id	sid	tid	token	lemma	pos
ajst-29613	1	1	academic	academic	ADJ
ajst-29613	1	2	journal	journal	NOUN
ajst-29613	1	3	of	of	ADP
ajst-29613	1	4	science	science	NOUN
ajst-29613	1	5	and	and	CCONJ
ajst-29613	1	6	technology	technology	NOUN
ajst-29613	1	7	issn	issn	NOUN
ajst-29613	1	8	:	:	PUNCT
ajst-29613	1	9	2771	2771	NUM
ajst-29613	1	10	-	-	SYM
ajst-29613	1	11	3032	3032	NUM
ajst-29613	1	12	|	|	NOUN
ajst-29613	1	13	vol	vol	NOUN
ajst-29613	1	14	.	.	PUNCT
ajst-29613	2	1	14	14	NUM
ajst-29613	2	2	,	,	PUNCT
ajst-29613	2	3	no	no	INTJ
ajst-29613	2	4	.	.	NOUN
ajst-29613	2	5	1	1	NUM
ajst-29613	2	6	,	,	PUNCT
ajst-29613	2	7	2025	2025	NUM
ajst-29613	2	8	227	227	NUM
ajst-29613	2	9	knowledge	knowledge	NOUN
ajst-29613	2	10	graph	graph	NOUN
ajst-29613	2	11	combined	combine	VERB
ajst-29613	2	12	with	with	ADP
ajst-29613	2	13	retrieval‐augmented	retrieval‐augmente	VERB
ajst-29613	2	14	generation	generation	NOUN
ajst-29613	2	15	for	for	ADP
ajst-29613	2	16	enhancing	enhance	VERB
ajst-29613	2	17	lms	lm	NOUN
ajst-29613	2	18	reasoning	reasoning	NOUN
ajst-29613	2	19	:	:	PUNCT
ajst-29613	2	20	a	a	DET
ajst-29613	2	21	survey	survey	NOUN
ajst-29613	2	22	haitao	haitao	PROPN
ajst-29613	2	23	wang	wang	PROPN
ajst-29613	2	24	,	,	PUNCT
ajst-29613	2	25	yangkun	yangkun	PROPN
ajst-29613	2	26	shi	shi	PROPN
ajst-29613	2	27	school	school	PROPN
ajst-29613	2	28	of	of	ADP
ajst-29613	2	29	computer	computer	NOUN
ajst-29613	2	30	science	science	NOUN
ajst-29613	2	31	and	and	CCONJ
ajst-29613	2	32	technology	technology	NOUN
ajst-29613	2	33	,	,	PUNCT
ajst-29613	2	34	henan	henan	PROPN
ajst-29613	2	35	polytechnic	polytechnic	PROPN
ajst-29613	2	36	university	university	PROPN
ajst-29613	2	37	,	,	PUNCT
ajst-29613	2	38	jiaozuo	jiaozuo	PROPN
ajst-29613	2	39	454003	454003	NUM
ajst-29613	2	40	,	,	PUNCT
ajst-29613	2	41	china	china	PROPN
ajst-29613	2	42	abstract	abstract	PROPN
ajst-29613	2	43	:	:	PUNCT
ajst-29613	2	44	large	large	ADJ
ajst-29613	2	45	language	language	NOUN
ajst-29613	2	46	models	model	NOUN
ajst-29613	2	47	(	(	PUNCT
ajst-29613	2	48	llms	llm	NOUN
ajst-29613	2	49	)	)	PUNCT
ajst-29613	2	50	have	have	AUX
ajst-29613	2	51	generated	generate	VERB
ajst-29613	2	52	significant	significant	ADJ
ajst-29613	2	53	waves	wave	NOUN
ajst-29613	2	54	in	in	ADP
ajst-29613	2	55	the	the	DET
ajst-29613	2	56	fields	field	NOUN
ajst-29613	2	57	of	of	ADP
ajst-29613	2	58	natural	natural	ADJ
ajst-29613	2	59	language	language	NOUN
ajst-29613	2	60	processing(nlp	processing(nlp	NOUN
ajst-29613	2	61	)	)	PUNCT
ajst-29613	2	62	and	and	CCONJ
ajst-29613	2	63	artificial	artificial	ADJ
ajst-29613	2	64	intelligence	intelligence	NOUN
ajst-29613	2	65	due	due	ADP
ajst-29613	2	66	to	to	ADP
ajst-29613	2	67	their	their	PRON
ajst-29613	2	68	remarkable	remarkable	ADJ
ajst-29613	2	69	capabilities	capability	NOUN
ajst-29613	2	70	and	and	CCONJ
ajst-29613	2	71	broad	broad	ADJ
ajst-29613	2	72	adaptability	adaptability	NOUN
ajst-29613	2	73	.	.	PUNCT
ajst-29613	3	1	retrieval	retrieval	NOUN
ajst-29613	3	2	-	-	PUNCT
ajst-29613	3	3	augmented	augment	VERB
ajst-29613	3	4	generation	generation	NOUN
ajst-29613	3	5	(	(	PUNCT
ajst-29613	3	6	rag	rag	NOUN
ajst-29613	3	7	)	)	PUNCT
ajst-29613	3	8	techniques	technique	NOUN
ajst-29613	3	9	have	have	AUX
ajst-29613	3	10	been	be	AUX
ajst-29613	3	11	widely	widely	ADV
ajst-29613	3	12	adopted	adopt	VERB
ajst-29613	3	13	,	,	PUNCT
ajst-29613	3	14	leveraging	leverage	VERB
ajst-29613	3	15	external	external	ADJ
ajst-29613	3	16	retrieval	retrieval	NOUN
ajst-29613	3	17	systems	system	NOUN
ajst-29613	3	18	to	to	PART
ajst-29613	3	19	significantly	significantly	ADV
ajst-29613	3	20	improve	improve	VERB
ajst-29613	3	21	the	the	DET
ajst-29613	3	22	timeliness	timeliness	NOUN
ajst-29613	3	23	of	of	ADP
ajst-29613	3	24	llms	llm	NOUN
ajst-29613	3	25	and	and	CCONJ
ajst-29613	3	26	substantially	substantially	ADV
ajst-29613	3	27	reduce	reduce	VERB
ajst-29613	3	28	hallucinations	hallucination	NOUN
ajst-29613	3	29	.	.	PUNCT
ajst-29613	4	1	although	although	SCONJ
ajst-29613	4	2	rag	rag	NOUN
ajst-29613	4	3	and	and	CCONJ
ajst-29613	4	4	its	its	PRON
ajst-29613	4	5	optimization	optimization	NOUN
ajst-29613	4	6	methods	method	NOUN
ajst-29613	4	7	have	have	AUX
ajst-29613	4	8	addressed	address	VERB
ajst-29613	4	9	most	most	ADJ
ajst-29613	4	10	hallucination	hallucination	NOUN
ajst-29613	4	11	issues	issue	NOUN
ajst-29613	4	12	caused	cause	VERB
ajst-29613	4	13	by	by	ADP
ajst-29613	4	14	knowledge	knowledge	NOUN
ajst-29613	4	15	gaps	gap	NOUN
ajst-29613	4	16	and	and	CCONJ
ajst-29613	4	17	outdated	outdated	ADJ
ajst-29613	4	18	information	information	NOUN
ajst-29613	4	19	,	,	PUNCT
ajst-29613	4	20	text	text	NOUN
ajst-29613	4	21	generation	generation	NOUN
ajst-29613	4	22	in	in	ADP
ajst-29613	4	23	specialized	specialized	ADJ
ajst-29613	4	24	domains	domain	NOUN
ajst-29613	4	25	such	such	ADJ
ajst-29613	4	26	as	as	ADP
ajst-29613	4	27	law	law	NOUN
ajst-29613	4	28	,	,	PUNCT
ajst-29613	4	29	medicine	medicine	NOUN
ajst-29613	4	30	,	,	PUNCT
ajst-29613	4	31	and	and	CCONJ
ajst-29613	4	32	science	science	NOUN
ajst-29613	4	33	—	—	PUNCT
ajst-29613	4	34	which	which	PRON
ajst-29613	4	35	require	require	VERB
ajst-29613	4	36	multi	multi	ADJ
ajst-29613	4	37	-	-	ADJ
ajst-29613	4	38	hop	hop	ADJ
ajst-29613	4	39	reasoning	reasoning	NOUN
ajst-29613	4	40	and	and	CCONJ
ajst-29613	4	41	analysis	analysis	NOUN
ajst-29613	4	42	—	—	PUNCT
ajst-29613	4	43	still	still	ADV
ajst-29613	4	44	suffers	suffer	VERB
ajst-29613	4	45	from	from	ADP
ajst-29613	4	46	a	a	DET
ajst-29613	4	47	lack	lack	NOUN
ajst-29613	4	48	of	of	ADP
ajst-29613	4	49	coherence	coherence	NOUN
ajst-29613	4	50	and	and	CCONJ
ajst-29613	4	51	logical	logical	ADJ
ajst-29613	4	52	consistency	consistency	NOUN
ajst-29613	4	53	,	,	PUNCT
ajst-29613	4	54	making	make	VERB
ajst-29613	4	55	it	it	PRON
ajst-29613	4	56	difficult	difficult	ADJ
ajst-29613	4	57	to	to	PART
ajst-29613	4	58	produce	produce	VERB
ajst-29613	4	59	correct	correct	ADJ
ajst-29613	4	60	and	and	CCONJ
ajst-29613	4	61	valuable	valuable	ADJ
ajst-29613	4	62	answers	answer	NOUN
ajst-29613	4	63	.	.	PUNCT
ajst-29613	5	1	to	to	PART
ajst-29613	5	2	address	address	VERB
ajst-29613	5	3	these	these	DET
ajst-29613	5	4	challenges	challenge	NOUN
ajst-29613	5	5	,	,	PUNCT
ajst-29613	5	6	related	related	ADJ
ajst-29613	5	7	research	research	NOUN
ajst-29613	5	8	has	have	AUX
ajst-29613	5	9	introduced	introduce	VERB
ajst-29613	5	10	knowledge	knowledge	NOUN
ajst-29613	5	11	graphs	graph	NOUN
ajst-29613	5	12	(	(	PUNCT
ajst-29613	5	13	kgs	kgs	NOUN
ajst-29613	5	14	)	)	PUNCT
ajst-29613	5	15	.	.	PUNCT
ajst-29613	6	1	this	this	DET
ajst-29613	6	2	integrated	integrate	VERB
ajst-29613	6	3	approach	approach	NOUN
ajst-29613	6	4	enhances	enhance	VERB
ajst-29613	6	5	rag	rag	NOUN
ajst-29613	6	6	’s	’s	PART
ajst-29613	6	7	capabilities	capability	NOUN
ajst-29613	6	8	by	by	ADP
ajst-29613	6	9	utilizing	utilize	VERB
ajst-29613	6	10	the	the	DET
ajst-29613	6	11	structured	structured	ADJ
ajst-29613	6	12	knowledge	knowledge	NOUN
ajst-29613	6	13	provided	provide	VERB
ajst-29613	6	14	by	by	ADP
ajst-29613	6	15	kgs	kg	NOUN
ajst-29613	6	16	,	,	PUNCT
ajst-29613	6	17	thereby	thereby	ADV
ajst-29613	6	18	improving	improve	VERB
ajst-29613	6	19	the	the	DET
ajst-29613	6	20	model	model	NOUN
ajst-29613	6	21	's	's	PART
ajst-29613	6	22	knowledge	knowledge	NOUN
ajst-29613	6	23	representation	representation	NOUN
ajst-29613	6	24	and	and	CCONJ
ajst-29613	6	25	reasoning	reasoning	NOUN
ajst-29613	6	26	abilities	ability	NOUN
ajst-29613	6	27	and	and	CCONJ
ajst-29613	6	28	enabling	enable	VERB
ajst-29613	6	29	the	the	DET
ajst-29613	6	30	generation	generation	NOUN
ajst-29613	6	31	of	of	ADP
ajst-29613	6	32	more	more	ADV
ajst-29613	6	33	accurate	accurate	ADJ
ajst-29613	6	34	answers	answer	NOUN
ajst-29613	6	35	.	.	PUNCT
ajst-29613	7	1	however	however	ADV
ajst-29613	7	2	,	,	PUNCT
ajst-29613	7	3	there	there	PRON
ajst-29613	7	4	remains	remain	VERB
ajst-29613	7	5	a	a	DET
ajst-29613	7	6	lack	lack	NOUN
ajst-29613	7	7	of	of	ADP
ajst-29613	7	8	systematic	systematic	ADJ
ajst-29613	7	9	reviews	review	NOUN
ajst-29613	7	10	in	in	ADP
ajst-29613	7	11	this	this	DET
ajst-29613	7	12	area	area	NOUN
ajst-29613	7	13	.	.	PUNCT
ajst-29613	8	1	therefore	therefore	ADV
ajst-29613	8	2	,	,	PUNCT
ajst-29613	8	3	this	this	DET
ajst-29613	8	4	paper	paper	NOUN
ajst-29613	8	5	provides	provide	VERB
ajst-29613	8	6	a	a	DET
ajst-29613	8	7	comprehensive	comprehensive	ADJ
ajst-29613	8	8	review	review	NOUN
ajst-29613	8	9	of	of	ADP
ajst-29613	8	10	studies	study	NOUN
ajst-29613	8	11	on	on	ADP
ajst-29613	8	12	enhancing	enhance	VERB
ajst-29613	8	13	llms	llm	NOUN
ajst-29613	8	14	reasoning	reason	VERB
ajst-29613	8	15	abilities	ability	NOUN
ajst-29613	8	16	by	by	ADP
ajst-29613	8	17	integrating	integrate	VERB
ajst-29613	8	18	kgs	kg	NOUN
ajst-29613	8	19	with	with	ADP
ajst-29613	8	20	rag	rag	NOUN
ajst-29613	8	21	.	.	PUNCT
ajst-29613	9	1	it	it	PRON
ajst-29613	9	2	first	first	ADV
ajst-29613	9	3	introduces	introduce	VERB
ajst-29613	9	4	the	the	DET
ajst-29613	9	5	basic	basic	ADJ
ajst-29613	9	6	concepts	concept	NOUN
ajst-29613	9	7	,	,	PUNCT
ajst-29613	9	8	followed	follow	VERB
ajst-29613	9	9	by	by	ADP
ajst-29613	9	10	an	an	DET
ajst-29613	9	11	overview	overview	NOUN
ajst-29613	9	12	of	of	ADP
ajst-29613	9	13	the	the	DET
ajst-29613	9	14	current	current	ADJ
ajst-29613	9	15	mainstream	mainstream	ADJ
ajst-29613	9	16	technical	technical	ADJ
ajst-29613	9	17	approaches	approach	NOUN
ajst-29613	9	18	,	,	PUNCT
ajst-29613	9	19	and	and	CCONJ
ajst-29613	9	20	concludes	conclude	VERB
ajst-29613	9	21	with	with	ADP
ajst-29613	9	22	a	a	DET
ajst-29613	9	23	discussion	discussion	NOUN
ajst-29613	9	24	of	of	ADP
ajst-29613	9	25	the	the	DET
ajst-29613	9	26	research	research	NOUN
ajst-29613	9	27	challenges	challenge	NOUN
ajst-29613	9	28	and	and	CCONJ
ajst-29613	9	29	future	future	ADJ
ajst-29613	9	30	development	development	NOUN
ajst-29613	9	31	trends	trend	NOUN
ajst-29613	9	32	in	in	ADP
ajst-29613	9	33	this	this	DET
ajst-29613	9	34	field	field	NOUN
ajst-29613	9	35	.	.	PUNCT
ajst-29613	10	1	keywords	keyword	NOUN
ajst-29613	10	2	:	:	PUNCT
ajst-29613	10	3	large	large	ADJ
ajst-29613	10	4	language	language	NOUN
ajst-29613	10	5	models	model	NOUN
ajst-29613	10	6	;	;	PUNCT
ajst-29613	10	7	natural	natural	ADJ
ajst-29613	10	8	language	language	NOUN
ajst-29613	10	9	processing	processing	NOUN
ajst-29613	10	10	;	;	PUNCT
ajst-29613	10	11	retrieval	retrieval	NOUN
ajst-29613	10	12	-	-	PUNCT
ajst-29613	10	13	augmented	augment	VERB
ajst-29613	10	14	generation	generation	NOUN
ajst-29613	10	15	;	;	PUNCT
ajst-29613	10	16	knowledge	knowledge	NOUN
ajst-29613	10	17	graphs	graph	NOUN
ajst-29613	10	18	;	;	PUNCT
ajst-29613	10	19	knowledge	knowledge	NOUN
ajst-29613	10	20	reasoning	reasoning	NOUN
ajst-29613	10	21	.	.	PUNCT
ajst-29613	11	1	1	1	X
ajst-29613	11	2	.	.	X
ajst-29613	11	3	introduction	introduction	NOUN
ajst-29613	11	4	in	in	ADP
ajst-29613	11	5	recent	recent	ADJ
ajst-29613	11	6	years	year	NOUN
ajst-29613	11	7	,	,	PUNCT
ajst-29613	11	8	large	large	ADJ
ajst-29613	11	9	language	language	NOUN
ajst-29613	11	10	models	model	NOUN
ajst-29613	11	11	(	(	PUNCT
ajst-29613	11	12	llms	llm	NOUN
ajst-29613	11	13	)	)	PUNCT
ajst-29613	11	14	such	such	ADJ
ajst-29613	11	15	as	as	ADP
ajst-29613	11	16	gpt-4[1	gpt-4[1	NOUN
ajst-29613	11	17	]	]	X
ajst-29613	11	18	,	,	PUNCT
ajst-29613	11	19	llama-3[2	llama-3[2	PROPN
ajst-29613	11	20	]	]	PUNCT
ajst-29613	11	21	,	,	PUNCT
ajst-29613	11	22	and	and	CCONJ
ajst-29613	11	23	gemini	gemini	NOUN
ajst-29613	11	24	[	[	X
ajst-29613	11	25	3	3	NUM
ajst-29613	11	26	]	]	PUNCT
ajst-29613	11	27	have	have	AUX
ajst-29613	11	28	achieved	achieve	VERB
ajst-29613	11	29	significant	significant	ADJ
ajst-29613	11	30	technological	technological	ADJ
ajst-29613	11	31	breakthroughs	breakthrough	NOUN
ajst-29613	11	32	in	in	ADP
ajst-29613	11	33	the	the	DET
ajst-29613	11	34	field	field	NOUN
ajst-29613	11	35	of	of	ADP
ajst-29613	11	36	natural	natural	ADJ
ajst-29613	11	37	language	language	NOUN
ajst-29613	11	38	processing	processing	NOUN
ajst-29613	11	39	,	,	PUNCT
ajst-29613	11	40	finding	find	VERB
ajst-29613	11	41	wide	wide	ADJ
ajst-29613	11	42	application	application	NOUN
ajst-29613	11	43	in	in	ADP
ajst-29613	11	44	tasks	task	NOUN
ajst-29613	11	45	such	such	ADJ
ajst-29613	11	46	as	as	ADP
ajst-29613	11	47	text	text	NOUN
ajst-29613	11	48	generation	generation	NOUN
ajst-29613	11	49	,	,	PUNCT
ajst-29613	11	50	dialogue	dialogue	NOUN
ajst-29613	11	51	systems	system	NOUN
ajst-29613	11	52	,	,	PUNCT
ajst-29613	11	53	translation	translation	NOUN
ajst-29613	11	54	,	,	PUNCT
ajst-29613	11	55	and	and	CCONJ
ajst-29613	11	56	summarization	summarization	NOUN
ajst-29613	11	57	.	.	PUNCT
ajst-29613	12	1	however	however	ADV
ajst-29613	12	2	,	,	PUNCT
ajst-29613	12	3	despite	despite	SCONJ
ajst-29613	12	4	their	their	PRON
ajst-29613	12	5	impressive	impressive	ADJ
ajst-29613	12	6	performance	performance	NOUN
ajst-29613	12	7	in	in	ADP
ajst-29613	12	8	general	general	ADJ
ajst-29613	12	9	contexts	contexts	NOUN
ajst-29613	12	10	,	,	PUNCT
ajst-29613	12	11	llms	llm	NOUN
ajst-29613	12	12	still	still	ADV
ajst-29613	12	13	face	face	VERB
ajst-29613	12	14	numerous	numerous	ADJ
ajst-29613	12	15	limitations	limitation	NOUN
ajst-29613	12	16	in	in	ADP
ajst-29613	12	17	specialized	specialized	ADJ
ajst-29613	12	18	domains	domain	NOUN
ajst-29613	12	19	or	or	CCONJ
ajst-29613	12	20	knowledge	knowledge	NOUN
ajst-29613	12	21	-	-	PUNCT
ajst-29613	12	22	intensive	intensive	ADJ
ajst-29613	12	23	tasks	task	NOUN
ajst-29613	12	24	[	[	X
ajst-29613	12	25	4	4	NUM
ajst-29613	12	26	]	]	PUNCT
ajst-29613	12	27	.	.	PUNCT
ajst-29613	13	1	specifically	specifically	ADV
ajst-29613	13	2	,	,	PUNCT
ajst-29613	13	3	when	when	SCONJ
ajst-29613	13	4	handling	handle	VERB
ajst-29613	13	5	queries	query	NOUN
ajst-29613	13	6	beyond	beyond	ADP
ajst-29613	13	7	the	the	DET
ajst-29613	13	8	scope	scope	NOUN
ajst-29613	13	9	of	of	ADP
ajst-29613	13	10	their	their	PRON
ajst-29613	13	11	training	training	NOUN
ajst-29613	13	12	data	datum	NOUN
ajst-29613	13	13	,	,	PUNCT
ajst-29613	13	14	these	these	DET
ajst-29613	13	15	models	model	NOUN
ajst-29613	13	16	often	often	ADV
ajst-29613	13	17	produce	produce	VERB
ajst-29613	13	18	factually	factually	ADV
ajst-29613	13	19	inaccurate	inaccurate	ADJ
ajst-29613	13	20	or	or	CCONJ
ajst-29613	13	21	fabricated	fabricate	VERB
ajst-29613	13	22	information	information	NOUN
ajst-29613	13	23	,	,	PUNCT
ajst-29613	13	24	a	a	DET
ajst-29613	13	25	phenomenon	phenomenon	NOUN
ajst-29613	13	26	known	know	VERB
ajst-29613	13	27	as	as	ADP
ajst-29613	13	28	"	"	PUNCT
ajst-29613	13	29	hallucination	hallucination	NOUN
ajst-29613	13	30	"	"	PUNCT
ajst-29613	14	1	[	[	X
ajst-29613	14	2	5	5	NUM
ajst-29613	14	3	]	]	PUNCT
ajst-29613	14	4	.	.	PUNCT
ajst-29613	15	1	this	this	DET
ajst-29613	15	2	hallucination	hallucination	NOUN
ajst-29613	15	3	not	not	PART
ajst-29613	15	4	only	only	ADV
ajst-29613	15	5	undermines	undermine	VERB
ajst-29613	15	6	the	the	DET
ajst-29613	15	7	reliability	reliability	NOUN
ajst-29613	15	8	of	of	ADP
ajst-29613	15	9	the	the	DET
ajst-29613	15	10	model	model	NOUN
ajst-29613	15	11	's	's	PART
ajst-29613	15	12	output	output	NOUN
ajst-29613	15	13	but	but	CCONJ
ajst-29613	15	14	also	also	ADV
ajst-29613	15	15	hinders	hinder	VERB
ajst-29613	15	16	its	its	PRON
ajst-29613	15	17	further	further	ADJ
ajst-29613	15	18	adoption	adoption	NOUN
ajst-29613	15	19	in	in	ADP
ajst-29613	15	20	critical	critical	ADJ
ajst-29613	15	21	application	application	NOUN
ajst-29613	15	22	areas	area	NOUN
ajst-29613	15	23	such	such	ADJ
ajst-29613	15	24	as	as	ADP
ajst-29613	15	25	healthcare	healthcare	NOUN
ajst-29613	15	26	,	,	PUNCT
ajst-29613	15	27	law	law	NOUN
ajst-29613	15	28	,	,	PUNCT
ajst-29613	15	29	and	and	CCONJ
ajst-29613	15	30	scientific	scientific	ADJ
ajst-29613	15	31	research	research	NOUN
ajst-29613	15	32	.	.	PUNCT
ajst-29613	16	1	moreover	moreover	ADV
ajst-29613	16	2	,	,	PUNCT
ajst-29613	16	3	since	since	SCONJ
ajst-29613	16	4	llms	llm	NOUN
ajst-29613	16	5	are	be	AUX
ajst-29613	16	6	typically	typically	ADV
ajst-29613	16	7	trained	train	VERB
ajst-29613	16	8	on	on	ADP
ajst-29613	16	9	large	large	ADJ
ajst-29613	16	10	-	-	PUNCT
ajst-29613	16	11	scale	scale	NOUN
ajst-29613	16	12	static	static	ADJ
ajst-29613	16	13	datasets	dataset	NOUN
ajst-29613	16	14	,	,	PUNCT
ajst-29613	16	15	their	their	PRON
ajst-29613	16	16	knowledge	knowledge	NOUN
ajst-29613	16	17	is	be	AUX
ajst-29613	16	18	limited	limit	VERB
ajst-29613	16	19	to	to	ADP
ajst-29613	16	20	a	a	DET
ajst-29613	16	21	specific	specific	ADJ
ajst-29613	16	22	temporal	temporal	ADJ
ajst-29613	16	23	cutoff	cutoff	NOUN
ajst-29613	16	24	,	,	PUNCT
ajst-29613	16	25	making	make	VERB
ajst-29613	16	26	it	it	PRON
ajst-29613	16	27	difficult	difficult	ADJ
ajst-29613	16	28	for	for	SCONJ
ajst-29613	16	29	them	they	PRON
ajst-29613	16	30	to	to	PART
ajst-29613	16	31	adapt	adapt	VERB
ajst-29613	16	32	to	to	ADP
ajst-29613	16	33	the	the	DET
ajst-29613	16	34	dynamic	dynamic	ADJ
ajst-29613	16	35	real	real	ADJ
ajst-29613	16	36	-	-	PUNCT
ajst-29613	16	37	world	world	NOUN
ajst-29613	16	38	environment	environment	NOUN
ajst-29613	16	39	or	or	CCONJ
ajst-29613	16	40	meet	meet	VERB
ajst-29613	16	41	the	the	DET
ajst-29613	16	42	rapidly	rapidly	ADV
ajst-29613	16	43	evolving	evolve	VERB
ajst-29613	16	44	demands	demand	NOUN
ajst-29613	16	45	for	for	ADP
ajst-29613	16	46	specialized	specialized	ADJ
ajst-29613	16	47	knowledge[6	knowledge[6	NOUN
ajst-29613	16	48	]	]	PUNCT
ajst-29613	16	49	.	.	PUNCT
ajst-29613	17	1	consequently	consequently	ADV
ajst-29613	17	2	,	,	PUNCT
ajst-29613	17	3	in	in	ADP
ajst-29613	17	4	many	many	ADJ
ajst-29613	17	5	cases	case	NOUN
ajst-29613	17	6	,	,	PUNCT
ajst-29613	17	7	users	user	NOUN
ajst-29613	17	8	require	require	VERB
ajst-29613	17	9	high	high	ADJ
ajst-29613	17	10	-	-	PUNCT
ajst-29613	17	11	quality	quality	NOUN
ajst-29613	17	12	information	information	NOUN
ajst-29613	17	13	related	relate	VERB
ajst-29613	17	14	to	to	ADP
ajst-29613	17	15	the	the	DET
ajst-29613	17	16	latest	late	ADJ
ajst-29613	17	17	literature	literature	NOUN
ajst-29613	17	18	,	,	PUNCT
ajst-29613	17	19	real	real	ADJ
ajst-29613	17	20	-	-	PUNCT
ajst-29613	17	21	time	time	NOUN
ajst-29613	17	22	data	datum	NOUN
ajst-29613	17	23	,	,	PUNCT
ajst-29613	17	24	or	or	CCONJ
ajst-29613	17	25	domain	domain	NOUN
ajst-29613	17	26	-	-	PUNCT
ajst-29613	17	27	specific	specific	ADJ
ajst-29613	17	28	knowledge	knowledge	NOUN
ajst-29613	17	29	,	,	PUNCT
ajst-29613	17	30	which	which	PRON
ajst-29613	17	31	may	may	AUX
ajst-29613	17	32	not	not	PART
ajst-29613	17	33	be	be	AUX
ajst-29613	17	34	covered	cover	VERB
ajst-29613	17	35	by	by	ADP
ajst-29613	17	36	the	the	DET
ajst-29613	17	37	model	model	NOUN
ajst-29613	17	38	's	's	PART
ajst-29613	17	39	training	training	NOUN
ajst-29613	17	40	data	datum	NOUN
ajst-29613	17	41	.	.	PUNCT
ajst-29613	18	1	to	to	PART
ajst-29613	18	2	address	address	VERB
ajst-29613	18	3	these	these	DET
ajst-29613	18	4	challenges	challenge	NOUN
ajst-29613	18	5	,	,	PUNCT
ajst-29613	18	6	researchers	researcher	NOUN
ajst-29613	18	7	have	have	AUX
ajst-29613	18	8	proposed	propose	VERB
ajst-29613	18	9	retrieval	retrieval	NOUN
ajst-29613	18	10	-	-	PUNCT
ajst-29613	18	11	augmented	augment	VERB
ajst-29613	18	12	generation	generation	NOUN
ajst-29613	18	13	(	(	PUNCT
ajst-29613	18	14	rag	rag	NOUN
ajst-29613	18	15	)	)	PUNCT
ajst-29613	18	16	techniques	technique	NOUN
ajst-29613	18	17	.	.	PUNCT
ajst-29613	19	1	rag	rag	NOUN
ajst-29613	19	2	integrates	integrate	VERB
ajst-29613	19	3	llms	llm	NOUN
ajst-29613	19	4	with	with	ADP
ajst-29613	19	5	external	external	ADJ
ajst-29613	19	6	knowledge	knowledge	NOUN
ajst-29613	19	7	bases	basis	NOUN
ajst-29613	19	8	or	or	CCONJ
ajst-29613	19	9	real	real	ADJ
ajst-29613	19	10	-	-	PUNCT
ajst-29613	19	11	time	time	NOUN
ajst-29613	19	12	retrieval	retrieval	NOUN
ajst-29613	19	13	systems	system	NOUN
ajst-29613	19	14	,	,	PUNCT
ajst-29613	19	15	significantly	significantly	ADV
ajst-29613	19	16	expanding	expand	VERB
ajst-29613	19	17	the	the	DET
ajst-29613	19	18	model	model	NOUN
ajst-29613	19	19	's	's	PART
ajst-29613	19	20	knowledge	knowledge	NOUN
ajst-29613	19	21	coverage	coverage	NOUN
ajst-29613	19	22	and	and	CCONJ
ajst-29613	19	23	improving	improve	VERB
ajst-29613	19	24	the	the	DET
ajst-29613	19	25	accuracy	accuracy	NOUN
ajst-29613	19	26	of	of	ADP
ajst-29613	19	27	its	its	PRON
ajst-29613	19	28	generated	generated	ADJ
ajst-29613	19	29	content[7	content[7	PROPN
ajst-29613	19	30	]	]	PUNCT
ajst-29613	19	31	.	.	PUNCT
ajst-29613	20	1	however	however	ADV
ajst-29613	20	2	,	,	PUNCT
ajst-29613	20	3	rag	rag	NOUN
ajst-29613	20	4	still	still	ADV
ajst-29613	20	5	exhibits	exhibit	VERB
ajst-29613	20	6	notable	notable	ADJ
ajst-29613	20	7	limitations	limitation	NOUN
ajst-29613	20	8	when	when	SCONJ
ajst-29613	20	9	dealing	deal	VERB
ajst-29613	20	10	with	with	ADP
ajst-29613	20	11	complex	complex	ADJ
ajst-29613	20	12	knowledge	knowledge	NOUN
ajst-29613	20	13	reasoning	reasoning	NOUN
ajst-29613	20	14	tasks	task	NOUN
ajst-29613	20	15	that	that	PRON
ajst-29613	20	16	require	require	VERB
ajst-29613	20	17	deep	deep	ADJ
ajst-29613	20	18	reasoning	reasoning	NOUN
ajst-29613	20	19	and	and	CCONJ
ajst-29613	20	20	responsible	responsible	ADJ
ajst-29613	20	21	decisionmaking	decisionmaking	NOUN
ajst-29613	20	22	.	.	PUNCT
ajst-29613	21	1	these	these	DET
ajst-29613	21	2	limitations	limitation	NOUN
ajst-29613	21	3	are	be	AUX
ajst-29613	21	4	particularly	particularly	ADV
ajst-29613	21	5	pronounced	pronounce	VERB
ajst-29613	21	6	in	in	ADP
ajst-29613	21	7	tasks	task	NOUN
ajst-29613	21	8	that	that	PRON
ajst-29613	21	9	demand	demand	VERB
ajst-29613	21	10	up	up	ADP
ajst-29613	21	11	-	-	PUNCT
ajst-29613	21	12	to	to	ADP
ajst-29613	21	13	-	-	PUNCT
ajst-29613	21	14	date	date	NOUN
ajst-29613	21	15	information	information	NOUN
ajst-29613	21	16	or	or	CCONJ
ajst-29613	21	17	highly	highly	ADV
ajst-29613	21	18	specialized	specialized	ADJ
ajst-29613	21	19	domain	domain	NOUN
ajst-29613	21	20	knowledge	knowledge	NOUN
ajst-29613	21	21	,	,	PUNCT
ajst-29613	21	22	such	such	ADJ
ajst-29613	21	23	as	as	ADP
ajst-29613	21	24	medical	medical	ADJ
ajst-29613	21	25	diagnosis	diagnosis	NOUN
ajst-29613	21	26	,	,	PUNCT
ajst-29613	21	27	legal	legal	ADJ
ajst-29613	21	28	consultation	consultation	NOUN
ajst-29613	21	29	,	,	PUNCT
ajst-29613	21	30	or	or	CCONJ
ajst-29613	21	31	scientific	scientific	ADJ
ajst-29613	21	32	research	research	NOUN
ajst-29613	21	33	.	.	PUNCT
ajst-29613	22	1	although	although	SCONJ
ajst-29613	22	2	rag	rag	NOUN
ajst-29613	22	3	has	have	AUX
ajst-29613	22	4	achieved	achieve	VERB
ajst-29613	22	5	significant	significant	ADJ
ajst-29613	22	6	results	result	NOUN
ajst-29613	22	7	and	and	CCONJ
ajst-29613	22	8	has	have	AUX
ajst-29613	22	9	been	be	AUX
ajst-29613	22	10	widely	widely	ADV
ajst-29613	22	11	applied	apply	VERB
ajst-29613	22	12	across	across	ADP
ajst-29613	22	13	various	various	ADJ
ajst-29613	22	14	fields	field	NOUN
ajst-29613	22	15	,	,	PUNCT
ajst-29613	22	16	it	it	PRON
ajst-29613	22	17	still	still	ADV
ajst-29613	22	18	faces	face	VERB
ajst-29613	22	19	several	several	ADJ
ajst-29613	22	20	key	key	ADJ
ajst-29613	22	21	limitations	limitation	NOUN
ajst-29613	22	22	in	in	ADP
ajst-29613	22	23	practical	practical	ADJ
ajst-29613	22	24	scenarios	scenario	NOUN
ajst-29613	22	25	:	:	PUNCT
ajst-29613	22	26	(	(	PUNCT
ajst-29613	22	27	i	i	NOUN
ajst-29613	22	28	)	)	PUNCT
ajst-29613	22	29	neglect	neglect	NOUN
ajst-29613	22	30	of	of	ADP
ajst-29613	22	31	structured	structured	ADJ
ajst-29613	22	32	relationships	relationship	NOUN
ajst-29613	22	33	.	.	PUNCT
ajst-29613	23	1	text	text	NOUN
ajst-29613	23	2	content	content	NOUN
ajst-29613	23	3	does	do	AUX
ajst-29613	23	4	not	not	PART
ajst-29613	23	5	exist	exist	VERB
ajst-29613	23	6	in	in	ADP
ajst-29613	23	7	isolation	isolation	NOUN
ajst-29613	23	8	but	but	CCONJ
ajst-29613	23	9	is	be	AUX
ajst-29613	23	10	interconnected	interconnect	VERB
ajst-29613	23	11	through	through	ADP
ajst-29613	23	12	complex	complex	ADJ
ajst-29613	23	13	relational	relational	ADJ
ajst-29613	23	14	networks	network	NOUN
ajst-29613	23	15	.	.	PUNCT
ajst-29613	24	1	traditional	traditional	ADJ
ajst-29613	24	2	rag	rag	NOUN
ajst-29613	24	3	methods	method	NOUN
ajst-29613	24	4	fail	fail	VERB
ajst-29613	24	5	to	to	PART
ajst-29613	24	6	effectively	effectively	ADV
ajst-29613	24	7	capture	capture	VERB
ajst-29613	24	8	such	such	ADJ
ajst-29613	24	9	structured	structured	ADJ
ajst-29613	24	10	relational	relational	ADJ
ajst-29613	24	11	knowledge	knowledge	NOUN
ajst-29613	24	12	,	,	PUNCT
ajst-29613	24	13	especially	especially	ADV
ajst-29613	24	14	knowledge	knowledge	NOUN
ajst-29613	24	15	that	that	PRON
ajst-29613	24	16	can	can	AUX
ajst-29613	24	17	not	not	PART
ajst-29613	24	18	be	be	AUX
ajst-29613	24	19	directly	directly	ADV
ajst-29613	24	20	expressed	express	VERB
ajst-29613	24	21	through	through	ADP
ajst-29613	24	22	semantic	semantic	ADJ
ajst-29613	24	23	similarity	similarity	NOUN
ajst-29613	24	24	.	.	PUNCT
ajst-29613	25	1	for	for	ADP
ajst-29613	25	2	example	example	NOUN
ajst-29613	25	3	,	,	PUNCT
ajst-29613	25	4	in	in	ADP
ajst-29613	25	5	the	the	DET
ajst-29613	25	6	citation	citation	NOUN
ajst-29613	25	7	network	network	NOUN
ajst-29613	25	8	of	of	ADP
ajst-29613	25	9	academic	academic	ADJ
ajst-29613	25	10	papers	paper	NOUN
ajst-29613	25	11	,	,	PUNCT
ajst-29613	25	12	the	the	DET
ajst-29613	25	13	citation	citation	NOUN
ajst-29613	25	14	relationships	relationship	NOUN
ajst-29613	25	15	between	between	ADP
ajst-29613	25	16	papers	paper	NOUN
ajst-29613	25	17	contain	contain	VERB
ajst-29613	25	18	rich	rich	ADJ
ajst-29613	25	19	semantic	semantic	ADJ
ajst-29613	25	20	information	information	NOUN
ajst-29613	25	21	.	.	PUNCT
ajst-29613	26	1	however	however	ADV
ajst-29613	26	2	,	,	PUNCT
ajst-29613	26	3	conventional	conventional	ADJ
ajst-29613	26	4	rag	rag	NOUN
ajst-29613	26	5	focuses	focus	VERB
ajst-29613	26	6	only	only	ADV
ajst-29613	26	7	on	on	ADP
ajst-29613	26	8	semantic	semantic	ADJ
ajst-29613	26	9	retrieval	retrieval	NOUN
ajst-29613	26	10	based	base	VERB
ajst-29613	26	11	on	on	ADP
ajst-29613	26	12	queries	query	NOUN
ajst-29613	26	13	,	,	PUNCT
ajst-29613	26	14	neglecting	neglect	VERB
ajst-29613	26	15	these	these	DET
ajst-29613	26	16	critical	critical	ADJ
ajst-29613	26	17	structured	structured	ADJ
ajst-29613	26	18	relationships	relationship	NOUN
ajst-29613	26	19	,	,	PUNCT
ajst-29613	26	20	resulting	result	VERB
ajst-29613	26	21	in	in	ADP
ajst-29613	26	22	underutilization	underutilization	NOUN
ajst-29613	26	23	of	of	ADP
ajst-29613	26	24	potential	potential	ADJ
ajst-29613	26	25	information	information	NOUN
ajst-29613	26	26	.	.	PUNCT
ajst-29613	27	1	(	(	PUNCT
ajst-29613	27	2	ii	ii	NOUN
ajst-29613	27	3	)	)	PUNCT
ajst-29613	27	4	introduction	introduction	NOUN
ajst-29613	27	5	of	of	ADP
ajst-29613	27	6	redundant	redundant	ADJ
ajst-29613	27	7	information	information	NOUN
ajst-29613	27	8	.	.	PUNCT
ajst-29613	28	1	when	when	SCONJ
ajst-29613	28	2	retrieved	retrieve	VERB
ajst-29613	28	3	text	text	NOUN
ajst-29613	28	4	fragments	fragment	NOUN
ajst-29613	28	5	are	be	AUX
ajst-29613	28	6	incorporated	incorporate	VERB
ajst-29613	28	7	into	into	ADP
ajst-29613	28	8	the	the	DET
ajst-29613	28	9	generation	generation	NOUN
ajst-29613	28	10	model	model	NOUN
ajst-29613	28	11	via	via	ADP
ajst-29613	28	12	prompts	prompt	NOUN
ajst-29613	28	13	,	,	PUNCT
ajst-29613	28	14	the	the	DET
ajst-29613	28	15	content	content	NOUN
ajst-29613	28	16	often	often	ADV
ajst-29613	28	17	appears	appear	VERB
ajst-29613	28	18	in	in	ADP
ajst-29613	28	19	a	a	DET
ajst-29613	28	20	repetitive	repetitive	ADJ
ajst-29613	28	21	manner	manner	NOUN
ajst-29613	28	22	,	,	PUNCT
ajst-29613	28	23	leading	lead	VERB
ajst-29613	28	24	to	to	ADP
ajst-29613	28	25	excessively	excessively	ADV
ajst-29613	28	26	lengthy	lengthy	ADJ
ajst-29613	28	27	prompt	prompt	ADJ
ajst-29613	28	28	contexts	contexts	NOUN
ajst-29613	28	29	.	.	PUNCT
ajst-29613	29	1	this	this	DET
ajst-29613	29	2	“	"	PUNCT
ajst-29613	29	3	verbosity	verbosity	ADJ
ajst-29613	29	4	effect	effect	NOUN
ajst-29613	29	5	”	"	PUNCT
ajst-29613	29	6	can	can	AUX
ajst-29613	29	7	cause	cause	VERB
ajst-29613	29	8	the	the	DET
ajst-29613	29	9	so	so	ADV
ajst-29613	29	10	-	-	PUNCT
ajst-29613	29	11	called	call	VERB
ajst-29613	29	12	"	"	PUNCT
ajst-29613	29	13	lost	lose	VERB
ajst-29613	29	14	in	in	ADP
ajst-29613	29	15	the	the	DET
ajst-29613	29	16	middle	middle	ADJ
ajst-29613	29	17	"	"	PUNCT
ajst-29613	29	18	phenomenon[8	phenomenon[8	NOUN
ajst-29613	29	19	]	]	PUNCT
ajst-29613	29	20	,	,	PUNCT
ajst-29613	29	21	where	where	SCONJ
ajst-29613	29	22	the	the	DET
ajst-29613	29	23	model	model	NOUN
ajst-29613	29	24	struggles	struggle	VERB
ajst-29613	29	25	to	to	PART
ajst-29613	29	26	focus	focus	VERB
ajst-29613	29	27	on	on	ADP
ajst-29613	29	28	core	core	ADJ
ajst-29613	29	29	information	information	NOUN
ajst-29613	29	30	due	due	ADP
ajst-29613	29	31	to	to	ADP
ajst-29613	29	32	overly	overly	ADV
ajst-29613	29	33	long	long	ADJ
ajst-29613	29	34	contexts	contexts	NOUN
ajst-29613	29	35	.	.	PUNCT
ajst-29613	30	1	(	(	PUNCT
ajst-29613	30	2	iii	iii	NOUN
ajst-29613	30	3	)	)	PUNCT
ajst-29613	30	4	lack	lack	NOUN
ajst-29613	30	5	of	of	ADP
ajst-29613	30	6	global	global	ADJ
ajst-29613	30	7	information	information	NOUN
ajst-29613	30	8	.	.	PUNCT
ajst-29613	31	1	since	since	SCONJ
ajst-29613	31	2	rag	rag	NOUN
ajst-29613	31	3	can	can	AUX
ajst-29613	31	4	only	only	ADV
ajst-29613	31	5	retrieve	retrieve	VERB
ajst-29613	31	6	a	a	DET
ajst-29613	31	7	limited	limited	ADJ
ajst-29613	31	8	number	number	NOUN
ajst-29613	31	9	of	of	ADP
ajst-29613	31	10	relevant	relevant	ADJ
ajst-29613	31	11	document	document	NOUN
ajst-29613	31	12	fragments	fragment	NOUN
ajst-29613	31	13	,	,	PUNCT
ajst-29613	31	14	the	the	DET
ajst-29613	31	15	system	system	NOUN
ajst-29613	31	16	lacks	lack	VERB
ajst-29613	31	17	a	a	DET
ajst-29613	31	18	global	global	ADJ
ajst-29613	31	19	perspective	perspective	NOUN
ajst-29613	31	20	of	of	ADP
ajst-29613	31	21	the	the	DET
ajst-29613	31	22	complete	complete	ADJ
ajst-29613	31	23	corpus	corpus	NOUN
ajst-29613	31	24	.	.	PUNCT
ajst-29613	32	1	this	this	DET
ajst-29613	32	2	deficiency	deficiency	NOUN
ajst-29613	32	3	leads	lead	VERB
ajst-29613	32	4	to	to	ADP
ajst-29613	32	5	suboptimal	suboptimal	ADJ
ajst-29613	32	6	performance	performance	NOUN
ajst-29613	32	7	in	in	ADP
ajst-29613	32	8	tasks	task	NOUN
ajst-29613	32	9	requiring	require	VERB
ajst-29613	32	10	long	long	ADJ
ajst-29613	32	11	-	-	PUNCT
ajst-29613	32	12	context	context	NOUN
ajst-29613	32	13	summarization	summarization	NOUN
ajst-29613	32	14	,	,	PUNCT
ajst-29613	32	15	such	such	ADJ
ajst-29613	32	16	as	as	ADP
ajst-29613	32	17	abstracting	abstract	VERB
ajst-29613	32	18	or	or	CCONJ
ajst-29613	32	19	summarizing	summarizing	NOUN
ajst-29613	32	20	content	content	NOUN
ajst-29613	32	21	.	.	PUNCT
ajst-29613	33	1	to	to	PART
ajst-29613	33	2	further	far	ADV
ajst-29613	33	3	optimize	optimize	VERB
ajst-29613	33	4	the	the	DET
ajst-29613	33	5	rag	rag	NOUN
ajst-29613	33	6	process	process	NOUN
ajst-29613	33	7	and	and	CCONJ
ajst-29613	33	8	enhance	enhance	VERB
ajst-29613	33	9	retrieval	retrieval	NOUN
ajst-29613	33	10	accuracy	accuracy	NOUN
ajst-29613	33	11	,	,	PUNCT
ajst-29613	33	12	a	a	DET
ajst-29613	33	13	natural	natural	ADJ
ajst-29613	33	14	and	and	CCONJ
ajst-29613	33	15	promising	promising	ADJ
ajst-29613	33	16	solution	solution	NOUN
ajst-29613	33	17	is	be	AUX
ajst-29613	33	18	to	to	PART
ajst-29613	33	19	integrate	integrate	VERB
ajst-29613	33	20	structured	structured	ADJ
ajst-29613	33	21	external	external	ADJ
ajst-29613	33	22	knowledge	knowledge	NOUN
ajst-29613	33	23	,	,	PUNCT
ajst-29613	33	24	such	such	ADJ
ajst-29613	33	25	as	as	ADP
ajst-29613	33	26	knowledge	knowledge	NOUN
ajst-29613	33	27	graphs	graph	NOUN
ajst-29613	33	28	(	(	PUNCT
ajst-29613	33	29	kgs	kgs	NOUN
ajst-29613	33	30	)	)	PUNCT
ajst-29613	33	31	,	,	PUNCT
ajst-29613	33	32	to	to	PART
ajst-29613	33	33	improve	improve	VERB
ajst-29613	33	34	llm	llm	PROPN
ajst-29613	33	35	reasoning	reasoning	NOUN
ajst-29613	33	36	.	.	PUNCT
ajst-29613	34	1	the	the	DET
ajst-29613	34	2	explicit	explicit	ADJ
ajst-29613	34	3	relational	relational	ADJ
ajst-29613	34	4	networks	network	NOUN
ajst-29613	34	5	in	in	ADP
ajst-29613	34	6	kgs	kgs	PROPN
ajst-29613	34	7	naturally	naturally	ADV
ajst-29613	34	8	support	support	VERB
ajst-29613	34	9	multi	multi	ADJ
ajst-29613	34	10	-	-	ADJ
ajst-29613	34	11	hop	hop	ADJ
ajst-29613	34	12	reasoning	reasoning	NOUN
ajst-29613	34	13	tasks	task	NOUN
ajst-29613	34	14	,	,	PUNCT
ajst-29613	34	15	where	where	SCONJ
ajst-29613	34	16	elements	element	NOUN
ajst-29613	34	17	such	such	ADJ
ajst-29613	34	18	as	as	ADP
ajst-29613	34	19	nodes	node	NOUN
ajst-29613	34	20	,	,	PUNCT
ajst-29613	34	21	triples	triple	NOUN
ajst-29613	34	22	,	,	PUNCT
ajst-29613	34	23	paths	path	NOUN
ajst-29613	34	24	,	,	PUNCT
ajst-29613	34	25	or	or	CCONJ
ajst-29613	34	26	subgraphs	subgraph	NOUN
ajst-29613	34	27	can	can	AUX
ajst-29613	34	28	228	228	NUM
ajst-29613	34	29	be	be	AUX
ajst-29613	34	30	utilized	utilize	VERB
ajst-29613	34	31	to	to	PART
ajst-29613	34	32	generate	generate	VERB
ajst-29613	34	33	responses	response	NOUN
ajst-29613	34	34	.	.	PUNCT
ajst-29613	35	1	kgs	kg	NOUN
ajst-29613	35	2	-	-	PUNCT
ajst-29613	35	3	rag	rag	NOUN
ajst-29613	35	4	considers	consider	VERB
ajst-29613	35	5	the	the	DET
ajst-29613	35	6	interconnections	interconnection	NOUN
ajst-29613	35	7	between	between	ADP
ajst-29613	35	8	texts	text	NOUN
ajst-29613	35	9	,	,	PUNCT
ajst-29613	35	10	enabling	enable	VERB
ajst-29613	35	11	more	more	ADV
ajst-29613	35	12	accurate	accurate	ADJ
ajst-29613	35	13	and	and	CCONJ
ajst-29613	35	14	comprehensive	comprehensive	ADJ
ajst-29613	35	15	retrieval	retrieval	NOUN
ajst-29613	35	16	of	of	ADP
ajst-29613	35	17	relational	relational	ADJ
ajst-29613	35	18	information	information	NOUN
ajst-29613	35	19	.	.	PUNCT
ajst-29613	36	1	when	when	SCONJ
ajst-29613	36	2	combined	combine	VERB
ajst-29613	36	3	with	with	ADP
ajst-29613	36	4	rag	rag	NOUN
ajst-29613	36	5	,	,	PUNCT
ajst-29613	36	6	kgs	kgs	PROPN
ajst-29613	36	7	can	can	AUX
ajst-29613	36	8	retrieve	retrieve	VERB
ajst-29613	36	9	relevant	relevant	ADJ
ajst-29613	36	10	entities	entity	NOUN
ajst-29613	36	11	and	and	CCONJ
ajst-29613	36	12	relationships	relationship	NOUN
ajst-29613	36	13	from	from	ADP
ajst-29613	36	14	the	the	DET
ajst-29613	36	15	graph	graph	NOUN
ajst-29613	36	16	,	,	PUNCT
ajst-29613	36	17	aiding	aid	VERB
ajst-29613	36	18	llms	llm	NOUN
ajst-29613	36	19	in	in	ADP
ajst-29613	36	20	performing	perform	VERB
ajst-29613	36	21	complex	complex	ADJ
ajst-29613	36	22	logical	logical	ADJ
ajst-29613	36	23	reasoning	reasoning	NOUN
ajst-29613	36	24	and	and	CCONJ
ajst-29613	36	25	causal	causal	ADJ
ajst-29613	36	26	analysis	analysis	NOUN
ajst-29613	36	27	.	.	PUNCT
ajst-29613	37	1	moreover	moreover	ADV
ajst-29613	37	2	,	,	PUNCT
ajst-29613	37	3	kgs	kgs	X
ajst-29613	37	4	dynamically	dynamically	ADV
ajst-29613	37	5	provide	provide	VERB
ajst-29613	37	6	high	high	ADJ
ajst-29613	37	7	-	-	PUNCT
ajst-29613	37	8	quality	quality	NOUN
ajst-29613	37	9	information	information	NOUN
ajst-29613	37	10	sources	source	NOUN
ajst-29613	37	11	,	,	PUNCT
ajst-29613	37	12	reducing	reduce	VERB
ajst-29613	37	13	the	the	DET
ajst-29613	37	14	risk	risk	NOUN
ajst-29613	37	15	of	of	ADP
ajst-29613	37	16	hallucinations	hallucination	NOUN
ajst-29613	37	17	in	in	ADP
ajst-29613	37	18	llm	llm	PROPN
ajst-29613	37	19	outputs	output	NOUN
ajst-29613	37	20	.	.	PUNCT
ajst-29613	38	1	this	this	DET
ajst-29613	38	2	integration	integration	NOUN
ajst-29613	38	3	also	also	ADV
ajst-29613	38	4	offers	offer	VERB
ajst-29613	38	5	high	high	ADJ
ajst-29613	38	6	interpretability	interpretability	NOUN
ajst-29613	38	7	,	,	PUNCT
ajst-29613	38	8	as	as	SCONJ
ajst-29613	38	9	knowledge	knowledge	NOUN
ajst-29613	38	10	graphs	graph	NOUN
ajst-29613	38	11	provide	provide	VERB
ajst-29613	38	12	clear	clear	ADJ
ajst-29613	38	13	knowledge	knowledge	NOUN
ajst-29613	38	14	sources	source	NOUN
ajst-29613	38	15	and	and	CCONJ
ajst-29613	38	16	reasoning	reasoning	NOUN
ajst-29613	38	17	paths	path	NOUN
ajst-29613	38	18	.	.	PUNCT
ajst-29613	39	1	together	together	ADV
ajst-29613	39	2	with	with	ADP
ajst-29613	39	3	documents	document	NOUN
ajst-29613	39	4	retrieved	retrieve	VERB
ajst-29613	39	5	by	by	ADP
ajst-29613	39	6	rag	rag	NOUN
ajst-29613	39	7	,	,	PUNCT
ajst-29613	39	8	they	they	PRON
ajst-29613	39	9	form	form	VERB
ajst-29613	39	10	a	a	DET
ajst-29613	39	11	solid	solid	ADJ
ajst-29613	39	12	basis	basis	NOUN
ajst-29613	39	13	for	for	ADP
ajst-29613	39	14	supporting	support	VERB
ajst-29613	39	15	the	the	DET
ajst-29613	39	16	generated	generate	VERB
ajst-29613	39	17	content	content	NOUN
ajst-29613	39	18	.	.	PUNCT
ajst-29613	40	1	this	this	DET
ajst-29613	40	2	combination	combination	NOUN
ajst-29613	40	3	improves	improve	VERB
ajst-29613	40	4	the	the	DET
ajst-29613	40	5	interpretability	interpretability	NOUN
ajst-29613	40	6	of	of	ADP
ajst-29613	40	7	the	the	DET
ajst-29613	40	8	model	model	NOUN
ajst-29613	40	9	’s	’s	PART
ajst-29613	40	10	output	output	NOUN
ajst-29613	40	11	,	,	PUNCT
ajst-29613	40	12	allowing	allow	VERB
ajst-29613	40	13	users	user	NOUN
ajst-29613	40	14	to	to	PART
ajst-29613	40	15	trace	trace	VERB
ajst-29613	40	16	the	the	DET
ajst-29613	40	17	sources	source	NOUN
ajst-29613	40	18	of	of	ADP
ajst-29613	40	19	information	information	NOUN
ajst-29613	40	20	.	.	PUNCT
ajst-29613	41	1	2	2	X
ajst-29613	41	2	.	.	X
ajst-29613	41	3	related	relate	VERB
ajst-29613	41	4	work	work	NOUN
ajst-29613	41	5	2.1	2.1	NUM
ajst-29613	41	6	.	.	PUNCT
ajst-29613	42	1	knowledge	knowledge	NOUN
ajst-29613	42	2	graphs	graph	VERB
ajst-29613	42	3	knowledge	knowledge	NOUN
ajst-29613	42	4	graph	graph	NOUN
ajst-29613	42	5	is	be	AUX
ajst-29613	42	6	a	a	DET
ajst-29613	42	7	framework	framework	NOUN
ajst-29613	42	8	for	for	ADP
ajst-29613	42	9	representing	represent	VERB
ajst-29613	42	10	knowledge	knowledge	NOUN
ajst-29613	42	11	in	in	ADP
ajst-29613	42	12	a	a	DET
ajst-29613	42	13	structured	structured	ADJ
ajst-29613	42	14	form	form	NOUN
ajst-29613	42	15	,	,	PUNCT
ajst-29613	42	16	capturing	capture	VERB
ajst-29613	42	17	entities	entity	NOUN
ajst-29613	42	18	and	and	CCONJ
ajst-29613	42	19	their	their	PRON
ajst-29613	42	20	relationships	relationship	NOUN
ajst-29613	42	21	to	to	PART
ajst-29613	42	22	reflect	reflect	VERB
ajst-29613	42	23	real	real	ADJ
ajst-29613	42	24	-	-	PUNCT
ajst-29613	42	25	world	world	NOUN
ajst-29613	42	26	objects	object	NOUN
ajst-29613	42	27	and	and	CCONJ
ajst-29613	42	28	their	their	PRON
ajst-29613	42	29	semantic	semantic	ADJ
ajst-29613	42	30	associations	association	NOUN
ajst-29613	42	31	.	.	PUNCT
ajst-29613	43	1	a	a	DET
ajst-29613	43	2	knowledge	knowledge	NOUN
ajst-29613	43	3	graph	graph	NOUN
ajst-29613	43	4	is	be	AUX
ajst-29613	43	5	composed	compose	VERB
ajst-29613	43	6	of	of	ADP
ajst-29613	43	7	nodes	node	NOUN
ajst-29613	43	8	and	and	CCONJ
ajst-29613	43	9	edges	edge	NOUN
ajst-29613	43	10	,	,	PUNCT
ajst-29613	43	11	where	where	SCONJ
ajst-29613	43	12	nodes	node	NOUN
ajst-29613	43	13	represent	represent	VERB
ajst-29613	43	14	entities	entity	NOUN
ajst-29613	43	15	,	,	PUNCT
ajst-29613	43	16	and	and	CCONJ
ajst-29613	43	17	edges	edge	NOUN
ajst-29613	43	18	denote	denote	VERB
ajst-29613	43	19	semantic	semantic	ADJ
ajst-29613	43	20	relationships	relationship	NOUN
ajst-29613	43	21	between	between	ADP
ajst-29613	43	22	these	these	DET
ajst-29613	43	23	entities	entity	NOUN
ajst-29613	43	24	.	.	PUNCT
ajst-29613	44	1	this	this	DET
ajst-29613	44	2	structured	structured	ADJ
ajst-29613	44	3	knowledge	knowledge	NOUN
ajst-29613	44	4	representation	representation	NOUN
ajst-29613	44	5	,	,	PUNCT
ajst-29613	44	6	based	base	VERB
ajst-29613	44	7	on	on	ADP
ajst-29613	44	8	triples	triple	NOUN
ajst-29613	44	9	<	<	X
ajst-29613	44	10	subject	subject	ADJ
ajst-29613	44	11	,	,	PUNCT
ajst-29613	44	12	predicate	predicate	NOUN
ajst-29613	44	13	,	,	PUNCT
ajst-29613	44	14	object	object	VERB
ajst-29613	44	15	>	>	PUNCT
ajst-29613	44	16	,	,	PUNCT
ajst-29613	44	17	effectively	effectively	ADV
ajst-29613	44	18	supports	support	VERB
ajst-29613	44	19	knowledge	knowledge	NOUN
ajst-29613	44	20	storage	storage	NOUN
ajst-29613	44	21	,	,	PUNCT
ajst-29613	44	22	retrieval	retrieval	NOUN
ajst-29613	44	23	,	,	PUNCT
ajst-29613	44	24	and	and	CCONJ
ajst-29613	44	25	reasoning	reasoning	NOUN
ajst-29613	44	26	.	.	PUNCT
ajst-29613	45	1	knowledge	knowledge	NOUN
ajst-29613	45	2	graphs	graph	NOUN
ajst-29613	45	3	were	be	AUX
ajst-29613	45	4	first	first	ADV
ajst-29613	45	5	formally	formally	ADV
ajst-29613	45	6	introduced	introduce	VERB
ajst-29613	45	7	by	by	ADP
ajst-29613	45	8	google	google	PROPN
ajst-29613	45	9	in	in	ADP
ajst-29613	45	10	2012	2012	NUM
ajst-29613	45	11	,	,	PUNCT
ajst-29613	45	12	aiming	aim	VERB
ajst-29613	45	13	to	to	PART
ajst-29613	45	14	enhance	enhance	VERB
ajst-29613	45	15	search	search	NOUN
ajst-29613	45	16	engine	engine	NOUN
ajst-29613	45	17	results	result	NOUN
ajst-29613	45	18	through	through	ADP
ajst-29613	45	19	structured	structured	ADJ
ajst-29613	45	20	information[9	information[9	ADV
ajst-29613	45	21	]	]	PUNCT
ajst-29613	45	22	.	.	PUNCT
ajst-29613	46	1	with	with	ADP
ajst-29613	46	2	further	further	ADJ
ajst-29613	46	3	research	research	NOUN
ajst-29613	46	4	,	,	PUNCT
ajst-29613	46	5	knowledge	knowledge	NOUN
ajst-29613	46	6	graphs	graph	NOUN
ajst-29613	46	7	have	have	AUX
ajst-29613	46	8	become	become	VERB
ajst-29613	46	9	a	a	DET
ajst-29613	46	10	cornerstone	cornerstone	NOUN
ajst-29613	46	11	of	of	ADP
ajst-29613	46	12	artificial	artificial	ADJ
ajst-29613	46	13	intelligence	intelligence	NOUN
ajst-29613	46	14	and	and	CCONJ
ajst-29613	46	15	are	be	AUX
ajst-29613	46	16	widely	widely	ADV
ajst-29613	46	17	applied	apply	VERB
ajst-29613	46	18	in	in	ADP
ajst-29613	46	19	semantic	semantic	ADJ
ajst-29613	46	20	search	search	NOUN
ajst-29613	46	21	,	,	PUNCT
ajst-29613	46	22	intelligent	intelligent	ADJ
ajst-29613	46	23	question	question	NOUN
ajst-29613	46	24	answering	answering	NOUN
ajst-29613	46	25	,	,	PUNCT
ajst-29613	46	26	recommendation	recommendation	NOUN
ajst-29613	46	27	systems	system	NOUN
ajst-29613	46	28	,	,	PUNCT
ajst-29613	46	29	and	and	CCONJ
ajst-29613	46	30	decision	decision	NOUN
ajst-29613	46	31	support	support	NOUN
ajst-29613	46	32	.	.	PUNCT
ajst-29613	47	1	in	in	ADP
ajst-29613	47	2	recent	recent	ADJ
ajst-29613	47	3	years	year	NOUN
ajst-29613	47	4	,	,	PUNCT
ajst-29613	47	5	multimodal	multimodal	ADJ
ajst-29613	47	6	knowledge	knowledge	NOUN
ajst-29613	47	7	graphs	graph	NOUN
ajst-29613	47	8	and	and	CCONJ
ajst-29613	47	9	dynamic	dynamic	ADJ
ajst-29613	47	10	knowledge	knowledge	NOUN
ajst-29613	47	11	graphs	graph	NOUN
ajst-29613	47	12	have	have	AUX
ajst-29613	47	13	emerged	emerge	VERB
ajst-29613	47	14	as	as	ADP
ajst-29613	47	15	research	research	NOUN
ajst-29613	47	16	hotspots	hotspot	NOUN
ajst-29613	47	17	.	.	PUNCT
ajst-29613	48	1	multimodal	multimodal	ADJ
ajst-29613	48	2	knowledge	knowledge	NOUN
ajst-29613	48	3	graphs	graph	NOUN
ajst-29613	48	4	integrate	integrate	VERB
ajst-29613	48	5	diverse	diverse	ADJ
ajst-29613	48	6	data	datum	NOUN
ajst-29613	48	7	forms	form	NOUN
ajst-29613	48	8	,	,	PUNCT
ajst-29613	48	9	such	such	ADJ
ajst-29613	48	10	as	as	ADP
ajst-29613	48	11	text	text	NOUN
ajst-29613	48	12	,	,	PUNCT
ajst-29613	48	13	images	image	NOUN
ajst-29613	48	14	,	,	PUNCT
ajst-29613	48	15	and	and	CCONJ
ajst-29613	48	16	videos	video	NOUN
ajst-29613	48	17	,	,	PUNCT
ajst-29613	48	18	to	to	PART
ajst-29613	48	19	enhance	enhance	VERB
ajst-29613	48	20	knowledge	knowledge	NOUN
ajst-29613	48	21	representation	representation	NOUN
ajst-29613	48	22	and	and	CCONJ
ajst-29613	48	23	semantic	semantic	ADJ
ajst-29613	48	24	richness[10	richness[10	NOUN
ajst-29613	48	25	]	]	PUNCT
ajst-29613	48	26	.	.	PUNCT
ajst-29613	49	1	dynamic	dynamic	ADJ
ajst-29613	49	2	knowledge	knowledge	NOUN
ajst-29613	49	3	graphs	graph	NOUN
ajst-29613	49	4	focus	focus	VERB
ajst-29613	49	5	on	on	ADP
ajst-29613	49	6	real	real	ADJ
ajst-29613	49	7	-	-	PUNCT
ajst-29613	49	8	time	time	NOUN
ajst-29613	49	9	updates	update	NOUN
ajst-29613	49	10	and	and	CCONJ
ajst-29613	49	11	incremental	incremental	ADJ
ajst-29613	49	12	reasoning	reasoning	NOUN
ajst-29613	49	13	in	in	ADP
ajst-29613	49	14	ever	ever	ADV
ajst-29613	49	15	-	-	PUNCT
ajst-29613	49	16	changing	change	VERB
ajst-29613	49	17	environments	environment	NOUN
ajst-29613	49	18	to	to	PART
ajst-29613	49	19	meet	meet	VERB
ajst-29613	49	20	rapidly	rapidly	ADV
ajst-29613	49	21	evolving	evolve	VERB
ajst-29613	49	22	business	business	NOUN
ajst-29613	49	23	needs[11	needs[11	PROPN
ajst-29613	49	24	]	]	PUNCT
ajst-29613	49	25	.	.	PUNCT
ajst-29613	50	1	knowledge	knowledge	NOUN
ajst-29613	50	2	graphs	graph	NOUN
ajst-29613	50	3	can	can	AUX
ajst-29613	50	4	significantly	significantly	ADV
ajst-29613	50	5	enhance	enhance	VERB
ajst-29613	50	6	the	the	DET
ajst-29613	50	7	multi	multi	ADJ
ajst-29613	50	8	-	-	ADJ
ajst-29613	50	9	hop	hop	ADJ
ajst-29613	50	10	reasoning	reasoning	NOUN
ajst-29613	50	11	capabilities	capability	NOUN
ajst-29613	50	12	of	of	ADP
ajst-29613	50	13	llms	llm	NOUN
ajst-29613	50	14	.	.	PUNCT
ajst-29613	51	1	by	by	ADP
ajst-29613	51	2	inherently	inherently	ADV
ajst-29613	51	3	supporting	support	VERB
ajst-29613	51	4	relationship	relationship	NOUN
ajst-29613	51	5	chain	chain	NOUN
ajst-29613	51	6	tracing	trace	VERB
ajst-29613	51	7	between	between	ADP
ajst-29613	51	8	entities	entity	NOUN
ajst-29613	51	9	,	,	PUNCT
ajst-29613	51	10	they	they	PRON
ajst-29613	51	11	excel	excel	VERB
ajst-29613	51	12	in	in	ADP
ajst-29613	51	13	complex	complex	ADJ
ajst-29613	51	14	reasoning	reasoning	NOUN
ajst-29613	51	15	tasks	task	NOUN
ajst-29613	51	16	.	.	PUNCT
ajst-29613	52	1	for	for	ADP
ajst-29613	52	2	instance	instance	NOUN
ajst-29613	52	3	,	,	PUNCT
ajst-29613	52	4	knowledge	knowledge	NOUN
ajst-29613	52	5	graphbased	graphbased	ADJ
ajst-29613	52	6	question	question	NOUN
ajst-29613	52	7	-	-	PUNCT
ajst-29613	52	8	answering	answer	VERB
ajst-29613	52	9	systems	system	NOUN
ajst-29613	52	10	can	can	AUX
ajst-29613	52	11	derive	derive	VERB
ajst-29613	52	12	answers	answer	NOUN
ajst-29613	52	13	through	through	ADP
ajst-29613	52	14	multi	multi	ADJ
ajst-29613	52	15	-	-	ADJ
ajst-29613	52	16	step	step	ADJ
ajst-29613	52	17	retrieval	retrieval	NOUN
ajst-29613	52	18	,	,	PUNCT
ajst-29613	52	19	a	a	DET
ajst-29613	52	20	capability	capability	NOUN
ajst-29613	52	21	relatively	relatively	ADV
ajst-29613	52	22	weak	weak	ADJ
ajst-29613	52	23	in	in	ADP
ajst-29613	52	24	traditional	traditional	ADJ
ajst-29613	52	25	llms	llm	NOUN
ajst-29613	52	26	.	.	PUNCT
ajst-29613	53	1	in	in	ADP
ajst-29613	53	2	terms	term	NOUN
ajst-29613	53	3	of	of	ADP
ajst-29613	53	4	content	content	NOUN
ajst-29613	53	5	generation	generation	NOUN
ajst-29613	53	6	interpretability	interpretability	NOUN
ajst-29613	53	7	,	,	PUNCT
ajst-29613	53	8	knowledge	knowledge	NOUN
ajst-29613	53	9	graphs	graph	NOUN
ajst-29613	53	10	provide	provide	VERB
ajst-29613	53	11	a	a	DET
ajst-29613	53	12	reliable	reliable	ADJ
ajst-29613	53	13	explanatory	explanatory	ADJ
ajst-29613	53	14	foundation	foundation	NOUN
ajst-29613	53	15	by	by	ADP
ajst-29613	53	16	explicitly	explicitly	ADV
ajst-29613	53	17	showcasing	showcase	VERB
ajst-29613	53	18	reasoning	reasoning	NOUN
ajst-29613	53	19	paths	path	NOUN
ajst-29613	53	20	and	and	CCONJ
ajst-29613	53	21	data	datum	NOUN
ajst-29613	53	22	sources	source	NOUN
ajst-29613	53	23	,	,	PUNCT
ajst-29613	53	24	which	which	PRON
ajst-29613	53	25	is	be	AUX
ajst-29613	53	26	especially	especially	ADV
ajst-29613	53	27	critical	critical	ADJ
ajst-29613	53	28	in	in	ADP
ajst-29613	53	29	highstakes	highstake	NOUN
ajst-29613	53	30	fields	field	NOUN
ajst-29613	53	31	requiring	require	VERB
ajst-29613	53	32	transparency	transparency	NOUN
ajst-29613	53	33	,	,	PUNCT
ajst-29613	53	34	such	such	ADJ
ajst-29613	53	35	as	as	ADP
ajst-29613	53	36	healthcare	healthcare	NOUN
ajst-29613	53	37	and	and	CCONJ
ajst-29613	53	38	finance	finance	NOUN
ajst-29613	53	39	.	.	PUNCT
ajst-29613	54	1	2.2	2.2	NUM
ajst-29613	54	2	.	.	PUNCT
ajst-29613	54	3	retrieval	retrieval	NOUN
ajst-29613	54	4	-	-	PUNCT
ajst-29613	54	5	augmented	augment	VERB
ajst-29613	54	6	generation	generation	NOUN
ajst-29613	54	7	retrieval	retrieval	NOUN
ajst-29613	54	8	-	-	PUNCT
ajst-29613	54	9	augmented	augment	VERB
ajst-29613	54	10	generation	generation	NOUN
ajst-29613	54	11	is	be	AUX
ajst-29613	54	12	a	a	DET
ajst-29613	54	13	technique	technique	NOUN
ajst-29613	54	14	that	that	PRON
ajst-29613	54	15	combines	combine	VERB
ajst-29613	54	16	information	information	NOUN
ajst-29613	54	17	retrieval	retrieval	NOUN
ajst-29613	54	18	and	and	CCONJ
ajst-29613	54	19	generative	generative	ADJ
ajst-29613	54	20	models	model	NOUN
ajst-29613	54	21	to	to	PART
ajst-29613	54	22	enhance	enhance	VERB
ajst-29613	54	23	the	the	DET
ajst-29613	54	24	capabilities	capability	NOUN
ajst-29613	54	25	of	of	ADP
ajst-29613	54	26	text	text	NOUN
ajst-29613	54	27	generation	generation	NOUN
ajst-29613	54	28	systems	system	NOUN
ajst-29613	54	29	.	.	PUNCT
ajst-29613	55	1	unlike	unlike	ADP
ajst-29613	55	2	traditional	traditional	ADJ
ajst-29613	55	3	generative	generative	ADJ
ajst-29613	55	4	models	model	NOUN
ajst-29613	55	5	that	that	PRON
ajst-29613	55	6	rely	rely	VERB
ajst-29613	55	7	solely	solely	ADV
ajst-29613	55	8	on	on	ADP
ajst-29613	55	9	training	train	VERB
ajst-29613	55	10	data	datum	NOUN
ajst-29613	55	11	to	to	PART
ajst-29613	55	12	produce	produce	VERB
ajst-29613	55	13	text	text	NOUN
ajst-29613	55	14	,	,	PUNCT
ajst-29613	55	15	rag	rag	NOUN
ajst-29613	55	16	incorporates	incorporate	VERB
ajst-29613	55	17	an	an	DET
ajst-29613	55	18	external	external	ADJ
ajst-29613	55	19	retrieval	retrieval	NOUN
ajst-29613	55	20	module	module	NOUN
ajst-29613	55	21	to	to	PART
ajst-29613	55	22	dynamically	dynamically	ADV
ajst-29613	55	23	fetch	fetch	VERB
ajst-29613	55	24	external	external	ADJ
ajst-29613	55	25	knowledge	knowledge	NOUN
ajst-29613	55	26	relevant	relevant	ADJ
ajst-29613	55	27	to	to	ADP
ajst-29613	55	28	the	the	DET
ajst-29613	55	29	user	user	NOUN
ajst-29613	55	30	input	input	NOUN
ajst-29613	55	31	.	.	PUNCT
ajst-29613	56	1	this	this	DET
ajst-29613	56	2	approach	approach	NOUN
ajst-29613	56	3	improves	improve	VERB
ajst-29613	56	4	the	the	DET
ajst-29613	56	5	accuracy	accuracy	NOUN
ajst-29613	56	6	and	and	CCONJ
ajst-29613	56	7	richness	richness	NOUN
ajst-29613	56	8	of	of	ADP
ajst-29613	56	9	the	the	DET
ajst-29613	56	10	generated	generate	VERB
ajst-29613	56	11	content	content	NOUN
ajst-29613	56	12	,	,	PUNCT
ajst-29613	56	13	making	make	VERB
ajst-29613	56	14	it	it	PRON
ajst-29613	56	15	particularly	particularly	ADV
ajst-29613	56	16	suitable	suitable	ADJ
ajst-29613	56	17	for	for	ADP
ajst-29613	56	18	tasks	task	NOUN
ajst-29613	56	19	requiring	require	VERB
ajst-29613	56	20	open	open	ADJ
ajst-29613	56	21	-	-	PUNCT
ajst-29613	56	22	domain	domain	NOUN
ajst-29613	56	23	knowledge	knowledge	NOUN
ajst-29613	56	24	,	,	PUNCT
ajst-29613	56	25	such	such	ADJ
ajst-29613	56	26	as	as	ADP
ajst-29613	56	27	question	question	NOUN
ajst-29613	56	28	answering	answering	NOUN
ajst-29613	56	29	,	,	PUNCT
ajst-29613	56	30	dialogue	dialogue	NOUN
ajst-29613	56	31	generation	generation	NOUN
ajst-29613	56	32	,	,	PUNCT
ajst-29613	56	33	and	and	CCONJ
ajst-29613	56	34	automated	automate	VERB
ajst-29613	56	35	writing	writing	NOUN
ajst-29613	56	36	.	.	PUNCT
ajst-29613	57	1	the	the	DET
ajst-29613	57	2	concept	concept	NOUN
ajst-29613	57	3	of	of	ADP
ajst-29613	57	4	rag	rag	NOUN
ajst-29613	57	5	originates	originate	NOUN
ajst-29613	57	6	from	from	ADP
ajst-29613	57	7	the	the	DET
ajst-29613	57	8	integration	integration	NOUN
ajst-29613	57	9	of	of	ADP
ajst-29613	57	10	information	information	NOUN
ajst-29613	57	11	retrieval	retrieval	NOUN
ajst-29613	57	12	and	and	CCONJ
ajst-29613	57	13	generative	generative	ADJ
ajst-29613	57	14	models	model	NOUN
ajst-29613	57	15	.	.	PUNCT
ajst-29613	58	1	early	early	ADV
ajst-29613	58	2	retrievalaugmented	retrievalaugmente	VERB
ajst-29613	58	3	methods	method	NOUN
ajst-29613	58	4	were	be	AUX
ajst-29613	58	5	primarily	primarily	ADV
ajst-29613	58	6	applied	apply	VERB
ajst-29613	58	7	in	in	ADP
ajst-29613	58	8	questionanswering	questionanswere	VERB
ajst-29613	58	9	systems	system	NOUN
ajst-29613	58	10	.	.	PUNCT
ajst-29613	59	1	for	for	ADP
ajst-29613	59	2	instance	instance	NOUN
ajst-29613	59	3	,	,	PUNCT
ajst-29613	59	4	drqa	drqa	VERB
ajst-29613	60	1	[	[	X
ajst-29613	60	2	12	12	NUM
ajst-29613	60	3	]	]	PUNCT
ajst-29613	60	4	combined	combine	VERB
ajst-29613	60	5	traditional	traditional	ADJ
ajst-29613	60	6	retrieval	retrieval	NOUN
ajst-29613	60	7	models	model	NOUN
ajst-29613	60	8	(	(	PUNCT
ajst-29613	60	9	e.g.	e.g.	ADV
ajst-29613	60	10	,	,	PUNCT
ajst-29613	60	11	tf	tf	PROPN
ajst-29613	60	12	-	-	PUNCT
ajst-29613	60	13	idf	idf	PROPN
ajst-29613	60	14	)	)	PUNCT
ajst-29613	60	15	with	with	ADP
ajst-29613	60	16	simple	simple	ADJ
ajst-29613	60	17	reader	reader	NOUN
ajst-29613	60	18	models	model	NOUN
ajst-29613	60	19	to	to	PART
ajst-29613	60	20	provide	provide	VERB
ajst-29613	60	21	robust	robust	ADJ
ajst-29613	60	22	support	support	NOUN
ajst-29613	60	23	for	for	ADP
ajst-29613	60	24	natural	natural	ADJ
ajst-29613	60	25	language	language	NOUN
ajst-29613	60	26	question	question	NOUN
ajst-29613	60	27	answering	answering	NOUN
ajst-29613	60	28	.	.	PUNCT
ajst-29613	61	1	with	with	ADP
ajst-29613	61	2	the	the	DET
ajst-29613	61	3	advancement	advancement	NOUN
ajst-29613	61	4	of	of	ADP
ajst-29613	61	5	deep	deep	ADJ
ajst-29613	61	6	learning	learning	NOUN
ajst-29613	61	7	,	,	PUNCT
ajst-29613	61	8	models	model	NOUN
ajst-29613	61	9	like	like	ADP
ajst-29613	61	10	dense	dense	ADJ
ajst-29613	61	11	passage	passage	NOUN
ajst-29613	61	12	retrieval	retrieval	NOUN
ajst-29613	61	13	(	(	PUNCT
ajst-29613	61	14	dpr	dpr	PROPN
ajst-29613	61	15	)	)	PUNCT
ajst-29613	62	1	[	[	X
ajst-29613	62	2	13	13	NUM
ajst-29613	62	3	]	]	PUNCT
ajst-29613	62	4	,	,	PUNCT
ajst-29613	62	5	which	which	PRON
ajst-29613	62	6	leverage	leverage	VERB
ajst-29613	62	7	vector	vector	NOUN
ajst-29613	62	8	-	-	PUNCT
ajst-29613	62	9	based	base	VERB
ajst-29613	62	10	semantic	semantic	ADJ
ajst-29613	62	11	retrieval	retrieval	NOUN
ajst-29613	62	12	,	,	PUNCT
ajst-29613	62	13	gradually	gradually	ADV
ajst-29613	62	14	replaced	replace	VERB
ajst-29613	62	15	traditional	traditional	ADJ
ajst-29613	62	16	sparse	sparse	ADJ
ajst-29613	62	17	retrieval	retrieval	NOUN
ajst-29613	62	18	methods	method	NOUN
ajst-29613	62	19	.	.	PUNCT
ajst-29613	63	1	these	these	DET
ajst-29613	63	2	approaches	approach	NOUN
ajst-29613	63	3	use	use	VERB
ajst-29613	63	4	pre	pre	ADJ
ajst-29613	63	5	-	-	ADJ
ajst-29613	63	6	trained	train	VERB
ajst-29613	63	7	models	model	NOUN
ajst-29613	63	8	(	(	PUNCT
ajst-29613	63	9	e.g.	e.g.	ADV
ajst-29613	63	10	,	,	PUNCT
ajst-29613	63	11	bert	bert	PROPN
ajst-29613	63	12	)	)	PUNCT
ajst-29613	63	13	to	to	PART
ajst-29613	63	14	generate	generate	VERB
ajst-29613	63	15	dense	dense	ADJ
ajst-29613	63	16	vector	vector	NOUN
ajst-29613	63	17	representations	representation	NOUN
ajst-29613	63	18	,	,	PUNCT
ajst-29613	63	19	significantly	significantly	ADV
ajst-29613	63	20	improving	improve	VERB
ajst-29613	63	21	retrieval	retrieval	ADJ
ajst-29613	63	22	accuracy	accuracy	NOUN
ajst-29613	63	23	.	.	PUNCT
ajst-29613	64	1	the	the	DET
ajst-29613	64	2	rag	rag	NOUN
ajst-29613	64	3	framework	framework	NOUN
ajst-29613	64	4	was	be	AUX
ajst-29613	64	5	subsequently	subsequently	ADV
ajst-29613	64	6	introduced	introduce	VERB
ajst-29613	64	7	by	by	ADP
ajst-29613	64	8	lewis	lewis	PROPN
ajst-29613	64	9	et	et	PROPN
ajst-29613	64	10	al	al	PROPN
ajst-29613	64	11	.	.	PROPN
ajst-29613	65	1	in	in	ADP
ajst-29613	65	2	2020	2020	NUM
ajst-29613	65	3	[	[	X
ajst-29613	65	4	14	14	NUM
ajst-29613	65	5	]	]	PUNCT
ajst-29613	65	6	,	,	PUNCT
ajst-29613	65	7	combining	combine	VERB
ajst-29613	65	8	a	a	DET
ajst-29613	65	9	retriever	retriever	NOUN
ajst-29613	65	10	to	to	PART
ajst-29613	65	11	fetch	fetch	VERB
ajst-29613	65	12	relevant	relevant	ADJ
ajst-29613	65	13	documents	document	NOUN
ajst-29613	65	14	with	with	ADP
ajst-29613	65	15	a	a	DET
ajst-29613	65	16	generative	generative	ADJ
ajst-29613	65	17	model	model	NOUN
ajst-29613	65	18	(	(	PUNCT
ajst-29613	65	19	e.g.	e.g.	ADV
ajst-29613	65	20	,	,	PUNCT
ajst-29613	65	21	bart	bart	PROPN
ajst-29613	65	22	or	or	CCONJ
ajst-29613	65	23	t5	t5	PROPN
ajst-29613	65	24	)	)	PUNCT
ajst-29613	65	25	for	for	ADP
ajst-29613	65	26	answer	answer	NOUN
ajst-29613	65	27	generation	generation	NOUN
ajst-29613	65	28	.	.	PUNCT
ajst-29613	66	1	this	this	DET
ajst-29613	66	2	framework	framework	NOUN
ajst-29613	66	3	pioneered	pioneer	VERB
ajst-29613	66	4	the	the	DET
ajst-29613	66	5	deep	deep	ADJ
ajst-29613	66	6	integration	integration	NOUN
ajst-29613	66	7	of	of	ADP
ajst-29613	66	8	retrieval	retrieval	NOUN
ajst-29613	66	9	and	and	CCONJ
ajst-29613	66	10	generation	generation	NOUN
ajst-29613	66	11	.	.	PUNCT
ajst-29613	67	1	naive	naive	ADJ
ajst-29613	67	2	rag	rag	NOUN
ajst-29613	67	3	follows	follow	VERB
ajst-29613	67	4	a	a	DET
ajst-29613	67	5	classic	classic	ADJ
ajst-29613	67	6	three	three	NUM
ajst-29613	67	7	-	-	PUNCT
ajst-29613	67	8	stage	stage	NOUN
ajst-29613	67	9	process	process	NOUN
ajst-29613	67	10	:	:	PUNCT
ajst-29613	67	11	indexing	indexing	NOUN
ajst-29613	67	12	,	,	PUNCT
ajst-29613	67	13	retrieval	retrieval	NOUN
ajst-29613	67	14	,	,	PUNCT
ajst-29613	67	15	and	and	CCONJ
ajst-29613	67	16	generation	generation	NOUN
ajst-29613	67	17	.	.	PUNCT
ajst-29613	68	1	(	(	PUNCT
ajst-29613	68	2	i	i	NOUN
ajst-29613	68	3	)	)	PUNCT
ajst-29613	68	4	indexing	indexing	NOUN
ajst-29613	68	5	:	:	PUNCT
ajst-29613	68	6	first	first	ADV
ajst-29613	68	7	,	,	PUNCT
ajst-29613	68	8	the	the	DET
ajst-29613	68	9	system	system	NOUN
ajst-29613	68	10	processes	process	VERB
ajst-29613	68	11	and	and	CCONJ
ajst-29613	68	12	cleans	clean	VERB
ajst-29613	68	13	raw	raw	ADJ
ajst-29613	68	14	data	datum	NOUN
ajst-29613	68	15	from	from	ADP
ajst-29613	68	16	diverse	diverse	ADJ
ajst-29613	68	17	sources	source	NOUN
ajst-29613	68	18	,	,	PUNCT
ajst-29613	68	19	which	which	PRON
ajst-29613	68	20	may	may	AUX
ajst-29613	68	21	be	be	AUX
ajst-29613	68	22	stored	store	VERB
ajst-29613	68	23	in	in	ADP
ajst-29613	68	24	formats	format	NOUN
ajst-29613	68	25	such	such	ADJ
ajst-29613	68	26	as	as	ADP
ajst-29613	68	27	pdf	pdf	NOUN
ajst-29613	68	28	,	,	PUNCT
ajst-29613	68	29	html	html	NOUN
ajst-29613	68	30	,	,	PUNCT
ajst-29613	68	31	word	word	NOUN
ajst-29613	68	32	,	,	PUNCT
ajst-29613	68	33	or	or	CCONJ
ajst-29613	68	34	markdown	markdown	ADJ
ajst-29613	68	35	.	.	PUNCT
ajst-29613	69	1	to	to	PART
ajst-29613	69	2	standardize	standardize	VERB
ajst-29613	69	3	knowledge	knowledge	NOUN
ajst-29613	69	4	representation	representation	NOUN
ajst-29613	69	5	,	,	PUNCT
ajst-29613	69	6	these	these	DET
ajst-29613	69	7	files	file	NOUN
ajst-29613	69	8	are	be	AUX
ajst-29613	69	9	converted	convert	VERB
ajst-29613	69	10	into	into	ADP
ajst-29613	69	11	plain	plain	ADJ
ajst-29613	69	12	text	text	NOUN
ajst-29613	69	13	format	format	NOUN
ajst-29613	69	14	.	.	PUNCT
ajst-29613	70	1	since	since	SCONJ
ajst-29613	70	2	llms	llm	NOUN
ajst-29613	70	3	typically	typically	ADV
ajst-29613	70	4	have	have	VERB
ajst-29613	70	5	limitations	limitation	NOUN
ajst-29613	70	6	on	on	ADP
ajst-29613	70	7	input	input	NOUN
ajst-29613	70	8	context	context	NOUN
ajst-29613	70	9	length	length	NOUN
ajst-29613	70	10	,	,	PUNCT
ajst-29613	70	11	long	long	ADJ
ajst-29613	70	12	documents	document	NOUN
ajst-29613	70	13	are	be	AUX
ajst-29613	70	14	further	far	ADV
ajst-29613	70	15	divided	divide	VERB
ajst-29613	70	16	into	into	ADP
ajst-29613	70	17	smaller	small	ADJ
ajst-29613	70	18	,	,	PUNCT
ajst-29613	70	19	semantically	semantically	ADV
ajst-29613	70	20	coherent	coherent	ADJ
ajst-29613	70	21	text	text	NOUN
ajst-29613	70	22	chunks	chunk	VERB
ajst-29613	70	23	to	to	PART
ajst-29613	70	24	ensure	ensure	VERB
ajst-29613	70	25	efficient	efficient	ADJ
ajst-29613	70	26	processing	processing	NOUN
ajst-29613	70	27	and	and	CCONJ
ajst-29613	70	28	understanding	understanding	NOUN
ajst-29613	70	29	by	by	ADP
ajst-29613	70	30	the	the	DET
ajst-29613	70	31	model	model	NOUN
ajst-29613	70	32	.	.	PUNCT
ajst-29613	71	1	these	these	DET
ajst-29613	71	2	chunks	chunk	NOUN
ajst-29613	71	3	are	be	AUX
ajst-29613	71	4	then	then	ADV
ajst-29613	71	5	encoded	encode	VERB
ajst-29613	71	6	into	into	ADP
ajst-29613	71	7	vector	vector	NOUN
ajst-29613	71	8	representations	representation	NOUN
ajst-29613	71	9	using	use	VERB
ajst-29613	71	10	an	an	DET
ajst-29613	71	11	embedding	embed	VERB
ajst-29613	71	12	model	model	NOUN
ajst-29613	71	13	to	to	PART
ajst-29613	71	14	capture	capture	VERB
ajst-29613	71	15	their	their	PRON
ajst-29613	71	16	semantic	semantic	ADJ
ajst-29613	71	17	features	feature	NOUN
ajst-29613	71	18	.	.	PUNCT
ajst-29613	72	1	the	the	DET
ajst-29613	72	2	resulting	result	VERB
ajst-29613	72	3	vectors	vector	NOUN
ajst-29613	72	4	are	be	AUX
ajst-29613	72	5	stored	store	VERB
ajst-29613	72	6	in	in	ADP
ajst-29613	72	7	an	an	DET
ajst-29613	72	8	efficient	efficient	ADJ
ajst-29613	72	9	vector	vector	NOUN
ajst-29613	72	10	database	database	NOUN
ajst-29613	72	11	,	,	PUNCT
ajst-29613	72	12	enabling	enable	VERB
ajst-29613	72	13	fast	fast	ADJ
ajst-29613	72	14	similarity	similarity	NOUN
ajst-29613	72	15	retrieval	retrieval	NOUN
ajst-29613	72	16	in	in	ADP
ajst-29613	72	17	subsequent	subsequent	ADJ
ajst-29613	72	18	stages	stage	NOUN
ajst-29613	72	19	.	.	PUNCT
ajst-29613	73	1	(	(	PUNCT
ajst-29613	73	2	ii	ii	NOUN
ajst-29613	73	3	)	)	PUNCT
ajst-29613	73	4	retrieval	retrieval	NOUN
ajst-29613	73	5	:	:	PUNCT
ajst-29613	73	6	when	when	SCONJ
ajst-29613	73	7	the	the	DET
ajst-29613	73	8	system	system	NOUN
ajst-29613	73	9	receives	receive	VERB
ajst-29613	73	10	a	a	DET
ajst-29613	73	11	user	user	NOUN
ajst-29613	73	12	query	query	NOUN
ajst-29613	73	13	,	,	PUNCT
ajst-29613	73	14	it	it	PRON
ajst-29613	73	15	uses	use	VERB
ajst-29613	73	16	the	the	DET
ajst-29613	73	17	same	same	ADJ
ajst-29613	73	18	embedding	embed	VERB
ajst-29613	73	19	model	model	NOUN
ajst-29613	73	20	from	from	ADP
ajst-29613	73	21	the	the	DET
ajst-29613	73	22	indexing	indexing	NOUN
ajst-29613	73	23	phase	phase	NOUN
ajst-29613	73	24	to	to	PART
ajst-29613	73	25	encode	encode	VERB
ajst-29613	73	26	the	the	DET
ajst-29613	73	27	query	query	NOUN
ajst-29613	73	28	into	into	ADP
ajst-29613	73	29	a	a	DET
ajst-29613	73	30	vector	vector	NOUN
ajst-29613	73	31	representation	representation	NOUN
ajst-29613	73	32	.	.	PUNCT
ajst-29613	74	1	by	by	ADP
ajst-29613	74	2	calculating	calculate	VERB
ajst-29613	74	3	similarity	similarity	NOUN
ajst-29613	74	4	scores	score	NOUN
ajst-29613	74	5	between	between	ADP
ajst-29613	74	6	the	the	DET
ajst-29613	74	7	query	query	NOUN
ajst-29613	74	8	vector	vector	NOUN
ajst-29613	74	9	and	and	CCONJ
ajst-29613	74	10	all	all	DET
ajst-29613	74	11	text	text	NOUN
ajst-29613	74	12	chunk	chunk	NOUN
ajst-29613	74	13	vectors	vector	NOUN
ajst-29613	74	14	in	in	ADP
ajst-29613	74	15	the	the	DET
ajst-29613	74	16	database	database	NOUN
ajst-29613	74	17	,	,	PUNCT
ajst-29613	74	18	the	the	DET
ajst-29613	74	19	system	system	NOUN
ajst-29613	74	20	effectively	effectively	ADV
ajst-29613	74	21	identifies	identify	VERB
ajst-29613	74	22	the	the	DET
ajst-29613	74	23	content	content	NOUN
ajst-29613	74	24	most	most	ADV
ajst-29613	74	25	relevant	relevant	ADJ
ajst-29613	74	26	to	to	ADP
ajst-29613	74	27	the	the	DET
ajst-29613	74	28	query	query	NOUN
ajst-29613	74	29	.	.	PUNCT
ajst-29613	75	1	to	to	PART
ajst-29613	75	2	enhance	enhance	VERB
ajst-29613	75	3	efficiency	efficiency	NOUN
ajst-29613	75	4	and	and	CCONJ
ajst-29613	75	5	relevance	relevance	NOUN
ajst-29613	75	6	,	,	PUNCT
ajst-29613	75	7	the	the	DET
ajst-29613	75	8	system	system	NOUN
ajst-29613	75	9	typically	typically	ADV
ajst-29613	75	10	retrieves	retrieve	VERB
ajst-29613	75	11	the	the	DET
ajst-29613	75	12	top	top	ADJ
ajst-29613	75	13	-	-	PUNCT
ajst-29613	75	14	k	k	NOUN
ajst-29613	75	15	text	text	NOUN
ajst-29613	75	16	chunks	chunk	VERB
ajst-29613	75	17	with	with	ADP
ajst-29613	75	18	the	the	DET
ajst-29613	75	19	highest	high	ADJ
ajst-29613	75	20	similarity	similarity	NOUN
ajst-29613	75	21	scores	score	NOUN
ajst-29613	75	22	.	.	PUNCT
ajst-29613	76	1	these	these	DET
ajst-29613	76	2	highly	highly	ADV
ajst-29613	76	3	relevant	relevant	ADJ
ajst-29613	76	4	chunks	chunk	NOUN
ajst-29613	76	5	serve	serve	VERB
ajst-29613	76	6	as	as	ADP
ajst-29613	76	7	extended	extended	ADJ
ajst-29613	76	8	context	context	NOUN
ajst-29613	76	9	,	,	PUNCT
ajst-29613	76	10	providing	provide	VERB
ajst-29613	76	11	semantic	semantic	ADJ
ajst-29613	76	12	support	support	NOUN
ajst-29613	76	13	for	for	ADP
ajst-29613	76	14	the	the	DET
ajst-29613	76	15	generation	generation	NOUN
ajst-29613	76	16	phase	phase	NOUN
ajst-29613	76	17	.	.	PUNCT
ajst-29613	77	1	(	(	PUNCT
ajst-29613	77	2	iii	iii	NOUN
ajst-29613	77	3	)	)	PUNCT
ajst-29613	77	4	generation	generation	NOUN
ajst-29613	77	5	:	:	PUNCT
ajst-29613	77	6	the	the	DET
ajst-29613	77	7	user	user	NOUN
ajst-29613	77	8	query	query	NOUN
ajst-29613	77	9	and	and	CCONJ
ajst-29613	77	10	the	the	DET
ajst-29613	77	11	relevant	relevant	ADJ
ajst-29613	77	12	document	document	NOUN
ajst-29613	77	13	chunks	chunk	NOUN
ajst-29613	77	14	retrieved	retrieve	VERB
ajst-29613	77	15	from	from	ADP
ajst-29613	77	16	the	the	DET
ajst-29613	77	17	index	index	NOUN
ajst-29613	77	18	are	be	AUX
ajst-29613	77	19	combined	combine	VERB
ajst-29613	77	20	into	into	ADP
ajst-29613	77	21	a	a	DET
ajst-29613	77	22	coherent	coherent	ADJ
ajst-29613	77	23	prompt	prompt	NOUN
ajst-29613	77	24	,	,	PUNCT
ajst-29613	77	25	which	which	PRON
ajst-29613	77	26	is	be	AUX
ajst-29613	77	27	then	then	ADV
ajst-29613	77	28	fed	feed	VERB
ajst-29613	77	29	into	into	ADP
ajst-29613	77	30	a	a	DET
ajst-29613	77	31	llm	llm	NOUN
ajst-29613	77	32	to	to	PART
ajst-29613	77	33	generate	generate	VERB
ajst-29613	77	34	the	the	DET
ajst-29613	77	35	final	final	ADJ
ajst-29613	77	36	response	response	NOUN
ajst-29613	77	37	.	.	PUNCT
ajst-29613	78	1	depending	depend	VERB
ajst-29613	78	2	on	on	ADP
ajst-29613	78	3	the	the	DET
ajst-29613	78	4	task	task	NOUN
ajst-29613	78	5	requirements	requirement	NOUN
ajst-29613	78	6	,	,	PUNCT
ajst-29613	78	7	llms	llm	NOUN
ajst-29613	78	8	can	can	AUX
ajst-29613	78	9	adopt	adopt	VERB
ajst-29613	78	10	flexible	flexible	ADJ
ajst-29613	78	11	response	response	NOUN
ajst-29613	78	12	strategies	strategy	NOUN
ajst-29613	78	13	.	.	PUNCT
ajst-29613	79	1	one	one	NUM
ajst-29613	79	2	approach	approach	NOUN
ajst-29613	79	3	leverages	leverage	VERB
ajst-29613	79	4	the	the	DET
ajst-29613	79	5	model	model	NOUN
ajst-29613	79	6	's	's	PART
ajst-29613	79	7	inherent	inherent	ADJ
ajst-29613	79	8	parametric	parametric	ADJ
ajst-29613	79	9	knowledge	knowledge	NOUN
ajst-29613	79	10	to	to	PART
ajst-29613	79	11	produce	produce	VERB
ajst-29613	79	12	answers	answer	NOUN
ajst-29613	79	13	,	,	PUNCT
ajst-29613	79	14	while	while	SCONJ
ajst-29613	79	15	another	another	PRON
ajst-29613	79	16	strictly	strictly	ADV
ajst-29613	79	17	constrains	constrain	VERB
ajst-29613	79	18	the	the	DET
ajst-29613	79	19	generated	generate	VERB
ajst-29613	79	20	content	content	NOUN
ajst-29613	79	21	to	to	ADP
ajst-29613	79	22	the	the	DET
ajst-29613	79	23	retrieved	retrieve	VERB
ajst-29613	79	24	document	document	NOUN
ajst-29613	79	25	information	information	NOUN
ajst-29613	79	26	to	to	PART
ajst-29613	79	27	ensure	ensure	VERB
ajst-29613	79	28	accuracy	accuracy	NOUN
ajst-29613	79	29	and	and	CCONJ
ajst-29613	79	30	verifiability	verifiability	NOUN
ajst-29613	79	31	.	.	PUNCT
ajst-29613	80	1	additionally	additionally	ADV
ajst-29613	80	2	,	,	PUNCT
ajst-29613	80	3	in	in	ADP
ajst-29613	80	4	dialogue	dialogue	NOUN
ajst-29613	80	5	scenarios	scenario	NOUN
ajst-29613	80	6	,	,	PUNCT
ajst-29613	80	7	rag	rag	NOUN
ajst-29613	80	8	systems	system	NOUN
ajst-29613	80	9	can	can	AUX
ajst-29613	80	10	incorporate	incorporate	VERB
ajst-29613	80	11	dialogue	dialogue	NOUN
ajst-29613	80	12	history	history	NOUN
ajst-29613	80	13	into	into	ADP
ajst-29613	80	14	the	the	DET
ajst-29613	80	15	prompt	prompt	NOUN
ajst-29613	80	16	,	,	PUNCT
ajst-29613	80	17	supporting	support	VERB
ajst-29613	80	18	multi	multi	ADJ
ajst-29613	80	19	-	-	ADJ
ajst-29613	80	20	turn	turn	VERB
ajst-29613	80	21	interactions	interaction	NOUN
ajst-29613	80	22	.	.	PUNCT
ajst-29613	81	1	this	this	DET
ajst-29613	81	2	design	design	NOUN
ajst-29613	81	3	not	not	PART
ajst-29613	81	4	only	only	ADV
ajst-29613	81	5	enhances	enhance	VERB
ajst-29613	81	6	the	the	DET
ajst-29613	81	7	contextual	contextual	ADJ
ajst-29613	81	8	coherence	coherence	NOUN
ajst-29613	81	9	of	of	ADP
ajst-29613	81	10	the	the	DET
ajst-29613	81	11	generated	generate	VERB
ajst-29613	81	12	content	content	NOUN
ajst-29613	81	13	but	but	CCONJ
ajst-29613	81	14	also	also	ADV
ajst-29613	81	15	enables	enable	VERB
ajst-29613	81	16	the	the	DET
ajst-29613	81	17	system	system	NOUN
ajst-29613	81	18	to	to	PART
ajst-29613	81	19	handle	handle	VERB
ajst-29613	81	20	complex	complex	ADJ
ajst-29613	81	21	and	and	CCONJ
ajst-29613	81	22	dynamic	dynamic	ADJ
ajst-29613	81	23	interaction	interaction	NOUN
ajst-29613	81	24	demands	demand	NOUN
ajst-29613	81	25	.	.	PUNCT
ajst-29613	82	1	in	in	ADP
ajst-29613	82	2	naive	naive	ADJ
ajst-29613	82	3	rag	rag	NOUN
ajst-29613	82	4	,	,	PUNCT
ajst-29613	82	5	retrieval	retrieval	NOUN
ajst-29613	82	6	is	be	AUX
ajst-29613	82	7	achieved	achieve	VERB
ajst-29613	82	8	by	by	ADP
ajst-29613	82	9	calculating	calculate	VERB
ajst-29613	82	10	the	the	DET
ajst-29613	82	11	similarity	similarity	NOUN
ajst-29613	82	12	(	(	PUNCT
ajst-29613	82	13	e.g.	e.g.	ADV
ajst-29613	82	14	,	,	PUNCT
ajst-29613	82	15	cosine	cosine	NOUN
ajst-29613	82	16	similarity	similarity	NOUN
ajst-29613	82	17	)	)	PUNCT
ajst-29613	82	18	between	between	ADP
ajst-29613	82	19	the	the	DET
ajst-29613	82	20	query	query	NOUN
ajst-29613	82	21	and	and	CCONJ
ajst-29613	82	22	document	document	NOUN
ajst-29613	82	23	chunk	chunk	NOUN
ajst-29613	82	24	embeddings[15	embeddings[15	ADV
ajst-29613	82	25	]	]	X
ajst-29613	82	26	,	,	PUNCT
ajst-29613	82	27	where	where	SCONJ
ajst-29613	82	28	the	the	DET
ajst-29613	82	29	semantic	semantic	ADJ
ajst-29613	82	30	representation	representation	NOUN
ajst-29613	82	31	capability	capability	NOUN
ajst-29613	82	32	of	of	ADP
ajst-29613	82	33	the	the	DET
ajst-29613	82	34	embedding	embed	VERB
ajst-29613	82	35	model	model	NOUN
ajst-29613	82	36	plays	play	VERB
ajst-29613	82	37	a	a	DET
ajst-29613	82	38	crucial	crucial	ADJ
ajst-29613	82	39	role	role	NOUN
ajst-29613	82	40	.	.	PUNCT
ajst-29613	83	1	this	this	PRON
ajst-29613	83	2	primarily	primarily	ADV
ajst-29613	83	3	includes	include	VERB
ajst-29613	83	4	sparse	sparse	ADJ
ajst-29613	83	5	encoders	encoder	NOUN
ajst-29613	83	6	(	(	PUNCT
ajst-29613	83	7	e.g.	e.g.	ADV
ajst-29613	83	8	,	,	PUNCT
ajst-29613	83	9	bm25)[16	bm25)[16	NOUN
ajst-29613	83	10	]	]	PUNCT
ajst-29613	83	11	and	and	CCONJ
ajst-29613	83	12	dense	dense	ADJ
ajst-29613	83	13	retrievers	retriever	NOUN
ajst-29613	83	14	(	(	PUNCT
ajst-29613	83	15	e.g.	e.g.	ADV
ajst-29613	83	16	,	,	PUNCT
ajst-29613	83	17	pre	pre	ADJ
ajst-29613	83	18	-	-	ADJ
ajst-29613	83	19	trained	train	VERB
ajst-29613	83	20	language	language	NOUN
ajst-29613	83	21	models	model	NOUN
ajst-29613	83	22	based	base	VERB
ajst-29613	83	23	on	on	ADP
ajst-29613	83	24	the	the	DET
ajst-29613	83	25	bert	bert	NOUN
ajst-29613	83	26	architecture	architecture	NOUN
ajst-29613	83	27	)	)	PUNCT
ajst-29613	83	28	.	.	PUNCT
ajst-29613	84	1	the	the	DET
ajst-29613	84	2	structure	structure	NOUN
ajst-29613	84	3	of	of	ADP
ajst-29613	84	4	the	the	DET
ajst-29613	84	5	229	229	NUM
ajst-29613	84	6	naive	naive	ADJ
ajst-29613	84	7	rag	rag	NOUN
ajst-29613	84	8	framework	framework	NOUN
ajst-29613	84	9	is	be	AUX
ajst-29613	84	10	illustrated	illustrate	VERB
ajst-29613	84	11	in	in	ADP
ajst-29613	84	12	figure	figure	NOUN
ajst-29613	84	13	1	1	NUM
ajst-29613	84	14	.	.	PUNCT
ajst-29613	84	15	figure	figure	NOUN
ajst-29613	84	16	1	1	NUM
ajst-29613	84	17	.	.	PUNCT
ajst-29613	84	18	naive	naive	ADJ
ajst-29613	84	19	rag	rag	NOUN
ajst-29613	84	20	framework	framework	NOUN
ajst-29613	84	21	3	3	NUM
ajst-29613	84	22	.	.	PUNCT
ajst-29613	85	1	kgs	kgs	NOUN
ajst-29613	85	2	-	-	PUNCT
ajst-29613	85	3	rag	rag	NOUN
ajst-29613	85	4	technical	technical	ADJ
ajst-29613	85	5	approach	approach	NOUN
ajst-29613	85	6	the	the	DET
ajst-29613	85	7	integration	integration	NOUN
ajst-29613	85	8	of	of	ADP
ajst-29613	85	9	kgs	kg	NOUN
ajst-29613	85	10	and	and	CCONJ
ajst-29613	85	11	rag	rag	NOUN
ajst-29613	85	12	represents	represent	VERB
ajst-29613	85	13	a	a	DET
ajst-29613	85	14	significant	significant	ADJ
ajst-29613	85	15	convergence	convergence	NOUN
ajst-29613	85	16	in	in	ADP
ajst-29613	85	17	modern	modern	ADJ
ajst-29613	85	18	artificial	artificial	ADJ
ajst-29613	85	19	intelligence	intelligence	NOUN
ajst-29613	85	20	technologies	technology	NOUN
ajst-29613	85	21	.	.	PUNCT
ajst-29613	86	1	kgs	kg	NOUN
ajst-29613	86	2	,	,	PUNCT
ajst-29613	86	3	with	with	ADP
ajst-29613	86	4	their	their	PRON
ajst-29613	86	5	structured	structured	ADJ
ajst-29613	86	6	knowledge	knowledge	NOUN
ajst-29613	86	7	representation	representation	NOUN
ajst-29613	86	8	and	and	CCONJ
ajst-29613	86	9	semantic	semantic	ADJ
ajst-29613	86	10	reasoning	reasoning	NOUN
ajst-29613	86	11	capabilities	capability	NOUN
ajst-29613	86	12	,	,	PUNCT
ajst-29613	86	13	address	address	VERB
ajst-29613	86	14	the	the	DET
ajst-29613	86	15	potential	potential	ADJ
ajst-29613	86	16	lack	lack	NOUN
ajst-29613	86	17	of	of	ADP
ajst-29613	86	18	relevance	relevance	NOUN
ajst-29613	86	19	in	in	ADP
ajst-29613	86	20	rag	rag	NOUN
ajst-29613	86	21	's	's	PART
ajst-29613	86	22	retrieval	retrieval	NOUN
ajst-29613	86	23	results	result	NOUN
ajst-29613	86	24	.	.	PUNCT
ajst-29613	87	1	meanwhile	meanwhile	ADV
ajst-29613	87	2	,	,	PUNCT
ajst-29613	87	3	rag	rag	NOUN
ajst-29613	87	4	's	's	PART
ajst-29613	87	5	dynamic	dynamic	ADJ
ajst-29613	87	6	retrieval	retrieval	NOUN
ajst-29613	87	7	and	and	CCONJ
ajst-29613	87	8	generation	generation	NOUN
ajst-29613	87	9	capabilities	capability	NOUN
ajst-29613	87	10	effectively	effectively	ADV
ajst-29613	87	11	broaden	broaden	VERB
ajst-29613	87	12	the	the	DET
ajst-29613	87	13	application	application	NOUN
ajst-29613	87	14	scenarios	scenario	NOUN
ajst-29613	87	15	of	of	ADP
ajst-29613	87	16	kgs	kgs	PROPN
ajst-29613	87	17	.	.	PUNCT
ajst-29613	88	1	this	this	DET
ajst-29613	88	2	combination	combination	NOUN
ajst-29613	88	3	provides	provide	VERB
ajst-29613	88	4	robust	robust	ADJ
ajst-29613	88	5	technical	technical	ADJ
ajst-29613	88	6	support	support	NOUN
ajst-29613	88	7	for	for	ADP
ajst-29613	88	8	fields	field	NOUN
ajst-29613	88	9	such	such	ADJ
ajst-29613	88	10	as	as	ADP
ajst-29613	88	11	opendomain	opendomain	NOUN
ajst-29613	88	12	question	question	NOUN
ajst-29613	88	13	answering	answering	NOUN
ajst-29613	88	14	,	,	PUNCT
ajst-29613	88	15	dialogue	dialogue	NOUN
ajst-29613	88	16	systems	system	NOUN
ajst-29613	88	17	,	,	PUNCT
ajst-29613	88	18	and	and	CCONJ
ajst-29613	88	19	recommendation	recommendation	NOUN
ajst-29613	88	20	systems	system	NOUN
ajst-29613	88	21	.	.	PUNCT
ajst-29613	89	1	the	the	DET
ajst-29613	89	2	overall	overall	ADJ
ajst-29613	89	3	framework	framework	NOUN
ajst-29613	89	4	of	of	ADP
ajst-29613	89	5	kgs	kgs	NOUN
ajst-29613	89	6	-	-	PUNCT
ajst-29613	89	7	rag	rag	NOUN
ajst-29613	89	8	is	be	AUX
ajst-29613	89	9	illustrated	illustrate	VERB
ajst-29613	89	10	in	in	ADP
ajst-29613	89	11	figure	figure	NOUN
ajst-29613	89	12	2	2	NUM
ajst-29613	89	13	.	.	PUNCT
ajst-29613	90	1	the	the	DET
ajst-29613	90	2	entire	entire	ADJ
ajst-29613	90	3	process	process	NOUN
ajst-29613	90	4	can	can	AUX
ajst-29613	90	5	be	be	AUX
ajst-29613	90	6	divided	divide	VERB
ajst-29613	90	7	into	into	ADP
ajst-29613	90	8	three	three	NUM
ajst-29613	90	9	key	key	ADJ
ajst-29613	90	10	steps	step	NOUN
ajst-29613	90	11	:	:	PUNCT
ajst-29613	90	12	index	index	NOUN
ajst-29613	90	13	construction	construction	NOUN
ajst-29613	90	14	,	,	PUNCT
ajst-29613	90	15	subgraph	subgraph	NOUN
ajst-29613	90	16	retrieval	retrieval	NOUN
ajst-29613	90	17	,	,	PUNCT
ajst-29613	90	18	and	and	CCONJ
ajst-29613	90	19	generation	generation	NOUN
ajst-29613	90	20	phase	phase	NOUN
ajst-29613	90	21	.	.	PUNCT
ajst-29613	91	1	in	in	ADP
ajst-29613	91	2	the	the	DET
ajst-29613	91	3	index	index	NOUN
ajst-29613	91	4	construction	construction	NOUN
ajst-29613	91	5	step	step	NOUN
ajst-29613	91	6	,	,	PUNCT
ajst-29613	91	7	nlp	nlp	NOUN
ajst-29613	91	8	techniques	technique	NOUN
ajst-29613	91	9	are	be	AUX
ajst-29613	91	10	used	use	VERB
ajst-29613	91	11	to	to	PART
ajst-29613	91	12	extract	extract	VERB
ajst-29613	91	13	entities	entity	NOUN
ajst-29613	91	14	,	,	PUNCT
ajst-29613	91	15	relationships	relationship	NOUN
ajst-29613	91	16	,	,	PUNCT
ajst-29613	91	17	and	and	CCONJ
ajst-29613	91	18	attributes	attribute	NOUN
ajst-29613	91	19	from	from	ADP
ajst-29613	91	20	raw	raw	ADJ
ajst-29613	91	21	data	datum	NOUN
ajst-29613	91	22	,	,	PUNCT
ajst-29613	91	23	while	while	SCONJ
ajst-29613	91	24	noise	noise	NOUN
ajst-29613	91	25	is	be	AUX
ajst-29613	91	26	removed	remove	VERB
ajst-29613	91	27	to	to	PART
ajst-29613	91	28	ensure	ensure	VERB
ajst-29613	91	29	the	the	DET
ajst-29613	91	30	semantic	semantic	ADJ
ajst-29613	91	31	consistency	consistency	NOUN
ajst-29613	91	32	of	of	ADP
ajst-29613	91	33	the	the	DET
ajst-29613	91	34	knowledge	knowledge	NOUN
ajst-29613	91	35	graph	graph	NOUN
ajst-29613	91	36	.	.	PUNCT
ajst-29613	92	1	for	for	ADP
ajst-29613	92	2	example	example	NOUN
ajst-29613	92	3	,	,	PUNCT
ajst-29613	92	4	the	the	DET
ajst-29613	92	5	sentence	sentence	NOUN
ajst-29613	92	6	“	"	PUNCT
ajst-29613	92	7	bill	bill	NOUN
ajst-29613	92	8	gates	gate	NOUN
ajst-29613	92	9	founded	found	VERB
ajst-29613	92	10	microsoft	microsoft	PROPN
ajst-29613	92	11	”	"	PUNCT
ajst-29613	92	12	can	can	AUX
ajst-29613	92	13	be	be	AUX
ajst-29613	92	14	converted	convert	VERB
ajst-29613	92	15	into	into	ADP
ajst-29613	92	16	the	the	DET
ajst-29613	92	17	triple	triple	ADJ
ajst-29613	92	18	<	<	NOUN
ajst-29613	92	19	bill	bill	NOUN
ajst-29613	92	20	gates	gate	NOUN
ajst-29613	92	21	,	,	PUNCT
ajst-29613	92	22	founded	found	VERB
ajst-29613	92	23	,	,	PUNCT
ajst-29613	92	24	microsoft	microsoft	PROPN
ajst-29613	92	25	>	>	PROPN
ajst-29613	92	26	.	.	PUNCT
ajst-29613	93	1	the	the	DET
ajst-29613	93	2	construction	construction	NOUN
ajst-29613	93	3	and	and	CCONJ
ajst-29613	93	4	indexing	indexing	NOUN
ajst-29613	93	5	of	of	ADP
ajst-29613	93	6	the	the	DET
ajst-29613	93	7	graph	graph	NOUN
ajst-29613	93	8	database	database	NOUN
ajst-29613	93	9	form	form	VERB
ajst-29613	93	10	the	the	DET
ajst-29613	93	11	foundation	foundation	NOUN
ajst-29613	93	12	of	of	ADP
ajst-29613	93	13	kgs	kgs	PROPN
ajst-29613	93	14	-	-	PUNCT
ajst-29613	93	15	rag	rag	NOUN
ajst-29613	93	16	,	,	PUNCT
ajst-29613	93	17	with	with	ADP
ajst-29613	93	18	the	the	DET
ajst-29613	93	19	quality	quality	NOUN
ajst-29613	93	20	of	of	ADP
ajst-29613	93	21	the	the	DET
ajst-29613	93	22	graph	graph	NOUN
ajst-29613	93	23	database	database	NOUN
ajst-29613	93	24	directly	directly	ADV
ajst-29613	93	25	influencing	influence	VERB
ajst-29613	93	26	its	its	PRON
ajst-29613	93	27	performance	performance	NOUN
ajst-29613	93	28	.	.	PUNCT
ajst-29613	94	1	in	in	ADP
ajst-29613	94	2	the	the	DET
ajst-29613	94	3	subgraph	subgraph	NOUN
ajst-29613	94	4	retrieval	retrieval	NOUN
ajst-29613	94	5	step	step	NOUN
ajst-29613	94	6	,	,	PUNCT
ajst-29613	94	7	the	the	DET
ajst-29613	94	8	system	system	NOUN
ajst-29613	94	9	parses	parse	VERB
ajst-29613	94	10	the	the	DET
ajst-29613	94	11	user	user	NOUN
ajst-29613	94	12	query	query	NOUN
ajst-29613	94	13	to	to	PART
ajst-29613	94	14	identify	identify	VERB
ajst-29613	94	15	core	core	ADJ
ajst-29613	94	16	entities	entity	NOUN
ajst-29613	94	17	or	or	CCONJ
ajst-29613	94	18	keywords	keyword	NOUN
ajst-29613	94	19	and	and	CCONJ
ajst-29613	94	20	rapidly	rapidly	ADV
ajst-29613	94	21	locates	locate	VERB
ajst-29613	94	22	the	the	DET
ajst-29613	94	23	subgraph	subgraph	NOUN
ajst-29613	94	24	relevant	relevant	ADJ
ajst-29613	94	25	to	to	ADP
ajst-29613	94	26	the	the	DET
ajst-29613	94	27	query	query	NOUN
ajst-29613	94	28	from	from	ADP
ajst-29613	94	29	the	the	DET
ajst-29613	94	30	large	large	ADJ
ajst-29613	94	31	-	-	PUNCT
ajst-29613	94	32	scale	scale	NOUN
ajst-29613	94	33	knowledge	knowledge	NOUN
ajst-29613	94	34	graph	graph	NOUN
ajst-29613	94	35	.	.	PUNCT
ajst-29613	95	1	finally	finally	ADV
ajst-29613	95	2	,	,	PUNCT
ajst-29613	95	3	in	in	ADP
ajst-29613	95	4	the	the	DET
ajst-29613	95	5	generation	generation	NOUN
ajst-29613	95	6	phase	phase	NOUN
ajst-29613	95	7	,	,	PUNCT
ajst-29613	95	8	the	the	DET
ajst-29613	95	9	retrieved	retrieve	VERB
ajst-29613	95	10	subgraph	subgraph	NOUN
ajst-29613	95	11	serves	serve	VERB
ajst-29613	95	12	as	as	ADP
ajst-29613	95	13	contextual	contextual	ADJ
ajst-29613	95	14	information	information	NOUN
ajst-29613	95	15	,	,	PUNCT
ajst-29613	95	16	which	which	PRON
ajst-29613	95	17	,	,	PUNCT
ajst-29613	95	18	combined	combine	VERB
ajst-29613	95	19	with	with	ADP
ajst-29613	95	20	the	the	DET
ajst-29613	95	21	user	user	NOUN
ajst-29613	95	22	query	query	NOUN
ajst-29613	95	23	,	,	PUNCT
ajst-29613	95	24	is	be	AUX
ajst-29613	95	25	fed	feed	VERB
ajst-29613	95	26	into	into	ADP
ajst-29613	95	27	the	the	DET
ajst-29613	95	28	generative	generative	ADJ
ajst-29613	95	29	model	model	NOUN
ajst-29613	95	30	to	to	PART
ajst-29613	95	31	produce	produce	VERB
ajst-29613	95	32	a	a	DET
ajst-29613	95	33	natural	natural	ADJ
ajst-29613	95	34	language	language	NOUN
ajst-29613	95	35	response	response	NOUN
ajst-29613	95	36	that	that	PRON
ajst-29613	95	37	aligns	align	VERB
ajst-29613	95	38	with	with	ADP
ajst-29613	95	39	the	the	DET
ajst-29613	95	40	semantic	semantic	ADJ
ajst-29613	95	41	requirements	requirement	NOUN
ajst-29613	95	42	.	.	PUNCT
ajst-29613	96	1	figure	figure	NOUN
ajst-29613	96	2	2	2	NUM
ajst-29613	96	3	.	.	PUNCT
ajst-29613	96	4	framework	framework	NOUN
ajst-29613	96	5	of	of	ADP
ajst-29613	96	6	kgs	kgs	NOUN
ajst-29613	96	7	-	-	PUNCT
ajst-29613	96	8	rag	rag	NOUN
ajst-29613	96	9	3.1	3.1	NUM
ajst-29613	96	10	.	.	PUNCT
ajst-29613	97	1	index	index	NOUN
ajst-29613	97	2	construction	construction	NOUN
ajst-29613	97	3	knowledge	knowledge	NOUN
ajst-29613	97	4	graphs	graph	NOUN
ajst-29613	97	5	can	can	AUX
ajst-29613	97	6	be	be	AUX
ajst-29613	97	7	classified	classify	VERB
ajst-29613	97	8	by	by	ADP
ajst-29613	97	9	data	datum	NOUN
ajst-29613	97	10	source	source	NOUN
ajst-29613	97	11	into	into	ADP
ajst-29613	97	12	open	open	ADJ
ajst-29613	97	13	-	-	PUNCT
ajst-29613	97	14	source	source	NOUN
ajst-29613	97	15	knowledge	knowledge	NOUN
ajst-29613	97	16	graphs	graph	NOUN
ajst-29613	97	17	and	and	CCONJ
ajst-29613	97	18	private	private	ADJ
ajst-29613	97	19	knowledge	knowledge	NOUN
ajst-29613	97	20	graphs	graph	NOUN
ajst-29613	97	21	.	.	PUNCT
ajst-29613	98	1	based	base	VERB
ajst-29613	98	2	on	on	ADP
ajst-29613	98	3	domain	domain	NOUN
ajst-29613	98	4	knowledge	knowledge	NOUN
ajst-29613	98	5	,	,	PUNCT
ajst-29613	98	6	they	they	PRON
ajst-29613	98	7	can	can	AUX
ajst-29613	98	8	be	be	AUX
ajst-29613	98	9	further	far	ADV
ajst-29613	98	10	divided	divide	VERB
ajst-29613	98	11	into	into	ADP
ajst-29613	98	12	general	general	ADJ
ajst-29613	98	13	knowledge	knowledge	NOUN
ajst-29613	98	14	graphs	graph	NOUN
ajst-29613	98	15	and	and	CCONJ
ajst-29613	98	16	domain	domain	NOUN
ajst-29613	98	17	-	-	PUNCT
ajst-29613	98	18	specific	specific	ADJ
ajst-29613	98	19	knowledge	knowledge	NOUN
ajst-29613	98	20	graphs	graph	NOUN
ajst-29613	98	21	.	.	PUNCT
ajst-29613	99	1	common	common	ADJ
ajst-29613	99	2	examples	example	NOUN
ajst-29613	99	3	of	of	ADP
ajst-29613	99	4	general	general	ADJ
ajst-29613	99	5	and	and	CCONJ
ajst-29613	99	6	domain	domain	NOUN
ajst-29613	99	7	-	-	PUNCT
ajst-29613	99	8	specific	specific	ADJ
ajst-29613	99	9	knowledge	knowledge	NOUN
ajst-29613	99	10	graphs	graph	NOUN
ajst-29613	99	11	are	be	AUX
ajst-29613	99	12	shown	show	VERB
ajst-29613	99	13	in	in	ADP
ajst-29613	99	14	figure	figure	NOUN
ajst-29613	99	15	3	3	NUM
ajst-29613	99	16	.	.	NOUN
ajst-29613	99	17	230	230	NUM
ajst-29613	99	18	figure	figure	NOUN
ajst-29613	99	19	3	3	NUM
ajst-29613	99	20	.	.	PUNCT
ajst-29613	100	1	classification	classification	NOUN
ajst-29613	100	2	of	of	ADP
ajst-29613	100	3	knowledge	knowledge	NOUN
ajst-29613	100	4	graphs	graph	NOUN
ajst-29613	100	5	indexing	indexing	NOUN
ajst-29613	100	6	in	in	ADP
ajst-29613	100	7	the	the	DET
ajst-29613	100	8	construction	construction	NOUN
ajst-29613	100	9	of	of	ADP
ajst-29613	100	10	kgs	kgs	PROPN
ajst-29613	100	11	is	be	AUX
ajst-29613	100	12	a	a	DET
ajst-29613	100	13	critical	critical	ADJ
ajst-29613	100	14	step	step	NOUN
ajst-29613	100	15	to	to	PART
ajst-29613	100	16	ensure	ensure	VERB
ajst-29613	100	17	efficient	efficient	ADJ
ajst-29613	100	18	data	datum	NOUN
ajst-29613	100	19	storage	storage	NOUN
ajst-29613	100	20	and	and	CCONJ
ajst-29613	100	21	retrieval	retrieval	NOUN
ajst-29613	100	22	,	,	PUNCT
ajst-29613	100	23	enabling	enable	VERB
ajst-29613	100	24	rapid	rapid	ADJ
ajst-29613	100	25	access	access	NOUN
ajst-29613	100	26	to	to	ADP
ajst-29613	100	27	entities	entity	NOUN
ajst-29613	100	28	,	,	PUNCT
ajst-29613	100	29	relationships	relationship	NOUN
ajst-29613	100	30	,	,	PUNCT
ajst-29613	100	31	and	and	CCONJ
ajst-29613	100	32	attributes	attribute	VERB
ajst-29613	100	33	.	.	PUNCT
ajst-29613	101	1	by	by	ADP
ajst-29613	101	2	establishing	establish	VERB
ajst-29613	101	3	a	a	DET
ajst-29613	101	4	fast	fast	ADJ
ajst-29613	101	5	localization	localization	NOUN
ajst-29613	101	6	mechanism	mechanism	NOUN
ajst-29613	101	7	,	,	PUNCT
ajst-29613	101	8	indexing	indexing	NOUN
ajst-29613	101	9	enhances	enhance	VERB
ajst-29613	101	10	retrieval	retrieval	NOUN
ajst-29613	101	11	accuracy	accuracy	NOUN
ajst-29613	101	12	and	and	CCONJ
ajst-29613	101	13	response	response	NOUN
ajst-29613	101	14	speed	speed	NOUN
ajst-29613	101	15	.	.	PUNCT
ajst-29613	102	1	common	common	ADJ
ajst-29613	102	2	indexing	indexing	NOUN
ajst-29613	102	3	structures	structure	NOUN
ajst-29613	102	4	include	include	VERB
ajst-29613	102	5	inverted	inverted	ADJ
ajst-29613	102	6	indexes	index	NOUN
ajst-29613	102	7	,	,	PUNCT
ajst-29613	102	8	hash	hash	NOUN
ajst-29613	102	9	indexes	index	NOUN
ajst-29613	102	10	,	,	PUNCT
ajst-29613	102	11	and	and	CCONJ
ajst-29613	102	12	graph	graph	NOUN
ajst-29613	102	13	database	database	ADJ
ajst-29613	102	14	-	-	PUNCT
ajst-29613	102	15	specific	specific	ADJ
ajst-29613	102	16	indexes	index	NOUN
ajst-29613	102	17	,	,	PUNCT
ajst-29613	102	18	which	which	PRON
ajst-29613	102	19	support	support	VERB
ajst-29613	102	20	semantic	semantic	ADJ
ajst-29613	102	21	queries	query	NOUN
ajst-29613	102	22	,	,	PUNCT
ajst-29613	102	23	path	path	NOUN
ajst-29613	102	24	searches	search	NOUN
ajst-29613	102	25	,	,	PUNCT
ajst-29613	102	26	and	and	CCONJ
ajst-29613	102	27	dynamic	dynamic	ADJ
ajst-29613	102	28	updates[17,18	updates[17,18	NOUN
ajst-29613	102	29	]	]	X
ajst-29613	102	30	.	.	PUNCT
ajst-29613	103	1	in	in	ADP
ajst-29613	103	2	recent	recent	ADJ
ajst-29613	103	3	years	year	NOUN
ajst-29613	103	4	,	,	PUNCT
ajst-29613	103	5	the	the	DET
ajst-29613	103	6	integration	integration	NOUN
ajst-29613	103	7	of	of	ADP
ajst-29613	103	8	deep	deep	ADJ
ajst-29613	103	9	learning	learning	NOUN
ajst-29613	103	10	techniques	technique	NOUN
ajst-29613	103	11	,	,	PUNCT
ajst-29613	103	12	such	such	ADJ
ajst-29613	103	13	as	as	ADP
ajst-29613	103	14	entity	entity	NOUN
ajst-29613	103	15	embeddings	embedding	NOUN
ajst-29613	103	16	and	and	CCONJ
ajst-29613	103	17	graph	graph	VERB
ajst-29613	103	18	neural	neural	ADJ
ajst-29613	103	19	networks	network	NOUN
ajst-29613	103	20	(	(	PUNCT
ajst-29613	103	21	gnns	gnns	NOUN
ajst-29613	103	22	)	)	PUNCT
ajst-29613	103	23	,	,	PUNCT
ajst-29613	103	24	has	have	AUX
ajst-29613	103	25	further	far	ADV
ajst-29613	103	26	optimized	optimize	VERB
ajst-29613	103	27	indexing	indexing	NOUN
ajst-29613	103	28	performance[19,20	performance[19,20	NOUN
ajst-29613	103	29	]	]	PUNCT
ajst-29613	103	30	.	.	PUNCT
ajst-29613	104	1	3.2	3.2	NUM
ajst-29613	104	2	.	.	PUNCT
ajst-29613	105	1	subgraph	subgraph	PROPN
ajst-29613	105	2	retrieval	retrieval	PROPN
ajst-29613	105	3	subgraph	subgraph	NOUN
ajst-29613	105	4	retrieval	retrieval	NOUN
ajst-29613	105	5	is	be	AUX
ajst-29613	105	6	a	a	DET
ajst-29613	105	7	key	key	ADJ
ajst-29613	105	8	step	step	NOUN
ajst-29613	105	9	in	in	ADP
ajst-29613	105	10	the	the	DET
ajst-29613	105	11	kgs	kgs	NOUN
ajst-29613	105	12	-	-	PUNCT
ajst-29613	105	13	rag	rag	NOUN
ajst-29613	105	14	framework	framework	NOUN
ajst-29613	105	15	,	,	PUNCT
ajst-29613	105	16	responsible	responsible	ADJ
ajst-29613	105	17	for	for	ADP
ajst-29613	105	18	efficiently	efficiently	ADV
ajst-29613	105	19	extracting	extract	VERB
ajst-29613	105	20	knowledge	knowledge	NOUN
ajst-29613	105	21	subgraphs	subgraph	NOUN
ajst-29613	105	22	relevant	relevant	ADJ
ajst-29613	105	23	to	to	ADP
ajst-29613	105	24	the	the	DET
ajst-29613	105	25	input	input	NOUN
ajst-29613	105	26	query	query	NOUN
ajst-29613	105	27	from	from	ADP
ajst-29613	105	28	large	large	ADJ
ajst-29613	105	29	-	-	PUNCT
ajst-29613	105	30	scale	scale	NOUN
ajst-29613	105	31	knowledge	knowledge	NOUN
ajst-29613	105	32	graphs	graph	NOUN
ajst-29613	105	33	,	,	PUNCT
ajst-29613	105	34	providing	provide	VERB
ajst-29613	105	35	a	a	DET
ajst-29613	105	36	structured	structured	ADJ
ajst-29613	105	37	knowledge	knowledge	NOUN
ajst-29613	105	38	foundation	foundation	NOUN
ajst-29613	105	39	for	for	ADP
ajst-29613	105	40	subsequent	subsequent	ADJ
ajst-29613	105	41	generation	generation	NOUN
ajst-29613	105	42	tasks	task	NOUN
ajst-29613	105	43	.	.	PUNCT
ajst-29613	106	1	(	(	PUNCT
ajst-29613	106	2	a	a	X
ajst-29613	106	3	)	)	PUNCT
ajst-29613	106	4	path	path	NOUN
ajst-29613	106	5	-	-	PUNCT
ajst-29613	106	6	based	base	VERB
ajst-29613	106	7	expansion	expansion	NOUN
ajst-29613	106	8	retrieval	retrieval	NOUN
ajst-29613	106	9	:	:	PUNCT
ajst-29613	106	10	this	this	DET
ajst-29613	106	11	approach	approach	NOUN
ajst-29613	106	12	constructs	construct	VERB
ajst-29613	106	13	subgraphs	subgraph	NOUN
ajst-29613	106	14	by	by	ADP
ajst-29613	106	15	expanding	expand	VERB
ajst-29613	106	16	from	from	ADP
ajst-29613	106	17	the	the	DET
ajst-29613	106	18	initial	initial	ADJ
ajst-29613	106	19	entity	entity	NOUN
ajst-29613	106	20	within	within	ADP
ajst-29613	106	21	a	a	DET
ajst-29613	106	22	fixed	fix	VERB
ajst-29613	106	23	number	number	NOUN
ajst-29613	106	24	of	of	ADP
ajst-29613	106	25	hops	hop	NOUN
ajst-29613	106	26	,	,	PUNCT
ajst-29613	106	27	including	include	VERB
ajst-29613	106	28	related	relate	VERB
ajst-29613	106	29	nodes	node	NOUN
ajst-29613	106	30	and	and	CCONJ
ajst-29613	106	31	relationships	relationship	NOUN
ajst-29613	106	32	.	.	PUNCT
ajst-29613	107	1	common	common	ADJ
ajst-29613	107	2	expansion	expansion	NOUN
ajst-29613	107	3	strategies	strategy	NOUN
ajst-29613	107	4	include	include	VERB
ajst-29613	107	5	:	:	PUNCT
ajst-29613	107	6	fixed	fix	VERB
ajst-29613	107	7	-	-	PUNCT
ajst-29613	107	8	hop	hop	NOUN
ajst-29613	107	9	expansion	expansion	NOUN
ajst-29613	107	10	:	:	PUNCT
ajst-29613	107	11	restricts	restrict	VERB
ajst-29613	107	12	the	the	DET
ajst-29613	107	13	expansion	expansion	NOUN
ajst-29613	107	14	to	to	ADP
ajst-29613	107	15	a	a	DET
ajst-29613	107	16	specific	specific	ADJ
ajst-29613	107	17	number	number	NOUN
ajst-29613	107	18	of	of	ADP
ajst-29613	107	19	hops	hop	NOUN
ajst-29613	107	20	,	,	PUNCT
ajst-29613	107	21	such	such	ADJ
ajst-29613	107	22	as	as	ADP
ajst-29613	107	23	retrieving	retrieve	VERB
ajst-29613	107	24	all	all	DET
ajst-29613	107	25	nodes	node	NOUN
ajst-29613	107	26	and	and	CCONJ
ajst-29613	107	27	relationships	relationship	NOUN
ajst-29613	107	28	within	within	ADP
ajst-29613	107	29	one	one	NUM
ajst-29613	107	30	to	to	ADP
ajst-29613	107	31	two	two	NUM
ajst-29613	107	32	hops	hop	NOUN
ajst-29613	107	33	from	from	ADP
ajst-29613	107	34	the	the	DET
ajst-29613	107	35	initial	initial	ADJ
ajst-29613	107	36	entity	entity	NOUN
ajst-29613	107	37	.	.	PUNCT
ajst-29613	108	1	semantic	semantic	ADJ
ajst-29613	108	2	-	-	PUNCT
ajst-29613	108	3	relevant	relevant	ADJ
ajst-29613	108	4	expansion	expansion	NOUN
ajst-29613	108	5	:	:	PUNCT
ajst-29613	108	6	calculates	calculate	VERB
ajst-29613	108	7	semantic	semantic	ADJ
ajst-29613	108	8	similarity	similarity	NOUN
ajst-29613	108	9	between	between	ADP
ajst-29613	108	10	nodes	node	NOUN
ajst-29613	108	11	and	and	CCONJ
ajst-29613	108	12	retains	retain	VERB
ajst-29613	108	13	only	only	ADV
ajst-29613	108	14	those	those	DET
ajst-29613	108	15	paths	path	NOUN
ajst-29613	108	16	that	that	PRON
ajst-29613	108	17	are	be	AUX
ajst-29613	108	18	semantically	semantically	ADV
ajst-29613	108	19	relevant	relevant	ADJ
ajst-29613	108	20	to	to	ADP
ajst-29613	108	21	the	the	DET
ajst-29613	108	22	query	query	NOUN
ajst-29613	108	23	.	.	PUNCT
ajst-29613	109	1	(	(	PUNCT
ajst-29613	109	2	b	b	X
ajst-29613	109	3	)	)	PUNCT
ajst-29613	109	4	query	query	NOUN
ajst-29613	109	5	-	-	PUNCT
ajst-29613	109	6	based	base	VERB
ajst-29613	109	7	subgraph	subgraph	NOUN
ajst-29613	109	8	pattern	pattern	NOUN
ajst-29613	109	9	matching	match	VERB
ajst-29613	109	10	:	:	PUNCT
ajst-29613	109	11	this	this	DET
ajst-29613	109	12	method	method	NOUN
ajst-29613	109	13	maps	map	VERB
ajst-29613	109	14	the	the	DET
ajst-29613	109	15	query	query	NOUN
ajst-29613	109	16	to	to	ADP
ajst-29613	109	17	a	a	DET
ajst-29613	109	18	graph	graph	NOUN
ajst-29613	109	19	pattern	pattern	NOUN
ajst-29613	109	20	and	and	CCONJ
ajst-29613	109	21	retrieves	retrieve	VERB
ajst-29613	109	22	subgraphs	subgraph	NOUN
ajst-29613	109	23	from	from	ADP
ajst-29613	109	24	the	the	DET
ajst-29613	109	25	knowledge	knowledge	NOUN
ajst-29613	109	26	graph	graph	NOUN
ajst-29613	109	27	that	that	PRON
ajst-29613	109	28	match	match	VERB
ajst-29613	109	29	the	the	DET
ajst-29613	109	30	pattern	pattern	NOUN
ajst-29613	109	31	.	.	PUNCT
ajst-29613	110	1	for	for	ADP
ajst-29613	110	2	example	example	NOUN
ajst-29613	110	3	,	,	PUNCT
ajst-29613	110	4	for	for	ADP
ajst-29613	110	5	the	the	DET
ajst-29613	110	6	query	query	NOUN
ajst-29613	110	7	"	"	PUNCT
ajst-29613	110	8	bill	bill	NOUN
ajst-29613	110	9	gates	gate	NOUN
ajst-29613	110	10	and	and	CCONJ
ajst-29613	110	11	microsoft	microsoft	PROPN
ajst-29613	110	12	collaboration	collaboration	PROPN
ajst-29613	110	13	,	,	PUNCT
ajst-29613	110	14	"	"	PUNCT
ajst-29613	110	15	a	a	DET
ajst-29613	110	16	graph	graph	NOUN
ajst-29613	110	17	pattern	pattern	NOUN
ajst-29613	110	18	containing	contain	VERB
ajst-29613	110	19	the	the	DET
ajst-29613	110	20	nodes	node	NOUN
ajst-29613	110	21	"	"	PUNCT
ajst-29613	110	22	bill	bill	NOUN
ajst-29613	110	23	gates	gate	NOUN
ajst-29613	110	24	"	"	PUNCT
ajst-29613	110	25	and	and	CCONJ
ajst-29613	110	26	"	"	PUNCT
ajst-29613	110	27	microsoft	microsoft	PROPN
ajst-29613	110	28	"	"	PUNCT
ajst-29613	110	29	and	and	CCONJ
ajst-29613	110	30	their	their	PRON
ajst-29613	110	31	possible	possible	ADJ
ajst-29613	110	32	relationships	relationship	NOUN
ajst-29613	110	33	is	be	AUX
ajst-29613	110	34	constructed	construct	VERB
ajst-29613	110	35	to	to	PART
ajst-29613	110	36	match	match	VERB
ajst-29613	110	37	relevant	relevant	ADJ
ajst-29613	110	38	subgraphs	subgraph	NOUN
ajst-29613	110	39	.	.	PUNCT
ajst-29613	111	1	(	(	PUNCT
ajst-29613	111	2	c	c	X
ajst-29613	111	3	)	)	PUNCT
ajst-29613	111	4	embedding	embed	VERB
ajst-29613	111	5	-	-	PUNCT
ajst-29613	111	6	based	base	VERB
ajst-29613	111	7	semantic	semantic	ADJ
ajst-29613	111	8	retrieval	retrieval	NOUN
ajst-29613	111	9	:	:	PUNCT
ajst-29613	111	10	this	this	DET
ajst-29613	111	11	approach	approach	NOUN
ajst-29613	111	12	embeds	embed	VERB
ajst-29613	111	13	the	the	DET
ajst-29613	111	14	nodes	node	NOUN
ajst-29613	111	15	and	and	CCONJ
ajst-29613	111	16	edges	edge	NOUN
ajst-29613	111	17	of	of	ADP
ajst-29613	111	18	the	the	DET
ajst-29613	111	19	knowledge	knowledge	NOUN
ajst-29613	111	20	graph	graph	NOUN
ajst-29613	111	21	into	into	ADP
ajst-29613	111	22	a	a	DET
ajst-29613	111	23	high	high	ADJ
ajst-29613	111	24	-	-	PUNCT
ajst-29613	111	25	dimensional	dimensional	ADJ
ajst-29613	111	26	vector	vector	NOUN
ajst-29613	111	27	space	space	NOUN
ajst-29613	111	28	,	,	PUNCT
ajst-29613	111	29	performing	perform	VERB
ajst-29613	111	30	subgraph	subgraph	NOUN
ajst-29613	111	31	retrieval	retrieval	NOUN
ajst-29613	111	32	based	base	VERB
ajst-29613	111	33	on	on	ADP
ajst-29613	111	34	vector	vector	NOUN
ajst-29613	111	35	similarity	similarity	NOUN
ajst-29613	111	36	.	.	PUNCT
ajst-29613	112	1	for	for	ADP
ajst-29613	112	2	instance	instance	NOUN
ajst-29613	112	3	,	,	PUNCT
ajst-29613	112	4	embedding	embed	VERB
ajst-29613	112	5	models	model	NOUN
ajst-29613	112	6	like	like	ADP
ajst-29613	112	7	transe	transe	NOUN
ajst-29613	112	8	and	and	CCONJ
ajst-29613	112	9	rotate	rotate	VERB
ajst-29613	112	10	can	can	AUX
ajst-29613	112	11	compute	compute	VERB
ajst-29613	112	12	the	the	DET
ajst-29613	112	13	similarity	similarity	NOUN
ajst-29613	112	14	between	between	ADP
ajst-29613	112	15	the	the	DET
ajst-29613	112	16	query	query	NOUN
ajst-29613	112	17	entity	entity	NOUN
ajst-29613	112	18	and	and	CCONJ
ajst-29613	112	19	other	other	ADJ
ajst-29613	112	20	entities	entity	NOUN
ajst-29613	112	21	in	in	ADP
ajst-29613	112	22	the	the	DET
ajst-29613	112	23	knowledge	knowledge	NOUN
ajst-29613	112	24	graph	graph	NOUN
ajst-29613	112	25	to	to	PART
ajst-29613	112	26	dynamically	dynamically	ADV
ajst-29613	112	27	retrieve	retrieve	VERB
ajst-29613	112	28	subgraphs	subgraphs	NOUN
ajst-29613	112	29	relevant	relevant	ADJ
ajst-29613	112	30	to	to	ADP
ajst-29613	112	31	the	the	DET
ajst-29613	112	32	query[21	query[21	NOUN
ajst-29613	112	33	]	]	PUNCT
ajst-29613	112	34	.	.	PUNCT
ajst-29613	113	1	subgraph	subgraph	PROPN
ajst-29613	113	2	retrieval	retrieval	PROPN
ajst-29613	113	3	is	be	AUX
ajst-29613	113	4	a	a	DET
ajst-29613	113	5	core	core	NOUN
ajst-29613	113	6	component	component	NOUN
ajst-29613	113	7	of	of	ADP
ajst-29613	113	8	the	the	DET
ajst-29613	113	9	kgs	kgs	NOUN
ajst-29613	113	10	-	-	PUNCT
ajst-29613	113	11	rag	rag	NOUN
ajst-29613	113	12	framework	framework	NOUN
ajst-29613	113	13	,	,	PUNCT
ajst-29613	113	14	responsible	responsible	ADJ
ajst-29613	113	15	for	for	ADP
ajst-29613	113	16	extracting	extract	VERB
ajst-29613	113	17	relevant	relevant	ADJ
ajst-29613	113	18	knowledge	knowledge	NOUN
ajst-29613	113	19	subgraphs	subgraphs	NOUN
ajst-29613	113	20	from	from	ADP
ajst-29613	113	21	large	large	ADJ
ajst-29613	113	22	-	-	PUNCT
ajst-29613	113	23	scale	scale	NOUN
ajst-29613	113	24	knowledge	knowledge	NOUN
ajst-29613	113	25	graphs	graph	NOUN
ajst-29613	113	26	to	to	PART
ajst-29613	113	27	provide	provide	VERB
ajst-29613	113	28	critical	critical	ADJ
ajst-29613	113	29	support	support	NOUN
ajst-29613	113	30	for	for	ADP
ajst-29613	113	31	generative	generative	ADJ
ajst-29613	113	32	models	model	NOUN
ajst-29613	113	33	.	.	PUNCT
ajst-29613	114	1	through	through	ADP
ajst-29613	114	2	methods	method	NOUN
ajst-29613	114	3	such	such	ADJ
ajst-29613	114	4	as	as	ADP
ajst-29613	114	5	entity	entity	NOUN
ajst-29613	114	6	matching	matching	NOUN
ajst-29613	114	7	,	,	PUNCT
ajst-29613	114	8	path	path	NOUN
ajst-29613	114	9	expansion	expansion	NOUN
ajst-29613	114	10	,	,	PUNCT
ajst-29613	114	11	and	and	CCONJ
ajst-29613	114	12	semantic	semantic	ADJ
ajst-29613	114	13	retrieval	retrieval	NOUN
ajst-29613	114	14	,	,	PUNCT
ajst-29613	114	15	subgraph	subgraph	NOUN
ajst-29613	114	16	retrieval	retrieval	NOUN
ajst-29613	114	17	significantly	significantly	ADV
ajst-29613	114	18	enhances	enhance	VERB
ajst-29613	114	19	the	the	DET
ajst-29613	114	20	accuracy	accuracy	NOUN
ajst-29613	114	21	and	and	CCONJ
ajst-29613	114	22	coherence	coherence	NOUN
ajst-29613	114	23	of	of	ADP
ajst-29613	114	24	generated	generate	VERB
ajst-29613	114	25	content	content	NOUN
ajst-29613	114	26	.	.	PUNCT
ajst-29613	115	1	as	as	SCONJ
ajst-29613	115	2	shown	show	VERB
ajst-29613	115	3	in	in	ADP
ajst-29613	115	4	figure	figure	NOUN
ajst-29613	115	5	4	4	NUM
ajst-29613	115	6	,	,	PUNCT
ajst-29613	115	7	various	various	ADJ
ajst-29613	115	8	retrievers	retriever	NOUN
ajst-29613	115	9	in	in	ADP
ajst-29613	115	10	kgs	kgs	NOUN
ajst-29613	115	11	-	-	PUNCT
ajst-29613	115	12	rag	rag	NOUN
ajst-29613	115	13	exhibit	exhibit	NOUN
ajst-29613	115	14	unique	unique	ADJ
ajst-29613	115	15	strengths	strength	NOUN
ajst-29613	115	16	in	in	ADP
ajst-29613	115	17	handling	handle	VERB
ajst-29613	115	18	different	different	ADJ
ajst-29613	115	19	aspects	aspect	NOUN
ajst-29613	115	20	of	of	ADP
ajst-29613	115	21	retrieval	retrieval	NOUN
ajst-29613	115	22	tasks	task	NOUN
ajst-29613	115	23	.	.	PUNCT
ajst-29613	116	1	based	base	VERB
ajst-29613	116	2	on	on	ADP
ajst-29613	116	3	their	their	PRON
ajst-29613	116	4	underlying	underlie	VERB
ajst-29613	116	5	models	model	NOUN
ajst-29613	116	6	,	,	PUNCT
ajst-29613	116	7	retrievers	retriever	NOUN
ajst-29613	116	8	are	be	AUX
ajst-29613	116	9	classified	classify	VERB
ajst-29613	116	10	into	into	ADP
ajst-29613	116	11	three	three	NUM
ajst-29613	116	12	types	type	NOUN
ajst-29613	116	13	:	:	PUNCT
ajst-29613	116	14	traditional	traditional	ADJ
ajst-29613	116	15	graph	graph	NOUN
ajst-29613	116	16	retrieval	retrieval	NOUN
ajst-29613	116	17	algorithm	algorithm	NOUN
ajst-29613	116	18	-	-	PUNCT
ajst-29613	116	19	based	base	VERB
ajst-29613	116	20	retrievers	retriever	NOUN
ajst-29613	116	21	,	,	PUNCT
ajst-29613	116	22	language	language	NOUN
ajst-29613	116	23	modelbased	modelbase	VERB
ajst-29613	116	24	retrievers	retriever	NOUN
ajst-29613	116	25	,	,	PUNCT
ajst-29613	116	26	and	and	CCONJ
ajst-29613	116	27	graph	graph	VERB
ajst-29613	116	28	neural	neural	ADJ
ajst-29613	116	29	network	network	NOUN
ajst-29613	116	30	(	(	PUNCT
ajst-29613	116	31	gnn)-based	gnn)-base	VERB
ajst-29613	116	32	retrievers	retriever	NOUN
ajst-29613	116	33	.	.	PUNCT
ajst-29613	117	1	figure	figure	NOUN
ajst-29613	117	2	4	4	NUM
ajst-29613	117	3	.	.	PUNCT
ajst-29613	118	1	retriever	retriever	NOUN
ajst-29613	118	2	of	of	ADP
ajst-29613	118	3	kgs	kgs	NOUN
ajst-29613	118	4	-	-	PUNCT
ajst-29613	118	5	rag	rag	NOUN
ajst-29613	118	6	231	231	NUM
ajst-29613	118	7	(	(	PUNCT
ajst-29613	118	8	a	a	NOUN
ajst-29613	118	9	)	)	PUNCT
ajst-29613	118	10	methods	method	NOUN
ajst-29613	118	11	based	base	VERB
ajst-29613	118	12	on	on	ADP
ajst-29613	118	13	heuristic	heuristic	ADJ
ajst-29613	118	14	rules	rule	NOUN
ajst-29613	118	15	or	or	CCONJ
ajst-29613	118	16	traditional	traditional	ADJ
ajst-29613	118	17	graph	graph	NOUN
ajst-29613	118	18	search	search	NOUN
ajst-29613	118	19	algorithms	algorithm	NOUN
ajst-29613	118	20	enable	enable	VERB
ajst-29613	118	21	efficient	efficient	ADJ
ajst-29613	118	22	retrieval	retrieval	NOUN
ajst-29613	118	23	without	without	ADP
ajst-29613	118	24	relying	rely	VERB
ajst-29613	118	25	on	on	ADP
ajst-29613	118	26	deep	deep	ADJ
ajst-29613	118	27	learning	learning	NOUN
ajst-29613	118	28	models	model	NOUN
ajst-29613	118	29	.	.	PUNCT
ajst-29613	119	1	for	for	ADP
ajst-29613	119	2	example	example	NOUN
ajst-29613	119	3	,	,	PUNCT
ajst-29613	119	4	qa	qa	NOUN
ajst-29613	119	5	-	-	PUNCT
ajst-29613	119	6	gnn[22	gnn[22	NOUN
ajst-29613	119	7	]	]	PUNCT
ajst-29613	119	8	and	and	CCONJ
ajst-29613	119	9	grapeqa[23	grapeqa[23	PROPN
ajst-29613	119	10	]	]	PUNCT
ajst-29613	119	11	proposed	propose	VERB
ajst-29613	119	12	retrieving	retrieve	VERB
ajst-29613	119	13	k	k	PROPN
ajst-29613	119	14	-	-	PUNCT
ajst-29613	119	15	hop	hop	NOUN
ajst-29613	119	16	paths	path	NOUN
ajst-29613	119	17	from	from	ADP
ajst-29613	119	18	the	the	DET
ajst-29613	119	19	topic	topic	NOUN
ajst-29613	119	20	entity	entity	NOUN
ajst-29613	119	21	for	for	ADP
ajst-29613	119	22	each	each	DET
ajst-29613	119	23	question	question	NOUN
ajst-29613	119	24	-	-	PUNCT
ajst-29613	119	25	option	option	NOUN
ajst-29613	119	26	pair	pair	NOUN
ajst-29613	119	27	to	to	PART
ajst-29613	119	28	achieve	achieve	VERB
ajst-29613	119	29	information	information	NOUN
ajst-29613	119	30	extraction	extraction	NOUN
ajst-29613	119	31	.	.	PUNCT
ajst-29613	120	1	g	g	NOUN
ajst-29613	120	2	-	-	PUNCT
ajst-29613	120	3	retriever[24	retriever[24	NOUN
ajst-29613	120	4	]	]	PUNCT
ajst-29613	120	5	improved	improve	VERB
ajst-29613	120	6	the	the	DET
ajst-29613	120	7	traditional	traditional	ADJ
ajst-29613	120	8	prizecollecting	prizecollecting	NOUN
ajst-29613	120	9	steiner	steiner	NOUN
ajst-29613	120	10	tree	tree	NOUN
ajst-29613	120	11	(	(	PUNCT
ajst-29613	120	12	pcst	pcst	ADJ
ajst-29613	120	13	)	)	PUNCT
ajst-29613	120	14	algorithm	algorithm	NOUN
ajst-29613	120	15	by	by	ADP
ajst-29613	120	16	introducing	introduce	VERB
ajst-29613	120	17	edge	edge	NOUN
ajst-29613	120	18	weight	weight	NOUN
ajst-29613	120	19	mechanisms	mechanism	NOUN
ajst-29613	120	20	and	and	CCONJ
ajst-29613	120	21	optimizing	optimize	VERB
ajst-29613	120	22	the	the	DET
ajst-29613	120	23	relevant	relevant	ADJ
ajst-29613	120	24	subgraph	subgraph	NOUN
ajst-29613	120	25	extraction	extraction	NOUN
ajst-29613	120	26	process	process	NOUN
ajst-29613	120	27	,	,	PUNCT
ajst-29613	120	28	thereby	thereby	ADV
ajst-29613	120	29	enhancing	enhance	VERB
ajst-29613	120	30	retrieval	retrieval	NOUN
ajst-29613	120	31	performance	performance	NOUN
ajst-29613	120	32	.	.	PUNCT
ajst-29613	121	1	additionally	additionally	ADV
ajst-29613	121	2	,	,	PUNCT
ajst-29613	121	3	methods	method	NOUN
ajst-29613	121	4	by	by	ADP
ajst-29613	121	5	delile	delile	PROPN
ajst-29613	121	6	et	et	PROPN
ajst-29613	121	7	al	al	PROPN
ajst-29613	121	8	.	.	PUNCT
ajst-29613	122	1	(	(	PUNCT
ajst-29613	122	2	2024)[25	2024)[25	NUM
ajst-29613	122	3	]	]	X
ajst-29613	122	4	and	and	CCONJ
ajst-29613	122	5	gnn	gnn	NOUN
ajst-29613	122	6	-	-	PUNCT
ajst-29613	122	7	rag[26	rag[26	NOUN
ajst-29613	122	8	]	]	PUNCT
ajst-29613	122	9	first	first	ADV
ajst-29613	122	10	identify	identify	VERB
ajst-29613	122	11	relevant	relevant	ADJ
ajst-29613	122	12	entities	entity	NOUN
ajst-29613	122	13	from	from	ADP
ajst-29613	122	14	the	the	DET
ajst-29613	122	15	query	query	NOUN
ajst-29613	122	16	and	and	CCONJ
ajst-29613	122	17	then	then	ADV
ajst-29613	122	18	retrieve	retrieve	VERB
ajst-29613	122	19	the	the	DET
ajst-29613	122	20	shortest	short	ADJ
ajst-29613	122	21	paths	path	NOUN
ajst-29613	122	22	associated	associate	VERB
ajst-29613	122	23	with	with	ADP
ajst-29613	122	24	these	these	DET
ajst-29613	122	25	entities	entity	NOUN
ajst-29613	122	26	.	.	PUNCT
ajst-29613	123	1	such	such	ADJ
ajst-29613	123	2	methods	method	NOUN
ajst-29613	123	3	typically	typically	ADV
ajst-29613	123	4	include	include	VERB
ajst-29613	123	5	an	an	DET
ajst-29613	123	6	entity	entity	NOUN
ajst-29613	123	7	linking	link	VERB
ajst-29613	123	8	preprocessing	preprocesse	VERB
ajst-29613	123	9	step	step	NOUN
ajst-29613	123	10	to	to	PART
ajst-29613	123	11	identify	identify	VERB
ajst-29613	123	12	target	target	NOUN
ajst-29613	123	13	nodes	node	NOUN
ajst-29613	123	14	in	in	ADP
ajst-29613	123	15	the	the	DET
ajst-29613	123	16	graph	graph	NOUN
ajst-29613	123	17	prior	prior	ADV
ajst-29613	123	18	to	to	ADP
ajst-29613	123	19	retrieval	retrieval	NOUN
ajst-29613	123	20	.	.	PUNCT
ajst-29613	124	1	(	(	PUNCT
ajst-29613	124	2	b	b	X
ajst-29613	124	3	)	)	PUNCT
ajst-29613	124	4	lms	lm	NOUN
ajst-29613	124	5	serve	serve	VERB
ajst-29613	124	6	as	as	ADP
ajst-29613	124	7	highly	highly	ADV
ajst-29613	124	8	effective	effective	ADJ
ajst-29613	124	9	retrievers	retriever	NOUN
ajst-29613	124	10	in	in	ADP
ajst-29613	124	11	the	the	DET
ajst-29613	124	12	kgsrag	kgsrag	NOUN
ajst-29613	124	13	framework	framework	NOUN
ajst-29613	124	14	due	due	ADP
ajst-29613	124	15	to	to	ADP
ajst-29613	124	16	their	their	PRON
ajst-29613	124	17	powerful	powerful	ADJ
ajst-29613	124	18	natural	natural	ADJ
ajst-29613	124	19	language	language	NOUN
ajst-29613	124	20	understanding	understanding	NOUN
ajst-29613	124	21	capabilities	capability	NOUN
ajst-29613	124	22	.	.	PUNCT
ajst-29613	125	1	these	these	DET
ajst-29613	125	2	models	model	NOUN
ajst-29613	125	3	excel	excel	VERB
ajst-29613	125	4	in	in	ADP
ajst-29613	125	5	processing	processing	NOUN
ajst-29613	125	6	and	and	CCONJ
ajst-29613	125	7	interpreting	interpret	VERB
ajst-29613	125	8	diverse	diverse	ADJ
ajst-29613	125	9	natural	natural	ADJ
ajst-29613	125	10	language	language	NOUN
ajst-29613	125	11	queries	query	NOUN
ajst-29613	125	12	,	,	PUNCT
ajst-29613	125	13	demonstrating	demonstrate	VERB
ajst-29613	125	14	broad	broad	ADJ
ajst-29613	125	15	applicability	applicability	NOUN
ajst-29613	125	16	in	in	ADP
ajst-29613	125	17	graph	graph	NOUN
ajst-29613	125	18	-	-	PUNCT
ajst-29613	125	19	based	base	VERB
ajst-29613	125	20	retrieval	retrieval	NOUN
ajst-29613	125	21	tasks	task	NOUN
ajst-29613	125	22	.	.	PUNCT
ajst-29613	126	1	lms	lm	NOUN
ajst-29613	126	2	are	be	AUX
ajst-29613	126	3	primarily	primarily	ADV
ajst-29613	126	4	categorized	categorize	VERB
ajst-29613	126	5	into	into	ADP
ajst-29613	126	6	two	two	NUM
ajst-29613	126	7	types	type	NOUN
ajst-29613	126	8	:	:	PUNCT
ajst-29613	126	9	discriminative	discriminative	NOUN
ajst-29613	126	10	language	language	NOUN
ajst-29613	126	11	models	model	NOUN
ajst-29613	126	12	and	and	CCONJ
ajst-29613	126	13	generative	generative	ADJ
ajst-29613	126	14	language	language	NOUN
ajst-29613	126	15	models	model	NOUN
ajst-29613	126	16	.	.	PUNCT
ajst-29613	127	1	subgraph	subgraph	PROPN
ajst-29613	127	2	retriever[27	retriever[27	PROPN
ajst-29613	127	3	]	]	PUNCT
ajst-29613	127	4	utilizes	utilize	VERB
ajst-29613	127	5	roberta[28	roberta[28	X
ajst-29613	127	6	]	]	PUNCT
ajst-29613	127	7	as	as	ADP
ajst-29613	127	8	a	a	DET
ajst-29613	127	9	trained	train	VERB
ajst-29613	127	10	retriever	retriever	NOUN
ajst-29613	127	11	to	to	PART
ajst-29613	127	12	expand	expand	VERB
ajst-29613	127	13	and	and	CCONJ
ajst-29613	127	14	retrieve	retrieve	VERB
ajst-29613	127	15	relevant	relevant	ADJ
ajst-29613	127	16	paths	path	NOUN
ajst-29613	127	17	from	from	ADP
ajst-29613	127	18	a	a	DET
ajst-29613	127	19	topic	topic	NOUN
ajst-29613	127	20	entity	entity	NOUN
ajst-29613	127	21	through	through	ADP
ajst-29613	127	22	a	a	DET
ajst-29613	127	23	sequential	sequential	ADJ
ajst-29613	127	24	decision	decision	NOUN
ajst-29613	127	25	-	-	PUNCT
ajst-29613	127	26	making	make	VERB
ajst-29613	127	27	process	process	NOUN
ajst-29613	127	28	.	.	PUNCT
ajst-29613	128	1	kg	kg	PROPN
ajst-29613	128	2	-	-	PUNCT
ajst-29613	128	3	gpt[29	gpt[29	NOUN
ajst-29613	128	4	]	]	PUNCT
ajst-29613	128	5	employs	employ	NOUN
ajst-29613	128	6	(	(	PUNCT
ajst-29613	128	7	llms	llm	NOUN
ajst-29613	128	8	to	to	PART
ajst-29613	128	9	generate	generate	VERB
ajst-29613	128	10	the	the	DET
ajst-29613	128	11	top	top	ADJ
ajst-29613	128	12	-	-	PUNCT
ajst-29613	128	13	k	k	NOUN
ajst-29613	128	14	sets	set	NOUN
ajst-29613	128	15	of	of	ADP
ajst-29613	128	16	relations	relation	NOUN
ajst-29613	128	17	associated	associate	VERB
ajst-29613	128	18	with	with	ADP
ajst-29613	128	19	specific	specific	ADJ
ajst-29613	128	20	entities	entity	NOUN
ajst-29613	128	21	.	.	PUNCT
ajst-29613	129	1	text	text	NOUN
ajst-29613	129	2	-	-	PUNCT
ajst-29613	129	3	to	to	ADP
ajst-29613	129	4	-	-	PUNCT
ajst-29613	129	5	kg[30	kg[30	NOUN
ajst-29613	129	6	]	]	X
ajst-29613	129	7	fine	fine	ADV
ajst-29613	129	8	-	-	PUNCT
ajst-29613	129	9	tuned	tune	VERB
ajst-29613	129	10	gpt-2	gpt-2	PUNCT
ajst-29613	129	11	to	to	PART
ajst-29613	129	12	generate	generate	VERB
ajst-29613	129	13	reasoning	reasoning	NOUN
ajst-29613	129	14	paths	path	NOUN
ajst-29613	129	15	.	.	PUNCT
ajst-29613	130	1	structgpt[31	structgpt[31	X
ajst-29613	130	2	]	]	PUNCT
ajst-29613	130	3	leverages	leverage	VERB
ajst-29613	130	4	llms	llm	NOUN
ajst-29613	130	5	to	to	PART
ajst-29613	130	6	automatically	automatically	ADV
ajst-29613	130	7	invoke	invoke	VERB
ajst-29613	130	8	multiple	multiple	ADJ
ajst-29613	130	9	predefined	predefine	VERB
ajst-29613	130	10	functions	function	NOUN
ajst-29613	130	11	,	,	PUNCT
ajst-29613	130	12	retrieving	retrieve	VERB
ajst-29613	130	13	and	and	CCONJ
ajst-29613	130	14	integrating	integrate	VERB
ajst-29613	130	15	relevant	relevant	ADJ
ajst-29613	130	16	information	information	NOUN
ajst-29613	130	17	to	to	PART
ajst-29613	130	18	support	support	VERB
ajst-29613	130	19	more	more	ADV
ajst-29613	130	20	complex	complex	ADJ
ajst-29613	130	21	reasoning	reasoning	NOUN
ajst-29613	130	22	tasks	task	NOUN
ajst-29613	130	23	.	.	PUNCT
ajst-29613	131	1	(	(	PUNCT
ajst-29613	131	2	c	c	X
ajst-29613	131	3	)	)	PUNCT
ajst-29613	131	4	graph	graph	NOUN
ajst-29613	131	5	neural	neural	ADJ
ajst-29613	131	6	networks	network	NOUN
ajst-29613	131	7	(	(	PUNCT
ajst-29613	131	8	gnns	gnns	ADJ
ajst-29613	131	9	)	)	PUNCT
ajst-29613	131	10	offer	offer	VERB
ajst-29613	131	11	significant	significant	ADJ
ajst-29613	131	12	advantages	advantage	NOUN
ajst-29613	131	13	in	in	ADP
ajst-29613	131	14	understanding	understanding	NOUN
ajst-29613	131	15	and	and	CCONJ
ajst-29613	131	16	leveraging	leverage	VERB
ajst-29613	131	17	complex	complex	ADJ
ajst-29613	131	18	graph	graph	NOUN
ajst-29613	131	19	structures	structure	NOUN
ajst-29613	131	20	.	.	PUNCT
ajst-29613	132	1	gnn	gnn	NOUN
ajst-29613	132	2	-	-	PUNCT
ajst-29613	132	3	based	base	VERB
ajst-29613	132	4	retrievers	retriever	NOUN
ajst-29613	132	5	typically	typically	ADV
ajst-29613	132	6	encode	encode	VERB
ajst-29613	132	7	graph	graph	NOUN
ajst-29613	132	8	data	datum	NOUN
ajst-29613	132	9	and	and	CCONJ
ajst-29613	132	10	score	score	VERB
ajst-29613	132	11	different	different	ADJ
ajst-29613	132	12	retrieval	retrieval	NOUN
ajst-29613	132	13	granularities	granularity	NOUN
ajst-29613	132	14	based	base	VERB
ajst-29613	132	15	on	on	ADP
ajst-29613	132	16	their	their	PRON
ajst-29613	132	17	similarity	similarity	NOUN
ajst-29613	132	18	to	to	ADP
ajst-29613	132	19	the	the	DET
ajst-29613	132	20	query	query	NOUN
ajst-29613	132	21	,	,	PUNCT
ajst-29613	132	22	enabling	enable	VERB
ajst-29613	132	23	efficient	efficient	ADJ
ajst-29613	132	24	information	information	NOUN
ajst-29613	132	25	extraction	extraction	NOUN
ajst-29613	132	26	.	.	PUNCT
ajst-29613	133	1	for	for	ADP
ajst-29613	133	2	instance	instance	NOUN
ajst-29613	133	3	,	,	PUNCT
ajst-29613	133	4	gnn	gnn	PROPN
ajst-29613	133	5	-	-	PUNCT
ajst-29613	133	6	rag[26	rag[26	NOUN
ajst-29613	133	7	]	]	PUNCT
ajst-29613	133	8	encodes	encode	VERB
ajst-29613	133	9	the	the	DET
ajst-29613	133	10	graph	graph	NOUN
ajst-29613	133	11	,	,	PUNCT
ajst-29613	133	12	assigns	assign	VERB
ajst-29613	133	13	relevance	relevance	NOUN
ajst-29613	133	14	scores	score	NOUN
ajst-29613	133	15	to	to	ADP
ajst-29613	133	16	each	each	DET
ajst-29613	133	17	entity	entity	NOUN
ajst-29613	133	18	,	,	PUNCT
ajst-29613	133	19	and	and	CCONJ
ajst-29613	133	20	retrieves	retrieve	VERB
ajst-29613	133	21	entities	entity	NOUN
ajst-29613	133	22	related	relate	VERB
ajst-29613	133	23	to	to	ADP
ajst-29613	133	24	the	the	DET
ajst-29613	133	25	query	query	NOUN
ajst-29613	133	26	based	base	VERB
ajst-29613	133	27	on	on	ADP
ajst-29613	133	28	a	a	DET
ajst-29613	133	29	predefined	predefine	VERB
ajst-29613	133	30	threshold	threshold	NOUN
ajst-29613	133	31	.	.	PUNCT
ajst-29613	134	1	etd[32	etd[32	NOUN
ajst-29613	134	2	]	]	PUNCT
ajst-29613	134	3	employs	employ	VERB
ajst-29613	134	4	an	an	DET
ajst-29613	134	5	iterative	iterative	NOUN
ajst-29613	134	6	approach	approach	NOUN
ajst-29613	134	7	to	to	PART
ajst-29613	134	8	retrieve	retrieve	VERB
ajst-29613	134	9	relevant	relevant	ADJ
ajst-29613	134	10	paths	path	NOUN
ajst-29613	134	11	.	.	PUNCT
ajst-29613	135	1	in	in	ADP
ajst-29613	135	2	each	each	DET
ajst-29613	135	3	iteration	iteration	NOUN
ajst-29613	135	4	,	,	PUNCT
ajst-29613	135	5	the	the	DET
ajst-29613	135	6	method	method	NOUN
ajst-29613	135	7	uses	use	VERB
ajst-29613	135	8	llama2[60	llama2[60	PROPN
ajst-29613	135	9	]	]	PUNCT
ajst-29613	135	10	to	to	PART
ajst-29613	135	11	select	select	VERB
ajst-29613	135	12	edges	edge	NOUN
ajst-29613	135	13	connected	connect	VERB
ajst-29613	135	14	to	to	ADP
ajst-29613	135	15	the	the	DET
ajst-29613	135	16	current	current	ADJ
ajst-29613	135	17	node	node	NOUN
ajst-29613	135	18	,	,	PUNCT
ajst-29613	135	19	then	then	ADV
ajst-29613	135	20	calculates	calculate	VERB
ajst-29613	135	21	embeddings	embedding	NOUN
ajst-29613	135	22	for	for	ADP
ajst-29613	135	23	the	the	DET
ajst-29613	135	24	next	next	ADJ
ajst-29613	135	25	layer	layer	NOUN
ajst-29613	135	26	of	of	ADP
ajst-29613	135	27	nodes	node	NOUN
ajst-29613	135	28	using	use	VERB
ajst-29613	135	29	gnns	gnn	NOUN
ajst-29613	135	30	,	,	PUNCT
ajst-29613	135	31	thereby	thereby	ADV
ajst-29613	135	32	supporting	support	VERB
ajst-29613	135	33	the	the	DET
ajst-29613	135	34	next	next	ADJ
ajst-29613	135	35	round	round	NOUN
ajst-29613	135	36	of	of	ADP
ajst-29613	135	37	llm	llm	NOUN
ajst-29613	135	38	-	-	PUNCT
ajst-29613	135	39	based	base	VERB
ajst-29613	135	40	selection	selection	NOUN
ajst-29613	135	41	.	.	PUNCT
ajst-29613	136	1	3.3	3.3	NUM
ajst-29613	136	2	.	.	PUNCT
ajst-29613	137	1	enhanced	enhance	VERB
ajst-29613	137	2	generation	generation	NOUN
ajst-29613	137	3	in	in	ADP
ajst-29613	137	4	the	the	DET
ajst-29613	137	5	kgs	kgs	NOUN
ajst-29613	137	6	-	-	PUNCT
ajst-29613	137	7	rag	rag	NOUN
ajst-29613	137	8	framework	framework	NOUN
ajst-29613	137	9	,	,	PUNCT
ajst-29613	137	10	the	the	DET
ajst-29613	137	11	generation	generation	NOUN
ajst-29613	137	12	phase	phase	NOUN
ajst-29613	137	13	is	be	AUX
ajst-29613	137	14	another	another	DET
ajst-29613	137	15	critical	critical	ADJ
ajst-29613	137	16	step	step	NOUN
ajst-29613	137	17	that	that	PRON
ajst-29613	137	18	deeply	deeply	ADV
ajst-29613	137	19	integrates	integrate	VERB
ajst-29613	137	20	the	the	DET
ajst-29613	137	21	retrieved	retrieve	VERB
ajst-29613	137	22	subgraph	subgraph	NOUN
ajst-29613	137	23	data	datum	NOUN
ajst-29613	137	24	from	from	ADP
ajst-29613	137	25	the	the	DET
ajst-29613	137	26	knowledge	knowledge	NOUN
ajst-29613	137	27	graph	graph	NOUN
ajst-29613	137	28	with	with	ADP
ajst-29613	137	29	the	the	DET
ajst-29613	137	30	user	user	NOUN
ajst-29613	137	31	query	query	NOUN
ajst-29613	137	32	,	,	PUNCT
ajst-29613	137	33	significantly	significantly	ADV
ajst-29613	137	34	enhancing	enhance	VERB
ajst-29613	137	35	the	the	DET
ajst-29613	137	36	accuracy	accuracy	NOUN
ajst-29613	137	37	,	,	PUNCT
ajst-29613	137	38	logic	logic	NOUN
ajst-29613	137	39	,	,	PUNCT
ajst-29613	137	40	and	and	CCONJ
ajst-29613	137	41	semantic	semantic	ADJ
ajst-29613	137	42	coherence	coherence	NOUN
ajst-29613	137	43	of	of	ADP
ajst-29613	137	44	the	the	DET
ajst-29613	137	45	generated	generate	VERB
ajst-29613	137	46	content	content	NOUN
ajst-29613	137	47	.	.	PUNCT
ajst-29613	138	1	this	this	DET
ajst-29613	138	2	phase	phase	NOUN
ajst-29613	138	3	not	not	PART
ajst-29613	138	4	only	only	ADV
ajst-29613	138	5	involves	involve	VERB
ajst-29613	138	6	selecting	select	VERB
ajst-29613	138	7	a	a	DET
ajst-29613	138	8	generation	generation	NOUN
ajst-29613	138	9	model	model	NOUN
ajst-29613	138	10	suitable	suitable	ADJ
ajst-29613	138	11	for	for	ADP
ajst-29613	138	12	the	the	DET
ajst-29613	138	13	downstream	downstream	ADJ
ajst-29613	138	14	task	task	NOUN
ajst-29613	138	15	but	but	CCONJ
ajst-29613	138	16	also	also	ADV
ajst-29613	138	17	requires	require	VERB
ajst-29613	138	18	effective	effective	ADJ
ajst-29613	138	19	preprocessing	preprocessing	NOUN
ajst-29613	138	20	and	and	CCONJ
ajst-29613	138	21	format	format	NOUN
ajst-29613	138	22	conversion	conversion	NOUN
ajst-29613	138	23	of	of	ADP
ajst-29613	138	24	the	the	DET
ajst-29613	138	25	retrieved	retrieve	VERB
ajst-29613	138	26	graph	graph	NOUN
ajst-29613	138	27	data	datum	NOUN
ajst-29613	138	28	to	to	PART
ajst-29613	138	29	seamlessly	seamlessly	ADV
ajst-29613	138	30	incorporate	incorporate	VERB
ajst-29613	138	31	it	it	PRON
ajst-29613	138	32	into	into	ADP
ajst-29613	138	33	the	the	DET
ajst-29613	138	34	generator	generator	NOUN
ajst-29613	138	35	's	's	PART
ajst-29613	138	36	input	input	NOUN
ajst-29613	138	37	.	.	PUNCT
ajst-29613	139	1	additionally	additionally	ADV
ajst-29613	139	2	,	,	PUNCT
ajst-29613	139	3	by	by	ADP
ajst-29613	139	4	introducing	introduce	VERB
ajst-29613	139	5	techniques	technique	NOUN
ajst-29613	139	6	for	for	ADP
ajst-29613	139	7	enhanced	enhanced	ADJ
ajst-29613	139	8	generation	generation	NOUN
ajst-29613	139	9	,	,	PUNCT
ajst-29613	139	10	the	the	DET
ajst-29613	139	11	interaction	interaction	NOUN
ajst-29613	139	12	between	between	ADP
ajst-29613	139	13	the	the	DET
ajst-29613	139	14	query	query	NOUN
ajst-29613	139	15	and	and	CCONJ
ajst-29613	139	16	graph	graph	NOUN
ajst-29613	139	17	data	datum	NOUN
ajst-29613	139	18	can	can	AUX
ajst-29613	139	19	be	be	AUX
ajst-29613	139	20	further	far	ADV
ajst-29613	139	21	optimized	optimize	VERB
ajst-29613	139	22	,	,	PUNCT
ajst-29613	139	23	enriching	enrich	VERB
ajst-29613	139	24	the	the	DET
ajst-29613	139	25	diversity	diversity	NOUN
ajst-29613	139	26	and	and	CCONJ
ajst-29613	139	27	depth	depth	NOUN
ajst-29613	139	28	of	of	ADP
ajst-29613	139	29	the	the	DET
ajst-29613	139	30	generated	generate	VERB
ajst-29613	139	31	content	content	NOUN
ajst-29613	139	32	,	,	PUNCT
ajst-29613	139	33	ultimately	ultimately	ADV
ajst-29613	139	34	improving	improve	VERB
ajst-29613	139	35	the	the	DET
ajst-29613	139	36	quality	quality	NOUN
ajst-29613	139	37	of	of	ADP
ajst-29613	139	38	the	the	DET
ajst-29613	139	39	output	output	NOUN
ajst-29613	139	40	and	and	CCONJ
ajst-29613	139	41	the	the	DET
ajst-29613	139	42	overall	overall	ADJ
ajst-29613	139	43	user	user	NOUN
ajst-29613	139	44	experience	experience	NOUN
ajst-29613	139	45	.	.	PUNCT
ajst-29613	140	1	3.3.1	3.3.1	X
ajst-29613	140	2	.	.	PUNCT
ajst-29613	140	3	selection	selection	NOUN
ajst-29613	140	4	of	of	ADP
ajst-29613	140	5	generation	generation	NOUN
ajst-29613	140	6	models	model	NOUN
ajst-29613	140	7	the	the	DET
ajst-29613	140	8	first	first	ADJ
ajst-29613	140	9	step	step	NOUN
ajst-29613	140	10	in	in	ADP
ajst-29613	140	11	the	the	DET
ajst-29613	140	12	generation	generation	NOUN
ajst-29613	140	13	phase	phase	NOUN
ajst-29613	140	14	is	be	AUX
ajst-29613	140	15	to	to	PART
ajst-29613	140	16	select	select	VERB
ajst-29613	140	17	an	an	DET
ajst-29613	140	18	appropriate	appropriate	ADJ
ajst-29613	140	19	generation	generation	NOUN
ajst-29613	140	20	model	model	NOUN
ajst-29613	140	21	based	base	VERB
ajst-29613	140	22	on	on	ADP
ajst-29613	140	23	the	the	DET
ajst-29613	140	24	specific	specific	ADJ
ajst-29613	140	25	requirements	requirement	NOUN
ajst-29613	140	26	of	of	ADP
ajst-29613	140	27	the	the	DET
ajst-29613	140	28	downstream	downstream	ADJ
ajst-29613	140	29	task	task	NOUN
ajst-29613	140	30	.	.	PUNCT
ajst-29613	141	1	these	these	DET
ajst-29613	141	2	models	model	NOUN
ajst-29613	141	3	are	be	AUX
ajst-29613	141	4	typically	typically	ADV
ajst-29613	141	5	built	build	VERB
ajst-29613	141	6	on	on	ADP
ajst-29613	141	7	pre	pre	ADJ
ajst-29613	141	8	-	-	ADJ
ajst-29613	141	9	trained	train	VERB
ajst-29613	141	10	deep	deep	ADJ
ajst-29613	141	11	learning	learning	NOUN
ajst-29613	141	12	architectures	architecture	NOUN
ajst-29613	141	13	,	,	PUNCT
ajst-29613	141	14	such	such	ADJ
ajst-29613	141	15	as	as	ADP
ajst-29613	141	16	bert[32	bert[32	PROPN
ajst-29613	141	17	]	]	X
ajst-29613	141	18	,	,	PUNCT
ajst-29613	141	19	t5[33	t5[33	PROPN
ajst-29613	141	20	]	]	PUNCT
ajst-29613	141	21	,	,	PUNCT
ajst-29613	141	22	the	the	DET
ajst-29613	141	23	gpt	gpt	NOUN
ajst-29613	141	24	series	series	NOUN
ajst-29613	141	25	,	,	PUNCT
ajst-29613	141	26	or	or	CCONJ
ajst-29613	141	27	more	more	ADV
ajst-29613	141	28	advanced	advanced	ADJ
ajst-29613	141	29	multimodal	multimodal	NOUN
ajst-29613	141	30	models	model	NOUN
ajst-29613	141	31	like	like	ADP
ajst-29613	141	32	palm	palm	NOUN
ajst-29613	141	33	-	-	PUNCT
ajst-29613	141	34	e[34	e[34	PROPN
ajst-29613	141	35	]	]	PUNCT
ajst-29613	141	36	,	,	PUNCT
ajst-29613	141	37	to	to	PART
ajst-29613	141	38	ensure	ensure	VERB
ajst-29613	141	39	that	that	SCONJ
ajst-29613	141	40	the	the	DET
ajst-29613	141	41	generated	generate	VERB
ajst-29613	141	42	output	output	NOUN
ajst-29613	141	43	can	can	AUX
ajst-29613	141	44	handle	handle	VERB
ajst-29613	141	45	various	various	ADJ
ajst-29613	141	46	task	task	NOUN
ajst-29613	141	47	types	type	NOUN
ajst-29613	141	48	,	,	PUNCT
ajst-29613	141	49	including	include	VERB
ajst-29613	141	50	question	question	NOUN
ajst-29613	141	51	answering	answering	NOUN
ajst-29613	141	52	,	,	PUNCT
ajst-29613	141	53	text	text	NOUN
ajst-29613	141	54	summarization	summarization	NOUN
ajst-29613	141	55	,	,	PUNCT
ajst-29613	141	56	and	and	CCONJ
ajst-29613	141	57	dialogue	dialogue	NOUN
ajst-29613	141	58	generation	generation	NOUN
ajst-29613	141	59	.	.	PUNCT
ajst-29613	142	1	the	the	DET
ajst-29613	142	2	selection	selection	NOUN
ajst-29613	142	3	of	of	ADP
ajst-29613	142	4	a	a	DET
ajst-29613	142	5	generation	generation	NOUN
ajst-29613	142	6	model	model	NOUN
ajst-29613	142	7	should	should	AUX
ajst-29613	142	8	consider	consider	VERB
ajst-29613	142	9	the	the	DET
ajst-29613	142	10	following	follow	VERB
ajst-29613	142	11	factors	factor	NOUN
ajst-29613	142	12	:	:	PUNCT
ajst-29613	142	13	task	task	NOUN
ajst-29613	142	14	suitability	suitability	NOUN
ajst-29613	142	15	different	different	ADJ
ajst-29613	142	16	tasks	task	NOUN
ajst-29613	142	17	have	have	VERB
ajst-29613	142	18	varying	vary	VERB
ajst-29613	142	19	requirements	requirement	NOUN
ajst-29613	142	20	for	for	ADP
ajst-29613	142	21	generation	generation	NOUN
ajst-29613	142	22	models	model	NOUN
ajst-29613	142	23	.	.	PUNCT
ajst-29613	143	1	for	for	ADP
ajst-29613	143	2	instance	instance	NOUN
ajst-29613	143	3	,	,	PUNCT
ajst-29613	143	4	in	in	ADP
ajst-29613	143	5	open	open	ADJ
ajst-29613	143	6	-	-	PUNCT
ajst-29613	143	7	domain	domain	NOUN
ajst-29613	143	8	question	question	NOUN
ajst-29613	143	9	answering	answering	NOUN
ajst-29613	143	10	,	,	PUNCT
ajst-29613	143	11	models	model	NOUN
ajst-29613	143	12	capable	capable	ADJ
ajst-29613	143	13	of	of	ADP
ajst-29613	143	14	handling	handle	VERB
ajst-29613	143	15	complex	complex	ADJ
ajst-29613	143	16	semantics	semantic	NOUN
ajst-29613	143	17	and	and	CCONJ
ajst-29613	143	18	producing	produce	VERB
ajst-29613	143	19	precise	precise	ADJ
ajst-29613	143	20	answers	answer	NOUN
ajst-29613	143	21	(	(	PUNCT
ajst-29613	143	22	e.g.	e.g.	ADV
ajst-29613	143	23	,	,	PUNCT
ajst-29613	143	24	gpt-4	gpt-4	NOUN
ajst-29613	143	25	)	)	PUNCT
ajst-29613	143	26	are	be	AUX
ajst-29613	143	27	preferred	prefer	VERB
ajst-29613	143	28	.	.	PUNCT
ajst-29613	144	1	in	in	ADP
ajst-29613	144	2	contrast	contrast	NOUN
ajst-29613	144	3	,	,	PUNCT
ajst-29613	144	4	for	for	ADP
ajst-29613	144	5	scientific	scientific	ADJ
ajst-29613	144	6	literature	literature	NOUN
ajst-29613	144	7	generation	generation	NOUN
ajst-29613	144	8	or	or	CCONJ
ajst-29613	144	9	technical	technical	ADJ
ajst-29613	144	10	document	document	NOUN
ajst-29613	144	11	drafting	drafting	NOUN
ajst-29613	144	12	,	,	PUNCT
ajst-29613	144	13	models	model	NOUN
ajst-29613	144	14	capable	capable	ADJ
ajst-29613	144	15	of	of	ADP
ajst-29613	144	16	generating	generate	VERB
ajst-29613	144	17	coherent	coherent	ADJ
ajst-29613	144	18	long	long	ADJ
ajst-29613	144	19	-	-	PUNCT
ajst-29613	144	20	form	form	NOUN
ajst-29613	144	21	text	text	NOUN
ajst-29613	144	22	with	with	ADP
ajst-29613	144	23	domain	domain	NOUN
ajst-29613	144	24	-	-	PUNCT
ajst-29613	144	25	specific	specific	ADJ
ajst-29613	144	26	knowledge	knowledge	NOUN
ajst-29613	144	27	(	(	PUNCT
ajst-29613	144	28	e.g.	e.g.	ADV
ajst-29613	144	29	,	,	PUNCT
ajst-29613	144	30	t5	t5	PROPN
ajst-29613	144	31	-	-	PUNCT
ajst-29613	144	32	11b	11b	NOUN
ajst-29613	144	33	)	)	PUNCT
ajst-29613	144	34	are	be	AUX
ajst-29613	144	35	more	more	ADV
ajst-29613	144	36	appropriate	appropriate	ADJ
ajst-29613	144	37	.	.	PUNCT
ajst-29613	145	1	model	model	NOUN
ajst-29613	145	2	scalability	scalability	NOUN
ajst-29613	145	3	and	and	CCONJ
ajst-29613	145	4	reasoning	reasoning	NOUN
ajst-29613	145	5	capability	capability	NOUN
ajst-29613	145	6	the	the	DET
ajst-29613	145	7	generation	generation	NOUN
ajst-29613	145	8	model	model	NOUN
ajst-29613	145	9	should	should	AUX
ajst-29613	145	10	possess	possess	VERB
ajst-29613	145	11	the	the	DET
ajst-29613	145	12	ability	ability	NOUN
ajst-29613	145	13	to	to	PART
ajst-29613	145	14	infer	infer	VERB
ajst-29613	145	15	implicit	implicit	ADJ
ajst-29613	145	16	relationships	relationship	NOUN
ajst-29613	145	17	from	from	ADP
ajst-29613	145	18	graph	graph	NOUN
ajst-29613	145	19	data	datum	NOUN
ajst-29613	145	20	.	.	PUNCT
ajst-29613	146	1	integrating	integrate	VERB
ajst-29613	146	2	graph	graph	NOUN
ajst-29613	146	3	neural	neural	ADJ
ajst-29613	146	4	networks	network	NOUN
ajst-29613	146	5	(	(	PUNCT
ajst-29613	146	6	gnns	gnns	NOUN
ajst-29613	146	7	)	)	PUNCT
ajst-29613	146	8	and	and	CCONJ
ajst-29613	146	9	knowledge	knowledge	NOUN
ajst-29613	146	10	graph	graph	NOUN
ajst-29613	146	11	embeddings	embedding	NOUN
ajst-29613	146	12	can	can	AUX
ajst-29613	146	13	enhance	enhance	VERB
ajst-29613	146	14	the	the	DET
ajst-29613	146	15	model	model	NOUN
ajst-29613	146	16	’s	’s	PART
ajst-29613	146	17	performance	performance	NOUN
ajst-29613	146	18	in	in	ADP
ajst-29613	146	19	generating	generating	NOUN
ajst-29613	146	20	responses	response	NOUN
ajst-29613	146	21	for	for	ADP
ajst-29613	146	22	semantically	semantically	ADV
ajst-29613	146	23	complex	complex	ADJ
ajst-29613	146	24	queries	query	NOUN
ajst-29613	146	25	.	.	PUNCT
ajst-29613	147	1	computational	computational	ADJ
ajst-29613	147	2	efficiency	efficiency	NOUN
ajst-29613	147	3	and	and	CCONJ
ajst-29613	147	4	real	real	ADJ
ajst-29613	147	5	-	-	PUNCT
ajst-29613	147	6	time	time	NOUN
ajst-29613	147	7	responsiveness	responsiveness	NOUN
ajst-29613	147	8	in	in	ADP
ajst-29613	147	9	scenarios	scenario	NOUN
ajst-29613	147	10	requiring	require	VERB
ajst-29613	147	11	real	real	ADJ
ajst-29613	147	12	-	-	PUNCT
ajst-29613	147	13	time	time	NOUN
ajst-29613	147	14	response	response	NOUN
ajst-29613	147	15	generation	generation	NOUN
ajst-29613	147	16	(	(	PUNCT
ajst-29613	147	17	e.g.	e.g.	ADV
ajst-29613	147	18	,	,	PUNCT
ajst-29613	147	19	dialogue	dialogue	NOUN
ajst-29613	147	20	systems	system	NOUN
ajst-29613	147	21	)	)	PUNCT
ajst-29613	147	22	,	,	PUNCT
ajst-29613	147	23	inference	inference	NOUN
ajst-29613	147	24	speed	speed	NOUN
ajst-29613	147	25	and	and	CCONJ
ajst-29613	147	26	computational	computational	ADJ
ajst-29613	147	27	cost	cost	NOUN
ajst-29613	147	28	are	be	AUX
ajst-29613	147	29	critical	critical	ADJ
ajst-29613	147	30	considerations	consideration	NOUN
ajst-29613	147	31	.	.	PUNCT
ajst-29613	148	1	lightweight	lightweight	ADJ
ajst-29613	148	2	generation	generation	NOUN
ajst-29613	148	3	models	model	NOUN
ajst-29613	148	4	or	or	CCONJ
ajst-29613	148	5	those	those	PRON
ajst-29613	148	6	optimized	optimize	VERB
ajst-29613	148	7	using	use	VERB
ajst-29613	148	8	knowledge	knowledge	NOUN
ajst-29613	148	9	distillation	distillation	NOUN
ajst-29613	148	10	techniques	technique	NOUN
ajst-29613	148	11	may	may	AUX
ajst-29613	148	12	offer	offer	VERB
ajst-29613	148	13	significant	significant	ADJ
ajst-29613	148	14	advantages	advantage	NOUN
ajst-29613	148	15	.	.	PUNCT
ajst-29613	149	1	3.3.2	3.3.2	X
ajst-29613	149	2	.	.	X
ajst-29613	149	3	preprocessing	preprocessing	NOUN
ajst-29613	149	4	and	and	CCONJ
ajst-29613	149	5	format	format	NOUN
ajst-29613	149	6	conversion	conversion	NOUN
ajst-29613	149	7	of	of	ADP
ajst-29613	149	8	graph	graph	NOUN
ajst-29613	149	9	data	datum	NOUN
ajst-29613	149	10	the	the	DET
ajst-29613	149	11	core	core	NOUN
ajst-29613	149	12	of	of	ADP
ajst-29613	149	13	the	the	DET
ajst-29613	149	14	generation	generation	NOUN
ajst-29613	149	15	phase	phase	NOUN
ajst-29613	149	16	lies	lie	VERB
ajst-29613	149	17	in	in	ADP
ajst-29613	149	18	converting	convert	VERB
ajst-29613	149	19	the	the	DET
ajst-29613	149	20	retrieved	retrieve	VERB
ajst-29613	149	21	graph	graph	NOUN
ajst-29613	149	22	data	datum	NOUN
ajst-29613	149	23	into	into	ADP
ajst-29613	149	24	an	an	DET
ajst-29613	149	25	input	input	NOUN
ajst-29613	149	26	format	format	NOUN
ajst-29613	149	27	directly	directly	ADV
ajst-29613	149	28	usable	usable	ADJ
ajst-29613	149	29	by	by	ADP
ajst-29613	149	30	the	the	DET
ajst-29613	149	31	generation	generation	NOUN
ajst-29613	149	32	model	model	NOUN
ajst-29613	149	33	.	.	PUNCT
ajst-29613	150	1	this	this	DET
ajst-29613	150	2	process	process	NOUN
ajst-29613	150	3	involves	involve	VERB
ajst-29613	150	4	the	the	DET
ajst-29613	150	5	following	follow	VERB
ajst-29613	150	6	key	key	ADJ
ajst-29613	150	7	steps	step	NOUN
ajst-29613	150	8	:	:	PUNCT
ajst-29613	150	9	linearization	linearization	NOUN
ajst-29613	150	10	of	of	ADP
ajst-29613	150	11	structured	structured	ADJ
ajst-29613	150	12	knowledge	knowledge	NOUN
ajst-29613	150	13	knowledge	knowledge	NOUN
ajst-29613	150	14	graphs	graph	NOUN
ajst-29613	150	15	are	be	AUX
ajst-29613	150	16	typically	typically	ADV
ajst-29613	150	17	represented	represent	VERB
ajst-29613	150	18	in	in	ADP
ajst-29613	150	19	the	the	DET
ajst-29613	150	20	form	form	NOUN
ajst-29613	150	21	of	of	ADP
ajst-29613	150	22	triples	triple	NOUN
ajst-29613	150	23	(	(	PUNCT
ajst-29613	150	24	e.g.	e.g.	ADV
ajst-29613	150	25	,	,	PUNCT
ajst-29613	150	26	<	<	X
ajst-29613	150	27	entity1	entity1	ADJ
ajst-29613	150	28	,	,	PUNCT
ajst-29613	150	29	relationship	relationship	NOUN
ajst-29613	150	30	,	,	PUNCT
ajst-29613	150	31	entity2	entity2	PROPN
ajst-29613	150	32	>	>	X
ajst-29613	150	33	)	)	PUNCT
ajst-29613	150	34	,	,	PUNCT
ajst-29613	150	35	while	while	SCONJ
ajst-29613	150	36	generation	generation	NOUN
ajst-29613	150	37	models	model	NOUN
ajst-29613	150	38	usually	usually	ADV
ajst-29613	150	39	process	process	VERB
ajst-29613	150	40	natural	natural	ADJ
ajst-29613	150	41	language	language	NOUN
ajst-29613	150	42	text	text	NOUN
ajst-29613	150	43	.	.	PUNCT
ajst-29613	151	1	therefore	therefore	ADV
ajst-29613	151	2	,	,	PUNCT
ajst-29613	151	3	the	the	DET
ajst-29613	151	4	graph	graph	NOUN
ajst-29613	151	5	data	datum	NOUN
ajst-29613	151	6	needs	need	VERB
ajst-29613	151	7	to	to	PART
ajst-29613	151	8	be	be	AUX
ajst-29613	151	9	linearized	linearize	VERB
ajst-29613	151	10	into	into	ADP
ajst-29613	151	11	a	a	DET
ajst-29613	151	12	natural	natural	ADJ
ajst-29613	151	13	language	language	NOUN
ajst-29613	151	14	format	format	NOUN
ajst-29613	151	15	.	.	PUNCT
ajst-29613	152	1	for	for	ADP
ajst-29613	152	2	example	example	NOUN
ajst-29613	152	3	,	,	PUNCT
ajst-29613	152	4	the	the	DET
ajst-29613	152	5	triple	triple	ADJ
ajst-29613	152	6	<	<	NOUN
ajst-29613	152	7	bill	bill	NOUN
ajst-29613	152	8	gates	gate	NOUN
ajst-29613	152	9	,	,	PUNCT
ajst-29613	152	10	founded	found	VERB
ajst-29613	152	11	,	,	PUNCT
ajst-29613	152	12	microsoft	microsoft	PROPN
ajst-29613	152	13	>	>	X
ajst-29613	152	14	can	can	AUX
ajst-29613	152	15	be	be	AUX
ajst-29613	152	16	transformed	transform	VERB
ajst-29613	152	17	into	into	ADP
ajst-29613	152	18	"	"	PUNCT
ajst-29613	152	19	bill	bill	NOUN
ajst-29613	152	20	gates	gate	NOUN
ajst-29613	152	21	founded	found	VERB
ajst-29613	152	22	microsoft	microsoft	PROPN
ajst-29613	152	23	.	.	PUNCT
ajst-29613	152	24	"	"	PUNCT
ajst-29613	153	1	semantic	semantic	ADJ
ajst-29613	153	2	expansion	expansion	NOUN
ajst-29613	153	3	of	of	ADP
ajst-29613	153	4	nodes	node	NOUN
ajst-29613	153	5	and	and	CCONJ
ajst-29613	153	6	relationships	relationship	NOUN
ajst-29613	153	7	during	during	ADP
ajst-29613	153	8	format	format	NOUN
ajst-29613	153	9	conversion	conversion	NOUN
ajst-29613	153	10	,	,	PUNCT
ajst-29613	153	11	semantic	semantic	ADJ
ajst-29613	153	12	descriptions	description	NOUN
ajst-29613	153	13	of	of	ADP
ajst-29613	153	14	nodes	node	NOUN
ajst-29613	153	15	and	and	CCONJ
ajst-29613	153	16	relationships	relationship	NOUN
ajst-29613	153	17	can	can	AUX
ajst-29613	153	18	be	be	AUX
ajst-29613	153	19	added	add	VERB
ajst-29613	153	20	to	to	PART
ajst-29613	153	21	enrich	enrich	VERB
ajst-29613	153	22	the	the	DET
ajst-29613	153	23	context	context	NOUN
ajst-29613	153	24	of	of	ADP
ajst-29613	153	25	the	the	DET
ajst-29613	153	26	graph	graph	NOUN
ajst-29613	153	27	data	datum	NOUN
ajst-29613	153	28	.	.	PUNCT
ajst-29613	154	1	for	for	ADP
ajst-29613	154	2	instance	instance	NOUN
ajst-29613	154	3	,	,	PUNCT
ajst-29613	154	4	the	the	DET
ajst-29613	154	5	node	node	ADJ
ajst-29613	154	6	"	"	PUNCT
ajst-29613	154	7	bill	bill	NOUN
ajst-29613	154	8	gates	gate	NOUN
ajst-29613	154	9	"	"	PUNCT
ajst-29613	154	10	could	could	AUX
ajst-29613	154	11	be	be	AUX
ajst-29613	154	12	expanded	expand	VERB
ajst-29613	154	13	to	to	ADP
ajst-29613	154	14	"	"	PUNCT
ajst-29613	154	15	bill	bill	NOUN
ajst-29613	154	16	gates	gate	NOUN
ajst-29613	154	17	,	,	PUNCT
ajst-29613	154	18	the	the	DET
ajst-29613	154	19	co	co	NOUN
ajst-29613	154	20	-	-	NOUN
ajst-29613	154	21	founder	founder	NOUN
ajst-29613	154	22	of	of	ADP
ajst-29613	154	23	microsoft	microsoft	PROPN
ajst-29613	154	24	.	.	PUNCT
ajst-29613	154	25	"	"	PUNCT
ajst-29613	155	1	context	context	NOUN
ajst-29613	155	2	concatenation	concatenation	NOUN
ajst-29613	155	3	and	and	CCONJ
ajst-29613	155	4	query	query	NOUN
ajst-29613	155	5	fusion	fusion	NOUN
ajst-29613	155	6	to	to	PART
ajst-29613	155	7	enhance	enhance	VERB
ajst-29613	155	8	the	the	DET
ajst-29613	155	9	relevance	relevance	NOUN
ajst-29613	155	10	between	between	ADP
ajst-29613	155	11	the	the	DET
ajst-29613	155	12	graph	graph	NOUN
ajst-29613	155	13	data	datum	NOUN
ajst-29613	155	14	and	and	CCONJ
ajst-29613	155	15	the	the	DET
ajst-29613	155	16	user	user	NOUN
ajst-29613	155	17	query	query	NOUN
ajst-29613	155	18	,	,	PUNCT
ajst-29613	155	19	the	the	DET
ajst-29613	155	20	graph	graph	NOUN
ajst-29613	155	21	data	datum	NOUN
ajst-29613	155	22	should	should	AUX
ajst-29613	155	23	be	be	AUX
ajst-29613	155	24	concatenated	concatenate	VERB
ajst-29613	155	25	with	with	ADP
ajst-29613	155	26	the	the	DET
ajst-29613	155	27	query	query	NOUN
ajst-29613	155	28	into	into	ADP
ajst-29613	155	29	a	a	DET
ajst-29613	155	30	complete	complete	ADJ
ajst-29613	155	31	input	input	NOUN
ajst-29613	155	32	sequence	sequence	NOUN
ajst-29613	155	33	.	.	PUNCT
ajst-29613	156	1	for	for	ADP
ajst-29613	156	2	example	example	NOUN
ajst-29613	156	3	,	,	PUNCT
ajst-29613	156	4	for	for	ADP
ajst-29613	156	5	the	the	DET
ajst-29613	156	6	query	query	NOUN
ajst-29613	156	7	"	"	PUNCT
ajst-29613	156	8	who	who	PRON
ajst-29613	156	9	founded	found	VERB
ajst-29613	156	10	microsoft	microsoft	PROPN
ajst-29613	156	11	?	?	PUNCT
ajst-29613	156	12	"	"	PUNCT
ajst-29613	156	13	,	,	PUNCT
ajst-29613	156	14	the	the	DET
ajst-29613	156	15	concatenated	concatenate	VERB
ajst-29613	156	16	input	input	NOUN
ajst-29613	156	17	could	could	AUX
ajst-29613	156	18	be	be	AUX
ajst-29613	156	19	:	:	PUNCT
ajst-29613	156	20	"	"	PUNCT
ajst-29613	156	21	query	query	NOUN
ajst-29613	156	22	:	:	PUNCT
ajst-29613	156	23	who	who	PRON
ajst-29613	156	24	founded	found	VERB
ajst-29613	156	25	microsoft	microsoft	PROPN
ajst-29613	156	26	?	?	PUNCT
ajst-29613	157	1	context	context	NOUN
ajst-29613	157	2	:	:	PUNCT
ajst-29613	158	1	bill	bill	PROPN
ajst-29613	158	2	gates	gate	NOUN
ajst-29613	158	3	founded	found	VERB
ajst-29613	158	4	microsoft	microsoft	PROPN
ajst-29613	158	5	;	;	PUNCT
ajst-29613	158	6	microsoft	microsoft	PROPN
ajst-29613	158	7	is	be	AUX
ajst-29613	158	8	a	a	DET
ajst-29613	158	9	technology	technology	NOUN
ajst-29613	158	10	company	company	NOUN
ajst-29613	158	11	.	.	PUNCT
ajst-29613	158	12	"	"	PUNCT
ajst-29613	159	1	integration	integration	NOUN
ajst-29613	159	2	of	of	ADP
ajst-29613	159	3	multimodal	multimodal	NOUN
ajst-29613	159	4	information	information	NOUN
ajst-29613	159	5	in	in	ADP
ajst-29613	159	6	some	some	DET
ajst-29613	159	7	applications	application	NOUN
ajst-29613	159	8	,	,	PUNCT
ajst-29613	159	9	nodes	node	NOUN
ajst-29613	159	10	in	in	ADP
ajst-29613	159	11	the	the	DET
ajst-29613	159	12	knowledge	knowledge	NOUN
ajst-29613	159	13	graph	graph	NOUN
ajst-29613	159	14	may	may	AUX
ajst-29613	159	15	include	include	VERB
ajst-29613	159	16	multimodal	multimodal	ADJ
ajst-29613	159	17	information	information	NOUN
ajst-29613	159	18	such	such	ADJ
ajst-29613	159	19	as	as	ADP
ajst-29613	159	20	images	image	NOUN
ajst-29613	159	21	,	,	PUNCT
ajst-29613	159	22	videos	video	NOUN
ajst-29613	159	23	,	,	PUNCT
ajst-29613	159	24	or	or	CCONJ
ajst-29613	159	25	audio	audio	NOUN
ajst-29613	159	26	.	.	PUNCT
ajst-29613	160	1	by	by	ADP
ajst-29613	160	2	introducing	introduce	VERB
ajst-29613	160	3	multimodal	multimodal	NOUN
ajst-29613	160	4	encoders	encoder	NOUN
ajst-29613	160	5	,	,	PUNCT
ajst-29613	160	6	this	this	DET
ajst-29613	160	7	information	information	NOUN
ajst-29613	160	8	can	can	AUX
ajst-29613	160	9	be	be	AUX
ajst-29613	160	10	converted	convert	VERB
ajst-29613	160	11	into	into	ADP
ajst-29613	160	12	embeddings	embedding	NOUN
ajst-29613	160	13	compatible	compatible	ADJ
ajst-29613	160	14	with	with	ADP
ajst-29613	160	15	the	the	DET
ajst-29613	160	16	generation	generation	NOUN
ajst-29613	160	17	model	model	NOUN
ajst-29613	160	18	,	,	PUNCT
ajst-29613	160	19	enabling	enable	VERB
ajst-29613	160	20	richer	rich	ADJ
ajst-29613	160	21	content	content	NOUN
ajst-29613	160	22	generation	generation	NOUN
ajst-29613	160	23	.	.	PUNCT
ajst-29613	161	1	232	232	NUM
ajst-29613	161	2	3.3.3	3.3.3	NUM
ajst-29613	161	3	.	.	PUNCT
ajst-29613	162	1	introduction	introduction	NOUN
ajst-29613	162	2	of	of	ADP
ajst-29613	162	3	enhanced	enhanced	ADJ
ajst-29613	162	4	generation	generation	NOUN
ajst-29613	162	5	techniques	technique	NOUN
ajst-29613	162	6	to	to	PART
ajst-29613	162	7	further	far	ADV
ajst-29613	162	8	improve	improve	VERB
ajst-29613	162	9	the	the	DET
ajst-29613	162	10	output	output	NOUN
ajst-29613	162	11	quality	quality	NOUN
ajst-29613	162	12	in	in	ADP
ajst-29613	162	13	the	the	DET
ajst-29613	162	14	generation	generation	NOUN
ajst-29613	162	15	phase	phase	NOUN
ajst-29613	162	16	,	,	PUNCT
ajst-29613	162	17	graphrag	graphrag	PROPN
ajst-29613	162	18	can	can	AUX
ajst-29613	162	19	leverage	leverage	VERB
ajst-29613	162	20	enhanced	enhanced	ADJ
ajst-29613	162	21	generation	generation	NOUN
ajst-29613	162	22	techniques	technique	NOUN
ajst-29613	162	23	to	to	PART
ajst-29613	162	24	optimize	optimize	VERB
ajst-29613	162	25	the	the	DET
ajst-29613	162	26	interaction	interaction	NOUN
ajst-29613	162	27	between	between	ADP
ajst-29613	162	28	the	the	DET
ajst-29613	162	29	query	query	NOUN
ajst-29613	162	30	and	and	CCONJ
ajst-29613	162	31	graph	graph	NOUN
ajst-29613	162	32	data	datum	NOUN
ajst-29613	162	33	while	while	SCONJ
ajst-29613	162	34	enriching	enrich	VERB
ajst-29613	162	35	the	the	DET
ajst-29613	162	36	generated	generate	VERB
ajst-29613	162	37	content	content	NOUN
ajst-29613	162	38	.	.	PUNCT
ajst-29613	163	1	these	these	DET
ajst-29613	163	2	techniques	technique	NOUN
ajst-29613	163	3	primarily	primarily	ADV
ajst-29613	163	4	include	include	VERB
ajst-29613	163	5	the	the	DET
ajst-29613	163	6	following	follow	VERB
ajst-29613	163	7	:	:	PUNCT
ajst-29613	163	8	generation	generation	NOUN
ajst-29613	163	9	optimization	optimization	NOUN
ajst-29613	163	10	via	via	ADP
ajst-29613	163	11	graph	graph	NOUN
ajst-29613	163	12	attention	attention	NOUN
ajst-29613	163	13	mechanism	mechanism	NOUN
ajst-29613	163	14	by	by	ADP
ajst-29613	163	15	incorporating	incorporate	VERB
ajst-29613	163	16	a	a	DET
ajst-29613	163	17	graph	graph	NOUN
ajst-29613	163	18	attention	attention	NOUN
ajst-29613	163	19	mechanism	mechanism	NOUN
ajst-29613	163	20	,	,	PUNCT
ajst-29613	163	21	the	the	DET
ajst-29613	163	22	generation	generation	NOUN
ajst-29613	163	23	model	model	NOUN
ajst-29613	163	24	can	can	AUX
ajst-29613	163	25	dynamically	dynamically	ADV
ajst-29613	163	26	focus	focus	VERB
ajst-29613	163	27	on	on	ADP
ajst-29613	163	28	key	key	ADJ
ajst-29613	163	29	nodes	node	NOUN
ajst-29613	163	30	and	and	CCONJ
ajst-29613	163	31	relationships	relationship	NOUN
ajst-29613	163	32	within	within	ADP
ajst-29613	163	33	the	the	DET
ajst-29613	163	34	retrieved	retrieve	VERB
ajst-29613	163	35	subgraph	subgraph	NOUN
ajst-29613	163	36	,	,	PUNCT
ajst-29613	163	37	improving	improve	VERB
ajst-29613	163	38	the	the	DET
ajst-29613	163	39	relevance	relevance	NOUN
ajst-29613	163	40	of	of	ADP
ajst-29613	163	41	the	the	DET
ajst-29613	163	42	generated	generate	VERB
ajst-29613	163	43	output	output	NOUN
ajst-29613	163	44	.	.	PUNCT
ajst-29613	164	1	for	for	ADP
ajst-29613	164	2	example	example	NOUN
ajst-29613	164	3	,	,	PUNCT
ajst-29613	164	4	in	in	ADP
ajst-29613	164	5	a	a	DET
ajst-29613	164	6	complex	complex	ADJ
ajst-29613	164	7	medical	medical	ADJ
ajst-29613	164	8	question	question	NOUN
ajst-29613	164	9	-	-	PUNCT
ajst-29613	164	10	answering	answer	VERB
ajst-29613	164	11	task	task	NOUN
ajst-29613	164	12	,	,	PUNCT
ajst-29613	164	13	the	the	DET
ajst-29613	164	14	attention	attention	NOUN
ajst-29613	164	15	mechanism	mechanism	NOUN
ajst-29613	164	16	can	can	AUX
ajst-29613	164	17	guide	guide	VERB
ajst-29613	164	18	the	the	DET
ajst-29613	164	19	model	model	NOUN
ajst-29613	164	20	to	to	PART
ajst-29613	164	21	focus	focus	VERB
ajst-29613	164	22	on	on	ADP
ajst-29613	164	23	the	the	DET
ajst-29613	164	24	core	core	NOUN
ajst-29613	164	25	connections	connection	NOUN
ajst-29613	164	26	between	between	ADP
ajst-29613	164	27	symptoms	symptom	NOUN
ajst-29613	164	28	and	and	CCONJ
ajst-29613	164	29	treatment	treatment	NOUN
ajst-29613	164	30	methods	method	NOUN
ajst-29613	164	31	.	.	PUNCT
ajst-29613	165	1	generation	generation	NOUN
ajst-29613	165	2	-	-	PUNCT
ajst-29613	165	3	guided	guide	VERB
ajst-29613	165	4	dynamic	dynamic	ADJ
ajst-29613	165	5	graph	graph	NOUN
ajst-29613	165	6	expansion	expansion	NOUN
ajst-29613	165	7	during	during	ADP
ajst-29613	165	8	the	the	DET
ajst-29613	165	9	generation	generation	NOUN
ajst-29613	165	10	process	process	NOUN
ajst-29613	165	11	,	,	PUNCT
ajst-29613	165	12	the	the	DET
ajst-29613	165	13	model	model	NOUN
ajst-29613	165	14	can	can	AUX
ajst-29613	165	15	dynamically	dynamically	ADV
ajst-29613	165	16	issue	issue	VERB
ajst-29613	165	17	supplementary	supplementary	ADJ
ajst-29613	165	18	queries	query	NOUN
ajst-29613	165	19	to	to	PART
ajst-29613	165	20	retrieve	retrieve	VERB
ajst-29613	165	21	additional	additional	ADJ
ajst-29613	165	22	relevant	relevant	ADJ
ajst-29613	165	23	graph	graph	NOUN
ajst-29613	165	24	data	datum	NOUN
ajst-29613	165	25	,	,	PUNCT
ajst-29613	165	26	enabling	enable	VERB
ajst-29613	165	27	continuous	continuous	ADJ
ajst-29613	165	28	content	content	NOUN
ajst-29613	165	29	optimization	optimization	NOUN
ajst-29613	165	30	.	.	PUNCT
ajst-29613	166	1	for	for	ADP
ajst-29613	166	2	instance	instance	NOUN
ajst-29613	166	3	,	,	PUNCT
ajst-29613	166	4	in	in	ADP
ajst-29613	166	5	technical	technical	ADJ
ajst-29613	166	6	document	document	NOUN
ajst-29613	166	7	generation	generation	NOUN
ajst-29613	166	8	,	,	PUNCT
ajst-29613	166	9	the	the	DET
ajst-29613	166	10	model	model	NOUN
ajst-29613	166	11	can	can	AUX
ajst-29613	166	12	perform	perform	VERB
ajst-29613	166	13	"	"	PUNCT
ajst-29613	166	14	supplementary	supplementary	ADJ
ajst-29613	166	15	retrieval	retrieval	NOUN
ajst-29613	166	16	"	"	PUNCT
ajst-29613	166	17	to	to	PART
ajst-29613	166	18	obtain	obtain	VERB
ajst-29613	166	19	the	the	DET
ajst-29613	166	20	latest	late	ADJ
ajst-29613	166	21	industry	industry	NOUN
ajst-29613	166	22	standards	standard	NOUN
ajst-29613	166	23	or	or	CCONJ
ajst-29613	166	24	research	research	NOUN
ajst-29613	166	25	advancements	advancement	NOUN
ajst-29613	166	26	.	.	PUNCT
ajst-29613	167	1	enhanced	enhance	VERB
ajst-29613	167	2	contrastive	contrastive	ADJ
ajst-29613	167	3	learning	learning	NOUN
ajst-29613	167	4	and	and	CCONJ
ajst-29613	167	5	content	content	NOUN
ajst-29613	167	6	calibration	calibration	NOUN
ajst-29613	167	7	contrastive	contrastive	ADJ
ajst-29613	167	8	learning	learning	NOUN
ajst-29613	167	9	allows	allow	VERB
ajst-29613	167	10	the	the	DET
ajst-29613	167	11	model	model	NOUN
ajst-29613	167	12	to	to	PART
ajst-29613	167	13	learn	learn	VERB
ajst-29613	167	14	to	to	PART
ajst-29613	167	15	distinguish	distinguish	VERB
ajst-29613	167	16	between	between	ADP
ajst-29613	167	17	high	high	ADJ
ajst-29613	167	18	-	-	PUNCT
ajst-29613	167	19	quality	quality	NOUN
ajst-29613	167	20	and	and	CCONJ
ajst-29613	167	21	low	low	ADJ
ajst-29613	167	22	-	-	PUNCT
ajst-29613	167	23	quality	quality	NOUN
ajst-29613	167	24	generated	generate	VERB
ajst-29613	167	25	content	content	NOUN
ajst-29613	167	26	,	,	PUNCT
ajst-29613	167	27	thereby	thereby	ADV
ajst-29613	167	28	improving	improve	VERB
ajst-29613	167	29	the	the	DET
ajst-29613	167	30	reliability	reliability	NOUN
ajst-29613	167	31	of	of	ADP
ajst-29613	167	32	the	the	DET
ajst-29613	167	33	output	output	NOUN
ajst-29613	167	34	.	.	PUNCT
ajst-29613	168	1	additionally	additionally	ADV
ajst-29613	168	2	,	,	PUNCT
ajst-29613	168	3	a	a	DET
ajst-29613	168	4	content	content	NOUN
ajst-29613	168	5	calibration	calibration	NOUN
ajst-29613	168	6	mechanism	mechanism	NOUN
ajst-29613	168	7	can	can	AUX
ajst-29613	168	8	be	be	AUX
ajst-29613	168	9	introduced	introduce	VERB
ajst-29613	168	10	to	to	PART
ajst-29613	168	11	align	align	VERB
ajst-29613	168	12	the	the	DET
ajst-29613	168	13	generated	generate	VERB
ajst-29613	168	14	content	content	NOUN
ajst-29613	168	15	with	with	ADP
ajst-29613	168	16	the	the	DET
ajst-29613	168	17	structured	structured	ADJ
ajst-29613	168	18	data	datum	NOUN
ajst-29613	168	19	in	in	ADP
ajst-29613	168	20	the	the	DET
ajst-29613	168	21	knowledge	knowledge	NOUN
ajst-29613	168	22	graph	graph	NOUN
ajst-29613	168	23	,	,	PUNCT
ajst-29613	168	24	ensuring	ensure	VERB
ajst-29613	168	25	accuracy	accuracy	NOUN
ajst-29613	168	26	and	and	CCONJ
ajst-29613	168	27	consistency	consistency	NOUN
ajst-29613	168	28	.	.	PUNCT
ajst-29613	169	1	stylized	stylize	VERB
ajst-29613	169	2	and	and	CCONJ
ajst-29613	169	3	personalized	personalized	ADJ
ajst-29613	169	4	generation	generation	NOUN
ajst-29613	169	5	depending	depend	VERB
ajst-29613	169	6	on	on	ADP
ajst-29613	169	7	user	user	NOUN
ajst-29613	169	8	needs	need	NOUN
ajst-29613	169	9	and	and	CCONJ
ajst-29613	169	10	task	task	NOUN
ajst-29613	169	11	contexts	contexts	PROPN
ajst-29613	169	12	,	,	PUNCT
ajst-29613	169	13	the	the	DET
ajst-29613	169	14	generated	generate	VERB
ajst-29613	169	15	content	content	NOUN
ajst-29613	169	16	can	can	AUX
ajst-29613	169	17	be	be	AUX
ajst-29613	169	18	stylistically	stylistically	ADV
ajst-29613	169	19	adjusted	adjust	VERB
ajst-29613	169	20	.	.	PUNCT
ajst-29613	170	1	for	for	ADP
ajst-29613	170	2	example	example	NOUN
ajst-29613	170	3	,	,	PUNCT
ajst-29613	170	4	in	in	ADP
ajst-29613	170	5	tasks	task	NOUN
ajst-29613	170	6	aimed	aim	VERB
ajst-29613	170	7	at	at	ADP
ajst-29613	170	8	general	general	ADJ
ajst-29613	170	9	users	user	NOUN
ajst-29613	170	10	,	,	PUNCT
ajst-29613	170	11	the	the	DET
ajst-29613	170	12	model	model	NOUN
ajst-29613	170	13	can	can	AUX
ajst-29613	170	14	adopt	adopt	VERB
ajst-29613	170	15	a	a	DET
ajst-29613	170	16	simple	simple	ADJ
ajst-29613	170	17	and	and	CCONJ
ajst-29613	170	18	easy	easy	ADJ
ajst-29613	170	19	-	-	PUNCT
ajst-29613	170	20	to	to	PART
ajst-29613	170	21	-	-	PUNCT
ajst-29613	170	22	understand	understand	VERB
ajst-29613	170	23	language	language	NOUN
ajst-29613	170	24	style	style	NOUN
ajst-29613	170	25	,	,	PUNCT
ajst-29613	170	26	whereas	whereas	SCONJ
ajst-29613	170	27	in	in	ADP
ajst-29613	170	28	academic	academic	ADJ
ajst-29613	170	29	scenarios	scenario	NOUN
ajst-29613	170	30	,	,	PUNCT
ajst-29613	170	31	it	it	PRON
ajst-29613	170	32	can	can	AUX
ajst-29613	170	33	generate	generate	VERB
ajst-29613	170	34	more	more	ADV
ajst-29613	170	35	specialized	specialized	ADJ
ajst-29613	170	36	and	and	CCONJ
ajst-29613	170	37	formal	formal	ADJ
ajst-29613	170	38	content	content	NOUN
ajst-29613	170	39	.	.	PUNCT
ajst-29613	171	1	4	4	X
ajst-29613	171	2	.	.	X
ajst-29613	171	3	optimization	optimization	NOUN
ajst-29613	171	4	4.1	4.1	NUM
ajst-29613	171	5	.	.	PUNCT
ajst-29613	172	1	query	query	NOUN
ajst-29613	172	2	expansion	expansion	NOUN
ajst-29613	172	3	since	since	SCONJ
ajst-29613	172	4	queries	query	NOUN
ajst-29613	172	5	are	be	AUX
ajst-29613	172	6	often	often	ADV
ajst-29613	172	7	short	short	ADJ
ajst-29613	172	8	and	and	CCONJ
ajst-29613	172	9	lack	lack	VERB
ajst-29613	172	10	sufficient	sufficient	ADJ
ajst-29613	172	11	information	information	NOUN
ajst-29613	172	12	,	,	PUNCT
ajst-29613	172	13	query	query	NOUN
ajst-29613	172	14	expansion	expansion	NOUN
ajst-29613	172	15	aims	aim	VERB
ajst-29613	172	16	to	to	PART
ajst-29613	172	17	improve	improve	VERB
ajst-29613	172	18	search	search	NOUN
ajst-29613	172	19	results	result	NOUN
ajst-29613	172	20	by	by	ADP
ajst-29613	172	21	supplementing	supplement	VERB
ajst-29613	172	22	or	or	CCONJ
ajst-29613	172	23	refining	refine	VERB
ajst-29613	172	24	relevant	relevant	ADJ
ajst-29613	172	25	terms	term	NOUN
ajst-29613	172	26	or	or	CCONJ
ajst-29613	172	27	concepts	concept	NOUN
ajst-29613	172	28	in	in	ADP
ajst-29613	172	29	the	the	DET
ajst-29613	172	30	original	original	ADJ
ajst-29613	172	31	query	query	NOUN
ajst-29613	172	32	.	.	PUNCT
ajst-29613	173	1	luo	luo	PROPN
ajst-29613	173	2	et	et	PROPN
ajst-29613	173	3	al	al	PROPN
ajst-29613	173	4	.	.	PUNCT
ajst-29613	174	1	[	[	X
ajst-29613	174	2	19	19	NUM
ajst-29613	174	3	]	]	PUNCT
ajst-29613	174	4	enhanced	enhanced	ADJ
ajst-29613	174	5	retrieval	retrieval	NOUN
ajst-29613	174	6	queries	query	NOUN
ajst-29613	174	7	by	by	ADP
ajst-29613	174	8	generating	generate	VERB
ajst-29613	174	9	relation	relation	NOUN
ajst-29613	174	10	paths	path	NOUN
ajst-29613	174	11	based	base	VERB
ajst-29613	174	12	on	on	ADP
ajst-29613	174	13	knowledge	knowledge	NOUN
ajst-29613	174	14	graphs	graph	NOUN
ajst-29613	174	15	(	(	PUNCT
ajst-29613	174	16	kgs	kgs	NOUN
ajst-29613	174	17	)	)	PUNCT
ajst-29613	174	18	.	.	PUNCT
ajst-29613	175	1	sparql[36	sparql[36	PROPN
ajst-29613	175	2	]	]	PUNCT
ajst-29613	175	3	is	be	AUX
ajst-29613	175	4	used	use	VERB
ajst-29613	175	5	to	to	PART
ajst-29613	175	6	retrieve	retrieve	VERB
ajst-29613	175	7	all	all	DET
ajst-29613	175	8	aliases	alias	NOUN
ajst-29613	175	9	of	of	ADP
ajst-29613	175	10	query	query	NOUN
ajst-29613	175	11	entities	entity	NOUN
ajst-29613	175	12	from	from	ADP
ajst-29613	175	13	wikidata	wikidata	NOUN
ajst-29613	175	14	,	,	PUNCT
ajst-29613	175	15	expanding	expand	VERB
ajst-29613	175	16	retrieval	retrieval	NOUN
ajst-29613	175	17	queries	query	NOUN
ajst-29613	175	18	to	to	PART
ajst-29613	175	19	capture	capture	VERB
ajst-29613	175	20	lexical	lexical	ADJ
ajst-29613	175	21	variations	variation	NOUN
ajst-29613	175	22	of	of	ADP
ajst-29613	175	23	the	the	DET
ajst-29613	175	24	same	same	ADJ
ajst-29613	175	25	entity	entity	NOUN
ajst-29613	175	26	.	.	PUNCT
ajst-29613	176	1	huang	huang	PROPN
ajst-29613	176	2	et	et	PROPN
ajst-29613	176	3	al	al	PROPN
ajst-29613	176	4	.	.	PUNCT
ajst-29613	177	1	(	(	PUNCT
ajst-29613	177	2	2023a)[37	2023a)[37	NOUN
ajst-29613	177	3	]	]	PUNCT
ajst-29613	177	4	proposed	propose	VERB
ajst-29613	177	5	a	a	DET
ajst-29613	177	6	consensus	consensus	NOUN
ajst-29613	177	7	-	-	PUNCT
ajst-29613	177	8	view	view	NOUN
ajst-29613	177	9	knowledge	knowledge	NOUN
ajst-29613	177	10	retrieval	retrieval	NOUN
ajst-29613	177	11	method	method	NOUN
ajst-29613	177	12	,	,	PUNCT
ajst-29613	177	13	which	which	PRON
ajst-29613	177	14	improves	improve	VERB
ajst-29613	177	15	retrieval	retrieval	NOUN
ajst-29613	177	16	performance	performance	NOUN
ajst-29613	177	17	and	and	CCONJ
ajst-29613	177	18	accuracy	accuracy	NOUN
ajst-29613	177	19	by	by	ADP
ajst-29613	177	20	discovering	discover	VERB
ajst-29613	177	21	semantically	semantically	ADV
ajst-29613	177	22	related	relate	VERB
ajst-29613	177	23	queries	query	NOUN
ajst-29613	177	24	and	and	CCONJ
ajst-29613	177	25	reweighting	reweighte	VERB
ajst-29613	177	26	original	original	ADJ
ajst-29613	177	27	query	query	NOUN
ajst-29613	177	28	terms	term	NOUN
ajst-29613	177	29	.	.	PUNCT
ajst-29613	178	1	hykge[17	hykge[17	PROPN
ajst-29613	178	2	]	]	X
ajst-29613	178	3	leverages	leverage	VERB
ajst-29613	178	4	large	large	ADJ
ajst-29613	178	5	models	model	NOUN
ajst-29613	178	6	to	to	PART
ajst-29613	178	7	generate	generate	VERB
ajst-29613	178	8	hypothetical	hypothetical	ADJ
ajst-29613	178	9	outputs	output	NOUN
ajst-29613	178	10	for	for	ADP
ajst-29613	178	11	the	the	DET
ajst-29613	178	12	query	query	NOUN
ajst-29613	178	13	and	and	CCONJ
ajst-29613	178	14	appends	append	VERB
ajst-29613	178	15	them	they	PRON
ajst-29613	178	16	to	to	ADP
ajst-29613	178	17	the	the	DET
ajst-29613	178	18	original	original	ADJ
ajst-29613	178	19	query	query	NOUN
ajst-29613	178	20	as	as	ADP
ajst-29613	178	21	input	input	NOUN
ajst-29613	178	22	for	for	ADP
ajst-29613	178	23	the	the	DET
ajst-29613	178	24	retriever	retriever	NOUN
ajst-29613	178	25	.	.	PUNCT
ajst-29613	179	1	goldenretriever	goldenretriever	PROPN
ajst-29613	179	2	(	(	PUNCT
ajst-29613	179	3	an	an	DET
ajst-29613	179	4	et	et	NOUN
ajst-29613	179	5	al	al	PROPN
ajst-29613	179	6	.	.	PROPN
ajst-29613	179	7	,	,	PUNCT
ajst-29613	179	8	2024)[2	2024)[2	NUM
ajst-29613	179	9	]	]	PUNCT
ajst-29613	179	10	identifies	identify	VERB
ajst-29613	179	11	technical	technical	ADJ
ajst-29613	179	12	terms	term	NOUN
ajst-29613	179	13	or	or	CCONJ
ajst-29613	179	14	jargon	jargon	NOUN
ajst-29613	179	15	in	in	ADP
ajst-29613	179	16	the	the	DET
ajst-29613	179	17	query	query	NOUN
ajst-29613	179	18	and	and	CCONJ
ajst-29613	179	19	retrieves	retrieve	VERB
ajst-29613	179	20	their	their	PRON
ajst-29613	179	21	explanations	explanation	NOUN
ajst-29613	179	22	to	to	PART
ajst-29613	179	23	supplement	supplement	VERB
ajst-29613	179	24	the	the	DET
ajst-29613	179	25	query	query	NOUN
ajst-29613	179	26	content	content	NOUN
ajst-29613	179	27	.	.	PUNCT
ajst-29613	180	1	4.2	4.2	NUM
ajst-29613	180	2	.	.	PUNCT
ajst-29613	181	1	query	query	NOUN
ajst-29613	181	2	decomposition	decomposition	NOUN
ajst-29613	181	3	query	query	NOUN
ajst-29613	181	4	decomposition	decomposition	NOUN
ajst-29613	181	5	techniques	technique	NOUN
ajst-29613	181	6	aim	aim	VERB
ajst-29613	181	7	to	to	PART
ajst-29613	181	8	reduce	reduce	VERB
ajst-29613	181	9	the	the	DET
ajst-29613	181	10	complexity	complexity	NOUN
ajst-29613	181	11	and	and	CCONJ
ajst-29613	181	12	ambiguity	ambiguity	NOUN
ajst-29613	181	13	of	of	ADP
ajst-29613	181	14	user	user	NOUN
ajst-29613	181	15	queries	query	NOUN
ajst-29613	181	16	by	by	ADP
ajst-29613	181	17	breaking	break	VERB
ajst-29613	181	18	them	they	PRON
ajst-29613	181	19	down	down	ADP
ajst-29613	181	20	into	into	ADP
ajst-29613	181	21	smaller	small	ADJ
ajst-29613	181	22	,	,	PUNCT
ajst-29613	181	23	more	more	ADV
ajst-29613	181	24	specific	specific	ADJ
ajst-29613	181	25	subqueries	subquerie	NOUN
ajst-29613	181	26	.	.	PUNCT
ajst-29613	182	1	each	each	DET
ajst-29613	182	2	subquery	subquery	NOUN
ajst-29613	182	3	typically	typically	ADV
ajst-29613	182	4	focuses	focus	VERB
ajst-29613	182	5	on	on	ADP
ajst-29613	182	6	a	a	DET
ajst-29613	182	7	particular	particular	ADJ
ajst-29613	182	8	aspect	aspect	NOUN
ajst-29613	182	9	or	or	CCONJ
ajst-29613	182	10	component	component	NOUN
ajst-29613	182	11	of	of	ADP
ajst-29613	182	12	the	the	DET
ajst-29613	182	13	original	original	ADJ
ajst-29613	182	14	query	query	NOUN
ajst-29613	182	15	,	,	PUNCT
ajst-29613	182	16	thereby	thereby	ADV
ajst-29613	182	17	capturing	capture	VERB
ajst-29613	182	18	user	user	NOUN
ajst-29613	182	19	intent	intent	NOUN
ajst-29613	182	20	more	more	ADV
ajst-29613	182	21	precisely	precisely	ADV
ajst-29613	182	22	and	and	CCONJ
ajst-29613	182	23	improving	improve	VERB
ajst-29613	182	24	retrieval	retrieval	NOUN
ajst-29613	182	25	performance	performance	NOUN
ajst-29613	182	26	.	.	PUNCT
ajst-29613	183	1	for	for	ADP
ajst-29613	183	2	example	example	NOUN
ajst-29613	183	3	,	,	PUNCT
ajst-29613	183	4	complexlogical	complexlogical	NOUN
ajst-29613	183	5	-	-	PUNCT
ajst-29613	183	6	reasoning[39	reasoning[39	NOUN
ajst-29613	183	7	]	]	PUNCT
ajst-29613	183	8	and	and	CCONJ
ajst-29613	183	9	kg	kg	PROPN
ajst-29613	183	10	-	-	PROPN
ajst-29613	183	11	gpt[40	gpt[40	NOUN
ajst-29613	183	12	]	]	PUNCT
ajst-29613	183	13	proposed	propose	VERB
ajst-29613	183	14	methods	method	NOUN
ajst-29613	183	15	to	to	PART
ajst-29613	183	16	decompose	decompose	VERB
ajst-29613	183	17	complex	complex	ADJ
ajst-29613	183	18	primary	primary	ADJ
ajst-29613	183	19	questions	question	NOUN
ajst-29613	183	20	into	into	ADP
ajst-29613	183	21	multiple	multiple	ADJ
ajst-29613	183	22	clauses	clause	NOUN
ajst-29613	183	23	,	,	PUNCT
ajst-29613	183	24	with	with	ADP
ajst-29613	183	25	each	each	DET
ajst-29613	183	26	clause	clause	NOUN
ajst-29613	183	27	corresponding	correspond	VERB
ajst-29613	183	28	to	to	ADP
ajst-29613	183	29	an	an	DET
ajst-29613	183	30	independent	independent	ADJ
ajst-29613	183	31	semantic	semantic	ADJ
ajst-29613	183	32	relation	relation	NOUN
ajst-29613	183	33	.	.	PUNCT
ajst-29613	184	1	relevant	relevant	ADJ
ajst-29613	184	2	knowledge	knowledge	NOUN
ajst-29613	184	3	graph	graph	NOUN
ajst-29613	184	4	triples	triple	NOUN
ajst-29613	184	5	are	be	AUX
ajst-29613	184	6	then	then	ADV
ajst-29613	184	7	retrieved	retrieve	VERB
ajst-29613	184	8	for	for	ADP
ajst-29613	184	9	each	each	DET
ajst-29613	184	10	clause	clause	NOUN
ajst-29613	184	11	separately	separately	ADV
ajst-29613	184	12	.	.	PUNCT
ajst-29613	185	1	this	this	DET
ajst-29613	185	2	process	process	NOUN
ajst-29613	185	3	not	not	PART
ajst-29613	185	4	only	only	ADV
ajst-29613	185	5	enhances	enhance	VERB
ajst-29613	185	6	the	the	DET
ajst-29613	185	7	specificity	specificity	NOUN
ajst-29613	185	8	and	and	CCONJ
ajst-29613	185	9	efficiency	efficiency	NOUN
ajst-29613	185	10	of	of	ADP
ajst-29613	185	11	retrieval	retrieval	NOUN
ajst-29613	185	12	but	but	CCONJ
ajst-29613	185	13	also	also	ADV
ajst-29613	185	14	lays	lay	VERB
ajst-29613	185	15	the	the	DET
ajst-29613	185	16	foundation	foundation	NOUN
ajst-29613	185	17	for	for	ADP
ajst-29613	185	18	subsequent	subsequent	ADJ
ajst-29613	185	19	knowledge	knowledge	NOUN
ajst-29613	185	20	integration	integration	NOUN
ajst-29613	185	21	and	and	CCONJ
ajst-29613	185	22	reasoning	reasoning	NOUN
ajst-29613	185	23	,	,	PUNCT
ajst-29613	185	24	demonstrating	demonstrate	VERB
ajst-29613	185	25	significant	significant	ADJ
ajst-29613	185	26	potential	potential	NOUN
ajst-29613	185	27	in	in	ADP
ajst-29613	185	28	handling	handle	VERB
ajst-29613	185	29	complex	complex	ADJ
ajst-29613	185	30	question	question	NOUN
ajst-29613	185	31	-	-	PUNCT
ajst-29613	185	32	answering	answer	VERB
ajst-29613	185	33	tasks	task	NOUN
ajst-29613	185	34	or	or	CCONJ
ajst-29613	185	35	multi	multi	ADJ
ajst-29613	185	36	-	-	ADJ
ajst-29613	185	37	hop	hop	ADJ
ajst-29613	185	38	reasoning	reasoning	NOUN
ajst-29613	185	39	scenarios	scenario	NOUN
ajst-29613	185	40	.	.	PUNCT
ajst-29613	186	1	figure	figure	NOUN
ajst-29613	186	2	5	5	NUM
ajst-29613	186	3	.	.	PUNCT
ajst-29613	186	4	query	query	PROPN
ajst-29613	186	5	rewrite	rewrite	VERB
ajst-29613	186	6	4.3	4.3	NUM
ajst-29613	186	7	.	.	PUNCT
ajst-29613	187	1	re	re	VERB
ajst-29613	187	2	-	-	ADJ
ajst-29613	187	3	ranking	ranking	ADJ
ajst-29613	187	4	module	module	NOUN
ajst-29613	187	5	knowledge	knowledge	NOUN
ajst-29613	187	6	pruning	pruning	NOUN
ajst-29613	187	7	techniques	technique	NOUN
ajst-29613	187	8	aim	aim	VERB
ajst-29613	187	9	to	to	PART
ajst-29613	187	10	refine	refine	VERB
ajst-29613	187	11	retrieval	retrieval	NOUN
ajst-29613	187	12	results	result	NOUN
ajst-29613	187	13	by	by	ADP
ajst-29613	187	14	filtering	filter	VERB
ajst-29613	187	15	out	out	ADP
ajst-29613	187	16	irrelevant	irrelevant	ADJ
ajst-29613	187	17	or	or	CCONJ
ajst-29613	187	18	redundant	redundant	ADJ
ajst-29613	187	19	information	information	NOUN
ajst-29613	187	20	,	,	PUNCT
ajst-29613	187	21	thereby	thereby	ADV
ajst-29613	187	22	enhancing	enhance	VERB
ajst-29613	187	23	system	system	NOUN
ajst-29613	187	24	performance	performance	NOUN
ajst-29613	187	25	.	.	PUNCT
ajst-29613	188	1	existing	exist	VERB
ajst-29613	188	2	knowledge	knowledge	NOUN
ajst-29613	188	3	pruning	prune	VERB
ajst-29613	188	4	methods	method	NOUN
ajst-29613	188	5	can	can	AUX
ajst-29613	188	6	be	be	AUX
ajst-29613	188	7	broadly	broadly	ADV
ajst-29613	188	8	categorized	categorize	VERB
ajst-29613	188	9	into	into	ADP
ajst-29613	188	10	two	two	NUM
ajst-29613	188	11	types	type	NOUN
ajst-29613	188	12	:	:	PUNCT
ajst-29613	188	13	re	re	VERB
ajst-29613	188	14	-	-	ADJ
ajst-29613	188	15	ranking	ranking	ADJ
ajst-29613	188	16	methods	method	NOUN
ajst-29613	188	17	and	and	CCONJ
ajst-29613	188	18	large	large	ADJ
ajst-29613	188	19	language	language	NOUN
ajst-29613	188	20	model	model	NOUN
ajst-29613	188	21	(	(	PUNCT
ajst-29613	188	22	llm)-based	llm)-based	ADJ
ajst-29613	188	23	methods	method	NOUN
ajst-29613	188	24	.	.	PUNCT
ajst-29613	189	1	re	re	VERB
ajst-29613	189	2	-	-	ADJ
ajst-29613	189	3	ranking	ranking	ADJ
ajst-29613	189	4	methods	method	NOUN
ajst-29613	189	5	improve	improve	VERB
ajst-29613	189	6	the	the	DET
ajst-29613	189	7	relevance	relevance	NOUN
ajst-29613	189	8	and	and	CCONJ
ajst-29613	189	9	quality	quality	NOUN
ajst-29613	189	10	of	of	ADP
ajst-29613	189	11	retrieval	retrieval	NOUN
ajst-29613	189	12	results	result	NOUN
ajst-29613	189	13	by	by	ADP
ajst-29613	189	14	applying	apply	VERB
ajst-29613	189	15	customized	customize	VERB
ajst-29613	189	16	metrics	metric	NOUN
ajst-29613	189	17	to	to	PART
ajst-29613	189	18	reorder	reorder	VERB
ajst-29613	189	19	or	or	CCONJ
ajst-29613	189	20	prioritize	prioritize	VERB
ajst-29613	189	21	the	the	DET
ajst-29613	189	22	retrieved	retrieve	VERB
ajst-29613	189	23	information	information	NOUN
ajst-29613	189	24	.	.	PUNCT
ajst-29613	190	1	one	one	NUM
ajst-29613	190	2	common	common	ADJ
ajst-29613	190	3	re	re	ADJ
ajst-29613	190	4	-	-	ADJ
ajst-29613	190	5	ranking	ranking	ADJ
ajst-29613	190	6	approach	approach	NOUN
ajst-29613	190	7	involves	involve	VERB
ajst-29613	190	8	using	use	VERB
ajst-29613	190	9	more	more	ADJ
ajst-29613	190	10	powerful	powerful	ADJ
ajst-29613	190	11	models	model	NOUN
ajst-29613	190	12	to	to	PART
ajst-29613	190	13	optimize	optimize	VERB
ajst-29613	190	14	the	the	DET
ajst-29613	190	15	ranking	ranking	NOUN
ajst-29613	190	16	of	of	ADP
ajst-29613	190	17	retrieval	retrieval	NOUN
ajst-29613	190	18	results	result	NOUN
ajst-29613	190	19	.	.	PUNCT
ajst-29613	191	1	for	for	ADP
ajst-29613	191	2	instance	instance	NOUN
ajst-29613	191	3	,	,	PUNCT
ajst-29613	191	4	li	li	PROPN
ajst-29613	191	5	et	et	PROPN
ajst-29613	191	6	al	al	PROPN
ajst-29613	191	7	.	.	PROPN
ajst-29613	192	1	(	(	PUNCT
ajst-29613	192	2	2023)[41	2023)[41	X
ajst-29613	192	3	]	]	PUNCT
ajst-29613	192	4	proposed	propose	VERB
ajst-29613	192	5	a	a	DET
ajst-29613	192	6	method	method	NOUN
ajst-29613	192	7	that	that	PRON
ajst-29613	192	8	concatenates	concatenate	VERB
ajst-29613	192	9	each	each	PRON
ajst-29613	192	10	retrieved	retrieve	VERB
ajst-29613	192	11	triple	triple	ADV
ajst-29613	192	12	with	with	ADP
ajst-29613	192	13	a	a	DET
ajst-29613	192	14	question	question	NOUN
ajst-29613	192	15	-	-	PUNCT
ajst-29613	192	16	option	option	NOUN
ajst-29613	192	17	pair	pair	NOUN
ajst-29613	192	18	and	and	CCONJ
ajst-29613	192	19	uses	use	VERB
ajst-29613	192	20	a	a	DET
ajst-29613	192	21	pre	pre	ADJ
ajst-29613	192	22	-	-	ADJ
ajst-29613	192	23	trained	train	VERB
ajst-29613	192	24	cross	cross	NOUN
ajst-29613	192	25	-	-	NOUN
ajst-29613	192	26	encoder	encoder	NOUN
ajst-29613	192	27	to	to	PART
ajst-29613	192	28	re	re	VERB
ajst-29613	192	29	-	-	VERB
ajst-29613	192	30	rank	rank	VERB
ajst-29613	192	31	the	the	DET
ajst-29613	192	32	triples	triple	NOUN
ajst-29613	192	33	,	,	PUNCT
ajst-29613	192	34	improving	improve	VERB
ajst-29613	192	35	the	the	DET
ajst-29613	192	36	accuracy	accuracy	NOUN
ajst-29613	192	37	of	of	ADP
ajst-29613	192	38	relevance	relevance	NOUN
ajst-29613	192	39	evaluation	evaluation	NOUN
ajst-29613	192	40	.	.	PUNCT
ajst-29613	193	1	another	another	DET
ajst-29613	193	2	233	233	NUM
ajst-29613	193	3	method	method	NOUN
ajst-29613	193	4	,	,	PUNCT
ajst-29613	193	5	hykge[17	hykge[17	X
ajst-29613	193	6	]	]	X
ajst-29613	193	7	,	,	PUNCT
ajst-29613	193	8	employs	employ	VERB
ajst-29613	193	9	flagembedding	flagembedde	VERB
ajst-29613	193	10	for	for	ADP
ajst-29613	193	11	efficient	efficient	ADJ
ajst-29613	193	12	text	text	NOUN
ajst-29613	193	13	encoding	encoding	NOUN
ajst-29613	193	14	,	,	PUNCT
ajst-29613	193	15	which	which	PRON
ajst-29613	193	16	reorders	reorder	VERB
ajst-29613	193	17	the	the	DET
ajst-29613	193	18	top	top	ADJ
ajst-29613	193	19	k	k	PROPN
ajst-29613	193	20	documents	document	NOUN
ajst-29613	193	21	returned	return	VERB
ajst-29613	193	22	by	by	ADP
ajst-29613	193	23	the	the	DET
ajst-29613	193	24	embedding	embed	VERB
ajst-29613	193	25	model	model	NOUN
ajst-29613	193	26	"	"	PUNCT
ajst-29613	193	27	bge_reranker_large	bge_reranker_large	NOUN
ajst-29613	193	28	,	,	PUNCT
ajst-29613	193	29	"	"	PUNCT
ajst-29613	193	30	significantly	significantly	ADV
ajst-29613	193	31	refining	refine	VERB
ajst-29613	193	32	the	the	DET
ajst-29613	193	33	retrieval	retrieval	NOUN
ajst-29613	193	34	results	result	NOUN
ajst-29613	193	35	.	.	PUNCT
ajst-29613	194	1	these	these	DET
ajst-29613	194	2	re	re	VERB
ajst-29613	194	3	-	-	ADJ
ajst-29613	194	4	ranking	ranking	ADJ
ajst-29613	194	5	methods	method	NOUN
ajst-29613	194	6	not	not	PART
ajst-29613	194	7	only	only	ADV
ajst-29613	194	8	enhance	enhance	VERB
ajst-29613	194	9	the	the	DET
ajst-29613	194	10	alignment	alignment	NOUN
ajst-29613	194	11	between	between	ADP
ajst-29613	194	12	queries	query	NOUN
ajst-29613	194	13	and	and	CCONJ
ajst-29613	194	14	retrieval	retrieval	NOUN
ajst-29613	194	15	results	result	NOUN
ajst-29613	194	16	but	but	CCONJ
ajst-29613	194	17	also	also	ADV
ajst-29613	194	18	provide	provide	VERB
ajst-29613	194	19	higherquality	higherquality	NOUN
ajst-29613	194	20	input	input	NOUN
ajst-29613	194	21	for	for	ADP
ajst-29613	194	22	complex	complex	ADJ
ajst-29613	194	23	reasoning	reasoning	NOUN
ajst-29613	194	24	tasks	task	NOUN
ajst-29613	194	25	,	,	PUNCT
ajst-29613	194	26	such	such	ADJ
ajst-29613	194	27	as	as	ADP
ajst-29613	194	28	knowledge	knowledge	NOUN
ajst-29613	194	29	graph	graph	NOUN
ajst-29613	194	30	-	-	PUNCT
ajst-29613	194	31	based	base	VERB
ajst-29613	194	32	question	question	NOUN
ajst-29613	194	33	answering	answering	NOUN
ajst-29613	194	34	and	and	CCONJ
ajst-29613	194	35	multimodal	multimodal	NOUN
ajst-29613	194	36	retrieval	retrieval	NOUN
ajst-29613	194	37	.	.	PUNCT
ajst-29613	195	1	this	this	DET
ajst-29613	195	2	highlights	highlight	VERB
ajst-29613	195	3	the	the	DET
ajst-29613	195	4	growing	grow	VERB
ajst-29613	195	5	importance	importance	NOUN
ajst-29613	195	6	of	of	ADP
ajst-29613	195	7	pruning	prune	VERB
ajst-29613	195	8	techniques	technique	NOUN
ajst-29613	195	9	in	in	ADP
ajst-29613	195	10	future	future	ADJ
ajst-29613	195	11	information	information	NOUN
ajst-29613	195	12	retrieval	retrieval	NOUN
ajst-29613	195	13	and	and	CCONJ
ajst-29613	195	14	knowledge	knowledge	NOUN
ajst-29613	195	15	management	management	NOUN
ajst-29613	195	16	systems	system	NOUN
ajst-29613	195	17	,	,	PUNCT
ajst-29613	195	18	particularly	particularly	ADV
ajst-29613	195	19	as	as	SCONJ
ajst-29613	195	20	model	model	NOUN
ajst-29613	195	21	capabilities	capability	NOUN
ajst-29613	195	22	continue	continue	VERB
ajst-29613	195	23	to	to	PART
ajst-29613	195	24	advance	advance	VERB
ajst-29613	195	25	.	.	PUNCT
ajst-29613	196	1	figure	figure	VERB
ajst-29613	196	2	6	6	NUM
ajst-29613	196	3	.	.	PUNCT
ajst-29613	196	4	framework	framework	NOUN
ajst-29613	196	5	of	of	ADP
ajst-29613	196	6	re	re	NOUN
ajst-29613	196	7	-	-	NOUN
ajst-29613	196	8	rank	rank	ADJ
ajst-29613	196	9	5	5	NUM
ajst-29613	196	10	.	.	PUNCT
ajst-29613	197	1	future	future	ADJ
ajst-29613	197	2	prospects	prospect	NOUN
ajst-29613	197	3	in	in	ADP
ajst-29613	197	4	the	the	DET
ajst-29613	197	5	future	future	NOUN
ajst-29613	197	6	,	,	PUNCT
ajst-29613	197	7	constructing	construct	VERB
ajst-29613	197	8	knowledge	knowledge	NOUN
ajst-29613	197	9	graphs	graph	NOUN
ajst-29613	197	10	that	that	PRON
ajst-29613	197	11	incorporate	incorporate	VERB
ajst-29613	197	12	multimodal	multimodal	NOUN
ajst-29613	197	13	information	information	NOUN
ajst-29613	197	14	will	will	AUX
ajst-29613	197	15	become	become	VERB
ajst-29613	197	16	a	a	DET
ajst-29613	197	17	significant	significant	ADJ
ajst-29613	197	18	development	development	NOUN
ajst-29613	197	19	direction	direction	NOUN
ajst-29613	197	20	.	.	PUNCT
ajst-29613	198	1	for	for	ADP
ajst-29613	198	2	instance	instance	NOUN
ajst-29613	198	3	,	,	PUNCT
ajst-29613	198	4	integrating	integrate	VERB
ajst-29613	198	5	information	information	NOUN
ajst-29613	198	6	from	from	ADP
ajst-29613	198	7	text	text	NOUN
ajst-29613	198	8	,	,	PUNCT
ajst-29613	198	9	images	image	NOUN
ajst-29613	198	10	,	,	PUNCT
ajst-29613	198	11	and	and	CCONJ
ajst-29613	198	12	videos	video	NOUN
ajst-29613	198	13	into	into	ADP
ajst-29613	198	14	a	a	DET
ajst-29613	198	15	unified	unified	ADJ
ajst-29613	198	16	knowledge	knowledge	NOUN
ajst-29613	198	17	graph	graph	NOUN
ajst-29613	198	18	can	can	AUX
ajst-29613	198	19	help	help	VERB
ajst-29613	198	20	provide	provide	VERB
ajst-29613	198	21	a	a	DET
ajst-29613	198	22	more	more	ADV
ajst-29613	198	23	comprehensive	comprehensive	ADJ
ajst-29613	198	24	description	description	NOUN
ajst-29613	198	25	of	of	ADP
ajst-29613	198	26	entities	entity	NOUN
ajst-29613	198	27	and	and	CCONJ
ajst-29613	198	28	their	their	PRON
ajst-29613	198	29	relationships	relationship	NOUN
ajst-29613	198	30	.	.	PUNCT
ajst-29613	199	1	this	this	PRON
ajst-29613	199	2	not	not	PART
ajst-29613	199	3	only	only	ADV
ajst-29613	199	4	expands	expand	VERB
ajst-29613	199	5	the	the	DET
ajst-29613	199	6	application	application	NOUN
ajst-29613	199	7	scenarios	scenario	NOUN
ajst-29613	199	8	of	of	ADP
ajst-29613	199	9	knowledge	knowledge	NOUN
ajst-29613	199	10	graphs	graph	NOUN
ajst-29613	199	11	,	,	PUNCT
ajst-29613	199	12	such	such	ADJ
ajst-29613	199	13	as	as	ADP
ajst-29613	199	14	visual	visual	ADJ
ajst-29613	199	15	question	question	NOUN
ajst-29613	199	16	answering	answering	NOUN
ajst-29613	199	17	and	and	CCONJ
ajst-29613	199	18	video	video	NOUN
ajst-29613	199	19	content	content	NOUN
ajst-29613	199	20	analysis	analysis	NOUN
ajst-29613	199	21	,	,	PUNCT
ajst-29613	199	22	but	but	CCONJ
ajst-29613	199	23	also	also	ADV
ajst-29613	199	24	provides	provide	VERB
ajst-29613	199	25	diverse	diverse	ADJ
ajst-29613	199	26	retrieval	retrieval	NOUN
ajst-29613	199	27	content	content	NOUN
ajst-29613	199	28	for	for	ADP
ajst-29613	199	29	rag	rag	NOUN
ajst-29613	199	30	models	model	NOUN
ajst-29613	199	31	,	,	PUNCT
ajst-29613	199	32	thereby	thereby	ADV
ajst-29613	199	33	enhancing	enhance	VERB
ajst-29613	199	34	their	their	PRON
ajst-29613	199	35	reasoning	reasoning	NOUN
ajst-29613	199	36	capabilities	capability	NOUN
ajst-29613	199	37	.	.	PUNCT
ajst-29613	200	1	achieving	achieve	VERB
ajst-29613	200	2	this	this	DET
ajst-29613	200	3	goal	goal	NOUN
ajst-29613	200	4	requires	require	VERB
ajst-29613	200	5	research	research	NOUN
ajst-29613	200	6	into	into	ADP
ajst-29613	200	7	designing	design	VERB
ajst-29613	200	8	effective	effective	ADJ
ajst-29613	200	9	multimodal	multimodal	NOUN
ajst-29613	200	10	feature	feature	NOUN
ajst-29613	200	11	representations	representation	NOUN
ajst-29613	200	12	,	,	PUNCT
ajst-29613	200	13	cross	cross	ADJ
ajst-29613	200	14	-	-	ADJ
ajst-29613	200	15	modal	modal	ADJ
ajst-29613	200	16	alignment	alignment	NOUN
ajst-29613	200	17	,	,	PUNCT
ajst-29613	200	18	and	and	CCONJ
ajst-29613	200	19	multimodal	multimodal	ADJ
ajst-29613	200	20	fusion	fusion	NOUN
ajst-29613	200	21	strategies	strategy	NOUN
ajst-29613	200	22	.	.	PUNCT
ajst-29613	201	1	additionally	additionally	ADV
ajst-29613	201	2	,	,	PUNCT
ajst-29613	201	3	the	the	DET
ajst-29613	201	4	construction	construction	NOUN
ajst-29613	201	5	of	of	ADP
ajst-29613	201	6	multimodal	multimodal	ADJ
ajst-29613	201	7	knowledge	knowledge	NOUN
ajst-29613	201	8	graphs	graph	NOUN
ajst-29613	201	9	must	must	AUX
ajst-29613	201	10	address	address	VERB
ajst-29613	201	11	technical	technical	ADJ
ajst-29613	201	12	challenges	challenge	NOUN
ajst-29613	201	13	such	such	ADJ
ajst-29613	201	14	as	as	ADP
ajst-29613	201	15	heterogeneous	heterogeneous	ADJ
ajst-29613	201	16	data	datum	NOUN
ajst-29613	201	17	processing	processing	NOUN
ajst-29613	201	18	and	and	CCONJ
ajst-29613	201	19	crossmodal	crossmodal	ADJ
ajst-29613	201	20	association	association	NOUN
ajst-29613	201	21	learning	learning	NOUN
ajst-29613	201	22	.	.	PUNCT
ajst-29613	202	1	the	the	DET
ajst-29613	202	2	static	static	ADJ
ajst-29613	202	3	nature	nature	NOUN
ajst-29613	202	4	of	of	ADP
ajst-29613	202	5	knowledge	knowledge	NOUN
ajst-29613	202	6	graphs	graph	NOUN
ajst-29613	202	7	limits	limit	VERB
ajst-29613	202	8	their	their	PRON
ajst-29613	202	9	applications	application	NOUN
ajst-29613	202	10	in	in	ADP
ajst-29613	202	11	rapidly	rapidly	ADV
ajst-29613	202	12	changing	change	VERB
ajst-29613	202	13	domains	domain	NOUN
ajst-29613	202	14	,	,	PUNCT
ajst-29613	202	15	such	such	ADJ
ajst-29613	202	16	as	as	ADP
ajst-29613	202	17	news	news	NOUN
ajst-29613	202	18	,	,	PUNCT
ajst-29613	202	19	social	social	ADJ
ajst-29613	202	20	media	medium	NOUN
ajst-29613	202	21	analysis	analysis	NOUN
ajst-29613	202	22	,	,	PUNCT
ajst-29613	202	23	and	and	CCONJ
ajst-29613	202	24	financial	financial	ADJ
ajst-29613	202	25	market	market	NOUN
ajst-29613	202	26	forecasting	forecasting	NOUN
ajst-29613	202	27	.	.	PUNCT
ajst-29613	203	1	therefore	therefore	ADV
ajst-29613	203	2	,	,	PUNCT
ajst-29613	203	3	enabling	enable	VERB
ajst-29613	203	4	dynamic	dynamic	ADJ
ajst-29613	203	5	updates	update	NOUN
ajst-29613	203	6	to	to	ADP
ajst-29613	203	7	knowledge	knowledge	NOUN
ajst-29613	203	8	graphs	graph	NOUN
ajst-29613	203	9	is	be	AUX
ajst-29613	203	10	key	key	ADJ
ajst-29613	203	11	to	to	ADP
ajst-29613	203	12	enhancing	enhance	VERB
ajst-29613	203	13	the	the	DET
ajst-29613	203	14	reasoning	reasoning	NOUN
ajst-29613	203	15	capabilities	capability	NOUN
ajst-29613	203	16	of	of	ADP
ajst-29613	203	17	large	large	ADJ
ajst-29613	203	18	models	model	NOUN
ajst-29613	203	19	.	.	PUNCT
ajst-29613	204	1	future	future	ADJ
ajst-29613	204	2	research	research	NOUN
ajst-29613	204	3	should	should	AUX
ajst-29613	204	4	focus	focus	VERB
ajst-29613	204	5	on	on	ADP
ajst-29613	204	6	developing	develop	VERB
ajst-29613	204	7	efficient	efficient	ADJ
ajst-29613	204	8	methods	method	NOUN
ajst-29613	204	9	for	for	ADP
ajst-29613	204	10	dynamic	dynamic	ADJ
ajst-29613	204	11	knowledge	knowledge	NOUN
ajst-29613	204	12	graph	graph	NOUN
ajst-29613	204	13	updates	update	NOUN
ajst-29613	204	14	,	,	PUNCT
ajst-29613	204	15	including	include	VERB
ajst-29613	204	16	automated	automate	VERB
ajst-29613	204	17	data	data	NOUN
ajst-29613	204	18	collection	collection	NOUN
ajst-29613	204	19	,	,	PUNCT
ajst-29613	204	20	knowledge	knowledge	NOUN
ajst-29613	204	21	extraction	extraction	NOUN
ajst-29613	204	22	,	,	PUNCT
ajst-29613	204	23	and	and	CCONJ
ajst-29613	204	24	update	update	NOUN
ajst-29613	204	25	mechanisms	mechanism	NOUN
ajst-29613	204	26	.	.	PUNCT
ajst-29613	205	1	at	at	ADP
ajst-29613	205	2	the	the	DET
ajst-29613	205	3	same	same	ADJ
ajst-29613	205	4	time	time	NOUN
ajst-29613	205	5	,	,	PUNCT
ajst-29613	205	6	dynamic	dynamic	ADJ
ajst-29613	205	7	updates	update	NOUN
ajst-29613	205	8	must	must	AUX
ajst-29613	205	9	balance	balance	VERB
ajst-29613	205	10	data	datum	NOUN
ajst-29613	205	11	timeliness	timeliness	NOUN
ajst-29613	205	12	with	with	ADP
ajst-29613	205	13	the	the	DET
ajst-29613	205	14	structural	structural	ADJ
ajst-29613	205	15	stability	stability	NOUN
ajst-29613	205	16	of	of	ADP
ajst-29613	205	17	the	the	DET
ajst-29613	205	18	knowledge	knowledge	NOUN
ajst-29613	205	19	graph	graph	NOUN
ajst-29613	205	20	to	to	PART
ajst-29613	205	21	ensure	ensure	VERB
ajst-29613	205	22	that	that	SCONJ
ajst-29613	205	23	newly	newly	ADV
ajst-29613	205	24	added	add	VERB
ajst-29613	205	25	knowledge	knowledge	NOUN
ajst-29613	205	26	does	do	AUX
ajst-29613	205	27	not	not	PART
ajst-29613	205	28	introduce	introduce	VERB
ajst-29613	205	29	redundancy	redundancy	NOUN
ajst-29613	205	30	or	or	CCONJ
ajst-29613	205	31	contradictions	contradiction	NOUN
ajst-29613	205	32	.	.	PUNCT
ajst-29613	206	1	additionally	additionally	ADV
ajst-29613	206	2	,	,	PUNCT
ajst-29613	206	3	to	to	PART
ajst-29613	206	4	support	support	VERB
ajst-29613	206	5	real	real	ADJ
ajst-29613	206	6	-	-	PUNCT
ajst-29613	206	7	time	time	NOUN
ajst-29613	206	8	reasoning	reasoning	NOUN
ajst-29613	206	9	,	,	PUNCT
ajst-29613	206	10	efficient	efficient	ADJ
ajst-29613	206	11	knowledge	knowledge	NOUN
ajst-29613	206	12	retrieval	retrieval	NOUN
ajst-29613	206	13	and	and	CCONJ
ajst-29613	206	14	reasoning	reason	VERB
ajst-29613	206	15	algorithms	algorithm	NOUN
ajst-29613	206	16	need	need	VERB
ajst-29613	206	17	to	to	PART
ajst-29613	206	18	be	be	AUX
ajst-29613	206	19	designed	design	VERB
ajst-29613	206	20	to	to	PART
ajst-29613	206	21	enable	enable	VERB
ajst-29613	206	22	models	model	NOUN
ajst-29613	206	23	to	to	PART
ajst-29613	206	24	quickly	quickly	ADV
ajst-29613	206	25	adapt	adapt	VERB
ajst-29613	206	26	to	to	ADP
ajst-29613	206	27	updated	update	VERB
ajst-29613	206	28	knowledge	knowledge	NOUN
ajst-29613	206	29	.	.	PUNCT
ajst-29613	207	1	this	this	PRON
ajst-29613	207	2	necessitates	necessitate	VERB
ajst-29613	207	3	the	the	DET
ajst-29613	207	4	integration	integration	NOUN
ajst-29613	207	5	of	of	ADP
ajst-29613	207	6	stream	stream	NOUN
ajst-29613	207	7	data	datum	NOUN
ajst-29613	207	8	processing	processing	NOUN
ajst-29613	207	9	techniques	technique	NOUN
ajst-29613	207	10	,	,	PUNCT
ajst-29613	207	11	online	online	ADJ
ajst-29613	207	12	learning	learning	NOUN
ajst-29613	207	13	algorithms	algorithm	NOUN
ajst-29613	207	14	,	,	PUNCT
ajst-29613	207	15	and	and	CCONJ
ajst-29613	207	16	optimization	optimization	NOUN
ajst-29613	207	17	strategies	strategy	NOUN
ajst-29613	207	18	for	for	ADP
ajst-29613	207	19	real	real	ADJ
ajst-29613	207	20	-	-	PUNCT
ajst-29613	207	21	time	time	NOUN
ajst-29613	207	22	reasoning	reasoning	NOUN
ajst-29613	207	23	in	in	ADP
ajst-29613	207	24	knowledge	knowledge	NOUN
ajst-29613	207	25	graph	graph	NOUN
ajst-29613	207	26	construction	construction	NOUN
ajst-29613	207	27	.	.	PUNCT
ajst-29613	208	1	6	6	X
ajst-29613	208	2	.	.	X
ajst-29613	208	3	conclusion	conclusion	NOUN
ajst-29613	208	4	kgs	kgs	PROPN
ajst-29613	208	5	-	-	PUNCT
ajst-29613	208	6	rag	rag	NOUN
ajst-29613	208	7	combines	combine	VERB
ajst-29613	208	8	graph	graph	NOUN
ajst-29613	208	9	-	-	PUNCT
ajst-29613	208	10	structured	structure	VERB
ajst-29613	208	11	data	datum	NOUN
ajst-29613	208	12	with	with	ADP
ajst-29613	208	13	the	the	DET
ajst-29613	208	14	rag	rag	NOUN
ajst-29613	208	15	framework	framework	NOUN
ajst-29613	208	16	,	,	PUNCT
ajst-29613	208	17	providing	provide	VERB
ajst-29613	208	18	an	an	DET
ajst-29613	208	19	efficient	efficient	ADJ
ajst-29613	208	20	and	and	CCONJ
ajst-29613	208	21	accurate	accurate	ADJ
ajst-29613	208	22	solution	solution	NOUN
ajst-29613	208	23	for	for	ADP
ajst-29613	208	24	knowledge	knowledge	NOUN
ajst-29613	208	25	-	-	PUNCT
ajst-29613	208	26	intensive	intensive	ADJ
ajst-29613	208	27	tasks	task	NOUN
ajst-29613	208	28	.	.	PUNCT
ajst-29613	209	1	compared	compare	VERB
ajst-29613	209	2	to	to	ADP
ajst-29613	209	3	traditional	traditional	ADJ
ajst-29613	209	4	retrieval	retrieval	NOUN
ajst-29613	209	5	methods	method	NOUN
ajst-29613	209	6	,	,	PUNCT
ajst-29613	209	7	kgs	kgs	NOUN
ajst-29613	209	8	-	-	PUNCT
ajst-29613	209	9	rag	rag	NOUN
ajst-29613	209	10	leverages	leverage	VERB
ajst-29613	209	11	the	the	DET
ajst-29613	209	12	rich	rich	ADJ
ajst-29613	209	13	semantic	semantic	ADJ
ajst-29613	209	14	and	and	CCONJ
ajst-29613	209	15	relational	relational	ADJ
ajst-29613	209	16	information	information	NOUN
ajst-29613	209	17	embedded	embed	VERB
ajst-29613	209	18	in	in	ADP
ajst-29613	209	19	graph	graph	NOUN
ajst-29613	209	20	structures	structure	NOUN
ajst-29613	209	21	,	,	PUNCT
ajst-29613	209	22	enabling	enable	VERB
ajst-29613	209	23	deeper	deep	ADJ
ajst-29613	209	24	query	query	NOUN
ajst-29613	209	25	analysis	analysis	NOUN
ajst-29613	209	26	and	and	CCONJ
ajst-29613	209	27	reasoning	reasoning	NOUN
ajst-29613	209	28	.	.	PUNCT
ajst-29613	210	1	this	this	DET
ajst-29613	210	2	framework	framework	NOUN
ajst-29613	210	3	integrates	integrate	VERB
ajst-29613	210	4	the	the	DET
ajst-29613	210	5	strengths	strength	NOUN
ajst-29613	210	6	of	of	ADP
ajst-29613	210	7	knowledge	knowledge	NOUN
ajst-29613	210	8	graphs	graph	NOUN
ajst-29613	210	9	,	,	PUNCT
ajst-29613	210	10	graph	graph	NOUN
ajst-29613	210	11	neural	neural	ADJ
ajst-29613	210	12	networks	network	NOUN
ajst-29613	210	13	(	(	PUNCT
ajst-29613	210	14	gnns	gnns	NOUN
ajst-29613	210	15	)	)	PUNCT
ajst-29613	210	16	,	,	PUNCT
ajst-29613	210	17	and	and	CCONJ
ajst-29613	210	18	large	large	ADJ
ajst-29613	210	19	language	language	NOUN
ajst-29613	210	20	models	model	NOUN
ajst-29613	210	21	(	(	PUNCT
ajst-29613	210	22	llms	llm	NOUN
ajst-29613	210	23	)	)	PUNCT
ajst-29613	210	24	,	,	PUNCT
ajst-29613	210	25	demonstrating	demonstrate	VERB
ajst-29613	210	26	strong	strong	ADJ
ajst-29613	210	27	adaptability	adaptability	NOUN
ajst-29613	210	28	and	and	CCONJ
ajst-29613	210	29	scalability	scalability	NOUN
ajst-29613	210	30	.	.	PUNCT
ajst-29613	211	1	in	in	ADP
ajst-29613	211	2	practical	practical	ADJ
ajst-29613	211	3	applications	application	NOUN
ajst-29613	211	4	,	,	PUNCT
ajst-29613	211	5	kgs	kgs	NOUN
ajst-29613	211	6	-	-	PUNCT
ajst-29613	211	7	rag	rag	NOUN
ajst-29613	211	8	has	have	AUX
ajst-29613	211	9	achieved	achieve	VERB
ajst-29613	211	10	remarkable	remarkable	ADJ
ajst-29613	211	11	progress	progress	NOUN
ajst-29613	211	12	.	.	PUNCT
ajst-29613	212	1	for	for	ADP
ajst-29613	212	2	instance	instance	NOUN
ajst-29613	212	3	,	,	PUNCT
ajst-29613	212	4	graph	graph	NOUN
ajst-29613	212	5	-	-	PUNCT
ajst-29613	212	6	based	base	VERB
ajst-29613	212	7	query	query	NOUN
ajst-29613	212	8	expansion	expansion	NOUN
ajst-29613	212	9	supplements	supplement	NOUN
ajst-29613	212	10	relevant	relevant	ADJ
ajst-29613	212	11	semantic	semantic	ADJ
ajst-29613	212	12	information	information	NOUN
ajst-29613	212	13	to	to	PART
ajst-29613	212	14	address	address	VERB
ajst-29613	212	15	issues	issue	NOUN
ajst-29613	212	16	of	of	ADP
ajst-29613	212	17	query	query	NOUN
ajst-29613	212	18	brevity	brevity	NOUN
ajst-29613	212	19	and	and	CCONJ
ajst-29613	212	20	ambiguity	ambiguity	NOUN
ajst-29613	212	21	.	.	PUNCT
ajst-29613	213	1	path	path	NOUN
ajst-29613	213	2	retrieval	retrieval	NOUN
ajst-29613	213	3	algorithms	algorithm	NOUN
ajst-29613	213	4	identify	identify	VERB
ajst-29613	213	5	the	the	DET
ajst-29613	213	6	most	most	ADV
ajst-29613	213	7	relevant	relevant	ADJ
ajst-29613	213	8	subgraphs	subgraph	NOUN
ajst-29613	213	9	or	or	CCONJ
ajst-29613	213	10	relational	relational	ADJ
ajst-29613	213	11	paths	path	NOUN
ajst-29613	213	12	,	,	PUNCT
ajst-29613	213	13	laying	lay	VERB
ajst-29613	213	14	a	a	DET
ajst-29613	213	15	solid	solid	ADJ
ajst-29613	213	16	foundation	foundation	NOUN
ajst-29613	213	17	for	for	ADP
ajst-29613	213	18	multi	multi	ADJ
ajst-29613	213	19	-	-	ADJ
ajst-29613	213	20	hop	hop	ADJ
ajst-29613	213	21	reasoning	reasoning	NOUN
ajst-29613	213	22	.	.	PUNCT
ajst-29613	214	1	knowledge	knowledge	NOUN
ajst-29613	214	2	pruning	prune	VERB
ajst-29613	214	3	effectively	effectively	ADV
ajst-29613	214	4	filters	filter	VERB
ajst-29613	214	5	out	out	ADP
ajst-29613	214	6	redundant	redundant	ADJ
ajst-29613	214	7	or	or	CCONJ
ajst-29613	214	8	irrelevant	irrelevant	ADJ
ajst-29613	214	9	information	information	NOUN
ajst-29613	214	10	,	,	PUNCT
ajst-29613	214	11	significantly	significantly	ADV
ajst-29613	214	12	improving	improve	VERB
ajst-29613	214	13	result	result	NOUN
ajst-29613	214	14	accuracy	accuracy	NOUN
ajst-29613	214	15	.	.	PUNCT
ajst-29613	215	1	additionally	additionally	ADV
ajst-29613	215	2	,	,	PUNCT
ajst-29613	215	3	generative	generative	ADJ
ajst-29613	215	4	reasoning	reasoning	NOUN
ajst-29613	215	5	leverages	leverage	NOUN
ajst-29613	215	6	retrieved	retrieve	VERB
ajst-29613	215	7	information	information	NOUN
ajst-29613	215	8	to	to	PART
ajst-29613	215	9	perform	perform	VERB
ajst-29613	215	10	complex	complex	ADJ
ajst-29613	215	11	questionanswering	questionanswering	NOUN
ajst-29613	215	12	and	and	CCONJ
ajst-29613	215	13	knowledge	knowledge	NOUN
ajst-29613	215	14	generation	generation	NOUN
ajst-29613	215	15	tasks	task	NOUN
ajst-29613	215	16	through	through	ADP
ajst-29613	215	17	powerful	powerful	ADJ
ajst-29613	215	18	language	language	NOUN
ajst-29613	215	19	generation	generation	NOUN
ajst-29613	215	20	capabilities	capability	NOUN
ajst-29613	215	21	.	.	PUNCT
ajst-29613	216	1	these	these	DET
ajst-29613	216	2	methods	method	NOUN
ajst-29613	216	3	work	work	VERB
ajst-29613	216	4	in	in	ADP
ajst-29613	216	5	tandem	tandem	NOUN
ajst-29613	216	6	to	to	PART
ajst-29613	216	7	greatly	greatly	ADV
ajst-29613	216	8	enhance	enhance	VERB
ajst-29613	216	9	the	the	DET
ajst-29613	216	10	ability	ability	NOUN
ajst-29613	216	11	to	to	PART
ajst-29613	216	12	acquire	acquire	VERB
ajst-29613	216	13	knowledge	knowledge	NOUN
ajst-29613	216	14	and	and	CCONJ
ajst-29613	216	15	solve	solve	VERB
ajst-29613	216	16	complex	complex	ADJ
ajst-29613	216	17	problems	problem	NOUN
ajst-29613	216	18	.	.	PUNCT
ajst-29613	217	1	despite	despite	SCONJ
ajst-29613	217	2	its	its	PRON
ajst-29613	217	3	significant	significant	ADJ
ajst-29613	217	4	advancements	advancement	NOUN
ajst-29613	217	5	,	,	PUNCT
ajst-29613	217	6	kgs	kgs	NOUN
ajst-29613	217	7	-	-	PUNCT
ajst-29613	217	8	rag	rag	NOUN
ajst-29613	217	9	faces	face	VERB
ajst-29613	217	10	several	several	ADJ
ajst-29613	217	11	challenges	challenge	NOUN
ajst-29613	217	12	and	and	CCONJ
ajst-29613	217	13	directions	direction	NOUN
ajst-29613	217	14	for	for	ADP
ajst-29613	217	15	future	future	ADJ
ajst-29613	217	16	research	research	NOUN
ajst-29613	217	17	.	.	PUNCT
ajst-29613	218	1	first	first	ADV
ajst-29613	218	2	,	,	PUNCT
ajst-29613	218	3	improving	improve	VERB
ajst-29613	218	4	the	the	DET
ajst-29613	218	5	efficiency	efficiency	NOUN
ajst-29613	218	6	and	and	CCONJ
ajst-29613	218	7	robustness	robustness	NOUN
ajst-29613	218	8	of	of	ADP
ajst-29613	218	9	retrieval	retrieval	NOUN
ajst-29613	218	10	and	and	CCONJ
ajst-29613	218	11	generation	generation	NOUN
ajst-29613	218	12	processes	process	NOUN
ajst-29613	218	13	,	,	PUNCT
ajst-29613	218	14	particularly	particularly	ADV
ajst-29613	218	15	when	when	SCONJ
ajst-29613	218	16	dealing	deal	VERB
ajst-29613	218	17	with	with	ADP
ajst-29613	218	18	largescale	largescale	NOUN
ajst-29613	218	19	graph	graph	NOUN
ajst-29613	218	20	-	-	PUNCT
ajst-29613	218	21	structured	structure	VERB
ajst-29613	218	22	data	datum	NOUN
ajst-29613	218	23	,	,	PUNCT
ajst-29613	218	24	remains	remain	VERB
ajst-29613	218	25	a	a	DET
ajst-29613	218	26	pressing	press	VERB
ajst-29613	218	27	issue	issue	NOUN
ajst-29613	218	28	.	.	PUNCT
ajst-29613	219	1	second	second	ADJ
ajst-29613	219	2	,	,	PUNCT
ajst-29613	219	3	the	the	DET
ajst-29613	219	4	dynamic	dynamic	ADJ
ajst-29613	219	5	and	and	CCONJ
ajst-29613	219	6	uncertain	uncertain	ADJ
ajst-29613	219	7	nature	nature	NOUN
ajst-29613	219	8	of	of	ADP
ajst-29613	219	9	graph	graph	NOUN
ajst-29613	219	10	data	datum	NOUN
ajst-29613	219	11	imposes	impose	VERB
ajst-29613	219	12	higher	high	ADJ
ajst-29613	219	13	demands	demand	NOUN
ajst-29613	219	14	for	for	ADP
ajst-29613	219	15	real	real	ADJ
ajst-29613	219	16	-	-	PUNCT
ajst-29613	219	17	time	time	NOUN
ajst-29613	219	18	retrieval	retrieval	NOUN
ajst-29613	219	19	and	and	CCONJ
ajst-29613	219	20	generation	generation	NOUN
ajst-29613	219	21	,	,	PUNCT
ajst-29613	219	22	requiring	require	VERB
ajst-29613	219	23	more	more	ADJ
ajst-29613	219	24	adaptive	adaptive	ADJ
ajst-29613	219	25	models	model	NOUN
ajst-29613	219	26	.	.	PUNCT
ajst-29613	220	1	additionally	additionally	ADV
ajst-29613	220	2	,	,	PUNCT
ajst-29613	220	3	balancing	balance	VERB
ajst-29613	220	4	the	the	DET
ajst-29613	220	5	roles	role	NOUN
ajst-29613	220	6	of	of	ADP
ajst-29613	220	7	retrieval	retrieval	NOUN
ajst-29613	220	8	,	,	PUNCT
ajst-29613	220	9	pruning	prune	VERB
ajst-29613	220	10	,	,	PUNCT
ajst-29613	220	11	and	and	CCONJ
ajst-29613	220	12	generation	generation	NOUN
ajst-29613	220	13	across	across	ADP
ajst-29613	220	14	different	different	ADJ
ajst-29613	220	15	tasks	task	NOUN
ajst-29613	220	16	to	to	PART
ajst-29613	220	17	optimize	optimize	VERB
ajst-29613	220	18	overall	overall	ADJ
ajst-29613	220	19	performance	performance	NOUN
ajst-29613	220	20	warrants	warrant	NOUN
ajst-29613	220	21	further	further	ADJ
ajst-29613	220	22	exploration	exploration	NOUN
ajst-29613	220	23	.	.	PUNCT
ajst-29613	221	1	lastly	lastly	ADV
ajst-29613	221	2	,	,	PUNCT
ajst-29613	221	3	the	the	DET
ajst-29613	221	4	application	application	NOUN
ajst-29613	221	5	of	of	ADP
ajst-29613	221	6	kgs	kgs	NOUN
ajst-29613	221	7	-	-	PUNCT
ajst-29613	221	8	rag	rag	NOUN
ajst-29613	221	9	in	in	ADP
ajst-29613	221	10	domain	domain	NOUN
ajst-29613	221	11	234	234	NUM
ajst-29613	221	12	specific	specific	ADJ
ajst-29613	221	13	tasks	task	NOUN
ajst-29613	221	14	,	,	PUNCT
ajst-29613	221	15	such	such	ADJ
ajst-29613	221	16	as	as	ADP
ajst-29613	221	17	medical	medical	ADJ
ajst-29613	221	18	diagnosis	diagnosis	NOUN
ajst-29613	221	19	,	,	PUNCT
ajst-29613	221	20	legal	legal	ADJ
ajst-29613	221	21	reasoning	reasoning	NOUN
ajst-29613	221	22	,	,	PUNCT
ajst-29613	221	23	and	and	CCONJ
ajst-29613	221	24	financial	financial	ADJ
ajst-29613	221	25	analysis	analysis	NOUN
ajst-29613	221	26	,	,	PUNCT
ajst-29613	221	27	requires	require	VERB
ajst-29613	221	28	further	further	ADJ
ajst-29613	221	29	research	research	NOUN
ajst-29613	221	30	to	to	AUX
ajst-29613	221	31	better	well	ADV
ajst-29613	221	32	address	address	VERB
ajst-29613	221	33	diverse	diverse	ADJ
ajst-29613	221	34	real	real	ADJ
ajst-29613	221	35	-	-	PUNCT
ajst-29613	221	36	world	world	NOUN
ajst-29613	221	37	needs	need	NOUN
ajst-29613	221	38	.	.	PUNCT
ajst-29613	222	1	references	reference	NOUN
ajst-29613	222	2	[	[	X
ajst-29613	222	3	1	1	X
ajst-29613	222	4	]	]	PUNCT
ajst-29613	222	5	achiam	achiam	PROPN
ajst-29613	222	6	j	j	PROPN
ajst-29613	222	7	,	,	PUNCT
ajst-29613	222	8	adler	adler	PROPN
ajst-29613	222	9	s	s	PROPN
ajst-29613	222	10	,	,	PUNCT
ajst-29613	222	11	agarwal	agarwal	PROPN
ajst-29613	222	12	s	s	PROPN
ajst-29613	222	13	,	,	PUNCT
ajst-29613	222	14	et	et	PROPN
ajst-29613	222	15	al	al	PROPN
ajst-29613	222	16	.	.	PROPN
ajst-29613	222	17	gpt-4	gpt-4	PART
ajst-29613	223	1	technical	technical	ADJ
ajst-29613	223	2	report[j	report[j	NOUN
ajst-29613	223	3	]	]	PUNCT
ajst-29613	223	4	.	.	PUNCT
ajst-29613	224	1	arxiv	arxiv	PROPN
ajst-29613	224	2	preprint	preprint	VERB
ajst-29613	224	3	arxiv:2303.08774	arxiv:2303.08774	NOUN
ajst-29613	224	4	,	,	PUNCT
ajst-29613	224	5	2023	2023	NUM
ajst-29613	224	6	.	.	PUNCT
ajst-29613	225	1	[	[	X
ajst-29613	225	2	2	2	NUM
ajst-29613	225	3	]	]	PUNCT
ajst-29613	225	4	dubey	dubey	NOUN
ajst-29613	225	5	a	a	PRON
ajst-29613	225	6	,	,	PUNCT
ajst-29613	225	7	jauhri	jauhri	PROPN
ajst-29613	225	8	a	a	PROPN
ajst-29613	225	9	,	,	PUNCT
ajst-29613	225	10	pandey	pandey	PROPN
ajst-29613	225	11	a	a	PROPN
ajst-29613	225	12	,	,	PUNCT
ajst-29613	225	13	et	et	PROPN
ajst-29613	225	14	al	al	PROPN
ajst-29613	225	15	.	.	PUNCT
ajst-29613	226	1	the	the	DET
ajst-29613	226	2	llama	llama	ADJ
ajst-29613	226	3	3	3	NUM
ajst-29613	226	4	herd	herd	NOUN
ajst-29613	226	5	of	of	ADP
ajst-29613	226	6	models[j	models[j	PROPN
ajst-29613	226	7	]	]	PUNCT
ajst-29613	226	8	.	.	PUNCT
ajst-29613	227	1	arxiv	arxiv	PROPN
ajst-29613	227	2	preprint	preprint	NOUN
ajst-29613	227	3	arxiv:2407.21783	arxiv:2407.21783	NOUN
ajst-29613	227	4	,	,	PUNCT
ajst-29613	227	5	2024	2024	NUM
ajst-29613	227	6	.	.	PUNCT
ajst-29613	228	1	[	[	X
ajst-29613	228	2	3	3	NUM
ajst-29613	228	3	]	]	PUNCT
ajst-29613	228	4	team	team	NOUN
ajst-29613	228	5	g	g	PROPN
ajst-29613	228	6	,	,	PUNCT
ajst-29613	228	7	anil	anil	PROPN
ajst-29613	228	8	r	r	PROPN
ajst-29613	228	9	,	,	PUNCT
ajst-29613	228	10	borgeaud	borgeaud	NOUN
ajst-29613	228	11	s	s	PROPN
ajst-29613	228	12	,	,	PUNCT
ajst-29613	228	13	et	et	PROPN
ajst-29613	228	14	al	al	PROPN
ajst-29613	228	15	.	.	PROPN
ajst-29613	228	16	gemini	gemini	PROPN
ajst-29613	228	17	:	:	PUNCT
ajst-29613	228	18	a	a	DET
ajst-29613	228	19	family	family	NOUN
ajst-29613	228	20	of	of	ADP
ajst-29613	228	21	highly	highly	ADV
ajst-29613	228	22	capable	capable	ADJ
ajst-29613	228	23	multimodal	multimodal	NOUN
ajst-29613	228	24	models[j	models[j	PROPN
ajst-29613	228	25	]	]	PUNCT
ajst-29613	228	26	.	.	PUNCT
ajst-29613	229	1	arxiv	arxiv	PROPN
ajst-29613	229	2	preprint	preprint	PROPN
ajst-29613	229	3	arxiv:2312.11805	arxiv:2312.11805	PROPN
ajst-29613	229	4	,	,	PUNCT
ajst-29613	229	5	2023	2023	NUM
ajst-29613	229	6	.	.	PUNCT
ajst-29613	230	1	[	[	X
ajst-29613	230	2	4	4	X
ajst-29613	230	3	]	]	X
ajst-29613	230	4	kandpal	kandpal	PROPN
ajst-29613	230	5	n	n	CCONJ
ajst-29613	230	6	,	,	PUNCT
ajst-29613	230	7	deng	deng	PROPN
ajst-29613	230	8	h	h	PROPN
ajst-29613	230	9	,	,	PUNCT
ajst-29613	230	10	roberts	roberts	PROPN
ajst-29613	230	11	a	a	PROPN
ajst-29613	230	12	,	,	PUNCT
ajst-29613	230	13	et	et	PROPN
ajst-29613	230	14	al	al	PROPN
ajst-29613	230	15	.	.	PUNCT
ajst-29613	231	1	large	large	ADJ
ajst-29613	231	2	language	language	NOUN
ajst-29613	231	3	models	model	NOUN
ajst-29613	231	4	struggle	struggle	VERB
ajst-29613	231	5	to	to	PART
ajst-29613	231	6	learn	learn	VERB
ajst-29613	231	7	long	long	ADJ
ajst-29613	231	8	-	-	PUNCT
ajst-29613	231	9	tail	tail	NOUN
ajst-29613	231	10	knowledge[c]//international	knowledge[c]//international	PROPN
ajst-29613	231	11	conference	conference	NOUN
ajst-29613	231	12	on	on	ADP
ajst-29613	231	13	machine	machine	NOUN
ajst-29613	231	14	learning	learning	NOUN
ajst-29613	231	15	.	.	PUNCT
ajst-29613	232	1	pmlr	pmlr	NOUN
ajst-29613	232	2	,	,	PUNCT
ajst-29613	232	3	2023	2023	NUM
ajst-29613	232	4	:	:	PUNCT
ajst-29613	232	5	15696	15696	NUM
ajst-29613	232	6	-	-	SYM
ajst-29613	232	7	15707	15707	NUM
ajst-29613	233	1	[	[	X
ajst-29613	233	2	5	5	NUM
ajst-29613	233	3	]	]	X
ajst-29613	233	4	zhang	zhang	PROPN
ajst-29613	233	5	y	y	PROPN
ajst-29613	233	6	,	,	PUNCT
ajst-29613	233	7	li	li	PROPN
ajst-29613	233	8	y	y	PROPN
ajst-29613	233	9	,	,	PUNCT
ajst-29613	233	10	cui	cui	PROPN
ajst-29613	233	11	l	l	PROPN
ajst-29613	233	12	,	,	PUNCT
ajst-29613	233	13	et	et	PROPN
ajst-29613	233	14	al	al	PROPN
ajst-29613	233	15	.	.	PUNCT
ajst-29613	233	16	siren	siren	PROPN
ajst-29613	233	17	's	's	PART
ajst-29613	233	18	song	song	NOUN
ajst-29613	233	19	in	in	ADP
ajst-29613	233	20	the	the	DET
ajst-29613	233	21	ai	ai	PROPN
ajst-29613	233	22	ocean	ocean	NOUN
ajst-29613	233	23	:	:	PUNCT
ajst-29613	233	24	a	a	DET
ajst-29613	233	25	survey	survey	NOUN
ajst-29613	233	26	on	on	ADP
ajst-29613	233	27	hallucination	hallucination	NOUN
ajst-29613	233	28	in	in	ADP
ajst-29613	233	29	large	large	ADJ
ajst-29613	233	30	language	language	NOUN
ajst-29613	233	31	models[j	models[j	PROPN
ajst-29613	233	32	]	]	PUNCT
ajst-29613	233	33	.	.	PUNCT
ajst-29613	234	1	arxiv	arxiv	PROPN
ajst-29613	234	2	preprint	preprint	PROPN
ajst-29613	234	3	arxiv:2309.01219	arxiv:2309.01219	ADJ
ajst-29613	234	4	,	,	PUNCT
ajst-29613	234	5	2023	2023	NUM
ajst-29613	234	6	.	.	PUNCT
ajst-29613	235	1	[	[	X
ajst-29613	235	2	6	6	NUM
ajst-29613	235	3	]	]	X
ajst-29613	235	4	kandpal	kandpal	PROPN
ajst-29613	235	5	n	n	CCONJ
ajst-29613	235	6	,	,	PUNCT
ajst-29613	235	7	deng	deng	PROPN
ajst-29613	235	8	h	h	PROPN
ajst-29613	235	9	,	,	PUNCT
ajst-29613	235	10	roberts	roberts	PROPN
ajst-29613	235	11	a	a	PROPN
ajst-29613	235	12	,	,	PUNCT
ajst-29613	235	13	et	et	PROPN
ajst-29613	235	14	al	al	PROPN
ajst-29613	235	15	.	.	PUNCT
ajst-29613	236	1	large	large	ADJ
ajst-29613	236	2	language	language	NOUN
ajst-29613	236	3	models	model	NOUN
ajst-29613	236	4	struggle	struggle	VERB
ajst-29613	236	5	to	to	PART
ajst-29613	236	6	learn	learn	VERB
ajst-29613	236	7	long	long	ADJ
ajst-29613	236	8	-	-	PUNCT
ajst-29613	236	9	tail	tail	NOUN
ajst-29613	236	10	knowledge[c]//international	knowledge[c]//international	PROPN
ajst-29613	236	11	conference	conference	NOUN
ajst-29613	236	12	on	on	ADP
ajst-29613	236	13	machine	machine	NOUN
ajst-29613	236	14	learning	learning	NOUN
ajst-29613	236	15	.	.	PUNCT
ajst-29613	237	1	pmlr	pmlr	NOUN
ajst-29613	237	2	,	,	PUNCT
ajst-29613	237	3	2023	2023	NUM
ajst-29613	237	4	:	:	PUNCT
ajst-29613	237	5	15696	15696	NUM
ajst-29613	237	6	-	-	SYM
ajst-29613	237	7	15707	15707	NUM
ajst-29613	237	8	.	.	PUNCT
ajst-29613	238	1	[	[	X
ajst-29613	238	2	7	7	NUM
ajst-29613	238	3	]	]	X
ajst-29613	238	4	cheng	cheng	PROPN
ajst-29613	238	5	x	x	PROPN
ajst-29613	238	6	,	,	PUNCT
ajst-29613	238	7	luo	luo	PROPN
ajst-29613	238	8	d	d	PROPN
ajst-29613	238	9	,	,	PUNCT
ajst-29613	238	10	chen	chen	PROPN
ajst-29613	238	11	x	x	PROPN
ajst-29613	238	12	,	,	PUNCT
ajst-29613	238	13	et	et	PROPN
ajst-29613	238	14	al	al	PROPN
ajst-29613	238	15	.	.	PROPN
ajst-29613	238	16	lift	lift	VERB
ajst-29613	238	17	yourself	yourself	PRON
ajst-29613	238	18	up	up	ADP
ajst-29613	238	19	:	:	PUNCT
ajst-29613	238	20	retrievalaugmented	retrievalaugmente	VERB
ajst-29613	238	21	text	text	NOUN
ajst-29613	238	22	generation	generation	NOUN
ajst-29613	238	23	with	with	ADP
ajst-29613	238	24	self	self	NOUN
ajst-29613	238	25	-	-	PUNCT
ajst-29613	238	26	memory[j	memory[j	NOUN
ajst-29613	238	27	]	]	PUNCT
ajst-29613	238	28	.	.	PUNCT
ajst-29613	239	1	advances	advance	NOUN
ajst-29613	239	2	in	in	ADP
ajst-29613	239	3	neural	neural	ADJ
ajst-29613	239	4	information	information	NOUN
ajst-29613	239	5	processing	processing	NOUN
ajst-29613	239	6	systems	system	NOUN
ajst-29613	239	7	,	,	PUNCT
ajst-29613	239	8	2024	2024	NUM
ajst-29613	239	9	,	,	PUNCT
ajst-29613	239	10	36	36	NUM
ajst-29613	239	11	.	.	PUNCT
ajst-29613	240	1	[	[	X
ajst-29613	240	2	8	8	NUM
ajst-29613	240	3	]	]	X
ajst-29613	240	4	liu	liu	PROPN
ajst-29613	240	5	n	n	PROPN
ajst-29613	240	6	f	f	PROPN
ajst-29613	240	7	,	,	PUNCT
ajst-29613	240	8	lin	lin	PROPN
ajst-29613	240	9	k	k	PROPN
ajst-29613	240	10	,	,	PUNCT
ajst-29613	240	11	hewitt	hewitt	PROPN
ajst-29613	240	12	j	j	PROPN
ajst-29613	240	13	,	,	PUNCT
ajst-29613	240	14	et	et	PROPN
ajst-29613	240	15	al	al	PROPN
ajst-29613	240	16	.	.	PROPN
ajst-29613	240	17	lost	lose	VERB
ajst-29613	240	18	in	in	ADP
ajst-29613	240	19	the	the	DET
ajst-29613	240	20	middle	middle	NOUN
ajst-29613	240	21	:	:	PUNCT
ajst-29613	240	22	how	how	SCONJ
ajst-29613	240	23	language	language	NOUN
ajst-29613	240	24	models	model	NOUN
ajst-29613	240	25	use	use	VERB
ajst-29613	240	26	long	long	ADJ
ajst-29613	240	27	contexts[j	contexts[j	PROPN
ajst-29613	240	28	]	]	PUNCT
ajst-29613	240	29	.	.	PUNCT
ajst-29613	241	1	transactions	transaction	NOUN
ajst-29613	241	2	of	of	ADP
ajst-29613	241	3	the	the	DET
ajst-29613	241	4	association	association	NOUN
ajst-29613	241	5	for	for	ADP
ajst-29613	241	6	computational	computational	ADJ
ajst-29613	241	7	linguistics	linguistic	NOUN
ajst-29613	241	8	,	,	PUNCT
ajst-29613	241	9	2024	2024	NUM
ajst-29613	241	10	,	,	PUNCT
ajst-29613	241	11	12	12	NUM
ajst-29613	241	12	:	:	SYM
ajst-29613	241	13	157	157	NUM
ajst-29613	241	14	-	-	SYM
ajst-29613	241	15	173	173	NUM
ajst-29613	241	16	.	.	PUNCT
ajst-29613	242	1	[	[	X
ajst-29613	242	2	9	9	NUM
ajst-29613	242	3	]	]	X
ajst-29613	242	4	singhal	singhal	ADJ
ajst-29613	242	5	a.	a.	NOUN
ajst-29613	242	6	introducing	introduce	VERB
ajst-29613	242	7	the	the	DET
ajst-29613	242	8	knowledge	knowledge	NOUN
ajst-29613	242	9	graph	graph	NOUN
ajst-29613	242	10	:	:	PUNCT
ajst-29613	242	11	things	thing	NOUN
ajst-29613	242	12	,	,	PUNCT
ajst-29613	242	13	not	not	PART
ajst-29613	242	14	strings[j	strings[j	NOUN
ajst-29613	242	15	]	]	PUNCT
ajst-29613	242	16	.	.	PUNCT
ajst-29613	243	1	official	official	ADJ
ajst-29613	243	2	google	google	PROPN
ajst-29613	243	3	blog	blog	PROPN
ajst-29613	243	4	,	,	PUNCT
ajst-29613	243	5	2012	2012	NUM
ajst-29613	243	6	,	,	PUNCT
ajst-29613	243	7	5(16	5(16	NUM
ajst-29613	243	8	):	):	PUNCT
ajst-29613	243	9	3	3	X
ajst-29613	243	10	.	.	PUNCT
ajst-29613	244	1	[	[	X
ajst-29613	244	2	10	10	NUM
ajst-29613	244	3	]	]	X
ajst-29613	244	4	ji	ji	PROPN
ajst-29613	244	5	s	s	PROPN
ajst-29613	244	6	,	,	PUNCT
ajst-29613	244	7	pan	pan	PROPN
ajst-29613	244	8	s	s	PROPN
ajst-29613	244	9	,	,	PUNCT
ajst-29613	244	10	cambria	cambria	PROPN
ajst-29613	244	11	e	e	PROPN
ajst-29613	244	12	,	,	PUNCT
ajst-29613	244	13	et	et	PROPN
ajst-29613	244	14	al	al	PROPN
ajst-29613	244	15	.	.	PUNCT
ajst-29613	245	1	a	a	DET
ajst-29613	245	2	survey	survey	NOUN
ajst-29613	245	3	on	on	ADP
ajst-29613	245	4	knowledge	knowledge	NOUN
ajst-29613	245	5	graphs	graph	NOUN
ajst-29613	245	6	:	:	PUNCT
ajst-29613	245	7	representation	representation	NOUN
ajst-29613	245	8	,	,	PUNCT
ajst-29613	245	9	acquisition	acquisition	NOUN
ajst-29613	245	10	,	,	PUNCT
ajst-29613	245	11	and	and	CCONJ
ajst-29613	245	12	applications[j	applications[j	PROPN
ajst-29613	245	13	]	]	PUNCT
ajst-29613	245	14	.	.	PUNCT
ajst-29613	246	1	ieee	ieee	NOUN
ajst-29613	246	2	transactions	transaction	NOUN
ajst-29613	246	3	on	on	ADP
ajst-29613	246	4	neural	neural	ADJ
ajst-29613	246	5	networks	network	NOUN
ajst-29613	246	6	and	and	CCONJ
ajst-29613	246	7	learning	learning	NOUN
ajst-29613	246	8	systems	system	NOUN
ajst-29613	246	9	,	,	PUNCT
ajst-29613	246	10	2021	2021	NUM
ajst-29613	246	11	,	,	PUNCT
ajst-29613	246	12	33(2	33(2	NUM
ajst-29613	246	13	):	):	PUNCT
ajst-29613	246	14	494	494	NUM
ajst-29613	246	15	-	-	SYM
ajst-29613	246	16	514	514	NUM
ajst-29613	246	17	.	.	PUNCT
ajst-29613	247	1	[	[	X
ajst-29613	247	2	11	11	NUM
ajst-29613	247	3	]	]	SYM
ajst-29613	247	4	tresp	tresp	NOUN
ajst-29613	247	5	,	,	PUNCT
ajst-29613	247	6	v.	v.	ADV
ajst-29613	247	7	,	,	PUNCT
ajst-29613	247	8	gärtner	gärtner	NOUN
ajst-29613	247	9	,	,	PUNCT
ajst-29613	247	10	t.	t.	PROPN
ajst-29613	247	11	,	,	PUNCT
ajst-29613	247	12	&	&	CCONJ
ajst-29613	247	13	nickel	nickel	NOUN
ajst-29613	247	14	,	,	PUNCT
ajst-29613	247	15	m.	m.	NOUN
ajst-29613	247	16	(	(	PUNCT
ajst-29613	247	17	2021	2021	NUM
ajst-29613	247	18	)	)	PUNCT
ajst-29613	247	19	.	.	PUNCT
ajst-29613	248	1	dynamic	dynamic	ADJ
ajst-29613	248	2	knowledge	knowledge	NOUN
ajst-29613	248	3	graphs	graph	NOUN
ajst-29613	248	4	:	:	PUNCT
ajst-29613	248	5	a	a	DET
ajst-29613	248	6	survey	survey	NOUN
ajst-29613	248	7	.	.	PUNCT
ajst-29613	249	1	artificial	artificial	ADJ
ajst-29613	249	2	intelligence	intelligence	NOUN
ajst-29613	249	3	review	review	NOUN
ajst-29613	249	4	,	,	PUNCT
ajst-29613	249	5	55	55	NUM
ajst-29613	249	6	,	,	PUNCT
ajst-29613	249	7	349–392	349–392	NUM
ajst-29613	249	8	.	.	PUNCT
ajst-29613	250	1	[	[	X
ajst-29613	250	2	12	12	NUM
ajst-29613	250	3	]	]	X
ajst-29613	250	4	chen	chen	PROPN
ajst-29613	250	5	d.	d.	PROPN
ajst-29613	250	6	reading	read	VERB
ajst-29613	250	7	wikipedia	wikipedia	PROPN
ajst-29613	250	8	to	to	PART
ajst-29613	250	9	answer	answer	VERB
ajst-29613	250	10	open‐domain	open‐domain	PROPN
ajst-29613	250	11	questions[j	questions[j	PROPN
ajst-29613	250	12	]	]	PUNCT
ajst-29613	250	13	.	.	PUNCT
ajst-29613	251	1	arxiv	arxiv	PROPN
ajst-29613	251	2	preprint	preprint	PROPN
ajst-29613	251	3	arxiv:1704.00051	arxiv:1704.00051	PROPN
ajst-29613	251	4	,	,	PUNCT
ajst-29613	251	5	2017	2017	NUM
ajst-29613	251	6	.	.	PUNCT
ajst-29613	252	1	[	[	X
ajst-29613	252	2	13	13	NUM
ajst-29613	252	3	]	]	X
ajst-29613	252	4	karpukhin	karpukhin	PROPN
ajst-29613	252	5	v	v	NOUN
ajst-29613	252	6	,	,	PUNCT
ajst-29613	252	7	oğuz	oğuz	PROPN
ajst-29613	252	8	b	b	PROPN
ajst-29613	252	9	,	,	PUNCT
ajst-29613	252	10	min	min	PROPN
ajst-29613	252	11	s	s	PROPN
ajst-29613	252	12	,	,	PUNCT
ajst-29613	252	13	et	et	PROPN
ajst-29613	252	14	al	al	PROPN
ajst-29613	252	15	.	.	PUNCT
ajst-29613	253	1	dense	dense	ADJ
ajst-29613	253	2	passage	passage	NOUN
ajst-29613	253	3	retrieval	retrieval	NOUN
ajst-29613	253	4	for	for	ADP
ajst-29613	253	5	open	open	ADJ
ajst-29613	253	6	-	-	PUNCT
ajst-29613	253	7	domain	domain	NOUN
ajst-29613	253	8	question	question	NOUN
ajst-29613	253	9	answering	answer	VERB
ajst-29613	253	10	[	[	X
ajst-29613	253	11	j	j	X
ajst-29613	253	12	]	]	X
ajst-29613	253	13	.	.	PUNCT
ajst-29613	254	1	arxiv	arxiv	PROPN
ajst-29613	254	2	preprint	preprint	PROPN
ajst-29613	254	3	arxiv	arxiv	PROPN
ajst-29613	254	4	:	:	PUNCT
ajst-29613	254	5	2004.04906	2004.04906	NUM
ajst-29613	254	6	,	,	PUNCT
ajst-29613	254	7	2020	2020	NUM
ajst-29613	254	8	.	.	PUNCT
ajst-29613	255	1	[	[	X
ajst-29613	255	2	14	14	NUM
ajst-29613	255	3	]	]	PUNCT
ajst-29613	255	4	lewis	lewis	PROPN
ajst-29613	255	5	p	p	PROPN
ajst-29613	255	6	,	,	PUNCT
ajst-29613	255	7	perez	perez	PROPN
ajst-29613	255	8	e	e	PROPN
ajst-29613	255	9	,	,	PUNCT
ajst-29613	255	10	piktus	piktus	PROPN
ajst-29613	255	11	a	a	X
ajst-29613	255	12	,	,	PUNCT
ajst-29613	255	13	et	et	PROPN
ajst-29613	255	14	al	al	PROPN
ajst-29613	255	15	.	.	PROPN
ajst-29613	255	16	retrieval	retrieval	NOUN
ajst-29613	255	17	-	-	PUNCT
ajst-29613	255	18	augmented	augment	VERB
ajst-29613	255	19	generation	generation	NOUN
ajst-29613	255	20	for	for	ADP
ajst-29613	255	21	knowledge	knowledge	NOUN
ajst-29613	255	22	-	-	PUNCT
ajst-29613	255	23	intensive	intensive	ADJ
ajst-29613	255	24	nlp	nlp	NOUN
ajst-29613	255	25	tasks[j	tasks[j	NOUN
ajst-29613	255	26	]	]	PUNCT
ajst-29613	255	27	.	.	PUNCT
ajst-29613	256	1	advances	advance	NOUN
ajst-29613	256	2	in	in	ADP
ajst-29613	256	3	neural	neural	ADJ
ajst-29613	256	4	information	information	NOUN
ajst-29613	256	5	processing	processing	NOUN
ajst-29613	256	6	systems	system	NOUN
ajst-29613	256	7	,	,	PUNCT
ajst-29613	256	8	2020	2020	NUM
ajst-29613	256	9	,	,	PUNCT
ajst-29613	256	10	33	33	NUM
ajst-29613	256	11	:	:	SYM
ajst-29613	256	12	9459	9459	NUM
ajst-29613	256	13	-	-	SYM
ajst-29613	256	14	9474	9474	NUM
ajst-29613	256	15	.	.	PUNCT
ajst-29613	257	1	[	[	X
ajst-29613	257	2	15	15	NUM
ajst-29613	257	3	]	]	X
ajst-29613	257	4	gunawan	gunawan	PROPN
ajst-29613	257	5	d	d	PROPN
ajst-29613	257	6	,	,	PUNCT
ajst-29613	257	7	sembiring	sembire	VERB
ajst-29613	257	8	c	c	PROPN
ajst-29613	257	9	a	a	PRON
ajst-29613	257	10	,	,	PUNCT
ajst-29613	257	11	budiman	budiman	NOUN
ajst-29613	257	12	m	m	PROPN
ajst-29613	257	13	a.	a.	NOUN
ajst-29613	257	14	the	the	DET
ajst-29613	257	15	implementation	implementation	NOUN
ajst-29613	257	16	of	of	ADP
ajst-29613	257	17	cosine	cosine	NOUN
ajst-29613	257	18	similarity	similarity	NOUN
ajst-29613	257	19	to	to	PART
ajst-29613	257	20	calculate	calculate	VERB
ajst-29613	257	21	text	text	NOUN
ajst-29613	257	22	relevance	relevance	NOUN
ajst-29613	257	23	between	between	ADP
ajst-29613	257	24	two	two	NUM
ajst-29613	257	25	documents[c]//journal	documents[c]//journal	PROPN
ajst-29613	257	26	of	of	ADP
ajst-29613	257	27	physics	physics	PROPN
ajst-29613	257	28	:	:	PUNCT
ajst-29613	257	29	conference	conference	NOUN
ajst-29613	257	30	series	series	NOUN
ajst-29613	257	31	.	.	PUNCT
ajst-29613	258	1	iop	iop	PROPN
ajst-29613	258	2	publishing	publishing	NOUN
ajst-29613	258	3	,	,	PUNCT
ajst-29613	258	4	2018	2018	NUM
ajst-29613	258	5	,	,	PUNCT
ajst-29613	258	6	978	978	NUM
ajst-29613	258	7	:	:	SYM
ajst-29613	258	8	012120	012120	NUM
ajst-29613	258	9	.	.	PUNCT
ajst-29613	259	1	[	[	X
ajst-29613	259	2	16	16	NUM
ajst-29613	259	3	]	]	PUNCT
ajst-29613	259	4	ramelan	ramelan	NOUN
ajst-29613	259	5	m.	m.	NOUN
ajst-29613	259	6	a	a	DET
ajst-29613	259	7	query	query	NOUN
ajst-29613	259	8	expansion	expansion	NOUN
ajst-29613	259	9	using	use	VERB
ajst-29613	259	10	support	support	NOUN
ajst-29613	259	11	vector	vector	NOUN
ajst-29613	259	12	machine	machine	NOUN
ajst-29613	259	13	(	(	PUNCT
ajst-29613	259	14	svm	svm	PROPN
ajst-29613	259	15	)	)	PUNCT
ajst-29613	259	16	and	and	CCONJ
ajst-29613	259	17	best	good	ADJ
ajst-29613	259	18	matching	match	VERB
ajst-29613	259	19	25	25	NUM
ajst-29613	259	20	(	(	PUNCT
ajst-29613	259	21	bm25)[j	bm25)[j	NOUN
ajst-29613	259	22	]	]	PUNCT
ajst-29613	259	23	.	.	PUNCT
ajst-29613	260	1	international	international	ADJ
ajst-29613	260	2	journal	journal	PROPN
ajst-29613	260	3	of	of	ADP
ajst-29613	260	4	computer	computer	NOUN
ajst-29613	260	5	and	and	CCONJ
ajst-29613	260	6	information	information	NOUN
ajst-29613	260	7	system	system	NOUN
ajst-29613	260	8	(	(	PUNCT
ajst-29613	260	9	ijcis	ijcis	NOUN
ajst-29613	260	10	)	)	PUNCT
ajst-29613	260	11	,	,	PUNCT
ajst-29613	260	12	2024	2024	NUM
ajst-29613	260	13	,	,	PUNCT
ajst-29613	260	14	5(2	5(2	NUM
ajst-29613	260	15	):	):	PUNCT
ajst-29613	260	16	73	73	NUM
ajst-29613	260	17	-	-	SYM
ajst-29613	260	18	77	77	NUM
ajst-29613	260	19	.	.	PUNCT
ajst-29613	261	1	[	[	X
ajst-29613	261	2	17	17	NUM
ajst-29613	261	3	]	]	X
ajst-29613	261	4	jiang	jiang	PROPN
ajst-29613	261	5	x	x	PROPN
ajst-29613	261	6	,	,	PUNCT
ajst-29613	261	7	zhang	zhang	PROPN
ajst-29613	261	8	r	r	PROPN
ajst-29613	261	9	,	,	PUNCT
ajst-29613	261	10	xu	xu	PROPN
ajst-29613	261	11	y	y	PROPN
ajst-29613	261	12	,	,	PUNCT
ajst-29613	261	13	et	et	PROPN
ajst-29613	261	14	al	al	PROPN
ajst-29613	261	15	.	.	PROPN
ajst-29613	261	16	hykge	hykge	PROPN
ajst-29613	261	17	:	:	PUNCT
ajst-29613	261	18	a	a	DET
ajst-29613	261	19	hypothesis	hypothesis	NOUN
ajst-29613	261	20	knowledge	knowledge	NOUN
ajst-29613	261	21	graph	graph	NOUN
ajst-29613	261	22	enhanced	enhance	VERB
ajst-29613	261	23	framework	framework	NOUN
ajst-29613	261	24	for	for	ADP
ajst-29613	261	25	accurate	accurate	ADJ
ajst-29613	261	26	and	and	CCONJ
ajst-29613	261	27	reliable	reliable	ADJ
ajst-29613	261	28	medical	medical	ADJ
ajst-29613	261	29	llms	llm	NOUN
ajst-29613	261	30	responses[j	responses[j	PROPN
ajst-29613	261	31	]	]	PUNCT
ajst-29613	261	32	.	.	PUNCT
ajst-29613	262	1	arxiv	arxiv	PROPN
ajst-29613	262	2	preprint	preprint	PROPN
ajst-29613	262	3	arxiv:2312.15883	arxiv:2312.15883	PROPN
ajst-29613	262	4	,	,	PUNCT
ajst-29613	262	5	2024	2024	NUM
ajst-29613	262	6	.	.	PUNCT
ajst-29613	263	1	[	[	X
ajst-29613	263	2	18	18	NUM
ajst-29613	263	3	]	]	PUNCT
ajst-29613	263	4	jin	jin	PROPN
ajst-29613	263	5	b	b	PROPN
ajst-29613	263	6	,	,	PUNCT
ajst-29613	263	7	xie	xie	PROPN
ajst-29613	263	8	c	c	PROPN
ajst-29613	263	9	,	,	PUNCT
ajst-29613	263	10	zhang	zhang	PROPN
ajst-29613	263	11	j	j	PROPN
ajst-29613	263	12	,	,	PUNCT
ajst-29613	263	13	et	et	PROPN
ajst-29613	263	14	al	al	PROPN
ajst-29613	263	15	.	.	PROPN
ajst-29613	264	1	graph	graph	NOUN
ajst-29613	264	2	chain	chain	NOUN
ajst-29613	264	3	-	-	PUNCT
ajst-29613	264	4	of	of	ADP
ajst-29613	264	5	-	-	PUNCT
ajst-29613	264	6	thought	thought	NOUN
ajst-29613	264	7	:	:	PUNCT
ajst-29613	264	8	augmenting	augment	VERB
ajst-29613	264	9	large	large	ADJ
ajst-29613	264	10	language	language	NOUN
ajst-29613	264	11	models	model	NOUN
ajst-29613	264	12	by	by	ADP
ajst-29613	264	13	reasoning	reason	VERB
ajst-29613	264	14	on	on	ADP
ajst-29613	264	15	graphs[j	graphs[j	PROPN
ajst-29613	264	16	]	]	PUNCT
ajst-29613	264	17	.	.	PUNCT
ajst-29613	265	1	arxiv	arxiv	PROPN
ajst-29613	265	2	preprint	preprint	PROPN
ajst-29613	265	3	arxiv:2404.07103	arxiv:2404.07103	NOUN
ajst-29613	265	4	,	,	PUNCT
ajst-29613	265	5	2024	2024	NUM
ajst-29613	265	6	.	.	PUNCT
ajst-29613	266	1	[	[	X
ajst-29613	266	2	19	19	NUM
ajst-29613	266	3	]	]	X
ajst-29613	266	4	luo	luo	PROPN
ajst-29613	266	5	l	l	PROPN
ajst-29613	266	6	,	,	PUNCT
ajst-29613	266	7	li	li	PROPN
ajst-29613	266	8	y	y	PROPN
ajst-29613	266	9	f	f	PROPN
ajst-29613	266	10	,	,	PUNCT
ajst-29613	266	11	haffari	haffari	PROPN
ajst-29613	266	12	g	g	PROPN
ajst-29613	266	13	,	,	PUNCT
ajst-29613	266	14	et	et	PROPN
ajst-29613	266	15	al	al	PROPN
ajst-29613	266	16	.	.	PUNCT
ajst-29613	266	17	reasoning	reason	VERB
ajst-29613	266	18	on	on	ADP
ajst-29613	266	19	graphs	graph	NOUN
ajst-29613	266	20	:	:	PUNCT
ajst-29613	266	21	faithful	faithful	ADJ
ajst-29613	266	22	and	and	CCONJ
ajst-29613	266	23	interpretable	interpretable	ADJ
ajst-29613	266	24	large	large	ADJ
ajst-29613	266	25	language	language	NOUN
ajst-29613	266	26	model	model	NOUN
ajst-29613	266	27	reasoning[j	reasoning[j	PROPN
ajst-29613	266	28	]	]	PUNCT
ajst-29613	266	29	.	.	PUNCT
ajst-29613	267	1	arxiv	arxiv	PROPN
ajst-29613	267	2	preprint	preprint	PROPN
ajst-29613	267	3	arxiv:2310.01061	arxiv:2310.01061	PROPN
ajst-29613	267	4	,	,	PUNCT
ajst-29613	267	5	2023	2023	NUM
ajst-29613	267	6	.	.	PUNCT
ajst-29613	268	1	[	[	X
ajst-29613	268	2	20	20	NUM
ajst-29613	268	3	]	]	X
ajst-29613	268	4	ma	ma	PROPN
ajst-29613	268	5	s	s	PROPN
ajst-29613	268	6	,	,	PUNCT
ajst-29613	268	7	xu	xu	PROPN
ajst-29613	268	8	c	c	X
ajst-29613	268	9	,	,	PUNCT
ajst-29613	268	10	jiang	jiang	PROPN
ajst-29613	268	11	x	x	PROPN
ajst-29613	268	12	,	,	PUNCT
ajst-29613	268	13	et	et	PROPN
ajst-29613	268	14	al	al	PROPN
ajst-29613	268	15	.	.	PROPN
ajst-29613	268	16	think	think	VERB
ajst-29613	268	17	-	-	PUNCT
ajst-29613	268	18	on	on	ADP
ajst-29613	268	19	-	-	PUNCT
ajst-29613	268	20	graph	graph	NOUN
ajst-29613	268	21	2.0	2.0	NUM
ajst-29613	268	22	:	:	PUNCT
ajst-29613	268	23	deep	deep	ADJ
ajst-29613	268	24	and	and	CCONJ
ajst-29613	268	25	interpretable	interpretable	ADJ
ajst-29613	268	26	large	large	ADJ
ajst-29613	268	27	language	language	NOUN
ajst-29613	268	28	model	model	NOUN
ajst-29613	268	29	reasoning	reasoning	NOUN
ajst-29613	268	30	with	with	ADP
ajst-29613	268	31	knowledge	knowledge	NOUN
ajst-29613	268	32	graph	graph	NOUN
ajst-29613	268	33	-	-	PUNCT
ajst-29613	268	34	guided	guide	VERB
ajst-29613	268	35	retrieval	retrieval	NOUN
ajst-29613	269	1	[	[	X
ajst-29613	269	2	j	j	X
ajst-29613	269	3	]	]	X
ajst-29613	269	4	.	.	PUNCT
ajst-29613	270	1	arxiv	arxiv	PROPN
ajst-29613	270	2	e	e	PROPN
ajst-29613	270	3	-	-	NOUN
ajst-29613	270	4	prints	print	NOUN
ajst-29613	270	5	,	,	PUNCT
ajst-29613	270	6	2024	2024	NUM
ajst-29613	270	7	:	:	PUNCT
ajst-29613	270	8	arxiv	arxiv	NOUN
ajst-29613	270	9	:	:	PUNCT
ajst-29613	271	1	2407.10805	2407.10805	NUM
ajst-29613	271	2	.	.	PUNCT
ajst-29613	272	1	[	[	X
ajst-29613	272	2	21	21	NUM
ajst-29613	272	3	]	]	X
ajst-29613	272	4	yang	yang	PROPN
ajst-29613	272	5	h	h	PROPN
ajst-29613	272	6	,	,	PUNCT
ajst-29613	272	7	liu	liu	PROPN
ajst-29613	272	8	j.	j.	PROPN
ajst-29613	272	9	knowledge	knowledge	PROPN
ajst-29613	272	10	graph	graph	NOUN
ajst-29613	272	11	representation	representation	NOUN
ajst-29613	272	12	learning	learning	NOUN
ajst-29613	272	13	as	as	ADP
ajst-29613	272	14	groupoid	groupoid	NOUN
ajst-29613	272	15	:	:	PUNCT
ajst-29613	272	16	unifying	unify	VERB
ajst-29613	272	17	transe	transe	NOUN
ajst-29613	272	18	,	,	PUNCT
ajst-29613	272	19	rotate	rotate	VERB
ajst-29613	272	20	,	,	PUNCT
ajst-29613	272	21	quate	quate	ADJ
ajst-29613	272	22	,	,	PUNCT
ajst-29613	272	23	complex	complex	ADJ
ajst-29613	272	24	[	[	X
ajst-29613	272	25	c]//	c]//	ADJ
ajst-29613	272	26	proceedings	proceeding	NOUN
ajst-29613	272	27	of	of	ADP
ajst-29613	272	28	the	the	DET
ajst-29613	272	29	30th	30th	ADJ
ajst-29613	272	30	acm	acm	PROPN
ajst-29613	272	31	international	international	ADJ
ajst-29613	272	32	conference	conference	NOUN
ajst-29613	272	33	on	on	ADP
ajst-29613	272	34	information	information	NOUN
ajst-29613	272	35	&	&	CCONJ
ajst-29613	272	36	knowledge	knowledge	PROPN
ajst-29613	272	37	management	management	PROPN
ajst-29613	272	38	.	.	PUNCT
ajst-29613	273	1	2021	2021	NUM
ajst-29613	273	2	:	:	PUNCT
ajst-29613	273	3	2311	2311	NUM
ajst-29613	273	4	-	-	SYM
ajst-29613	273	5	2320	2320	NUM
ajst-29613	273	6	.	.	PUNCT
ajst-29613	274	1	[	[	X
ajst-29613	274	2	22	22	NUM
ajst-29613	274	3	]	]	X
ajst-29613	274	4	yasunaga	yasunaga	PROPN
ajst-29613	274	5	m	m	PROPN
ajst-29613	274	6	,	,	PUNCT
ajst-29613	274	7	ren	ren	PROPN
ajst-29613	274	8	h	h	NOUN
ajst-29613	274	9	,	,	PUNCT
ajst-29613	274	10	bosselut	bosselut	VERB
ajst-29613	274	11	a	a	PRON
ajst-29613	274	12	,	,	PUNCT
ajst-29613	274	13	et	et	PROPN
ajst-29613	274	14	al	al	PROPN
ajst-29613	274	15	.	.	PROPN
ajst-29613	275	1	qa	qa	PROPN
ajst-29613	275	2	-	-	PUNCT
ajst-29613	275	3	gnn	gnn	NOUN
ajst-29613	275	4	:	:	PUNCT
ajst-29613	275	5	reasoning	reason	VERB
ajst-29613	275	6	with	with	ADP
ajst-29613	275	7	language	language	NOUN
ajst-29613	275	8	models	model	NOUN
ajst-29613	275	9	and	and	CCONJ
ajst-29613	275	10	knowledge	knowledge	NOUN
ajst-29613	275	11	graphs	graph	NOUN
ajst-29613	275	12	for	for	ADP
ajst-29613	275	13	question	question	NOUN
ajst-29613	275	14	answering[j	answering[j	NOUN
ajst-29613	275	15	]	]	PUNCT
ajst-29613	275	16	.	.	PUNCT
ajst-29613	276	1	arxiv	arxiv	PROPN
ajst-29613	276	2	preprint	preprint	NOUN
ajst-29613	276	3	arxiv:2104.06378	arxiv:2104.06378	NOUN
ajst-29613	276	4	,	,	PUNCT
ajst-29613	276	5	2021	2021	NUM
ajst-29613	276	6	.	.	PUNCT
ajst-29613	277	1	[	[	X
ajst-29613	277	2	23	23	NUM
ajst-29613	277	3	]	]	X
ajst-29613	277	4	taunk	taunk	NOUN
ajst-29613	277	5	d	d	PROPN
ajst-29613	277	6	,	,	PUNCT
ajst-29613	277	7	khanna	khanna	PROPN
ajst-29613	277	8	l	l	PROPN
ajst-29613	277	9	,	,	PUNCT
ajst-29613	277	10	kandru	kandru	PROPN
ajst-29613	277	11	s	s	PROPN
ajst-29613	277	12	v	v	PROPN
ajst-29613	277	13	p	p	PROPN
ajst-29613	277	14	k	k	PROPN
ajst-29613	277	15	,	,	PUNCT
ajst-29613	277	16	et	et	PROPN
ajst-29613	277	17	al	al	PROPN
ajst-29613	277	18	.	.	PUNCT
ajst-29613	277	19	grapeqa	grapeqa	PROPN
ajst-29613	277	20	:	:	PUNCT
ajst-29613	277	21	graph	graph	NOUN
ajst-29613	277	22	augmentation	augmentation	NOUN
ajst-29613	277	23	and	and	CCONJ
ajst-29613	277	24	pruning	prune	VERB
ajst-29613	277	25	to	to	PART
ajst-29613	277	26	enhance	enhance	VERB
ajst-29613	277	27	questionanswering[c]//companion	questionanswering[c]//companion	NOUN
ajst-29613	277	28	proceedings	proceeding	NOUN
ajst-29613	277	29	of	of	ADP
ajst-29613	277	30	the	the	DET
ajst-29613	277	31	acm	acm	PROPN
ajst-29613	277	32	web	web	NOUN
ajst-29613	277	33	conference	conference	NOUN
ajst-29613	277	34	2023	2023	NUM
ajst-29613	277	35	.	.	PUNCT
ajst-29613	278	1	2023	2023	NUM
ajst-29613	278	2	:	:	PUNCT
ajst-29613	278	3	1138	1138	NUM
ajst-29613	278	4	-	-	SYM
ajst-29613	278	5	1144	1144	NUM
ajst-29613	278	6	.	.	PUNCT
ajst-29613	279	1	[	[	X
ajst-29613	279	2	24	24	NUM
ajst-29613	279	3	]	]	PUNCT
ajst-29613	279	4	he	he	PRON
ajst-29613	279	5	x	x	PROPN
ajst-29613	279	6	,	,	PUNCT
ajst-29613	279	7	tian	tian	PROPN
ajst-29613	279	8	y	y	PROPN
ajst-29613	279	9	,	,	PUNCT
ajst-29613	279	10	sun	sun	PROPN
ajst-29613	279	11	y	y	PROPN
ajst-29613	279	12	,	,	PUNCT
ajst-29613	279	13	et	et	PROPN
ajst-29613	279	14	al	al	PROPN
ajst-29613	279	15	.	.	PUNCT
ajst-29613	280	1	g	g	NOUN
ajst-29613	280	2	-	-	PUNCT
ajst-29613	280	3	retriever	retriever	NOUN
ajst-29613	280	4	:	:	PUNCT
ajst-29613	280	5	retrieval	retrieval	NOUN
ajst-29613	280	6	-	-	PUNCT
ajst-29613	280	7	augmented	augment	VERB
ajst-29613	280	8	generation	generation	NOUN
ajst-29613	280	9	for	for	ADP
ajst-29613	280	10	textual	textual	ADJ
ajst-29613	280	11	graph	graph	NOUN
ajst-29613	280	12	understanding	understanding	NOUN
ajst-29613	280	13	and	and	CCONJ
ajst-29613	280	14	question	question	NOUN
ajst-29613	280	15	answering[j	answering[j	NOUN
ajst-29613	280	16	]	]	PUNCT
ajst-29613	280	17	.	.	PUNCT
ajst-29613	281	1	arxiv	arxiv	PROPN
ajst-29613	281	2	preprint	preprint	NOUN
ajst-29613	281	3	arxiv:2402.07630	arxiv:2402.07630	PROPN
ajst-29613	281	4	,	,	PUNCT
ajst-29613	281	5	2024	2024	NUM
ajst-29613	281	6	.	.	PUNCT
ajst-29613	282	1	[	[	X
ajst-29613	282	2	25	25	NUM
ajst-29613	282	3	]	]	X
ajst-29613	282	4	delile	delile	PROPN
ajst-29613	282	5	j	j	PROPN
ajst-29613	282	6	,	,	PUNCT
ajst-29613	282	7	mukherjee	mukherjee	PROPN
ajst-29613	282	8	s	s	PROPN
ajst-29613	282	9	,	,	PUNCT
ajst-29613	282	10	van	van	PROPN
ajst-29613	282	11	pamel	pamel	PROPN
ajst-29613	282	12	a	a	PRON
ajst-29613	282	13	,	,	PUNCT
ajst-29613	282	14	et	et	PROPN
ajst-29613	282	15	al	al	PROPN
ajst-29613	282	16	.	.	PROPN
ajst-29613	282	17	graph	graph	NOUN
ajst-29613	282	18	-	-	PUNCT
ajst-29613	282	19	based	base	VERB
ajst-29613	282	20	retriever	retriever	NOUN
ajst-29613	282	21	captures	capture	VERB
ajst-29613	282	22	the	the	DET
ajst-29613	282	23	long	long	ADJ
ajst-29613	282	24	tail	tail	NOUN
ajst-29613	282	25	of	of	ADP
ajst-29613	282	26	biomedical	biomedical	ADJ
ajst-29613	282	27	knowledge[j	knowledge[j	NOUN
ajst-29613	282	28	]	]	PUNCT
ajst-29613	282	29	.	.	PUNCT
ajst-29613	283	1	arxiv	arxiv	PROPN
ajst-29613	283	2	preprint	preprint	NOUN
ajst-29613	283	3	arxiv:2402.12352	arxiv:2402.12352	VERB
ajst-29613	283	4	,	,	PUNCT
ajst-29613	283	5	2024	2024	NUM
ajst-29613	283	6	.	.	PUNCT
ajst-29613	284	1	[	[	X
ajst-29613	284	2	26	26	NUM
ajst-29613	284	3	]	]	X
ajst-29613	284	4	mavromatis	mavromatis	NOUN
ajst-29613	284	5	c	c	NOUN
ajst-29613	284	6	,	,	PUNCT
ajst-29613	284	7	karypis	karypi	VERB
ajst-29613	284	8	g.	g.	PROPN
ajst-29613	284	9	gnn	gnn	PROPN
ajst-29613	284	10	-	-	PUNCT
ajst-29613	284	11	rag	rag	NOUN
ajst-29613	284	12	:	:	PUNCT
ajst-29613	284	13	graph	graph	NOUN
ajst-29613	284	14	neural	neural	ADJ
ajst-29613	284	15	retrieval	retrieval	NOUN
ajst-29613	284	16	for	for	ADP
ajst-29613	284	17	large	large	ADJ
ajst-29613	284	18	language	language	NOUN
ajst-29613	284	19	model	model	NOUN
ajst-29613	284	20	reasoning[j	reasoning[j	PROPN
ajst-29613	284	21	]	]	PUNCT
ajst-29613	284	22	.	.	PUNCT
ajst-29613	285	1	arxiv	arxiv	PROPN
ajst-29613	285	2	preprint	preprint	VERB
ajst-29613	285	3	arxiv:2405.20139	arxiv:2405.20139	NOUN
ajst-29613	285	4	,	,	PUNCT
ajst-29613	285	5	2024	2024	NUM
ajst-29613	285	6	.	.	PUNCT
ajst-29613	286	1	[	[	X
ajst-29613	286	2	27	27	NUM
ajst-29613	286	3	]	]	X
ajst-29613	286	4	zhang	zhang	PROPN
ajst-29613	286	5	j	j	PROPN
ajst-29613	286	6	,	,	PUNCT
ajst-29613	286	7	zhang	zhang	PROPN
ajst-29613	286	8	x	x	PROPN
ajst-29613	286	9	,	,	PUNCT
ajst-29613	286	10	yu	yu	PROPN
ajst-29613	286	11	j	j	PROPN
ajst-29613	286	12	,	,	PUNCT
ajst-29613	286	13	et	et	PROPN
ajst-29613	286	14	al	al	PROPN
ajst-29613	286	15	.	.	PROPN
ajst-29613	286	16	subgraph	subgraph	PROPN
ajst-29613	286	17	retrieval	retrieval	PROPN
ajst-29613	286	18	enhanced	enhance	VERB
ajst-29613	286	19	model	model	NOUN
ajst-29613	286	20	for	for	ADP
ajst-29613	286	21	multi	multi	ADJ
ajst-29613	286	22	-	-	ADJ
ajst-29613	286	23	hop	hop	ADJ
ajst-29613	286	24	knowledge	knowledge	NOUN
ajst-29613	286	25	base	base	NOUN
ajst-29613	286	26	question	question	NOUN
ajst-29613	286	27	answering[j	answering[j	NOUN
ajst-29613	286	28	]	]	PUNCT
ajst-29613	286	29	.	.	PUNCT
ajst-29613	287	1	arxiv	arxiv	PROPN
ajst-29613	287	2	preprint	preprint	PROPN
ajst-29613	287	3	arxiv:2202.13296	arxiv:2202.13296	PROPN
ajst-29613	287	4	,	,	PUNCT
ajst-29613	287	5	2022	2022	NUM
ajst-29613	287	6	.	.	PUNCT
ajst-29613	288	1	[	[	X
ajst-29613	288	2	28	28	NUM
ajst-29613	288	3	]	]	X
ajst-29613	288	4	liu	liu	PROPN
ajst-29613	288	5	y.	y.	PROPN
ajst-29613	288	6	roberta	roberta	PROPN
ajst-29613	288	7	:	:	PUNCT
ajst-29613	288	8	a	a	DET
ajst-29613	288	9	robustly	robustly	ADV
ajst-29613	288	10	optimized	optimize	VERB
ajst-29613	288	11	bert	bert	NOUN
ajst-29613	288	12	pretraining	pretraine	VERB
ajst-29613	288	13	approach[j	approach[j	PROPN
ajst-29613	288	14	]	]	PUNCT
ajst-29613	288	15	.	.	PUNCT
ajst-29613	289	1	arxiv	arxiv	PROPN
ajst-29613	289	2	preprint	preprint	NOUN
ajst-29613	289	3	arxiv:1907.11692	arxiv:1907.11692	NOUN
ajst-29613	289	4	,	,	PUNCT
ajst-29613	289	5	2019	2019	NUM
ajst-29613	289	6	,	,	PUNCT
ajst-29613	289	7	364	364	NUM
ajst-29613	289	8	.	.	PUNCT
ajst-29613	290	1	[	[	X
ajst-29613	290	2	29	29	NUM
ajst-29613	290	3	]	]	X
ajst-29613	290	4	kim	kim	PROPN
ajst-29613	290	5	j	j	PROPN
ajst-29613	290	6	,	,	PUNCT
ajst-29613	290	7	kwon	kwon	VERB
ajst-29613	290	8	y	y	PROPN
ajst-29613	290	9	,	,	PUNCT
ajst-29613	290	10	jo	jo	PROPN
ajst-29613	290	11	y	y	PROPN
ajst-29613	290	12	,	,	PUNCT
ajst-29613	290	13	et	et	PROPN
ajst-29613	290	14	al	al	PROPN
ajst-29613	290	15	.	.	PUNCT
ajst-29613	291	1	kg	kg	PROPN
ajst-29613	291	2	-	-	PUNCT
ajst-29613	291	3	gpt	gpt	NOUN
ajst-29613	291	4	:	:	PUNCT
ajst-29613	291	5	a	a	DET
ajst-29613	291	6	general	general	ADJ
ajst-29613	291	7	framework	framework	NOUN
ajst-29613	291	8	for	for	ADP
ajst-29613	291	9	reasoning	reasoning	NOUN
ajst-29613	291	10	on	on	ADP
ajst-29613	291	11	knowledge	knowledge	NOUN
ajst-29613	291	12	graphs	graph	NOUN
ajst-29613	291	13	using	use	VERB
ajst-29613	291	14	large	large	ADJ
ajst-29613	291	15	language	language	NOUN
ajst-29613	291	16	models[j	models[j	PROPN
ajst-29613	291	17	]	]	PUNCT
ajst-29613	291	18	.	.	PUNCT
ajst-29613	292	1	arxiv	arxiv	PROPN
ajst-29613	292	2	preprint	preprint	NOUN
ajst-29613	292	3	arxiv:2310.11220	arxiv:2310.11220	NOUN
ajst-29613	292	4	,	,	PUNCT
ajst-29613	292	5	2023	2023	NUM
ajst-29613	292	6	.	.	PUNCT
ajst-29613	293	1	[	[	X
ajst-29613	293	2	30	30	NUM
ajst-29613	293	3	]	]	X
ajst-29613	293	4	wold	wold	PROPN
ajst-29613	293	5	s	s	PROPN
ajst-29613	293	6	,	,	PUNCT
ajst-29613	293	7	øvrelid	øvrelid	ADJ
ajst-29613	293	8	l	l	NOUN
ajst-29613	293	9	,	,	PUNCT
ajst-29613	293	10	velldal	velldal	ADJ
ajst-29613	293	11	e.	e.	PROPN
ajst-29613	293	12	text	text	PROPN
ajst-29613	293	13	-	-	PUNCT
ajst-29613	293	14	to	to	ADP
ajst-29613	293	15	-	-	PUNCT
ajst-29613	293	16	kg	kg	NOUN
ajst-29613	293	17	alignment	alignment	NOUN
ajst-29613	293	18	:	:	PUNCT
ajst-29613	293	19	comparing	compare	VERB
ajst-29613	293	20	current	current	ADJ
ajst-29613	293	21	methods	method	NOUN
ajst-29613	293	22	on	on	ADP
ajst-29613	293	23	classification	classification	NOUN
ajst-29613	293	24	tasks[j	tasks[j	NOUN
ajst-29613	293	25	]	]	PUNCT
ajst-29613	293	26	.	.	PUNCT
ajst-29613	294	1	arxiv	arxiv	PROPN
ajst-29613	294	2	preprint	preprint	PROPN
ajst-29613	294	3	arxiv:2306.02871	arxiv:2306.02871	PROPN
ajst-29613	294	4	,	,	PUNCT
ajst-29613	294	5	2023	2023	NUM
ajst-29613	294	6	.	.	PUNCT
ajst-29613	295	1	[	[	X
ajst-29613	295	2	31	31	NUM
ajst-29613	295	3	]	]	X
ajst-29613	295	4	jiang	jiang	PROPN
ajst-29613	295	5	j	j	PROPN
ajst-29613	295	6	,	,	PUNCT
ajst-29613	295	7	zhou	zhou	PROPN
ajst-29613	295	8	k	k	PROPN
ajst-29613	295	9	,	,	PUNCT
ajst-29613	295	10	dong	dong	PROPN
ajst-29613	295	11	z	z	PROPN
ajst-29613	295	12	,	,	PUNCT
ajst-29613	295	13	et	et	PROPN
ajst-29613	295	14	al	al	PROPN
ajst-29613	295	15	.	.	PROPN
ajst-29613	295	16	structgpt	structgpt	PROPN
ajst-29613	295	17	:	:	PUNCT
ajst-29613	295	18	a	a	DET
ajst-29613	295	19	general	general	ADJ
ajst-29613	295	20	framework	framework	NOUN
ajst-29613	295	21	for	for	ADP
ajst-29613	295	22	large	large	ADJ
ajst-29613	295	23	language	language	NOUN
ajst-29613	295	24	model	model	NOUN
ajst-29613	295	25	to	to	PART
ajst-29613	295	26	reason	reason	NOUN
ajst-29613	295	27	over	over	ADP
ajst-29613	295	28	structured	structured	ADJ
ajst-29613	295	29	data[j	data[j	NOUN
ajst-29613	295	30	]	]	PUNCT
ajst-29613	295	31	.	.	PUNCT
ajst-29613	296	1	arxiv	arxiv	PROPN
ajst-29613	296	2	preprint	preprint	PROPN
ajst-29613	296	3	arxiv:2305.09645	arxiv:2305.09645	PROPN
ajst-29613	296	4	,	,	PUNCT
ajst-29613	296	5	2023	2023	NUM
ajst-29613	296	6	.	.	PUNCT
ajst-29613	297	1	[	[	X
ajst-29613	297	2	32	32	NUM
ajst-29613	297	3	]	]	X
ajst-29613	297	4	liu	liu	PROPN
ajst-29613	297	5	g	g	PROPN
ajst-29613	297	6	,	,	PUNCT
ajst-29613	297	7	zhang	zhang	PROPN
ajst-29613	297	8	y	y	PROPN
ajst-29613	297	9	,	,	PUNCT
ajst-29613	297	10	li	li	PROPN
ajst-29613	297	11	y	y	PROPN
ajst-29613	297	12	,	,	PUNCT
ajst-29613	297	13	et	et	PROPN
ajst-29613	297	14	al	al	PROPN
ajst-29613	297	15	.	.	PROPN
ajst-29613	297	16	explore	explore	VERB
ajst-29613	297	17	then	then	ADV
ajst-29613	297	18	determine	determine	VERB
ajst-29613	297	19	:	:	PUNCT
ajst-29613	297	20	a	a	DET
ajst-29613	297	21	gnnllm	gnnllm	ADJ
ajst-29613	297	22	synergy	synergy	NOUN
ajst-29613	297	23	framework	framework	NOUN
ajst-29613	297	24	for	for	ADP
ajst-29613	297	25	reasoning	reason	VERB
ajst-29613	297	26	over	over	ADP
ajst-29613	297	27	knowledge	knowledge	NOUN
ajst-29613	297	28	graph[j	graph[j	NOUN
ajst-29613	297	29	]	]	PUNCT
ajst-29613	297	30	.	.	PUNCT
ajst-29613	298	1	arxiv	arxiv	PROPN
ajst-29613	298	2	preprint	preprint	VERB
ajst-29613	298	3	arxiv:2406.01145	arxiv:2406.01145	NUM
ajst-29613	298	4	,	,	PUNCT
ajst-29613	298	5	2024	2024	NUM
ajst-29613	298	6	.	.	PUNCT
ajst-29613	299	1	[	[	X
ajst-29613	299	2	33	33	NUM
ajst-29613	299	3	]	]	PUNCT
ajst-29613	299	4	kenton	kenton	PROPN
ajst-29613	299	5	j	j	PROPN
ajst-29613	299	6	d	d	PROPN
ajst-29613	299	7	m	m	PROPN
ajst-29613	299	8	w	w	PROPN
ajst-29613	299	9	c	c	PROPN
ajst-29613	299	10	,	,	PUNCT
ajst-29613	299	11	toutanova	toutanova	PROPN
ajst-29613	299	12	l	l	PROPN
ajst-29613	299	13	k.	k.	PROPN
ajst-29613	299	14	bert	bert	PROPN
ajst-29613	299	15	:	:	PUNCT
ajst-29613	299	16	pre	pre	ADJ
ajst-29613	299	17	-	-	NOUN
ajst-29613	299	18	training	training	NOUN
ajst-29613	299	19	of	of	ADP
ajst-29613	299	20	deep	deep	ADJ
ajst-29613	299	21	bidirectional	bidirectional	ADJ
ajst-29613	299	22	transformers	transformer	NOUN
ajst-29613	299	23	for	for	ADP
ajst-29613	299	24	language	language	NOUN
ajst-29613	299	25	understanding	understanding	NOUN
ajst-29613	299	26	[	[	X
ajst-29613	299	27	c	c	X
ajst-29613	299	28	]	]	X
ajst-29613	299	29	//	//	NUM
ajst-29613	299	30	proceedings	proceeding	NOUN
ajst-29613	299	31	of	of	ADP
ajst-29613	299	32	naacl	naacl	PROPN
ajst-29613	299	33	-	-	PUNCT
ajst-29613	299	34	hlt	hlt	NOUN
ajst-29613	299	35	.	.	PROPN
ajst-29613	299	36	2019	2019	NUM
ajst-29613	299	37	,	,	PUNCT
ajst-29613	299	38	1	1	NUM
ajst-29613	299	39	:	:	SYM
ajst-29613	299	40	2	2	NUM
ajst-29613	299	41	.	.	PUNCT
ajst-29613	300	1	[	[	X
ajst-29613	300	2	34	34	NUM
ajst-29613	300	3	]	]	X
ajst-29613	300	4	raffel	raffel	NOUN
ajst-29613	300	5	c	c	NOUN
ajst-29613	300	6	,	,	PUNCT
ajst-29613	300	7	shazeer	shazeer	NOUN
ajst-29613	300	8	n	n	CCONJ
ajst-29613	300	9	,	,	PUNCT
ajst-29613	300	10	roberts	roberts	PROPN
ajst-29613	300	11	a	a	PROPN
ajst-29613	300	12	,	,	PUNCT
ajst-29613	300	13	et	et	PROPN
ajst-29613	300	14	al	al	PROPN
ajst-29613	300	15	.	.	PUNCT
ajst-29613	301	1	exploring	explore	VERB
ajst-29613	301	2	the	the	DET
ajst-29613	301	3	limits	limit	NOUN
ajst-29613	301	4	of	of	ADP
ajst-29613	301	5	transfer	transfer	NOUN
ajst-29613	301	6	learning	learn	VERB
ajst-29613	301	7	with	with	ADP
ajst-29613	301	8	a	a	DET
ajst-29613	301	9	unified	unified	ADJ
ajst-29613	301	10	text	text	NOUN
ajst-29613	301	11	-	-	PUNCT
ajst-29613	301	12	to	to	ADP
ajst-29613	301	13	-	-	PUNCT
ajst-29613	301	14	text	text	NOUN
ajst-29613	301	15	transformer[j	transformer[j	NOUN
ajst-29613	301	16	]	]	PUNCT
ajst-29613	301	17	.	.	PUNCT
ajst-29613	302	1	journal	journal	PROPN
ajst-29613	302	2	of	of	ADP
ajst-29613	302	3	machine	machine	NOUN
ajst-29613	302	4	learning	learn	VERB
ajst-29613	302	5	research	research	NOUN
ajst-29613	302	6	,	,	PUNCT
ajst-29613	302	7	2020	2020	NUM
ajst-29613	302	8	,	,	PUNCT
ajst-29613	302	9	21(140	21(140	NUM
ajst-29613	302	10	):	):	PUNCT
ajst-29613	302	11	1	1	NUM
ajst-29613	302	12	-	-	SYM
ajst-29613	302	13	67	67	NUM
ajst-29613	302	14	.	.	PUNCT
ajst-29613	303	1	[	[	X
ajst-29613	303	2	35	35	NUM
ajst-29613	303	3	]	]	X
ajst-29613	303	4	driess	driess	NOUN
ajst-29613	303	5	d	d	PROPN
ajst-29613	303	6	,	,	PUNCT
ajst-29613	303	7	xia	xia	PROPN
ajst-29613	303	8	f	f	PROPN
ajst-29613	303	9	,	,	PUNCT
ajst-29613	303	10	sajjadi	sajjadi	PROPN
ajst-29613	303	11	m	m	PROPN
ajst-29613	303	12	s	s	NOUN
ajst-29613	303	13	m	m	NOUN
ajst-29613	303	14	,	,	PUNCT
ajst-29613	303	15	et	et	PROPN
ajst-29613	303	16	al	al	PROPN
ajst-29613	303	17	.	.	PUNCT
ajst-29613	303	18	palm	palm	NOUN
ajst-29613	303	19	-	-	PUNCT
ajst-29613	303	20	e	e	NOUN
ajst-29613	303	21	:	:	PUNCT
ajst-29613	303	22	an	an	DET
ajst-29613	303	23	embodied	embody	VERB
ajst-29613	303	24	multimodal	multimodal	NOUN
ajst-29613	303	25	language	language	NOUN
ajst-29613	303	26	model	model	NOUN
ajst-29613	304	1	[	[	X
ajst-29613	304	2	j	j	X
ajst-29613	304	3	]	]	X
ajst-29613	304	4	.	.	PUNCT
ajst-29613	305	1	arxiv	arxiv	PROPN
ajst-29613	305	2	preprint	preprint	PROPN
ajst-29613	305	3	arxiv	arxiv	PROPN
ajst-29613	305	4	:	:	PUNCT
ajst-29613	305	5	2303.03378	2303.03378	NUM
ajst-29613	305	6	,	,	PUNCT
ajst-29613	305	7	2023	2023	NUM
ajst-29613	305	8	.	.	PUNCT
ajst-29613	306	1	[	[	X
ajst-29613	306	2	36	36	NUM
ajst-29613	306	3	]	]	X
ajst-29613	306	4	cheng	cheng	PROPN
ajst-29613	306	5	k	k	PROPN
ajst-29613	306	6	,	,	PUNCT
ajst-29613	306	7	lin	lin	PROPN
ajst-29613	306	8	g	g	PROPN
ajst-29613	306	9	,	,	PUNCT
ajst-29613	306	10	fei	fei	PROPN
ajst-29613	306	11	h	h	PROPN
ajst-29613	306	12	,	,	PUNCT
ajst-29613	306	13	et	et	PROPN
ajst-29613	306	14	al	al	PROPN
ajst-29613	306	15	.	.	PUNCT
ajst-29613	306	16	multi	multi	ADJ
ajst-29613	306	17	-	-	ADJ
ajst-29613	306	18	hop	hop	ADJ
ajst-29613	306	19	question	question	NOUN
ajst-29613	306	20	answering	answer	VERB
ajst-29613	306	21	under	under	ADP
ajst-29613	306	22	temporal	temporal	ADJ
ajst-29613	306	23	knowledge	knowledge	NOUN
ajst-29613	306	24	editing	edit	VERB
ajst-29613	306	25	[	[	X
ajst-29613	306	26	j	j	X
ajst-29613	306	27	]	]	X
ajst-29613	306	28	.	.	PUNCT
ajst-29613	307	1	arxiv	arxiv	PROPN
ajst-29613	307	2	preprint	preprint	PROPN
ajst-29613	307	3	arxiv	arxiv	PROPN
ajst-29613	307	4	:	:	PUNCT
ajst-29613	307	5	2404.00492	2404.00492	NUM
ajst-29613	307	6	,	,	PUNCT
ajst-29613	307	7	2024	2024	NUM
ajst-29613	307	8	.	.	PUNCT
ajst-29613	308	1	[	[	X
ajst-29613	308	2	37	37	NUM
ajst-29613	308	3	]	]	X
ajst-29613	308	4	huang	huang	PROPN
ajst-29613	308	5	y	y	PROPN
ajst-29613	308	6	,	,	PUNCT
ajst-29613	308	7	li	li	PROPN
ajst-29613	308	8	y	y	PROPN
ajst-29613	308	9	,	,	PUNCT
ajst-29613	308	10	xu	xu	PROPN
ajst-29613	308	11	y	y	PROPN
ajst-29613	308	12	,	,	PUNCT
ajst-29613	308	13	et	et	PROPN
ajst-29613	308	14	al	al	PROPN
ajst-29613	308	15	.	.	PUNCT
ajst-29613	309	1	mvp	mvp	PROPN
ajst-29613	309	2	-	-	PUNCT
ajst-29613	309	3	tuning	tuning	NOUN
ajst-29613	309	4	:	:	PUNCT
ajst-29613	309	5	multi	multi	ADJ
ajst-29613	309	6	-	-	ADJ
ajst-29613	309	7	view	view	ADJ
ajst-29613	309	8	knowledge	knowledge	NOUN
ajst-29613	309	9	retrieval	retrieval	NOUN
ajst-29613	309	10	with	with	ADP
ajst-29613	309	11	prompt	prompt	ADJ
ajst-29613	309	12	tuning	tuning	NOUN
ajst-29613	309	13	for	for	ADP
ajst-29613	309	14	commonsense	commonsense	ADJ
ajst-29613	309	15	reasoning[c]//proceedings	reasoning[c]//proceeding	NOUN
ajst-29613	309	16	of	of	ADP
ajst-29613	309	17	the	the	DET
ajst-29613	309	18	61st	61st	ADJ
ajst-29613	309	19	annual	annual	ADJ
ajst-29613	309	20	meeting	meeting	NOUN
ajst-29613	309	21	of	of	ADP
ajst-29613	309	22	the	the	DET
ajst-29613	309	23	association	association	NOUN
ajst-29613	309	24	for	for	ADP
ajst-29613	309	25	computational	computational	ADJ
ajst-29613	309	26	linguistics	linguistic	NOUN
ajst-29613	309	27	(	(	PUNCT
ajst-29613	309	28	volume	volume	NOUN
ajst-29613	309	29	1	1	NUM
ajst-29613	309	30	:	:	PUNCT
ajst-29613	309	31	long	long	ADJ
ajst-29613	309	32	papers	paper	NOUN
ajst-29613	309	33	)	)	PUNCT
ajst-29613	309	34	.	.	PUNCT
ajst-29613	310	1	2023	2023	NUM
ajst-29613	310	2	:	:	PUNCT
ajst-29613	310	3	13417	13417	NUM
ajst-29613	310	4	-	-	SYM
ajst-29613	310	5	13432	13432	NUM
ajst-29613	310	6	.	.	PUNCT
ajst-29613	311	1	[	[	X
ajst-29613	311	2	38	38	NUM
ajst-29613	311	3	]	]	PUNCT
ajst-29613	311	4	an	an	DET
ajst-29613	311	5	z	z	NOUN
ajst-29613	311	6	,	,	PUNCT
ajst-29613	311	7	ding	ding	NOUN
ajst-29613	311	8	x	x	NOUN
ajst-29613	311	9	,	,	PUNCT
ajst-29613	311	10	fu	fu	PROPN
ajst-29613	311	11	y	y	PROPN
ajst-29613	311	12	c	c	PROPN
ajst-29613	311	13	,	,	PUNCT
ajst-29613	311	14	et	et	PROPN
ajst-29613	311	15	al	al	PROPN
ajst-29613	311	16	.	.	PUNCT
ajst-29613	311	17	golden	golden	PROPN
ajst-29613	311	18	-	-	PUNCT
ajst-29613	311	19	retriever	retriever	NOUN
ajst-29613	311	20	:	:	PUNCT
ajst-29613	311	21	high	high	ADJ
ajst-29613	311	22	-	-	PUNCT
ajst-29613	311	23	fidelity	fidelity	NOUN
ajst-29613	311	24	agentic	agentic	ADJ
ajst-29613	311	25	retrieval	retrieval	NOUN
ajst-29613	311	26	augmented	augment	VERB
ajst-29613	311	27	generation	generation	NOUN
ajst-29613	311	28	for	for	ADP
ajst-29613	311	29	industrial	industrial	ADJ
ajst-29613	311	30	knowledge	knowledge	NOUN
ajst-29613	311	31	base[j	base[j	PROPN
ajst-29613	311	32	]	]	PUNCT
ajst-29613	311	33	.	.	PUNCT
ajst-29613	312	1	arxiv	arxiv	PROPN
ajst-29613	312	2	preprint	preprint	PROPN
ajst-29613	312	3	arxiv:2408.00798	arxiv:2408.00798	PROPN
ajst-29613	312	4	,	,	PUNCT
ajst-29613	312	5	2024	2024	NUM
ajst-29613	312	6	.	.	PUNCT
ajst-29613	313	1	235	235	NUM
ajst-29613	314	1	[	[	SYM
ajst-29613	314	2	39	39	NUM
ajst-29613	314	3	]	]	PUNCT
ajst-29613	314	4	choudhary	choudhary	PROPN
ajst-29613	314	5	n	n	PROPN
ajst-29613	314	6	,	,	PUNCT
ajst-29613	314	7	reddy	reddy	PROPN
ajst-29613	314	8	c	c	PROPN
ajst-29613	314	9	k.	k.	PROPN
ajst-29613	314	10	complex	complex	ADJ
ajst-29613	314	11	logical	logical	ADJ
ajst-29613	314	12	reasoning	reasoning	NOUN
ajst-29613	314	13	over	over	ADP
ajst-29613	314	14	knowledge	knowledge	NOUN
ajst-29613	314	15	graphs	graph	NOUN
ajst-29613	314	16	using	use	VERB
ajst-29613	314	17	large	large	ADJ
ajst-29613	314	18	language	language	NOUN
ajst-29613	314	19	models[j	models[j	PROPN
ajst-29613	314	20	]	]	PUNCT
ajst-29613	314	21	.	.	PUNCT
ajst-29613	315	1	arxiv	arxiv	PROPN
ajst-29613	315	2	preprint	preprint	VERB
ajst-29613	315	3	arxiv:2305.01157	arxiv:2305.01157	ADV
ajst-29613	315	4	,	,	PUNCT
ajst-29613	315	5	2023	2023	NUM
ajst-29613	315	6	.	.	PUNCT
ajst-29613	316	1	[	[	X
ajst-29613	316	2	40	40	NUM
ajst-29613	316	3	]	]	X
ajst-29613	316	4	kim	kim	PROPN
ajst-29613	316	5	j	j	PROPN
ajst-29613	316	6	,	,	PUNCT
ajst-29613	316	7	kwon	kwon	VERB
ajst-29613	316	8	y	y	PROPN
ajst-29613	316	9	,	,	PUNCT
ajst-29613	316	10	jo	jo	PROPN
ajst-29613	316	11	y	y	PROPN
ajst-29613	316	12	,	,	PUNCT
ajst-29613	316	13	et	et	PROPN
ajst-29613	316	14	al	al	PROPN
ajst-29613	316	15	.	.	PUNCT
ajst-29613	317	1	kg	kg	PROPN
ajst-29613	317	2	-	-	PUNCT
ajst-29613	317	3	gpt	gpt	NOUN
ajst-29613	317	4	:	:	PUNCT
ajst-29613	317	5	a	a	DET
ajst-29613	317	6	general	general	ADJ
ajst-29613	317	7	framework	framework	NOUN
ajst-29613	317	8	for	for	ADP
ajst-29613	317	9	reasoning	reasoning	NOUN
ajst-29613	317	10	on	on	ADP
ajst-29613	317	11	knowledge	knowledge	NOUN
ajst-29613	317	12	graphs	graph	NOUN
ajst-29613	317	13	using	use	VERB
ajst-29613	317	14	large	large	ADJ
ajst-29613	317	15	language	language	NOUN
ajst-29613	317	16	models[j	models[j	PROPN
ajst-29613	317	17	]	]	PUNCT
ajst-29613	317	18	.	.	PUNCT
ajst-29613	318	1	arxiv	arxiv	PROPN
ajst-29613	318	2	preprint	preprint	NOUN
ajst-29613	318	3	arxiv:2310.11220	arxiv:2310.11220	NOUN
ajst-29613	318	4	,	,	PUNCT
ajst-29613	318	5	2023	2023	NUM
ajst-29613	318	6	.	.	PUNCT
ajst-29613	319	1	[	[	X
ajst-29613	319	2	41	41	NUM
ajst-29613	319	3	]	]	X
ajst-29613	319	4	li	li	PROPN
ajst-29613	319	5	s	s	PROPN
ajst-29613	319	6	,	,	PUNCT
ajst-29613	319	7	gao	gao	PROPN
ajst-29613	319	8	y	y	PROPN
ajst-29613	319	9	,	,	PUNCT
ajst-29613	319	10	jiang	jiang	PROPN
ajst-29613	319	11	h	h	PROPN
ajst-29613	319	12	,	,	PUNCT
ajst-29613	319	13	et	et	PROPN
ajst-29613	319	14	al	al	PROPN
ajst-29613	319	15	.	.	PROPN
ajst-29613	320	1	graph	graph	NOUN
ajst-29613	320	2	reasoning	reasoning	NOUN
ajst-29613	320	3	for	for	ADP
ajst-29613	320	4	question	question	NOUN
ajst-29613	320	5	answering	answer	VERB
ajst-29613	320	6	with	with	ADP
ajst-29613	320	7	triplet	triplet	NOUN
ajst-29613	320	8	retrieval[j	retrieval[j	PROPN
ajst-29613	320	9	]	]	PUNCT
ajst-29613	320	10	.	.	PUNCT
ajst-29613	321	1	arxiv	arxiv	PROPN
ajst-29613	321	2	preprint	preprint	PROPN
ajst-29613	321	3	arxiv	arxiv	PROPN
ajst-29613	321	4	:	:	PUNCT
ajst-29613	321	5	2305.18742	2305.18742	NUM
ajst-29613	321	6	,	,	PUNCT
ajst-29613	321	7	2023	2023	NUM
ajst-29613	321	8	.	.	PUNCT