+1	wt+1	X
fcis-5214	83	1	=	=	PUNCT
fcis-5214	83	2	∑	∑	PUNCT
fcis-5214	83	3	aggi	aggi	PROPN
fcis-5214	83	4	∑	∑	PUNCT
fcis-5214	83	5	aggi	aggi	PROPN
fcis-5214	83	6	i∈u	i∈u	PROPN
fcis-5214	83	7	i	i	PRON
fcis-5214	83	8	(	(	PUNCT
fcis-5214	83	9	k)i∈u	k)i∈u	PROPN
fcis-5214	83	10	i	i	PRON
fcis-5214	83	11	(	(	PUNCT
fcis-5214	83	12	k	k	NOUN
fcis-5214	83	13	)	)	PUNCT
fcis-5214	83	14	wi	wi	PROPN
fcis-5214	83	15	t+1	t+1	PROPN
fcis-5214	83	16	(	(	PUNCT
fcis-5214	83	17	11	11	NUM
fcis-5214	83	18	)	)	PUNCT
fcis-5214	83	19	where	where	SCONJ
fcis-5214	83	20	wt+1	wt+1	PROPN
fcis-5214	83	21	is	be	AUX
fcis-5214	83	22	the	the	DET
fcis-5214	83	23	global	global	ADJ
fcis-5214	83	24	model	model	NOUN
fcis-5214	83	25	updated	update	VERB
fcis-5214	83	26	by	by	ADP
fcis-5214	83	27	t	t	PROPN
fcis-5214	83	28	+	+	CCONJ
fcis-5214	83	29	1	1	NUM
fcis-5214	83	30	rounds	round	NOUN
fcis-5214	83	31	of	of	ADP
fcis-5214	83	32	aggregation	aggregation	NOUN
fcis-5214	83	33	,	,	PUNCT
fcis-5214	83	34	α	α	PROPN
fcis-5214	83	35	and	and	CCONJ
fcis-5214	83	36	β	β	X
fcis-5214	83	37	are	be	AUX
fcis-5214	83	38	two	two	NUM
fcis-5214	83	39	hyperparameters	hyperparameter	NOUN
fcis-5214	83	40	?	?	PUNCT
fcis-5214	84	1	as	as	SCONJ
fcis-5214	84	2	the	the	DET
fcis-5214	84	3	model	model	NOUN
fcis-5214	84	4	is	be	AUX
fcis-5214	84	5	updated	update	VERB
fcis-5214	84	6	by	by	ADP
fcis-5214	84	7	multiple	multiple	ADJ
fcis-5214	84	8	rounds	round	NOUN
fcis-5214	84	9	of	of	ADP
fcis-5214	84	10	aggregation	aggregation	NOUN
fcis-5214	84	11	,	,	PUNCT
fcis-5214	84	12	the	the	DET
fcis-5214	84	13	performance	performance	NOUN
fcis-5214	84	14	of	of	ADP
fcis-5214	84	15	each	each	DET
fcis-5214	84	16	local	local	ADJ
fcis-5214	84	17	model	model	NOUN
fcis-5214	84	18	will	will	AUX
fcis-5214	84	19	be	be	AUX
fcis-5214	84	20	improved	improve	VERB
fcis-5214	84	21	,	,	PUNCT
fcis-5214	84	22	and	and	CCONJ
fcis-5214	84	23	then	then	ADV
fcis-5214	84	24	the	the	DET
fcis-5214	84	25	main	main	ADJ
fcis-5214	84	26	factor	factor	NOUN
fcis-5214	84	27	affecting	affect	VERB
fcis-5214	84	28	the	the	DET
fcis-5214	84	29	aggregation	aggregation	NOUN
fcis-5214	84	30	of	of	ADP
fcis-5214	84	31	the	the	DET
fcis-5214	84	32	model	model	NOUN
fcis-5214	84	33	will	will	AUX
fcis-5214	84	34	be	be	AUX
fcis-5214	84	35	shifted	shift	VERB
fcis-5214	84	36	to	to	ADP
fcis-5214	84	37	the	the	DET
fcis-5214	84	38	number	number	NOUN
fcis-5214	84	39	of	of	ADP
fcis-5214	84	40	samples	sample	NOUN
fcis-5214	84	41	of	of	ADP
fcis-5214	84	42	each	each	DET
fcis-5214	84	43	client	client	NOUN
fcis-5214	84	44	,	,	PUNCT
fcis-5214	84	45	so	so	SCONJ
fcis-5214	84	46	we	we	PRON
fcis-5214	84	47	gradually	gradually	ADV
fcis-5214	84	48	weaken	weaken	VERB
fcis-5214	84	49	the	the	DET
fcis-5214	84	50	influence	influence	NOUN
fcis-5214	84	51	of	of	ADP
fcis-5214	84	52	the	the	DET
fcis-5214	84	53	performance	performance	NOUN
fcis-5214	84	54	of	of	ADP
fcis-5214	84	55	each	each	DET
fcis-5214	84	56	local	local	ADJ
fcis-5214	84	57	model	model	NOUN
fcis-5214	84	58	when	when	SCONJ
fcis-5214	84	59	re	re	VERB
fcis-5214	84	60	-	-	VERB
fcis-5214	84	61	aggregating	aggregate	VERB
fcis-5214	84	62	the	the	DET
fcis-5214	84	63	parameters	parameter	NOUN
fcis-5214	84	64	,	,	PUNCT
fcis-5214	84	65	i.e.	i.e.	X
fcis-5214	84	66	,	,	PUNCT
fcis-5214	84	67	the	the	DET
fcis-5214	84	68	parameter	parameter	NOUN
fcis-5214	84	69	αt+1	αt+1	PROPN
fcis-5214	84	70	used	use	VERB
fcis-5214	84	71	in	in	ADP
fcis-5214	84	72	the	the	DET
fcis-5214	84	73	t	t	NOUN
fcis-5214	84	74	+	+	CCONJ
fcis-5214	84	75	1	1	NUM
fcis-5214	84	76	round	round	NOUN
fcis-5214	84	77	of	of	ADP
fcis-5214	84	78	aggregation	aggregation	NOUN
fcis-5214	84	79	is	be	AUX
fcis-5214	84	80	obtained	obtain	VERB
fcis-5214	84	81	by	by	ADP
fcis-5214	84	82	multiplying	multiply	VERB
fcis-5214	84	83	the	the	DET
fcis-5214	84	84	parameter	parameter	NOUN
fcis-5214	84	85	αt	αt	PROPN
fcis-5214	84	86	used	use	VERB
fcis-5214	84	87	in	in	ADP
fcis-5214	84	88	the	the	DET
fcis-5214	84	89	t	t	PROPN
fcis-5214	84	90	round	round	NOUN
fcis-5214	84	91	of	of	ADP
fcis-5214	84	92	aggregation	aggregation	NOUN
fcis-5214	84	93	by	by	ADP
fcis-5214	84	94	the	the	DET
fcis-5214	84	95	discount	discount	NOUN
fcis-5214	84	96	factor	factor	NOUN
fcis-5214	84	97	γ	γ	NOUN
fcis-5214	84	98	,	,	PUNCT
fcis-5214	84	99	so	so	SCONJ
fcis-5214	84	100	that	that	SCONJ
fcis-5214	84	101	the	the	DET
fcis-5214	84	102	value	value	NOUN
fcis-5214	84	103	of	of	ADP
fcis-5214	84	104	α	α	PROPN
fcis-5214	84	105	decreases	decrease	VERB
fcis-5214	84	106	with	with	ADP
fcis-5214	84	107	each	each	DET
fcis-5214	84	108	round	round	NOUN
fcis-5214	84	109	of	of	ADP
fcis-5214	84	110	the	the	DET
fcis-5214	84	111	aggregation	aggregation	NOUN
fcis-5214	84	112	process	process	NOUN
fcis-5214	84	113	.	.	PUNCT
fcis-5214	85	1	4	4	X
fcis-5214	85	2	.	.	X
fcis-5214	85	3	experiments	experiment	NOUN
fcis-5214	85	4	4.1	4.1	NUM
fcis-5214	85	5	.	.	PUNCT
fcis-5214	86	1	data	datum	NOUN
fcis-5214	86	2	experiments	experiment	NOUN
fcis-5214	86	3	are	be	AUX
fcis-5214	86	4	run	run	VERB
fcis-5214	86	5	on	on	ADP
fcis-5214	86	6	movielens	movielen	NOUN
fcis-5214	86	7	100k(ml-100k	100k(ml-100k	PROPN
fcis-5214	86	8	)	)	PUNCT
fcis-5214	86	9	and	and	CCONJ
fcis-5214	86	10	movielens	movielen	VERB
fcis-5214	86	11	1m(ml-1	1m(ml-1	NUM
fcis-5214	86	12	m	m	NOUN
fcis-5214	86	13	)	)	PUNCT
fcis-5214	86	14	.	.	PUNCT
fcis-5214	87	1	all	all	DET
fcis-5214	87	2	the	the	DET
fcis-5214	87	3	dataset	dataset	NOUN
fcis-5214	87	4	have	have	AUX
fcis-5214	87	5	been	be	AUX
fcis-5214	87	6	widely	widely	ADV
fcis-5214	87	7	used	use	VERB
fcis-5214	87	8	in	in	ADP
fcis-5214	87	9	recommender	recommender	NOUN
fcis-5214	87	10	systems	system	NOUN
fcis-5214	87	11	literature	literature	NOUN
fcis-5214	87	12	to	to	PART
fcis-5214	87	13	evaluate	evaluate	VERB
fcis-5214	87	14	collaborative	collaborative	ADJ
fcis-5214	87	15	filtering	filtering	NOUN
fcis-5214	87	16	algorithms	algorithm	NOUN
fcis-5214	87	17	.	.	PUNCT
fcis-5214	88	1	these	these	DET
fcis-5214	88	2	datasets	dataset	NOUN
fcis-5214	88	3	are	be	AUX
fcis-5214	88	4	commonly	commonly	ADV
fcis-5214	88	5	used	use	VERB
fcis-5214	88	6	benchmark	benchmark	NOUN
fcis-5214	88	7	datasets	dataset	NOUN
fcis-5214	88	8	in	in	ADP
fcis-5214	88	9	recommendation	recommendation	NOUN
fcis-5214	88	10	system	system	NOUN
fcis-5214	88	11	research	research	NOUN
fcis-5214	88	12	.	.	PUNCT
fcis-5214	89	1	4.2	4.2	NUM
fcis-5214	89	2	.	.	PUNCT
fcis-5214	89	3	evaluation	evaluation	NOUN
fcis-5214	89	4	for	for	ADP
fcis-5214	89	5	the	the	DET
fcis-5214	89	6	evaluation	evaluation	NOUN
fcis-5214	89	7	metric	metric	NOUN
fcis-5214	89	8	,	,	PUNCT
fcis-5214	89	9	we	we	PRON
fcis-5214	89	10	adopt	adopt	VERB
fcis-5214	89	11	the	the	DET
fcis-5214	89	12	root	root	NOUN
fcis-5214	89	13	mean	mean	VERB
fcis-5214	89	14	square	square	ADJ
fcis-5214	89	15	error	error	NOUN
fcis-5214	89	16	(	(	PUNCT
fcis-5214	89	17	rmse	rmse	NOUN
fcis-5214	89	18	)	)	PUNCT
fcis-5214	89	19	as	as	ADP
fcis-5214	89	20	performance	performance	NOUN
fcis-5214	89	21	metrics	metric	NOUN
fcis-5214	89	22	,	,	PUNCT
fcis-5214	89	23	which	which	PRON
fcis-5214	89	24	is	be	AUX
fcis-5214	89	25	wide	wide	ADV
fcis-5214	89	26	adopted	adopt	VERB
fcis-5214	89	27	in	in	ADP
fcis-5214	89	28	many	many	ADJ
fcis-5214	89	29	related	relate	VERB
fcis-5214	89	30	studies	study	NOUN
fcis-5214	89	31	for	for	ADP
fcis-5214	89	32	performance	performance	NOUN
fcis-5214	89	33	evaluation	evaluation	NOUN
fcis-5214	89	34	.	.	PUNCT
fcis-5214	90	1	we	we	PRON
fcis-5214	90	2	group	group	VERB
fcis-5214	90	3	the	the	DET
fcis-5214	90	4	users	user	NOUN
fcis-5214	90	5	after	after	ADP
fcis-5214	90	6	randomly	randomly	ADV
fcis-5214	90	7	removing	remove	VERB
fcis-5214	90	8	one	one	NUM
fcis-5214	90	9	local	local	ADJ
fcis-5214	90	10	rating	rating	NOUN
fcis-5214	90	11	record	record	NOUN
fcis-5214	90	12	,	,	PUNCT
fcis-5214	90	13	and	and	CCONJ
fcis-5214	90	14	the	the	DET
fcis-5214	90	15	users	user	NOUN
fcis-5214	90	16	within	within	ADP
fcis-5214	90	17	the	the	DET
fcis-5214	90	18	group	group	NOUN
fcis-5214	90	19	are	be	AUX
fcis-5214	90	20	trained	train	VERB
fcis-5214	90	21	for	for	ADP
fcis-5214	90	22	the	the	DET
fcis-5214	90	23	federal	federal	ADJ
fcis-5214	90	24	model	model	NOUN
fcis-5214	90	25	,	,	PUNCT
fcis-5214	90	26	and	and	CCONJ
fcis-5214	90	27	estimate	estimate	VERB
fcis-5214	90	28	the	the	DET
fcis-5214	90	29	rmse	rmse	NOUN
fcis-5214	90	30	between	between	ADP
fcis-5214	90	31	the	the	DET
fcis-5214	90	32	prediction	prediction	NOUN
fcis-5214	90	33	score	score	NOUN
fcis-5214	90	34	and	and	CCONJ
fcis-5214	90	35	the	the	DET
fcis-5214	90	36	original	original	ADJ
fcis-5214	90	37	score	score	NOUN
fcis-5214	90	38	for	for	ADP
fcis-5214	90	39	recommendation	recommendation	NOUN
fcis-5214	90	40	performance	performance	NOUN
fcis-5214	90	41	evaluation	evaluation	NOUN
fcis-5214	90	42	.	.	PUNCT
fcis-5214	91	1	rmse	rmse	PROPN
fcis-5214	91	2	is	be	AUX
fcis-5214	91	3	calculated	calculate	VERB
fcis-5214	91	4	as	as	SCONJ
fcis-5214	91	5	follows	follow	VERB
fcis-5214	91	6	:	:	PUNCT
fcis-5214	92	1	rmse	rmse	PROPN
fcis-5214	92	2	=	=	SYM
fcis-5214	93	1	√	√	PROPN
fcis-5214	93	2	1	1	NUM
fcis-5214	93	3	n	n	NOUN
fcis-5214	93	4	∑	∑	PUNCT
fcis-5214	93	5	(	(	PUNCT
fcis-5214	93	6	ri	ri	NOUN
fcis-5214	93	7	−	−	PROPN
fcis-5214	93	8	r̂i	r̂i	PROPN
fcis-5214	93	9	)	)	PUNCT
fcis-5214	93	10	n	n	CCONJ
fcis-5214	93	11	i=1	i=1	PROPN
fcis-5214	93	12	(	(	PUNCT
fcis-5214	93	13	12	12	NUM
fcis-5214	93	14	)	)	PUNCT
fcis-5214	93	15	where	where	SCONJ
fcis-5214	93	16	n	n	NOUN
fcis-5214	93	17	=	=	SYM
fcis-5214	93	18	total	total	ADJ
fcis-5214	93	19	number	number	NOUN
fcis-5214	93	20	of	of	ADP
fcis-5214	93	21	rating	rating	NOUN
fcis-5214	93	22	data	datum	NOUN
fcis-5214	93	23	records	record	NOUN
fcis-5214	93	24	for	for	ADP
fcis-5214	93	25	users	user	NOUN
fcis-5214	93	26	in	in	ADP
fcis-5214	93	27	the	the	DET
fcis-5214	93	28	current	current	ADJ
fcis-5214	93	29	group	group	NOUN
fcis-5214	93	30	in	in	ADP
fcis-5214	93	31	the	the	DET
fcis-5214	93	32	test	test	NOUN
fcis-5214	93	33	data	datum	NOUN
fcis-5214	93	34	,	,	PUNCT
fcis-5214	93	35	ri	ri	NOUN
fcis-5214	93	36	=	=	PUNCT
fcis-5214	93	37	original	original	ADJ
fcis-5214	93	38	rating	rating	NOUN
fcis-5214	93	39	information	information	NOUN
fcis-5214	93	40	,	,	PUNCT
fcis-5214	93	41	r̂i	r̂i	NOUN
fcis-5214	93	42	=	=	NOUN
fcis-5214	93	43	predictive	predictive	ADJ
fcis-5214	93	44	rating	rating	NOUN
fcis-5214	93	45	information	information	NOUN
fcis-5214	93	46	.	.	PUNCT
fcis-5214	94	1	we	we	PRON
fcis-5214	94	2	compare	compare	VERB
fcis-5214	94	3	our	our	PRON
fcis-5214	94	4	proposed	propose	VERB
fcis-5214	94	5	cffr	cffr	NOUN
fcis-5214	94	6	with	with	ADP
fcis-5214	94	7	the	the	DET
fcis-5214	94	8	popular	popular	ADJ
fcis-5214	94	9	fedavg	fedavg	NOUN
fcis-5214	94	10	method	method	NOUN
fcis-5214	94	11	using	use	VERB
fcis-5214	94	12	gmf	gmf	NOUN
fcis-5214	94	13	as	as	ADP
fcis-5214	94	14	recommendation	recommendation	NOUN
fcis-5214	94	15	model	model	NOUN
fcis-5214	94	16	.	.	PUNCT
fcis-5214	95	1	the	the	DET
fcis-5214	95	2	goal	goal	NOUN
fcis-5214	95	3	of	of	ADP
fcis-5214	95	4	our	our	PRON
fcis-5214	95	5	evaluation	evaluation	NOUN
fcis-5214	95	6	is	be	AUX
fcis-5214	95	7	to	to	PART
fcis-5214	95	8	compare	compare	VERB
fcis-5214	95	9	the	the	DET
fcis-5214	95	10	recommendation	recommendation	NOUN
fcis-5214	95	11	quality	quality	NOUN
fcis-5214	95	12	and	and	CCONJ
fcis-5214	95	13	convergence	convergence	NOUN
fcis-5214	95	14	speed	speed	NOUN
fcis-5214	95	15	in	in	ADP
fcis-5214	95	16	a	a	DET
fcis-5214	95	17	federated	federated	ADJ
fcis-5214	95	18	setting	setting	NOUN
fcis-5214	95	19	,	,	PUNCT
fcis-5214	95	20	so	so	SCONJ
fcis-5214	95	21	we	we	PRON
fcis-5214	95	22	omit	omit	VERB
fcis-5214	95	23	comparisons	comparison	NOUN
fcis-5214	95	24	with	with	ADP
fcis-5214	95	25	other	other	ADJ
fcis-5214	95	26	classical	classical	ADJ
fcis-5214	95	27	centralized	centralized	ADJ
fcis-5214	95	28	algorithms	algorithm	NOUN
fcis-5214	95	29	.	.	PUNCT
fcis-5214	96	1	users	user	NOUN
fcis-5214	96	2	are	be	AUX
fcis-5214	96	3	grouped	group	VERB
fcis-5214	96	4	after	after	ADP
fcis-5214	96	5	randomly	randomly	ADV
fcis-5214	96	6	deleting	delete	VERB
fcis-5214	96	7	a	a	DET
fcis-5214	96	8	local	local	ADJ
fcis-5214	96	9	rating	rating	NOUN
fcis-5214	96	10	record	record	NOUN
fcis-5214	96	11	,	,	PUNCT
fcis-5214	96	12	and	and	CCONJ
fcis-5214	96	13	the	the	DET
fcis-5214	96	14	users	user	NOUN
fcis-5214	96	15	in	in	ADP
fcis-5214	96	16	the	the	DET
fcis-5214	96	17	group	group	NOUN
fcis-5214	96	18	are	be	AUX
fcis-5214	96	19	trained	train	VERB
fcis-5214	96	20	on	on	ADP
fcis-5214	96	21	the	the	DET
fcis-5214	96	22	federation	federation	NOUN
fcis-5214	96	23	model	model	NOUN
fcis-5214	96	24	and	and	CCONJ
fcis-5214	96	25	the	the	DET
fcis-5214	96	26	error	error	NOUN
fcis-5214	96	27	between	between	ADP
fcis-5214	96	28	the	the	DET
fcis-5214	96	29	predicted	predict	VERB
fcis-5214	96	30	and	and	CCONJ
fcis-5214	96	31	original	original	ADJ
fcis-5214	96	32	scores	score	NOUN
fcis-5214	96	33	is	be	AUX
fcis-5214	96	34	estimated	estimate	VERB
fcis-5214	96	35	for	for	ADP
fcis-5214	96	36	recommendation	recommendation	NOUN
fcis-5214	96	37	performance	performance	NOUN
fcis-5214	96	38	evaluation	evaluation	NOUN
fcis-5214	96	39	.	.	PUNCT
fcis-5214	97	1	the	the	DET
fcis-5214	97	2	number	number	NOUN
fcis-5214	97	3	of	of	ADP
fcis-5214	97	4	federation	federation	NOUN
fcis-5214	97	5	training	training	NOUN
fcis-5214	97	6	rounds	round	NOUN
fcis-5214	97	7	is	be	AUX
fcis-5214	97	8	set	set	VERB
fcis-5214	97	9	to	to	ADP
fcis-5214	97	10	50	50	NUM
fcis-5214	97	11	,	,	PUNCT
fcis-5214	97	12	and	and	CCONJ
fcis-5214	97	13	the	the	DET
fcis-5214	97	14	relevant	relevant	ADJ
fcis-5214	97	15	parameters	parameter	NOUN
fcis-5214	97	16	are	be	AUX
fcis-5214	97	17	set	set	VERB
fcis-5214	97	18	by	by	ADP
fcis-5214	97	19	default	default	NOUN
fcis-5214	97	20	as	as	ADP
fcis-5214	97	21	follows:𝑙𝑟=	follows:𝑙𝑟=	ADJ
fcis-5214	97	22	0.1	0.1	NUM
fcis-5214	97	23	，	，	NOUN
fcis-5214	97	24	𝛼	𝛼	X
fcis-5214	98	1	=	=	NOUN
fcis-5214	98	2	1.0，𝛾	1.0，𝛾	NUM
fcis-5214	98	3	=	=	SYM
fcis-5214	98	4	0.98，𝛽	0.98，𝛽	NUM
fcis-5214	98	5	=	=	SYM
fcis-5214	99	1	1	1	NUM
fcis-5214	99	2	−	−	NOUN
fcis-5214	99	3	𝛼，𝑘	𝛼，𝑘	NOUN
fcis-5214	99	4	=	=	SYM
fcis-5214	99	5	200，δ=0.01，𝜆=0.03	200，δ=0.01，𝜆=0.03	NUM
fcis-5214	99	6	.	.	PUNCT
fcis-5214	100	1	4.3	4.3	NUM
fcis-5214	100	2	.	.	PUNCT
fcis-5214	100	3	prediction	prediction	NOUN
fcis-5214	100	4	accuracy	accuracy	NOUN
fcis-5214	100	5	in	in	ADP
fcis-5214	100	6	this	this	DET
fcis-5214	100	7	section	section	NOUN
fcis-5214	100	8	,	,	PUNCT
fcis-5214	100	9	we	we	PRON
fcis-5214	100	10	identify	identify	VERB
fcis-5214	100	11	the	the	DET
fcis-5214	100	12	improvement	improvement	NOUN
fcis-5214	100	13	of	of	ADP
fcis-5214	100	14	recommendation	recommendation	NOUN
fcis-5214	100	15	performance	performance	NOUN
fcis-5214	100	16	by	by	ADP
fcis-5214	100	17	grouping	group	VERB
fcis-5214	100	18	.	.	PUNCT
fcis-5214	101	1	first	first	ADV
fcis-5214	101	2	,	,	PUNCT
fcis-5214	101	3	let	let	VERB
fcis-5214	101	4	us	we	PRON
fcis-5214	101	5	consider	consider	VERB
fcis-5214	101	6	the	the	DET
fcis-5214	101	7	effect	effect	NOUN
fcis-5214	101	8	of	of	ADP
fcis-5214	101	9	group	group	NOUN
fcis-5214	101	10	size	size	NOUN
fcis-5214	101	11	on	on	ADP
fcis-5214	101	12	recommendation	recommendation	NOUN
fcis-5214	101	13	performance	performance	NOUN
fcis-5214	101	14	.	.	PUNCT
fcis-5214	102	1	the	the	PRON
fcis-5214	102	2	smaller	small	ADJ
fcis-5214	102	3	the	the	DET
fcis-5214	102	4	group	group	NOUN
fcis-5214	102	5	size	size	NOUN
fcis-5214	102	6	,	,	PUNCT
fcis-5214	102	7	the	the	PRON
fcis-5214	102	8	higher	high	ADJ
fcis-5214	102	9	the	the	DET
fcis-5214	102	10	density	density	NOUN
fcis-5214	102	11	.	.	PUNCT
fcis-5214	103	1	however	however	ADV
fcis-5214	103	2	,	,	PUNCT
fcis-5214	103	3	the	the	DET
fcis-5214	103	4	information	information	NOUN
fcis-5214	103	5	about	about	ADP
fcis-5214	103	6	the	the	DET
fcis-5214	103	7	rating	rating	NOUN
fcis-5214	103	8	patterns	pattern	NOUN
fcis-5214	103	9	is	be	AUX
fcis-5214	103	10	reduced	reduce	VERB
fcis-5214	103	11	.	.	PUNCT
fcis-5214	104	1	conversely	conversely	ADV
fcis-5214	104	2	,	,	PUNCT
fcis-5214	104	3	if	if	SCONJ
fcis-5214	104	4	the	the	DET
fcis-5214	104	5	group	group	NOUN
fcis-5214	104	6	size	size	NOUN
fcis-5214	104	7	increases	increase	NOUN
fcis-5214	104	8	,	,	PUNCT
fcis-5214	104	9	the	the	DET
fcis-5214	104	10	information	information	NOUN
fcis-5214	104	11	about	about	ADP
fcis-5214	104	12	the	the	DET
fcis-5214	104	13	rating	rating	NOUN
fcis-5214	104	14	patterns	pattern	NOUN
fcis-5214	104	15	increases	increase	NOUN
fcis-5214	104	16	,	,	PUNCT
fcis-5214	104	17	but	but	CCONJ
fcis-5214	104	18	the	the	DET
fcis-5214	104	19	density	density	NOUN
fcis-5214	104	20	decreases	decrease	VERB
fcis-5214	104	21	.	.	PUNCT
fcis-5214	105	1	therefore	therefore	ADV
fcis-5214	105	2	,	,	PUNCT
fcis-5214	105	3	the	the	DET
fcis-5214	105	4	appropriate	appropriate	ADJ
fcis-5214	105	5	group	group	NOUN
fcis-5214	105	6	size	size	NOUN
fcis-5214	105	7	𝑘	𝑘	NOUN
fcis-5214	105	8	for	for	ADP
fcis-5214	105	9	grouping	grouping	NOUN
fcis-5214	105	10	is	be	AUX
fcis-5214	105	11	important	important	ADJ
fcis-5214	105	12	to	to	PART
fcis-5214	105	13	improve	improve	VERB
fcis-5214	105	14	the	the	DET
fcis-5214	105	15	recommendation	recommendation	NOUN
fcis-5214	105	16	performance	performance	NOUN
fcis-5214	105	17	.	.	PUNCT
fcis-5214	106	1	figure	figure	NOUN
fcis-5214	106	2	1	1	NUM
fcis-5214	106	3	.	.	PUNCT
fcis-5214	106	4	recommender	recommender	NOUN
fcis-5214	106	5	performance	performance	NOUN
fcis-5214	106	6	comparison	comparison	NOUN
fcis-5214	106	7	by	by	ADP
fcis-5214	106	8	group	group	NOUN
fcis-5214	106	9	size	size	NOUN
fcis-5214	106	10	in	in	ADP
fcis-5214	106	11	this	this	DET
fcis-5214	106	12	section	section	NOUN
fcis-5214	106	13	,	,	PUNCT
fcis-5214	106	14	we	we	PRON
fcis-5214	106	15	determine	determine	VERB
fcis-5214	106	16	the	the	DET
fcis-5214	106	17	improvement	improvement	NOUN
fcis-5214	106	18	of	of	ADP
fcis-5214	106	19	grouping	group	VERB
fcis-5214	106	20	on	on	ADP
fcis-5214	106	21	recommendation	recommendation	NOUN
fcis-5214	106	22	performance	performance	NOUN
fcis-5214	106	23	.	.	PUNCT
fcis-5214	107	1	first	first	ADV
fcis-5214	107	2	,	,	PUNCT
fcis-5214	107	3	we	we	PRON
fcis-5214	107	4	analyze	analyze	VERB
fcis-5214	107	5	the	the	DET
fcis-5214	107	6	effect	effect	NOUN
fcis-5214	107	7	of	of	ADP
fcis-5214	107	8	grouping	group	VERB
fcis-5214	107	9	size	size	NOUN
fcis-5214	107	10	on	on	ADP
fcis-5214	107	11	recommendation	recommendation	NOUN
fcis-5214	107	12	performance	performance	NOUN
fcis-5214	107	13	.	.	PUNCT
fcis-5214	108	1	the	the	PRON
fcis-5214	108	2	smaller	small	ADJ
fcis-5214	108	3	the	the	DET
fcis-5214	108	4	group	group	NOUN
fcis-5214	108	5	size	size	NOUN
fcis-5214	108	6	,	,	PUNCT
fcis-5214	108	7	the	the	PRON
fcis-5214	108	8	higher	high	ADJ
fcis-5214	108	9	the	the	DET
fcis-5214	108	10	data	data	NOUN
fcis-5214	108	11	density	density	NOUN
fcis-5214	108	12	and	and	CCONJ
fcis-5214	108	13	the	the	DET
fcis-5214	108	14	less	less	ADJ
fcis-5214	108	15	information	information	NOUN
fcis-5214	108	16	about	about	ADP
fcis-5214	108	17	user	user	NOUN
fcis-5214	108	18	rating	rating	NOUN
fcis-5214	108	19	patterns	pattern	NOUN
fcis-5214	108	20	,	,	PUNCT
fcis-5214	108	21	and	and	CCONJ
fcis-5214	108	22	conversely	conversely	ADV
fcis-5214	108	23	,	,	PUNCT
fcis-5214	108	24	if	if	SCONJ
fcis-5214	108	25	the	the	DET
fcis-5214	108	26	group	group	NOUN
fcis-5214	108	27	size	size	NOUN
fcis-5214	108	28	increases	increase	NOUN
fcis-5214	108	29	,	,	PUNCT
fcis-5214	108	30	the	the	DET
fcis-5214	108	31	information	information	NOUN
fcis-5214	108	32	about	about	ADP
fcis-5214	108	33	user	user	NOUN
fcis-5214	108	34	rating	rating	NOUN
fcis-5214	108	35	patterns	pattern	NOUN
fcis-5214	108	36	increases	increase	NOUN
fcis-5214	108	37	,	,	PUNCT
fcis-5214	108	38	but	but	CCONJ
fcis-5214	108	39	the	the	DET
fcis-5214	108	40	data	datum	NOUN
fcis-5214	108	41	density	density	NOUN
fcis-5214	108	42	decreases	decrease	VERB
fcis-5214	108	43	.	.	PUNCT
fcis-5214	109	1	figure	figure	NOUN
fcis-5214	109	2	1	1	NUM
fcis-5214	109	3	shows	show	VERB
fcis-5214	109	4	that	that	SCONJ
fcis-5214	109	5	the	the	DET
fcis-5214	109	6	performance	performance	NOUN
fcis-5214	109	7	of	of	ADP
fcis-5214	109	8	recommendation	recommendation	NOUN
fcis-5214	109	9	prediction	prediction	NOUN
fcis-5214	109	10	is	be	AUX
fcis-5214	109	11	not	not	PART
fcis-5214	109	12	satisfactory	satisfactory	ADJ
fcis-5214	109	13	when	when	SCONJ
fcis-5214	109	14	the	the	DET
fcis-5214	109	15	group	group	NOUN
fcis-5214	109	16	size	size	NOUN
fcis-5214	109	17	is	be	AUX
fcis-5214	109	18	very	very	ADV
fcis-5214	109	19	small	small	ADJ
fcis-5214	109	20	or	or	CCONJ
fcis-5214	109	21	very	very	ADV
fcis-5214	109	22	large	large	ADJ
fcis-5214	109	23	.	.	PUNCT
fcis-5214	110	1	the	the	DET
fcis-5214	110	2	proposed	propose	VERB
fcis-5214	110	3	method	method	NOUN
fcis-5214	110	4	in	in	ADP
fcis-5214	110	5	this	this	DET
fcis-5214	110	6	paper	paper	NOUN
fcis-5214	110	7	performs	perform	VERB
fcis-5214	110	8	better	well	ADV
fcis-5214	110	9	in	in	ADP
fcis-5214	110	10	the	the	DET
fcis-5214	110	11	ml100k	ml100k	PROPN
fcis-5214	110	12	dataset	dataset	VERB
fcis-5214	110	13	when	when	SCONJ
fcis-5214	110	14	the	the	DET
fcis-5214	110	15	group	group	NOUN
fcis-5214	110	16	size	size	NOUN
fcis-5214	110	17	k	k	PROPN
fcis-5214	110	18	is	be	AUX
fcis-5214	110	19	from	from	ADP
fcis-5214	110	20	20	20	NUM
fcis-5214	110	21	to	to	PART
fcis-5214	110	22	80	80	NUM
fcis-5214	110	23	.	.	PUNCT
fcis-5214	111	1	among	among	ADP
fcis-5214	111	2	them	they	PRON
fcis-5214	111	3	,	,	PUNCT
fcis-5214	111	4	the	the	DET
fcis-5214	111	5	best	good	ADJ
fcis-5214	111	6	performance	performance	NOUN
fcis-5214	111	7	is	be	AUX
fcis-5214	111	8	achieved	achieve	VERB
fcis-5214	111	9	at	at	ADP
fcis-5214	111	10	k	k	PROPN
fcis-5214	111	11	=	=	PROPN
fcis-5214	111	12	40	40	NUM
fcis-5214	111	13	.	.	PUNCT
fcis-5214	112	1	in	in	ADP
fcis-5214	112	2	the	the	DET
fcis-5214	112	3	ml1	ml1	PROPN
fcis-5214	112	4	m	m	NOUN
fcis-5214	112	5	dataset	dataset	NOUN
fcis-5214	112	6	,	,	PUNCT
fcis-5214	112	7	the	the	DET
fcis-5214	112	8	recommended	recommend	VERB
fcis-5214	112	9	prediction	prediction	NOUN
fcis-5214	112	10	performs	perform	VERB
fcis-5214	112	11	better	well	ADV
fcis-5214	112	12	when	when	SCONJ
fcis-5214	112	13	k	k	PROPN
fcis-5214	112	14	is	be	AUX
fcis-5214	112	15	80	80	NUM
fcis-5214	112	16	.	.	PUNCT
fcis-5214	113	1	this	this	DET
fcis-5214	113	2	result	result	NOUN
fcis-5214	113	3	indicates	indicate	VERB
fcis-5214	113	4	that	that	SCONJ
fcis-5214	113	5	an	an	DET
fcis-5214	113	6	appropriate	appropriate	ADJ
fcis-5214	113	7	grouping	grouping	NOUN
fcis-5214	113	8	size	size	NOUN
fcis-5214	113	9	k	k	PROPN
fcis-5214	113	10	can	can	AUX
fcis-5214	113	11	effectively	effectively	ADV
fcis-5214	113	12	improve	improve	VERB
fcis-5214	113	13	the	the	DET
fcis-5214	113	14	recommendation	recommendation	NOUN
fcis-5214	113	15	performance	performance	NOUN
fcis-5214	113	16	.	.	PUNCT
fcis-5214	114	1	4.4	4.4	NUM
fcis-5214	114	2	.	.	PUNCT
fcis-5214	115	1	analysis	analysis	NOUN
fcis-5214	115	2	of	of	ADP
fcis-5214	115	3	model	model	NOUN
fcis-5214	115	4	convergence	convergence	NOUN
fcis-5214	115	5	figure	figure	NOUN
fcis-5214	115	6	2	2	NUM
fcis-5214	115	7	.	.	PUNCT
fcis-5214	115	8	convergence	convergence	NOUN
fcis-5214	115	9	speed	speed	NOUN
fcis-5214	115	10	between	between	ADP
fcis-5214	115	11	cffr	cffr	PROPN
fcis-5214	115	12	and	and	CCONJ
fcis-5214	115	13	fedavg	fedavg	NOUN
fcis-5214	115	14	in	in	ADP
fcis-5214	115	15	addition	addition	NOUN
fcis-5214	115	16	to	to	ADP
fcis-5214	115	17	the	the	DET
fcis-5214	115	18	comparison	comparison	NOUN
fcis-5214	115	19	of	of	ADP
fcis-5214	115	20	the	the	DET
fcis-5214	115	21	effect	effect	NOUN
fcis-5214	115	22	of	of	ADP
fcis-5214	115	23	group	group	NOUN
fcis-5214	115	24	size	size	NOUN
fcis-5214	115	25	on	on	ADP
fcis-5214	115	26	performance	performance	NOUN
fcis-5214	115	27	,	,	PUNCT
fcis-5214	115	28	this	this	DET
fcis-5214	115	29	paper	paper	NOUN
fcis-5214	115	30	also	also	ADV
fcis-5214	115	31	analyzes	analyze	VERB
fcis-5214	115	32	the	the	DET
fcis-5214	115	33	convergence	convergence	NOUN
fcis-5214	115	34	speed	speed	NOUN
fcis-5214	115	35	difference	difference	NOUN
fcis-5214	115	36	between	between	ADP
fcis-5214	115	37	the	the	DET
fcis-5214	115	38	proposed	propose	VERB
fcis-5214	115	39	method	method	NOUN
fcis-5214	115	40	and	and	CCONJ
fcis-5214	115	41	the	the	DET
fcis-5214	115	42	comparison	comparison	NOUN
fcis-5214	115	43	method	method	NOUN
fcis-5214	115	44	fedavg	fedavg	NOUN
fcis-5214	115	45	[	[	X
fcis-5214	115	46	12	12	NUM
fcis-5214	115	47	]	]	PUNCT
fcis-5214	115	48	by	by	ADP
fcis-5214	115	49	comparing	compare	VERB
fcis-5214	115	50	the	the	DET
fcis-5214	115	51	variation	variation	NOUN
fcis-5214	115	52	of	of	ADP
fcis-5214	115	53	rmse	rmse	ADJ
fcis-5214	115	54	metrics	metric	NOUN
fcis-5214	115	55	with	with	ADP
fcis-5214	115	56	the	the	DET
fcis-5214	115	57	number	number	NOUN
fcis-5214	115	58	of	of	ADP
fcis-5214	115	59	model	model	NOUN
fcis-5214	115	60	aggregations	aggregation	NOUN
fcis-5214	115	61	under	under	ADP
fcis-5214	115	62	different	different	ADJ
fcis-5214	115	63	learning	learning	NOUN
fcis-5214	115	64	rate	rate	NOUN
fcis-5214	115	65	settings	setting	NOUN
fcis-5214	115	66	.	.	PUNCT
fcis-5214	116	1	since	since	SCONJ
fcis-5214	116	2	the	the	DET
fcis-5214	116	3	federated	federated	ADJ
fcis-5214	116	4	recommender	recommender	NOUN
fcis-5214	116	5	system	system	NOUN
fcis-5214	116	6	needs	need	VERB
fcis-5214	116	7	to	to	PART
fcis-5214	116	8	aggregate	aggregate	VERB
fcis-5214	116	9	updates	update	NOUN
fcis-5214	116	10	uploaded	upload	VERB
fcis-5214	116	11	by	by	ADP
fcis-5214	116	12	multiple	multiple	ADJ
fcis-5214	116	13	clients	client	NOUN
fcis-5214	116	14	to	to	PART
fcis-5214	116	15	perform	perform	VERB
fcis-5214	116	16	global	global	ADJ
fcis-5214	116	17	model	model	NOUN
fcis-5214	116	18	updates	update	NOUN
fcis-5214	116	19	,	,	PUNCT
fcis-5214	116	20	the	the	DET
fcis-5214	116	21	convergence	convergence	NOUN
fcis-5214	116	22	speed	speed	NOUN
fcis-5214	116	23	of	of	ADP
fcis-5214	116	24	the	the	DET
fcis-5214	116	25	model	model	NOUN
fcis-5214	116	26	is	be	AUX
fcis-5214	116	27	the	the	DET
fcis-5214	116	28	focus	focus	NOUN
fcis-5214	116	29	of	of	ADP
fcis-5214	116	30	the	the	DET
fcis-5214	116	31	scheme	scheme	NOUN
fcis-5214	116	32	design	design	NOUN
fcis-5214	116	33	.	.	PUNCT
fcis-5214	117	1	therefore	therefore	ADV
fcis-5214	117	2	,	,	PUNCT
fcis-5214	117	3	this	this	DET
fcis-5214	117	4	section	section	NOUN
fcis-5214	117	5	compares	compare	VERB
fcis-5214	117	6	the	the	DET
fcis-5214	117	7	changes	change	NOUN
fcis-5214	117	8	in	in	ADP
fcis-5214	117	9	model	model	NOUN
fcis-5214	117	10	accuracy	accuracy	NOUN
fcis-5214	117	11	between	between	ADP
fcis-5214	117	12	the	the	DET
fcis-5214	117	13	proposed	propose	VERB
fcis-5214	117	14	method	method	NOUN
fcis-5214	117	15	and	and	CCONJ
fcis-5214	117	16	fedavg	fedavg	NOUN
fcis-5214	117	17	with	with	ADP
fcis-5214	117	18	different	different	ADJ
fcis-5214	117	19	parameter	parameter	NOUN
fcis-5214	117	20	settings	setting	NOUN
fcis-5214	117	21	as	as	ADP
fcis-5214	117	22	the	the	DET
fcis-5214	117	23	number	number	NOUN
fcis-5214	117	24	of	of	ADP
fcis-5214	117	25	aggregations	aggregation	NOUN
fcis-5214	117	26	66	66	NUM
fcis-5214	117	27	increases	increase	NOUN
fcis-5214	117	28	.	.	PUNCT
fcis-5214	118	1	set	set	VERB
fcis-5214	118	2	the	the	DET
fcis-5214	118	3	group	group	NOUN
fcis-5214	118	4	size	size	NOUN
fcis-5214	118	5	k	k	PROPN
fcis-5214	118	6	=	=	PROPN
fcis-5214	118	7	200，𝛼	200，𝛼	PROPN
fcis-5214	118	8	=	=	SYM
fcis-5214	119	1	1.0，𝛾	1.0，𝛾	NUM
fcis-5214	120	1	=	=	SYM
fcis-5214	120	2	0.98，𝛽	0.98，𝛽	NUM
fcis-5214	120	3	=	=	SYM
fcis-5214	120	4	1	1	NUM
fcis-5214	120	5	−	−	NUM
fcis-5214	120	6	𝛼,and	𝛼,and	NUM
fcis-5214	120	7	the	the	DET
fcis-5214	120	8	learning	learning	NOUN
fcis-5214	120	9	rate	rate	NOUN
fcis-5214	120	10	is	be	AUX
fcis-5214	120	11	set	set	VERB
fcis-5214	120	12	to	to	ADP
fcis-5214	120	13	0.05	0.05	NUM
fcis-5214	120	14	and	and	CCONJ
fcis-5214	120	15	0.1	0.1	NUM
fcis-5214	120	16	.	.	PUNCT
fcis-5214	121	1	the	the	DET
fcis-5214	121	2	experimental	experimental	ADJ
fcis-5214	121	3	results	result	NOUN
fcis-5214	121	4	are	be	AUX
fcis-5214	121	5	shown	show	VERB
fcis-5214	121	6	in	in	ADP
fcis-5214	121	7	figure	figure	NOUN
fcis-5214	121	8	2	2	NUM
fcis-5214	121	9	.	.	PUNCT
fcis-5214	122	1	under	under	ADP
fcis-5214	122	2	different	different	ADJ
fcis-5214	122	3	learning	learning	NOUN
fcis-5214	122	4	rate	rate	NOUN
fcis-5214	122	5	settings	setting	NOUN
fcis-5214	122	6	,	,	PUNCT
fcis-5214	122	7	this	this	DET
fcis-5214	122	8	scheme	scheme	NOUN
fcis-5214	122	9	has	have	VERB
fcis-5214	122	10	a	a	DET
fcis-5214	122	11	faster	fast	ADJ
fcis-5214	122	12	convergence	convergence	NOUN
fcis-5214	122	13	speed	speed	NOUN
fcis-5214	122	14	than	than	ADP
fcis-5214	122	15	the	the	DET
fcis-5214	122	16	fedavg	fedavg	NOUN
fcis-5214	122	17	method	method	NOUN
fcis-5214	122	18	,	,	PUNCT
fcis-5214	122	19	and	and	CCONJ
fcis-5214	122	20	the	the	DET
fcis-5214	122	21	effect	effect	NOUN
fcis-5214	122	22	is	be	AUX
fcis-5214	122	23	obvious	obvious	ADJ
fcis-5214	122	24	at	at	ADP
fcis-5214	122	25	larger	large	ADJ
fcis-5214	122	26	learning	learning	NOUN
fcis-5214	122	27	rates	rate	NOUN
fcis-5214	122	28	.	.	PUNCT
fcis-5214	123	1	this	this	PRON
fcis-5214	123	2	is	be	AUX
fcis-5214	123	3	due	due	ADJ
fcis-5214	123	4	to	to	ADP
fcis-5214	123	5	the	the	DET
fcis-5214	123	6	fact	fact	NOUN
fcis-5214	123	7	that	that	SCONJ
fcis-5214	123	8	in	in	ADP
fcis-5214	123	9	the	the	DET
fcis-5214	123	10	early	early	ADJ
fcis-5214	123	11	stage	stage	NOUN
fcis-5214	123	12	of	of	ADP
fcis-5214	123	13	federation	federation	NOUN
fcis-5214	123	14	model	model	NOUN
fcis-5214	123	15	training	training	NOUN
fcis-5214	123	16	,	,	PUNCT
fcis-5214	123	17	the	the	DET
fcis-5214	123	18	performance	performance	NOUN
fcis-5214	123	19	of	of	ADP
fcis-5214	123	20	the	the	DET
fcis-5214	123	21	locally	locally	ADV
fcis-5214	123	22	trained	train	VERB
fcis-5214	123	23	model	model	NOUN
fcis-5214	123	24	varies	vary	VERB
fcis-5214	123	25	among	among	ADP
fcis-5214	123	26	clients	client	NOUN
fcis-5214	123	27	,	,	PUNCT
fcis-5214	123	28	and	and	CCONJ
fcis-5214	123	29	the	the	DET
fcis-5214	123	30	aggregated	aggregated	ADJ
fcis-5214	123	31	model	model	NOUN
fcis-5214	123	32	is	be	AUX
fcis-5214	123	33	more	more	ADV
fcis-5214	123	34	affected	affect	VERB
fcis-5214	123	35	by	by	ADP
fcis-5214	123	36	the	the	DET
fcis-5214	123	37	performance	performance	NOUN
fcis-5214	123	38	of	of	ADP
fcis-5214	123	39	each	each	DET
fcis-5214	123	40	client	client	NOUN
fcis-5214	123	41	when	when	SCONJ
fcis-5214	123	42	the	the	DET
fcis-5214	123	43	learning	learning	NOUN
fcis-5214	123	44	rate	rate	NOUN
fcis-5214	123	45	is	be	AUX
fcis-5214	123	46	larger	large	ADJ
fcis-5214	123	47	.	.	PUNCT
fcis-5214	124	1	the	the	DET
fcis-5214	124	2	proposed	propose	VERB
fcis-5214	124	3	method	method	NOUN
fcis-5214	124	4	can	can	AUX
fcis-5214	124	5	accelerate	accelerate	VERB
fcis-5214	124	6	the	the	DET
fcis-5214	124	7	convergence	convergence	NOUN
fcis-5214	124	8	of	of	ADP
fcis-5214	124	9	the	the	DET
fcis-5214	124	10	model	model	NOUN
fcis-5214	124	11	by	by	ADP
fcis-5214	124	12	weakening	weaken	VERB
fcis-5214	124	13	the	the	DET
fcis-5214	124	14	impact	impact	NOUN
fcis-5214	124	15	of	of	ADP
fcis-5214	124	16	updates	update	NOUN
fcis-5214	124	17	of	of	ADP
fcis-5214	124	18	the	the	DET
fcis-5214	124	19	poorly	poorly	ADV
fcis-5214	124	20	performing	perform	VERB
fcis-5214	124	21	local	local	ADJ
fcis-5214	124	22	models	model	NOUN
fcis-5214	124	23	in	in	ADP
fcis-5214	124	24	the	the	DET
fcis-5214	124	25	early	early	ADJ
fcis-5214	124	26	stage	stage	NOUN
fcis-5214	124	27	of	of	ADP
fcis-5214	124	28	training	training	NOUN
fcis-5214	124	29	.	.	PUNCT
fcis-5214	125	1	4.5	4.5	NUM
fcis-5214	125	2	.	.	PUNCT
fcis-5214	125	3	ablation	ablation	NOUN
fcis-5214	125	4	study	study	NOUN
fcis-5214	125	5	figure	figure	NOUN
fcis-5214	125	6	3	3	NUM
fcis-5214	125	7	.	.	PUNCT
fcis-5214	126	1	contributions	contribution	NOUN
fcis-5214	126	2	of	of	ADP
fcis-5214	126	3	the	the	DET
fcis-5214	126	4	individual	individual	ADJ
fcis-5214	126	5	components	component	NOUN
fcis-5214	126	6	of	of	ADP
fcis-5214	126	7	cffr	cffr	PROPN
fcis-5214	126	8	the	the	DET
fcis-5214	126	9	proposed	propose	VERB
fcis-5214	126	10	federal	federal	ADJ
fcis-5214	126	11	recommendation	recommendation	NOUN
fcis-5214	126	12	algorithm	algorithm	NOUN
fcis-5214	126	13	in	in	ADP
fcis-5214	126	14	this	this	DET
fcis-5214	126	15	paper	paper	NOUN
fcis-5214	126	16	contains	contain	VERB
fcis-5214	126	17	two	two	NUM
fcis-5214	126	18	main	main	ADJ
fcis-5214	126	19	modules	module	NOUN
fcis-5214	126	20	,	,	PUNCT
fcis-5214	126	21	the	the	DET
fcis-5214	126	22	first	first	ADJ
fcis-5214	126	23	part	part	NOUN
fcis-5214	126	24	is	be	AUX
fcis-5214	126	25	the	the	DET
fcis-5214	126	26	grouping	group	VERB
fcis-5214	126	27	strategy	strategy	NOUN
fcis-5214	126	28	,	,	PUNCT
fcis-5214	126	29	and	and	CCONJ
fcis-5214	126	30	the	the	DET
fcis-5214	126	31	second	second	ADJ
fcis-5214	126	32	part	part	NOUN
fcis-5214	126	33	is	be	AUX
fcis-5214	126	34	the	the	DET
fcis-5214	126	35	parameter	parameter	NOUN
fcis-5214	126	36	aggregation	aggregation	NOUN
fcis-5214	126	37	mechanism	mechanism	NOUN
fcis-5214	126	38	that	that	PRON
fcis-5214	126	39	incorporates	incorporate	VERB
fcis-5214	126	40	the	the	DET
fcis-5214	126	41	performance	performance	NOUN
fcis-5214	126	42	differences	difference	NOUN
fcis-5214	126	43	and	and	CCONJ
fcis-5214	126	44	data	datum	NOUN
fcis-5214	126	45	differences	difference	NOUN
fcis-5214	126	46	among	among	ADP
fcis-5214	126	47	client	client	NOUN
fcis-5214	126	48	models	model	NOUN
fcis-5214	126	49	.	.	PUNCT
fcis-5214	127	1	in	in	ADP
fcis-5214	127	2	order	order	NOUN
fcis-5214	127	3	to	to	PART
fcis-5214	127	4	explore	explore	VERB
fcis-5214	127	5	the	the	DET
fcis-5214	127	6	performance	performance	NOUN
fcis-5214	127	7	impact	impact	NOUN
fcis-5214	127	8	of	of	ADP
fcis-5214	127	9	the	the	DET
fcis-5214	127	10	two	two	NUM
fcis-5214	127	11	modules	module	NOUN
fcis-5214	127	12	on	on	ADP
fcis-5214	127	13	the	the	DET
fcis-5214	127	14	final	final	ADJ
fcis-5214	127	15	solution	solution	NOUN
fcis-5214	127	16	,	,	PUNCT
fcis-5214	127	17	this	this	DET
fcis-5214	127	18	section	section	NOUN
fcis-5214	127	19	conducts	conduct	VERB
fcis-5214	127	20	ablation	ablation	NOUN
fcis-5214	127	21	experiments	experiment	NOUN
fcis-5214	127	22	on	on	ADP
fcis-5214	127	23	two	two	NUM
fcis-5214	127	24	datasets	dataset	NOUN
fcis-5214	127	25	,	,	PUNCT
fcis-5214	127	26	ml100k	ml100k	PROPN
fcis-5214	127	27	and	and	CCONJ
fcis-5214	127	28	ml1	ml1	PROPN
fcis-5214	127	29	m	m	PROPN
fcis-5214	127	30	,	,	PUNCT
fcis-5214	127	31	and	and	CCONJ
fcis-5214	127	32	analyzes	analyze	VERB
fcis-5214	127	33	the	the	DET
fcis-5214	127	34	final	final	ADJ
fcis-5214	127	35	results	result	NOUN
fcis-5214	127	36	,	,	PUNCT
fcis-5214	127	37	which	which	PRON
fcis-5214	127	38	are	be	AUX
fcis-5214	127	39	shown	show	VERB
fcis-5214	127	40	in	in	ADP
fcis-5214	127	41	figure	figure	NOUN
fcis-5214	127	42	3	3	NUM
fcis-5214	127	43	.	.	PUNCT
fcis-5214	128	1	in	in	ADP
fcis-5214	128	2	fig	fig	NOUN
fcis-5214	128	3	.	.	PUNCT
fcis-5214	129	1	3	3	NUM
fcis-5214	129	2	,	,	PUNCT
fcis-5214	129	3	(	(	PUNCT
fcis-5214	129	4	-grouping	-grouping	NOUN
fcis-5214	129	5	)	)	PUNCT
fcis-5214	129	6	is	be	AUX
fcis-5214	129	7	the	the	DET
fcis-5214	129	8	removal	removal	NOUN
fcis-5214	129	9	of	of	ADP
fcis-5214	129	10	user	user	NOUN
fcis-5214	129	11	grouping	grouping	NOUN
fcis-5214	129	12	module	module	NOUN
fcis-5214	129	13	and	and	CCONJ
fcis-5214	129	14	(	(	PUNCT
fcis-5214	129	15	-pa	-pa	NOUN
fcis-5214	129	16	)	)	PUNCT
fcis-5214	129	17	is	be	AUX
fcis-5214	129	18	the	the	DET
fcis-5214	129	19	removal	removal	NOUN
fcis-5214	129	20	of	of	ADP
fcis-5214	129	21	parameter	parameter	NOUN
fcis-5214	129	22	aggregation	aggregation	NOUN
fcis-5214	129	23	module	module	NOUN
fcis-5214	129	24	.	.	PUNCT
fcis-5214	130	1	from	from	ADP
fcis-5214	130	2	fig	fig	NOUN
fcis-5214	130	3	.	.	PUNCT
fcis-5214	131	1	3	3	NUM
fcis-5214	131	2	,	,	PUNCT
fcis-5214	131	3	it	it	PRON
fcis-5214	131	4	can	can	AUX
fcis-5214	131	5	be	be	AUX
fcis-5214	131	6	observed	observe	VERB
fcis-5214	131	7	that	that	SCONJ
fcis-5214	131	8	the	the	DET
fcis-5214	131	9	scheme	scheme	NOUN
fcis-5214	131	10	in	in	ADP
fcis-5214	131	11	this	this	DET
fcis-5214	131	12	paper	paper	NOUN
fcis-5214	131	13	achieves	achieve	VERB
fcis-5214	131	14	the	the	DET
fcis-5214	131	15	best	good	ADJ
fcis-5214	131	16	results	result	NOUN
fcis-5214	131	17	on	on	ADP
fcis-5214	131	18	both	both	DET
fcis-5214	131	19	datasets	dataset	NOUN
fcis-5214	131	20	,	,	PUNCT
fcis-5214	131	21	which	which	PRON
fcis-5214	131	22	verifies	verify	VERB
fcis-5214	131	23	the	the	DET
fcis-5214	131	24	importance	importance	NOUN
fcis-5214	131	25	of	of	ADP
fcis-5214	131	26	the	the	DET
fcis-5214	131	27	two	two	NUM
fcis-5214	131	28	key	key	ADJ
fcis-5214	131	29	modules	module	NOUN
fcis-5214	131	30	.	.	PUNCT
fcis-5214	132	1	among	among	ADP
fcis-5214	132	2	them	they	PRON
fcis-5214	132	3	,	,	PUNCT
fcis-5214	132	4	the	the	DET
fcis-5214	132	5	model	model	NOUN
fcis-5214	132	6	performance	performance	NOUN
fcis-5214	132	7	degrades	degrade	VERB
fcis-5214	132	8	significantly	significantly	ADV
fcis-5214	132	9	in	in	ADP
fcis-5214	132	10	the	the	DET
fcis-5214	132	11	experiments	experiment	NOUN
fcis-5214	132	12	where	where	SCONJ
fcis-5214	132	13	the	the	DET
fcis-5214	132	14	user	user	NOUN
fcis-5214	132	15	grouping	grouping	NOUN
fcis-5214	132	16	module	module	NOUN
fcis-5214	132	17	is	be	AUX
fcis-5214	132	18	removed	remove	VERB
fcis-5214	132	19	,	,	PUNCT
fcis-5214	132	20	which	which	PRON
fcis-5214	132	21	proves	prove	VERB
fcis-5214	132	22	that	that	SCONJ
fcis-5214	132	23	the	the	DET
fcis-5214	132	24	lack	lack	NOUN
fcis-5214	132	25	of	of	ADP
fcis-5214	132	26	personalized	personalized	ADJ
fcis-5214	132	27	grouping	grouping	NOUN
fcis-5214	132	28	significantly	significantly	ADV
fcis-5214	132	29	reduces	reduce	VERB
fcis-5214	132	30	the	the	DET
fcis-5214	132	31	learning	learning	NOUN
fcis-5214	132	32	capability	capability	NOUN
fcis-5214	132	33	of	of	ADP
fcis-5214	132	34	the	the	DET
fcis-5214	132	35	framework	framework	NOUN
fcis-5214	132	36	.	.	PUNCT
fcis-5214	133	1	the	the	DET
fcis-5214	133	2	absence	absence	NOUN
fcis-5214	133	3	of	of	ADP
fcis-5214	133	4	the	the	DET
fcis-5214	133	5	parameter	parameter	NOUN
fcis-5214	133	6	aggregation	aggregation	NOUN
fcis-5214	133	7	method	method	NOUN
fcis-5214	133	8	still	still	ADV
fcis-5214	133	9	significantly	significantly	ADV
fcis-5214	133	10	degrades	degrade	VERB
fcis-5214	133	11	the	the	DET
fcis-5214	133	12	model	model	NOUN
fcis-5214	133	13	performance	performance	NOUN
fcis-5214	133	14	for	for	ADP
fcis-5214	133	15	the	the	DET
fcis-5214	133	16	same	same	ADJ
fcis-5214	133	17	number	number	NOUN
fcis-5214	133	18	of	of	ADP
fcis-5214	133	19	aggregation	aggregation	NOUN
fcis-5214	133	20	rounds	round	NOUN
fcis-5214	133	21	,	,	PUNCT
fcis-5214	133	22	demonstrating	demonstrate	VERB
fcis-5214	133	23	the	the	DET
fcis-5214	133	24	importance	importance	NOUN
fcis-5214	133	25	of	of	ADP
fcis-5214	133	26	the	the	DET
fcis-5214	133	27	impact	impact	NOUN
fcis-5214	133	28	of	of	ADP
fcis-5214	133	29	local	local	ADJ
fcis-5214	133	30	model	model	NOUN
fcis-5214	133	31	performance	performance	NOUN
fcis-5214	133	32	on	on	ADP
fcis-5214	133	33	the	the	DET
fcis-5214	133	34	aggregated	aggregate	VERB
fcis-5214	133	35	global	global	ADJ
fcis-5214	133	36	model	model	NOUN
fcis-5214	133	37	when	when	SCONJ
fcis-5214	133	38	the	the	DET
fcis-5214	133	39	model	model	NOUN
fcis-5214	133	40	is	be	AUX
fcis-5214	133	41	aggregated	aggregate	VERB
fcis-5214	133	42	.	.	PUNCT
fcis-5214	134	1	each	each	PRON
fcis-5214	134	2	of	of	ADP
fcis-5214	134	3	the	the	DET
fcis-5214	134	4	modules	module	NOUN
fcis-5214	134	5	demonstrated	demonstrate	VERB
fcis-5214	134	6	by	by	ADP
fcis-5214	134	7	the	the	DET
fcis-5214	134	8	ablation	ablation	NOUN
fcis-5214	134	9	experiments	experiment	NOUN
fcis-5214	134	10	improves	improve	VERB
fcis-5214	134	11	the	the	DET
fcis-5214	134	12	model	model	NOUN
fcis-5214	134	13	performance	performance	NOUN
fcis-5214	134	14	in	in	ADP
fcis-5214	134	15	a	a	DET
fcis-5214	134	16	different	different	ADJ
fcis-5214	134	17	way	way	NOUN
fcis-5214	134	18	and	and	CCONJ
fcis-5214	134	19	significantly	significantly	ADV
fcis-5214	134	20	.	.	PUNCT
fcis-5214	135	1	figure	figure	VERB
fcis-5214	135	2	4	4	NUM
fcis-5214	135	3	.	.	PUNCT
fcis-5214	135	4	comparison	comparison	NOUN
fcis-5214	135	5	of	of	ADP
fcis-5214	135	6	model	model	NOUN
fcis-5214	135	7	convergence	convergence	NOUN
fcis-5214	135	8	performance	performance	NOUN
fcis-5214	135	9	figure	figure	NOUN
fcis-5214	135	10	4	4	NUM
fcis-5214	135	11	shows	show	VERB
fcis-5214	135	12	the	the	DET
fcis-5214	135	13	change	change	NOUN
fcis-5214	135	14	of	of	ADP
fcis-5214	135	15	model	model	NOUN
fcis-5214	135	16	convergence	convergence	NOUN
fcis-5214	135	17	after	after	ADP
fcis-5214	135	18	removing	remove	VERB
fcis-5214	135	19	each	each	DET
fcis-5214	135	20	module	module	NOUN
fcis-5214	135	21	on	on	ADP
fcis-5214	135	22	the	the	DET
fcis-5214	135	23	ml-100k	ml-100k	PROPN
fcis-5214	135	24	dataset	dataset	NOUN
fcis-5214	135	25	.	.	PUNCT
fcis-5214	136	1	when	when	SCONJ
fcis-5214	136	2	the	the	DET
fcis-5214	136	3	parameter	parameter	NOUN
fcis-5214	136	4	aggregation	aggregation	NOUN
fcis-5214	136	5	module	module	NOUN
fcis-5214	136	6	(	(	PUNCT
fcis-5214	136	7	-pa	-pa	NOUN
fcis-5214	136	8	)	)	PUNCT
fcis-5214	136	9	proposed	propose	VERB
fcis-5214	136	10	in	in	ADP
fcis-5214	136	11	this	this	DET
fcis-5214	136	12	scheme	scheme	NOUN
fcis-5214	136	13	is	be	AUX
fcis-5214	136	14	removed	remove	VERB
fcis-5214	136	15	,	,	PUNCT
fcis-5214	136	16	the	the	DET
fcis-5214	136	17	convergence	convergence	NOUN
fcis-5214	136	18	of	of	ADP
fcis-5214	136	19	the	the	DET
fcis-5214	136	20	model	model	NOUN
fcis-5214	136	21	slows	slow	VERB
fcis-5214	136	22	down	down	ADP
fcis-5214	136	23	slightly	slightly	ADV
fcis-5214	136	24	,	,	PUNCT
fcis-5214	136	25	but	but	CCONJ
fcis-5214	136	26	when	when	SCONJ
fcis-5214	136	27	the	the	DET
fcis-5214	136	28	user	user	NOUN
fcis-5214	136	29	grouping	grouping	NOUN
fcis-5214	136	30	module	module	NOUN
fcis-5214	136	31	(	(	PUNCT
fcis-5214	136	32	-grouping	-grouping	NOUN
fcis-5214	136	33	)	)	PUNCT
fcis-5214	136	34	is	be	AUX
fcis-5214	136	35	removed	remove	VERB
fcis-5214	136	36	,	,	PUNCT
fcis-5214	136	37	i.e.	i.e.	X
fcis-5214	136	38	,	,	PUNCT
fcis-5214	136	39	the	the	DET
fcis-5214	136	40	same	same	ADJ
fcis-5214	136	41	number	number	NOUN
fcis-5214	136	42	of	of	ADP
fcis-5214	136	43	users	user	NOUN
fcis-5214	136	44	are	be	AUX
fcis-5214	136	45	randomly	randomly	ADV
fcis-5214	136	46	selected	select	VERB
fcis-5214	136	47	to	to	PART
fcis-5214	136	48	collaborate	collaborate	VERB
fcis-5214	136	49	in	in	ADP
fcis-5214	136	50	model	model	NOUN
fcis-5214	136	51	training	training	NOUN
fcis-5214	136	52	,	,	PUNCT
fcis-5214	136	53	the	the	DET
fcis-5214	136	54	convergence	convergence	NOUN
fcis-5214	136	55	of	of	ADP
fcis-5214	136	56	the	the	DET
fcis-5214	136	57	model	model	NOUN
fcis-5214	136	58	slows	slow	VERB
fcis-5214	136	59	down	down	ADP
fcis-5214	136	60	significantly	significantly	ADV
fcis-5214	136	61	,	,	PUNCT
fcis-5214	136	62	and	and	CCONJ
fcis-5214	136	63	at	at	ADP
fcis-5214	136	64	this	this	DET
fcis-5214	136	65	time	time	NOUN
fcis-5214	136	66	there	there	PRON
fcis-5214	136	67	are	be	VERB
fcis-5214	136	68	large	large	ADJ
fcis-5214	136	69	differences	difference	NOUN
fcis-5214	136	70	in	in	ADP
fcis-5214	136	71	the	the	DET
fcis-5214	136	72	preferences	preference	NOUN
fcis-5214	136	73	and	and	CCONJ
fcis-5214	136	74	scoring	score	VERB
fcis-5214	136	75	patterns	pattern	NOUN
fcis-5214	136	76	of	of	ADP
fcis-5214	136	77	the	the	DET
fcis-5214	136	78	users	user	NOUN
fcis-5214	136	79	participating	participate	VERB
fcis-5214	136	80	in	in	ADP
fcis-5214	136	81	the	the	DET
fcis-5214	136	82	federal	federal	ADJ
fcis-5214	136	83	training	training	NOUN
fcis-5214	136	84	,	,	PUNCT
fcis-5214	136	85	resulting	result	VERB
fcis-5214	136	86	in	in	ADP
fcis-5214	136	87	the	the	DET
fcis-5214	136	88	inability	inability	NOUN
fcis-5214	136	89	to	to	PART
fcis-5214	136	90	get	get	VERB
fcis-5214	136	91	the	the	DET
fcis-5214	136	92	maximum	maximum	ADJ
fcis-5214	136	93	benefit	benefit	NOUN
fcis-5214	136	94	from	from	ADP
fcis-5214	136	95	the	the	DET
fcis-5214	136	96	training	training	NOUN
fcis-5214	136	97	of	of	ADP
fcis-5214	136	98	other	other	ADJ
fcis-5214	136	99	users	user	NOUN
fcis-5214	136	100	,	,	PUNCT
fcis-5214	136	101	thus	thus	ADV
fcis-5214	136	102	affecting	affect	VERB
fcis-5214	136	103	the	the	DET
fcis-5214	136	104	global	global	ADJ
fcis-5214	136	105	model	model	NOUN
fcis-5214	136	106	this	this	PRON
fcis-5214	136	107	has	have	VERB
fcis-5214	136	108	an	an	DET
fcis-5214	136	109	impact	impact	NOUN
fcis-5214	136	110	on	on	ADP
fcis-5214	136	111	the	the	DET
fcis-5214	136	112	convergence	convergence	NOUN
fcis-5214	136	113	speed	speed	NOUN
fcis-5214	136	114	of	of	ADP
fcis-5214	136	115	the	the	DET
fcis-5214	136	116	global	global	ADJ
fcis-5214	136	117	model	model	NOUN
fcis-5214	136	118	.	.	PUNCT
fcis-5214	137	1	compared	compare	VERB
fcis-5214	137	2	with	with	ADP
fcis-5214	137	3	traditional	traditional	ADJ
fcis-5214	137	4	training	training	NOUN
fcis-5214	137	5	methods	method	NOUN
fcis-5214	137	6	,	,	PUNCT
fcis-5214	137	7	i.e.	i.e.	X
fcis-5214	137	8	,	,	PUNCT
fcis-5214	137	9	removing	remove	VERB
fcis-5214	137	10	two	two	NUM
fcis-5214	137	11	modules	module	NOUN
fcis-5214	137	12	(	(	PUNCT
fcis-5214	137	13	grouping	grouping	NOUN
fcis-5214	137	14	and	and	CCONJ
fcis-5214	137	15	parameter	parameter	NOUN
fcis-5214	137	16	aggregation	aggregation	NOUN
fcis-5214	137	17	)	)	PUNCT
fcis-5214	137	18	,	,	PUNCT
fcis-5214	137	19	the	the	DET
fcis-5214	137	20	proposed	propose	VERB
fcis-5214	137	21	method	method	NOUN
fcis-5214	137	22	has	have	VERB
fcis-5214	137	23	significant	significant	ADJ
fcis-5214	137	24	advantages	advantage	NOUN
fcis-5214	137	25	in	in	ADP
fcis-5214	137	26	model	model	NOUN
fcis-5214	137	27	convergence	convergence	NOUN
fcis-5214	137	28	speed	speed	NOUN
fcis-5214	137	29	and	and	CCONJ
fcis-5214	137	30	prediction	prediction	NOUN
fcis-5214	137	31	accuracy	accuracy	NOUN
fcis-5214	137	32	5	5	NUM
fcis-5214	137	33	.	.	PUNCT
fcis-5214	137	34	conclusion	conclusion	NOUN
fcis-5214	137	35	in	in	ADP
fcis-5214	137	36	this	this	DET
fcis-5214	137	37	paper	paper	NOUN
fcis-5214	137	38	,	,	PUNCT
fcis-5214	137	39	we	we	PRON
fcis-5214	137	40	propose	propose	VERB
fcis-5214	137	41	a	a	DET
fcis-5214	137	42	communication	communication	NOUN
fcis-5214	137	43	-	-	PUNCT
fcis-5214	137	44	efficient	efficient	ADJ
fcis-5214	137	45	personalized	personalized	ADJ
fcis-5214	137	46	federation	federation	NOUN
fcis-5214	137	47	recommendation	recommendation	NOUN
fcis-5214	137	48	method	method	NOUN
fcis-5214	137	49	that	that	PRON
fcis-5214	137	50	uses	use	VERB
fcis-5214	137	51	adaptive	adaptive	ADJ
fcis-5214	137	52	client	client	NOUN
fcis-5214	137	53	grouping	group	VERB
fcis-5214	137	54	to	to	PART
fcis-5214	137	55	enable	enable	VERB
fcis-5214	137	56	users	user	NOUN
fcis-5214	137	57	with	with	ADP
fcis-5214	137	58	the	the	DET
fcis-5214	137	59	same	same	ADJ
fcis-5214	137	60	interests	interest	NOUN
fcis-5214	137	61	to	to	PART
fcis-5214	137	62	collaborate	collaborate	VERB
fcis-5214	137	63	in	in	ADP
fcis-5214	137	64	training	train	VERB
fcis-5214	137	65	the	the	DET
fcis-5214	137	66	model	model	NOUN
fcis-5214	137	67	,	,	PUNCT
fcis-5214	137	68	and	and	CCONJ
fcis-5214	137	69	assigns	assign	VERB
fcis-5214	137	70	different	different	ADJ
fcis-5214	137	71	aggregation	aggregation	NOUN
fcis-5214	137	72	weights	weight	NOUN
fcis-5214	137	73	to	to	ADP
fcis-5214	137	74	the	the	DET
fcis-5214	137	75	current	current	ADJ
fcis-5214	137	76	round	round	NOUN
fcis-5214	137	77	of	of	ADP
fcis-5214	137	78	weighted	weighted	ADJ
fcis-5214	137	79	aggregation	aggregation	NOUN
fcis-5214	137	80	of	of	ADP
fcis-5214	137	81	each	each	DET
fcis-5214	137	82	client	client	NOUN
fcis-5214	137	83	's	's	PART
fcis-5214	137	84	updated	update	VERB
fcis-5214	137	85	model	model	NOUN
fcis-5214	137	86	according	accord	VERB
fcis-5214	137	87	to	to	ADP
fcis-5214	137	88	the	the	DET
fcis-5214	137	89	amount	amount	NOUN
fcis-5214	137	90	of	of	ADP
fcis-5214	137	91	data	datum	NOUN
fcis-5214	137	92	and	and	CCONJ
fcis-5214	137	93	performance	performance	NOUN
fcis-5214	137	94	status	status	NOUN
fcis-5214	137	95	among	among	ADP
fcis-5214	137	96	different	different	ADJ
fcis-5214	137	97	clients	client	NOUN
fcis-5214	137	98	in	in	ADP
fcis-5214	137	99	training	training	NOUN
fcis-5214	137	100	,	,	PUNCT
fcis-5214	137	101	which	which	PRON
fcis-5214	137	102	reduces	reduce	VERB
fcis-5214	137	103	the	the	DET
fcis-5214	137	104	communication	communication	NOUN
fcis-5214	137	105	burden	burden	NOUN
fcis-5214	137	106	and	and	CCONJ
fcis-5214	137	107	accelerates	accelerate	VERB
fcis-5214	137	108	the	the	DET
fcis-5214	137	109	model	model	NOUN
fcis-5214	137	110	training	training	NOUN
fcis-5214	137	111	at	at	ADP
fcis-5214	137	112	the	the	DET
fcis-5214	137	113	same	same	ADJ
fcis-5214	137	114	time	time	NOUN
fcis-5214	137	115	.	.	PUNCT
fcis-5214	138	1	we	we	PRON
fcis-5214	138	2	have	have	AUX
fcis-5214	138	3	conducted	conduct	VERB
fcis-5214	138	4	experiments	experiment	NOUN
fcis-5214	138	5	on	on	ADP
fcis-5214	138	6	real	real	ADJ
fcis-5214	138	7	datasets	dataset	NOUN
fcis-5214	138	8	,	,	PUNCT
fcis-5214	138	9	and	and	CCONJ
fcis-5214	138	10	the	the	DET
fcis-5214	138	11	experimental	experimental	ADJ
fcis-5214	138	12	results	result	NOUN
fcis-5214	138	13	demonstrate	demonstrate	VERB
fcis-5214	138	14	the	the	DET
fcis-5214	138	15	effectiveness	effectiveness	NOUN
fcis-5214	138	16	of	of	ADP
fcis-5214	138	17	the	the	DET
fcis-5214	138	18	proposed	propose	VERB
fcis-5214	138	19	method	method	NOUN
fcis-5214	138	20	in	in	ADP
fcis-5214	138	21	this	this	DET
fcis-5214	138	22	paper	paper	NOUN
fcis-5214	138	23	.	.	PUNCT
fcis-5214	139	1	references	reference	NOUN
fcis-5214	139	2	[	[	X
fcis-5214	139	3	1	1	X
fcis-5214	139	4	]	]	PUNCT
fcis-5214	139	5	j.	j.	PROPN
fcis-5214	139	6	koneˇcn`y	koneˇcn`y	PROPN
fcis-5214	139	7	,	,	PUNCT
fcis-5214	139	8	h.b	h.b	PROPN
fcis-5214	139	9	.	.	PROPN
fcis-5214	139	10	mcmahan	mcmahan	PROPN
fcis-5214	139	11	,	,	PUNCT
fcis-5214	139	12	et	et	PROPN
fcis-5214	139	13	al	al	PROPN
fcis-5214	139	14	.	.	PROPN
fcis-5214	139	15	:	:	PUNCT
fcis-5214	139	16	federated	federated	ADJ
fcis-5214	139	17	learning	learning	NOUN
fcis-5214	139	18	:	:	PUNCT
fcis-5214	139	19	strategies	strategy	NOUN
fcis-5214	139	20	for	for	ADP
fcis-5214	139	21	improving	improve	VERB
fcis-5214	139	22	communication	communication	NOUN
fcis-5214	139	23	efficiency	efficiency	NOUN
fcis-5214	139	24	.	.	PUNCT
fcis-5214	140	1	arxiv	arxiv	PROPN
fcis-5214	140	2	preprint	preprint	VERB
fcis-5214	140	3	arxiv:1610.05492	arxiv:1610.05492	PROPN
fcis-5214	140	4	,	,	PUNCT
fcis-5214	140	5	2016	2016	NUM
fcis-5214	140	6	.	.	PUNCT
fcis-5214	141	1	[	[	X
fcis-5214	141	2	2	2	NUM
fcis-5214	141	3	]	]	PUNCT
fcis-5214	141	4	m.	m.	NOUN
fcis-5214	141	5	ammad	ammad	PROPN
fcis-5214	141	6	-	-	PUNCT
fcis-5214	141	7	ud	ud	ADP
fcis-5214	141	8	-	-	PUNCT
fcis-5214	141	9	din	din	NOUN
fcis-5214	141	10	,	,	PUNCT
fcis-5214	141	11	e.	e.	PROPN
fcis-5214	141	12	ivannikova	ivannikova	PROPN
fcis-5214	141	13	,	,	PUNCT
fcis-5214	141	14	et	et	PROPN
fcis-5214	141	15	al	al	PROPN
fcis-5214	141	16	.	.	PROPN
fcis-5214	141	17	:	:	PUNCT
fcis-5214	141	18	federated	federate	VERB
fcis-5214	141	19	collaborative	collaborative	ADJ
fcis-5214	141	20	filtering	filtering	NOUN
fcis-5214	141	21	for	for	ADP
fcis-5214	141	22	privacy	privacy	NOUN
fcis-5214	141	23	-	-	PUNCT
fcis-5214	141	24	preserving	preserve	VERB
fcis-5214	141	25	personalized	personalized	ADJ
fcis-5214	141	26	recommendation	recommendation	NOUN
fcis-5214	141	27	system	system	NOUN
fcis-5214	141	28	.	.	PUNCT
fcis-5214	142	1	arxiv	arxiv	PROPN
fcis-5214	142	2	preprint	preprint	PROPN
fcis-5214	142	3	arxiv:1901.09888	arxiv:1901.09888	PROPN
fcis-5214	142	4	,	,	PUNCT
fcis-5214	142	5	2019	2019	NUM
fcis-5214	142	6	.	.	PUNCT
fcis-5214	143	1	[	[	X
fcis-5214	143	2	3	3	NUM
fcis-5214	143	3	]	]	PUNCT
fcis-5214	143	4	a.	a.	NOUN
fcis-5214	143	5	flanagan	flanagan	PROPN
fcis-5214	143	6	,	,	PUNCT
fcis-5214	143	7	w.	w.	PROPN
fcis-5214	143	8	oyomno	oyomno	PROPN
fcis-5214	143	9	,	,	PUNCT
fcis-5214	143	10	et	et	PROPN
fcis-5214	143	11	a.l	a.l	NOUN
fcis-5214	143	12	:	:	PUNCT
fcis-5214	143	13	federated	federated	ADJ
fcis-5214	143	14	multi	multi	ADJ
fcis-5214	143	15	-	-	ADJ
fcis-5214	143	16	view	view	ADJ
fcis-5214	143	17	matrix	matrix	NOUN
fcis-5214	143	18	factorization	factorization	NOUN
fcis-5214	143	19	for	for	ADP
fcis-5214	143	20	personalized	personalized	ADJ
fcis-5214	143	21	recommendations	recommendation	NOUN
fcis-5214	143	22	.	.	PUNCT
fcis-5214	144	1	arxiv	arxiv	PROPN
fcis-5214	144	2	preprint	preprint	VERB
fcis-5214	144	3	arxiv:2004.04256	arxiv:2004.04256	PROPN
fcis-5214	144	4	,	,	PUNCT
fcis-5214	144	5	2020	2020	NUM
fcis-5214	144	6	.	.	PUNCT
fcis-5214	145	1	[	[	X
fcis-5214	145	2	4	4	NUM
fcis-5214	145	3	]	]	PUNCT
fcis-5214	145	4	d.	d.	NOUN
fcis-5214	145	5	chai	chai	PROPN
fcis-5214	145	6	,	,	PUNCT
fcis-5214	145	7	l.	l.	PROPN
fcis-5214	145	8	wang	wang	PROPN
fcis-5214	145	9	,	,	PUNCT
fcis-5214	145	10	k.	k.	PROPN
fcis-5214	145	11	chen	chen	PROPN
fcis-5214	145	12	,	,	PUNCT
fcis-5214	145	13	and	and	CCONJ
fcis-5214	145	14	q.	q.	PROPN
fcis-5214	145	15	yang	yang	PROPN
fcis-5214	145	16	:	:	PUNCT
fcis-5214	145	17	secure	secure	VERB
fcis-5214	145	18	federated	federated	ADJ
fcis-5214	145	19	matrix	matrix	NOUN
fcis-5214	145	20	factorization	factorization	NOUN
fcis-5214	145	21	.	.	PUNCT
fcis-5214	146	1	ieee	ieee	NOUN
fcis-5214	146	2	intelligent	intelligent	ADJ
fcis-5214	146	3	systems	system	NOUN
fcis-5214	146	4	,	,	PUNCT
fcis-5214	146	5	vol	vol	NOUN
fcis-5214	146	6	.	.	PROPN
fcis-5214	146	7	36	36	NUM
fcis-5214	146	8	(	(	PUNCT
fcis-5214	146	9	2020	2020	NUM
fcis-5214	146	10	)	)	PUNCT
fcis-5214	147	1	no	no	NOUN
fcis-5214	147	2	.	.	NOUN
fcis-5214	147	3	5	5	NUM
fcis-5214	147	4	p.11	p.11	NOUN
fcis-5214	147	5	-	-	X
fcis-5214	147	6	20	20	NUM
fcis-5214	147	7	.	.	PUNCT
fcis-5214	147	8	67	67	NUM
fcis-5214	148	1	[	[	X
fcis-5214	148	2	5	5	NUM
fcis-5214	148	3	]	]	PUNCT
fcis-5214	148	4	g.	g.	PROPN
fcis-5214	148	5	lin	lin	PROPN
fcis-5214	148	6	,	,	PUNCT
fcis-5214	148	7	f.	f.	PROPN
fcis-5214	148	8	liang	liang	PROPN
fcis-5214	148	9	,	,	PUNCT
fcis-5214	148	10	w.	w.	PROPN
fcis-5214	148	11	pan	pan	PROPN
fcis-5214	148	12	,	,	PUNCT
fcis-5214	148	13	and	and	CCONJ
fcis-5214	148	14	z.	z.	PROPN
fcis-5214	148	15	ming	ming	PROPN
fcis-5214	148	16	:	:	PUNCT
fcis-5214	148	17	fedrec	fedrec	NOUN
fcis-5214	148	18	:	:	PUNCT
fcis-5214	148	19	federated	federated	ADJ
fcis-5214	148	20	recommendation	recommendation	NOUN
fcis-5214	148	21	with	with	ADP
fcis-5214	148	22	explicit	explicit	ADJ
fcis-5214	148	23	feedback	feedback	NOUN
fcis-5214	148	24	.	.	PUNCT
fcis-5214	149	1	ieee	ieee	NOUN
fcis-5214	149	2	intelligent	intelligent	ADJ
fcis-5214	149	3	systems	system	NOUN
fcis-5214	149	4	,	,	PUNCT
fcis-5214	149	5	vol	vol	NOUN
fcis-5214	149	6	.	.	PROPN
fcis-5214	149	7	36	36	NUM
fcis-5214	149	8	(	(	PUNCT
fcis-5214	149	9	2021	2021	NUM
fcis-5214	149	10	)	)	PUNCT
fcis-5214	150	1	no	no	NOUN
fcis-5214	150	2	.	.	NOUN
fcis-5214	150	3	5	5	NUM
fcis-5214	150	4	p.21	p.21	NOUN
fcis-5214	150	5	-	-	PUNCT
fcis-5214	150	6	30	30	NUM
fcis-5214	150	7	.	.	PUNCT
fcis-5214	151	1	[	[	X
fcis-5214	151	2	6	6	NUM
fcis-5214	151	3	]	]	PUNCT
fcis-5214	151	4	f.	f.	PROPN
fcis-5214	151	5	liang	liang	PROPN
fcis-5214	151	6	,	,	PUNCT
fcis-5214	151	7	w.	w.	PROPN
fcis-5214	151	8	pan	pan	PROPN
fcis-5214	151	9	,	,	PUNCT
fcis-5214	151	10	and	and	CCONJ
fcis-5214	151	11	z.	z.	PROPN
fcis-5214	151	12	ming	ming	PROPN
fcis-5214	151	13	.	.	PUNCT
fcis-5214	152	1	fedrec++	fedrec++	PROPN
fcis-5214	152	2	:	:	PUNCT
fcis-5214	152	3	lossless	lossless	NOUN
fcis-5214	152	4	federated	federated	ADJ
fcis-5214	152	5	recommendation	recommendation	NOUN
fcis-5214	152	6	with	with	ADP
fcis-5214	152	7	explicit	explicit	ADJ
fcis-5214	152	8	feedback	feedback	NOUN
fcis-5214	152	9	.	.	PUNCT
fcis-5214	153	1	in	in	ADP
fcis-5214	153	2	proceedings	proceeding	NOUN
fcis-5214	153	3	of	of	ADP
fcis-5214	153	4	the	the	DET
fcis-5214	153	5	aaai	aaai	PROPN
fcis-5214	153	6	conference	conference	NOUN
fcis-5214	153	7	on	on	ADP
fcis-5214	153	8	artificial	artificial	ADJ
fcis-5214	153	9	intelligence	intelligence	NOUN
fcis-5214	153	10	(	(	PUNCT
fcis-5214	153	11	online	online	ADJ
fcis-5214	153	12	february	february	PROPN
fcis-5214	153	13	2	2	NUM
fcis-5214	153	14	-	-	SYM
fcis-5214	153	15	9	9	NUM
fcis-5214	153	16	,	,	PUNCT
fcis-5214	153	17	2021	2021	NUM
fcis-5214	153	18	)	)	PUNCT
fcis-5214	153	19	.	.	PUNCT
fcis-5214	154	1	vol.35	vol.35	NOUN
fcis-5214	154	2	,	,	PUNCT
fcis-5214	154	3	p.	p.	NOUN
fcis-5214	154	4	4224	4224	NUM
fcis-5214	154	5	.	.	PUNCT
fcis-5214	155	1	[	[	X
fcis-5214	155	2	7	7	X
fcis-5214	155	3	]	]	PUNCT
fcis-5214	155	4	x.	x.	NOUN
fcis-5214	155	5	ren	ren	PROPN
fcis-5214	155	6	,	,	PUNCT
fcis-5214	155	7	c.-m	c.-m	PROPN
fcis-5214	155	8	.	.	PUNCT
fcis-5214	156	1	yu	yu	PROPN
fcis-5214	156	2	,	,	PUNCT
fcis-5214	156	3	w.	w.	PROPN
fcis-5214	156	4	yu	yu	PROPN
fcis-5214	156	5	,	,	PUNCT
fcis-5214	156	6	s.	s.	PROPN
fcis-5214	156	7	yang	yang	PROPN
fcis-5214	156	8	,	,	PUNCT
fcis-5214	156	9	x.	x.	PROPN
fcis-5214	156	10	yang	yang	PROPN
fcis-5214	156	11	,	,	PUNCT
fcis-5214	156	12	j.a	j.a	PROPN
fcis-5214	156	13	.	.	PROPN
fcis-5214	156	14	mccann	mccann	PROPN
fcis-5214	156	15	,	,	PUNCT
fcis-5214	157	1	s.y	s.y	PROPN
fcis-5214	157	2	.	.	PROPN
fcis-5214	157	3	philip	philip	PROPN
fcis-5214	157	4	:	:	PUNCT
fcis-5214	157	5	high	high	ADJ
fcis-5214	157	6	-	-	PUNCT
fcis-5214	157	7	dimensional	dimensional	ADJ
fcis-5214	157	8	crowdsourced	crowdsourced	ADJ
fcis-5214	157	9	data	datum	NOUN
fcis-5214	157	10	publication	publication	NOUN
fcis-5214	157	11	with	with	ADP
fcis-5214	157	12	local	local	ADJ
fcis-5214	157	13	differential	differential	NOUN
fcis-5214	157	14	privacy	privacy	NOUN
fcis-5214	157	15	,	,	PUNCT
fcis-5214	157	16	ieee	ieee	NOUN
fcis-5214	157	17	trans	trans	PROPN
fcis-5214	157	18	.	.	PROPN
fcis-5214	158	1	inf	inf	PROPN
fcis-5214	158	2	.	.	PUNCT
fcis-5214	158	3	forensics	forensic	NOUN
fcis-5214	158	4	secur	secur	PROPN
fcis-5214	158	5	.	.	PUNCT
fcis-5214	159	1	vol	vol	NOUN
fcis-5214	159	2	.	.	PROPN
fcis-5214	159	3	13	13	NUM
fcis-5214	159	4	(	(	PUNCT
fcis-5214	159	5	2018	2018	NUM
fcis-5214	159	6	)	)	PUNCT
fcis-5214	159	7	no.9	no.9	NOUN
fcis-5214	159	8	p.2151	p.2151	NOUN
fcis-5214	159	9	–	–	PUNCT
fcis-5214	159	10	2166	2166	NUM
fcis-5214	159	11	.	.	PUNCT
fcis-5214	160	1	[	[	X
fcis-5214	160	2	8	8	NUM
fcis-5214	160	3	]	]	X
fcis-5214	160	4	y.	y.	PROPN
fcis-5214	160	5	koren	koren	PROPN
fcis-5214	160	6	,	,	PUNCT
fcis-5214	160	7	r.	r.	PROPN
fcis-5214	160	8	bell	bell	PROPN
fcis-5214	160	9	,	,	PUNCT
fcis-5214	160	10	c.	c.	PROPN
fcis-5214	160	11	volinsky	volinsky	NOUN
fcis-5214	160	12	:	:	PUNCT
fcis-5214	160	13	matrix	matrix	NOUN
fcis-5214	160	14	factorization	factorization	NOUN
fcis-5214	160	15	techniques	technique	NOUN
fcis-5214	160	16	for	for	ADP
fcis-5214	160	17	recommender	recommender	NOUN
fcis-5214	160	18	systems	system	NOUN
fcis-5214	160	19	,	,	PUNCT
fcis-5214	160	20	computer	computer	NOUN
fcis-5214	160	21	,	,	PUNCT
fcis-5214	160	22	vol.42	vol.42	NOUN
fcis-5214	160	23	(	(	PUNCT
fcis-5214	160	24	2009	2009	NUM
fcis-5214	160	25	)	)	PUNCT
fcis-5214	160	26	no.8	no.8	PROPN
fcis-5214	160	27	p.30	p.30	NOUN
fcis-5214	160	28	–	–	PUNCT
fcis-5214	160	29	37	37	NUM
fcis-5214	160	30	.	.	PUNCT
fcis-5214	161	1	[	[	X
fcis-5214	161	2	9	9	NUM
fcis-5214	161	3	]	]	X
fcis-5214	161	4	s.	s.	PROPN
fcis-5214	161	5	rendle	rendle	PROPN
fcis-5214	161	6	,	,	PUNCT
fcis-5214	161	7	c.	c.	NOUN
fcis-5214	161	8	freudenthaler	freudenthaler	NOUN
fcis-5214	161	9	,	,	PUNCT
fcis-5214	161	10	z.	z.	PROPN
fcis-5214	161	11	gantner	gantner	PROPN
fcis-5214	161	12	,	,	PUNCT
fcis-5214	161	13	l.	l.	PROPN
fcis-5214	161	14	schmidt	schmidt	PROPN
fcis-5214	161	15	-	-	PUNCT
fcis-5214	161	16	thieme	thieme	NOUN
fcis-5214	161	17	:	:	PUNCT
fcis-5214	161	18	bpr	bpr	PROPN
fcis-5214	161	19	:	:	PUNCT
fcis-5214	161	20	bayesian	bayesian	NOUN
fcis-5214	161	21	personalized	personalized	AUX
fcis-5214	161	22	ranking	rank	VERB
fcis-5214	161	23	from	from	ADP
fcis-5214	161	24	implicit	implicit	ADJ
fcis-5214	161	25	feedback	feedback	NOUN
fcis-5214	161	26	:	:	PUNCT
fcis-5214	161	27	proceedings	proceeding	NOUN
fcis-5214	161	28	of	of	ADP
fcis-5214	161	29	the	the	DET
fcis-5214	161	30	twenty	twenty	NUM
fcis-5214	161	31	-	-	PUNCT
fcis-5214	161	32	fifth	fifth	ADJ
fcis-5214	161	33	conference	conference	NOUN
fcis-5214	161	34	on	on	ADP
fcis-5214	161	35	uncertainty	uncertainty	NOUN
fcis-5214	161	36	in	in	ADP
fcis-5214	161	37	artificial	artificial	ADJ
fcis-5214	161	38	intelligence	intelligence	NOUN
fcis-5214	161	39	(	(	PUNCT
fcis-5214	161	40	montreal	montreal	PROPN
fcis-5214	161	41	,	,	PUNCT
fcis-5214	161	42	canada	canada	PROPN
fcis-5214	161	43	,	,	PUNCT
fcis-5214	161	44	june	june	PROPN
fcis-5214	161	45	18	18	NUM
fcis-5214	161	46	-	-	SYM
fcis-5214	161	47	21	21	NUM
fcis-5214	161	48	,	,	PUNCT
fcis-5214	161	49	2009	2009	NUM
fcis-5214	161	50	)	)	PUNCT
fcis-5214	161	51	.	.	PUNCT
fcis-5214	162	1	p.	p.	NOUN
fcis-5214	162	2	452	452	NUM
fcis-5214	162	3	.	.	PUNCT
fcis-5214	163	1	[	[	X
fcis-5214	163	2	10	10	NUM
fcis-5214	163	3	]	]	PUNCT
fcis-5214	163	4	x.	x.	NOUN
fcis-5214	164	1	he	he	PRON
fcis-5214	164	2	,	,	PUNCT
fcis-5214	164	3	l.	l.	PROPN
fcis-5214	164	4	liao	liao	PROPN
fcis-5214	164	5	,	,	PUNCT
fcis-5214	164	6	h.	h.	PROPN
fcis-5214	164	7	zhang	zhang	PROPN
fcis-5214	164	8	,	,	PUNCT
fcis-5214	164	9	l.	l.	PROPN
fcis-5214	164	10	nie	nie	PROPN
fcis-5214	164	11	,	,	PUNCT
fcis-5214	164	12	x.	x.	PROPN
fcis-5214	164	13	hu	hu	PROPN
fcis-5214	164	14	,	,	PUNCT
fcis-5214	164	15	t.-s	t.-s	PROPN
fcis-5214	164	16	.	.	PUNCT
fcis-5214	165	1	chua	chua	PROPN
fcis-5214	165	2	:	:	PUNCT
fcis-5214	165	3	neural	neural	ADJ
fcis-5214	165	4	collaborative	collaborative	ADJ
fcis-5214	165	5	filtering	filtering	NOUN
fcis-5214	165	6	,	,	PUNCT
fcis-5214	165	7	proceedings	proceeding	NOUN
fcis-5214	165	8	of	of	ADP
fcis-5214	165	9	the	the	DET
fcis-5214	165	10	26th	26th	ADJ
fcis-5214	165	11	international	international	ADJ
fcis-5214	165	12	conference	conference	NOUN
fcis-5214	165	13	on	on	ADP
fcis-5214	165	14	world	world	NOUN
fcis-5214	165	15	wide	wide	ADJ
fcis-5214	165	16	web	web	NOUN
fcis-5214	165	17	(	(	PUNCT
fcis-5214	165	18	perth	perth	PROPN
fcis-5214	165	19	,	,	PUNCT
fcis-5214	165	20	australia	australia	PROPN
fcis-5214	165	21	,	,	PUNCT
fcis-5214	165	22	april	april	PROPN
fcis-5214	165	23	3	3	NUM
fcis-5214	165	24	-	-	SYM
fcis-5214	165	25	7	7	NUM
fcis-5214	165	26	,	,	PUNCT
fcis-5214	165	27	2017	2017	NUM
fcis-5214	165	28	)	)	PUNCT
fcis-5214	165	29	.	.	PUNCT
fcis-5214	166	1	p.173	p.173	NOUN
fcis-5214	166	2	.	.	PUNCT
fcis-5214	167	1	[	[	X
fcis-5214	167	2	11	11	NUM
fcis-5214	167	3	]	]	PUNCT
fcis-5214	167	4	s.	s.	PROPN
fcis-5214	167	5	niwattanakul	niwattanakul	PROPN
fcis-5214	167	6	,	,	PUNCT
fcis-5214	167	7	et	et	PROPN
fcis-5214	167	8	al	al	PROPN
fcis-5214	167	9	.	.	PROPN
fcis-5214	167	10	:	:	PUNCT
fcis-5214	167	11	using	use	VERB
fcis-5214	167	12	of	of	ADP
fcis-5214	167	13	jaccard	jaccard	ADJ
fcis-5214	167	14	coefficient	coefficient	NOUN
fcis-5214	167	15	for	for	ADP
fcis-5214	167	16	keywords	keyword	NOUN
fcis-5214	167	17	similarity	similarity	NOUN
fcis-5214	167	18	,	,	PUNCT
fcis-5214	167	19	proc	proc	NOUN
fcis-5214	167	20	.	.	PUNCT
fcis-5214	168	1	international	international	ADJ
fcis-5214	168	2	multi	multi	PROPN
fcis-5214	168	3	conference	conference	NOUN
fcis-5214	168	4	of	of	ADP
fcis-5214	168	5	engineers	engineer	NOUN
fcis-5214	168	6	and	and	CCONJ
fcis-5214	168	7	computer	computer	NOUN
fcis-5214	168	8	scientists	scientist	NOUN
fcis-5214	168	9	(	(	PUNCT
fcis-5214	168	10	hong	hong	PROPN
fcis-5214	168	11	kong	kong	PROPN
fcis-5214	168	12	,	,	PUNCT
fcis-5214	168	13	china	china	PROPN
fcis-5214	168	14	,	,	PUNCT
fcis-5214	168	15	march	march	PROPN
fcis-5214	168	16	13	13	NUM
fcis-5214	168	17	-	-	SYM
fcis-5214	168	18	15	15	NUM
fcis-5214	168	19	,	,	PUNCT
fcis-5214	168	20	2013	2013	NUM
fcis-5214	168	21	)	)	PUNCT
fcis-5214	168	22	.	.	PUNCT
fcis-5214	169	1	vol.1	vol.1	ADV
fcis-5214	169	2	,	,	PUNCT
fcis-5214	169	3	p.380	p.380	X
fcis-5214	169	4	.	.	PUNCT
fcis-5214	170	1	[	[	X
fcis-5214	170	2	12	12	NUM
fcis-5214	170	3	]	]	X
fcis-5214	170	4	mcmahan	mcmahan	PROPN
fcis-5214	170	5	,	,	PUNCT
fcis-5214	170	6	h.	h.	PROPN
fcis-5214	170	7	b	b	PROPN
fcis-5214	170	8	,	,	PUNCT
fcis-5214	170	9	et	et	PROPN
fcis-5214	170	10	al	al	PROPN
fcis-5214	170	11	.	.	PUNCT
fcis-5214	171	1	"	"	PUNCT
fcis-5214	171	2	communication	communication	NOUN
fcis-5214	171	3	-	-	PUNCT
fcis-5214	171	4	efficient	efficient	ADJ
fcis-5214	171	5	learning	learning	NOUN
fcis-5214	171	6	of	of	ADP
fcis-5214	171	7	deep	deep	ADJ
fcis-5214	171	8	networks	network	NOUN
fcis-5214	171	9	from	from	ADP
fcis-5214	171	10	decentralized	decentralized	ADJ
fcis-5214	171	11	data	datum	NOUN
fcis-5214	171	12	.	.	PUNCT
fcis-5214	171	13	"	"	PUNCT
fcis-5214	172	1	2016	2016	NUM
fcis-5214	172	2	.	.	PUNCT