id	sid	tid	token	lemma	pos
fpq-14292	1	1	the	the	DET
fpq-14292	1	2	bias	bias	NOUN
fpq-14292	1	3	dilemma	dilemma	NOUN
fpq-14292	1	4	:	:	PUNCT
fpq-14292	1	5	the	the	DET
fpq-14292	1	6	ethics	ethic	NOUN
fpq-14292	1	7	of	of	ADP
fpq-14292	1	8	algorithmic	algorithmic	ADJ
fpq-14292	1	9	bias	bias	NOUN
fpq-14292	1	10	in	in	ADP
fpq-14292	1	11	natural	natural	ADJ
fpq-14292	1	12	-	-	PUNCT
fpq-14292	1	13	language	language	NOUN
fpq-14292	1	14	processing	processing	NOUN
fpq-14292	1	15	feminist	feminist	ADJ
fpq-14292	1	16	philosophy	philosophy	NOUN
fpq-14292	1	17	quarterly	quarterly	ADJ
fpq-14292	1	18	volume	volume	NOUN
fpq-14292	1	19	8	8	NUM
fpq-14292	1	20	|	|	NOUN
fpq-14292	1	21	issue	issue	VERB
fpq-14292	1	22	3/4	3/4	NUM
fpq-14292	1	23	article	article	NOUN
fpq-14292	1	24	1	1	NUM
fpq-14292	1	25	recommended	recommend	VERB
fpq-14292	1	26	citation	citation	NOUN
fpq-14292	1	27	deery	deery	NOUN
fpq-14292	1	28	,	,	PUNCT
fpq-14292	1	29	oisín	oisín	NOUN
fpq-14292	1	30	,	,	PUNCT
fpq-14292	1	31	and	and	CCONJ
fpq-14292	1	32	katherine	katherine	PROPN
fpq-14292	1	33	bailey	bailey	PROPN
fpq-14292	1	34	.	.	PUNCT
fpq-14292	1	35	2022	2022	NUM
fpq-14292	1	36	.	.	PUNCT
fpq-14292	2	1	“	"	PUNCT
fpq-14292	2	2	the	the	DET
fpq-14292	2	3	bias	bias	NOUN
fpq-14292	2	4	dilemma	dilemma	NOUN
fpq-14292	2	5	:	:	PUNCT
fpq-14292	2	6	the	the	DET
fpq-14292	2	7	ethics	ethic	NOUN
fpq-14292	2	8	of	of	ADP
fpq-14292	2	9	algorithmic	algorithmic	ADJ
fpq-14292	2	10	bias	bias	NOUN
fpq-14292	2	11	in	in	ADP
fpq-14292	2	12	natural	natural	ADJ
fpq-14292	2	13	-	-	PUNCT
fpq-14292	2	14	language	language	NOUN
fpq-14292	2	15	processing	processing	NOUN
fpq-14292	2	16	.	.	PUNCT
fpq-14292	2	17	”	"	PUNCT
fpq-14292	3	1	feminist	feminist	ADJ
fpq-14292	3	2	philosophy	philosophy	NOUN
fpq-14292	3	3	quarterly	quarterly	ADJ
fpq-14292	3	4	8	8	NUM
fpq-14292	3	5	(	(	PUNCT
fpq-14292	3	6	3/4	3/4	NUM
fpq-14292	3	7	)	)	PUNCT
fpq-14292	3	8	.	.	PUNCT
fpq-14292	4	1	article	article	NOUN
fpq-14292	4	2	1	1	NUM
fpq-14292	4	3	.	.	NOUN
fpq-14292	4	4	2022	2022	NUM
fpq-14292	4	5	the	the	DET
fpq-14292	4	6	bias	bias	NOUN
fpq-14292	4	7	dilemma	dilemma	NOUN
fpq-14292	4	8	:	:	PUNCT
fpq-14292	4	9	the	the	DET
fpq-14292	4	10	ethics	ethic	NOUN
fpq-14292	4	11	of	of	ADP
fpq-14292	4	12	algorithmic	algorithmic	ADJ
fpq-14292	4	13	bias	bias	NOUN
fpq-14292	4	14	in	in	ADP
fpq-14292	4	15	natural	natural	ADJ
fpq-14292	4	16	-	-	PUNCT
fpq-14292	4	17	language	language	NOUN
fpq-14292	4	18	processing	processing	NOUN
fpq-14292	4	19	oisín	oisín	PROPN
fpq-14292	4	20	deery	deery	PROPN
fpq-14292	4	21	york	york	PROPN
fpq-14292	4	22	university	university	PROPN
fpq-14292	4	23	and	and	CCONJ
fpq-14292	4	24	macquarie	macquarie	PROPN
fpq-14292	4	25	university	university	PROPN
fpq-14292	4	26	oisin@oisindeery.com	oisin@oisindeery.com	PROPN
fpq-14292	5	1	katherine	katherine	PROPN
fpq-14292	5	2	bailey	bailey	PROPN
fpq-14292	5	3	shopify	shopify	PROPN
fpq-14292	5	4	inc	inc	PROPN
fpq-14292	5	5	.	.	PUNCT
fpq-14292	6	1	katherine@katbailey.net	katherine@katbailey.net	PROPN
fpq-14292	6	2	deery	deery	NOUN
fpq-14292	6	3	and	and	CCONJ
fpq-14292	6	4	bailey	bailey	PROPN
fpq-14292	6	5	–	–	PUNCT
fpq-14292	6	6	the	the	DET
fpq-14292	6	7	bias	bias	NOUN
fpq-14292	6	8	dilemma	dilemma	NOUN
fpq-14292	6	9	:	:	PUNCT
fpq-14292	6	10	the	the	DET
fpq-14292	6	11	ethics	ethic	NOUN
fpq-14292	6	12	of	of	ADP
fpq-14292	6	13	algorithmic	algorithmic	ADJ
fpq-14292	6	14	bias	bias	NOUN
fpq-14292	6	15	in	in	ADP
fpq-14292	6	16	natural	natural	ADJ
fpq-14292	6	17	-	-	PUNCT
fpq-14292	6	18	language	language	NOUN
fpq-14292	6	19	processing	processing	NOUN
fpq-14292	6	20	published	publish	VERB
fpq-14292	6	21	by	by	ADP
fpq-14292	6	22	scholarship@western	scholarship@western	PROPN
fpq-14292	6	23	,	,	PUNCT
fpq-14292	6	24	2022	2022	NUM
fpq-14292	6	25	1	1	NUM
fpq-14292	6	26	the	the	DET
fpq-14292	6	27	bias	bias	NOUN
fpq-14292	6	28	dilemma	dilemma	NOUN
fpq-14292	6	29	:	:	PUNCT
fpq-14292	6	30	the	the	DET
fpq-14292	6	31	ethics	ethic	NOUN
fpq-14292	6	32	of	of	ADP
fpq-14292	6	33	algorithmic	algorithmic	ADJ
fpq-14292	6	34	bias	bias	NOUN
fpq-14292	6	35	in	in	ADP
fpq-14292	6	36	natural	natural	ADJ
fpq-14292	6	37	-	-	PUNCT
fpq-14292	6	38	language	language	NOUN
fpq-14292	6	39	processing	processing	NOUN
fpq-14292	6	40	oisín	oisín	NOUN
fpq-14292	6	41	deery	deery	NOUN
fpq-14292	6	42	and	and	CCONJ
fpq-14292	6	43	katherine	katherine	PROPN
fpq-14292	6	44	bailey	bailey	PROPN
fpq-14292	6	45	*	*	AUX
fpq-14292	6	46	abstract	abstract	ADJ
fpq-14292	6	47	addressing	address	VERB
fpq-14292	6	48	biases	bias	NOUN
fpq-14292	6	49	in	in	ADP
fpq-14292	6	50	natural	natural	ADJ
fpq-14292	6	51	-	-	PUNCT
fpq-14292	6	52	language	language	NOUN
fpq-14292	6	53	processing	processing	NOUN
fpq-14292	6	54	(	(	PUNCT
fpq-14292	6	55	nlp	nlp	NOUN
fpq-14292	6	56	)	)	PUNCT
fpq-14292	6	57	systems	system	NOUN
fpq-14292	6	58	presents	present	VERB
fpq-14292	6	59	an	an	DET
fpq-14292	6	60	underappreciated	underappreciated	ADJ
fpq-14292	6	61	ethical	ethical	ADJ
fpq-14292	6	62	dilemma	dilemma	NOUN
fpq-14292	6	63	,	,	PUNCT
fpq-14292	6	64	which	which	PRON
fpq-14292	6	65	we	we	PRON
fpq-14292	6	66	think	think	VERB
fpq-14292	6	67	underlies	underlie	VERB
fpq-14292	6	68	recent	recent	ADJ
fpq-14292	6	69	debates	debate	NOUN
fpq-14292	6	70	about	about	ADP
fpq-14292	6	71	bias	bias	NOUN
fpq-14292	6	72	in	in	ADP
fpq-14292	6	73	nlp	nlp	NOUN
fpq-14292	6	74	models	model	NOUN
fpq-14292	6	75	.	.	PUNCT
fpq-14292	7	1	in	in	ADP
fpq-14292	7	2	brief	brief	ADJ
fpq-14292	7	3	,	,	PUNCT
fpq-14292	7	4	even	even	ADV
fpq-14292	7	5	if	if	SCONJ
fpq-14292	7	6	we	we	PRON
fpq-14292	7	7	could	could	AUX
fpq-14292	7	8	eliminate	eliminate	VERB
fpq-14292	7	9	bias	bias	NOUN
fpq-14292	7	10	from	from	ADP
fpq-14292	7	11	language	language	NOUN
fpq-14292	7	12	models	model	NOUN
fpq-14292	7	13	or	or	CCONJ
fpq-14292	7	14	their	their	PRON
fpq-14292	7	15	outputs	output	NOUN
fpq-14292	7	16	,	,	PUNCT
fpq-14292	7	17	we	we	PRON
fpq-14292	7	18	would	would	AUX
fpq-14292	7	19	thereby	thereby	ADV
fpq-14292	7	20	often	often	ADV
fpq-14292	7	21	withhold	withhold	VERB
fpq-14292	7	22	descriptively	descriptively	ADV
fpq-14292	7	23	or	or	CCONJ
fpq-14292	7	24	ethically	ethically	ADV
fpq-14292	7	25	useful	useful	ADJ
fpq-14292	7	26	information	information	NOUN
fpq-14292	7	27	,	,	PUNCT
fpq-14292	7	28	despite	despite	SCONJ
fpq-14292	7	29	avoiding	avoid	VERB
fpq-14292	7	30	perpetuating	perpetuate	VERB
fpq-14292	7	31	or	or	CCONJ
fpq-14292	7	32	amplifying	amplifying	ADJ
fpq-14292	7	33	bias	bias	NOUN
fpq-14292	7	34	.	.	PUNCT
fpq-14292	8	1	yet	yet	ADV
fpq-14292	8	2	if	if	SCONJ
fpq-14292	8	3	we	we	PRON
fpq-14292	8	4	do	do	AUX
fpq-14292	8	5	not	not	PART
fpq-14292	8	6	debias	debia	NOUN
fpq-14292	8	7	,	,	PUNCT
fpq-14292	8	8	we	we	PRON
fpq-14292	8	9	can	can	AUX
fpq-14292	8	10	perpetuate	perpetuate	VERB
fpq-14292	8	11	or	or	CCONJ
fpq-14292	8	12	amplify	amplify	VERB
fpq-14292	8	13	bias	bias	NOUN
fpq-14292	8	14	,	,	PUNCT
fpq-14292	8	15	even	even	ADV
fpq-14292	8	16	if	if	SCONJ
fpq-14292	8	17	we	we	PRON
fpq-14292	8	18	retain	retain	VERB
fpq-14292	8	19	relevant	relevant	ADJ
fpq-14292	8	20	descriptively	descriptively	ADV
fpq-14292	8	21	or	or	CCONJ
fpq-14292	8	22	ethically	ethically	ADV
fpq-14292	8	23	useful	useful	ADJ
fpq-14292	8	24	information	information	NOUN
fpq-14292	8	25	.	.	PUNCT
fpq-14292	9	1	understanding	understand	VERB
fpq-14292	9	2	this	this	DET
fpq-14292	9	3	dilemma	dilemma	NOUN
fpq-14292	9	4	provides	provide	VERB
fpq-14292	9	5	for	for	ADP
fpq-14292	9	6	a	a	DET
fpq-14292	9	7	useful	useful	ADJ
fpq-14292	9	8	way	way	NOUN
fpq-14292	9	9	of	of	ADP
fpq-14292	9	10	rethinking	rethink	VERB
fpq-14292	9	11	the	the	DET
fpq-14292	9	12	ethics	ethic	NOUN
fpq-14292	9	13	of	of	ADP
fpq-14292	9	14	algorithmic	algorithmic	ADJ
fpq-14292	9	15	bias	bias	NOUN
fpq-14292	9	16	in	in	ADP
fpq-14292	9	17	nlp	nlp	NOUN
fpq-14292	9	18	.	.	PUNCT
fpq-14292	10	1	keywords	keyword	NOUN
fpq-14292	10	2	:	:	PUNCT
fpq-14292	10	3	artificial	artificial	ADJ
fpq-14292	10	4	intelligence	intelligence	NOUN
fpq-14292	10	5	,	,	PUNCT
fpq-14292	10	6	algorithms	algorithm	NOUN
fpq-14292	10	7	,	,	PUNCT
fpq-14292	10	8	bias	bias	NOUN
fpq-14292	10	9	1	1	NUM
fpq-14292	10	10	.	.	PUNCT
fpq-14292	11	1	introduction	introduction	NOUN
fpq-14292	11	2	bias	bias	NOUN
fpq-14292	11	3	in	in	ADP
fpq-14292	11	4	statistical	statistical	ADJ
fpq-14292	11	5	models	model	NOUN
fpq-14292	11	6	or	or	CCONJ
fpq-14292	11	7	algorithms	algorithm	NOUN
fpq-14292	11	8	can	can	AUX
fpq-14292	11	9	result	result	VERB
fpq-14292	11	10	in	in	ADP
fpq-14292	11	11	moral	moral	ADJ
fpq-14292	11	12	harms	harm	NOUN
fpq-14292	11	13	.	.	PUNCT
fpq-14292	12	1	a	a	DET
fpq-14292	12	2	model	model	NOUN
fpq-14292	12	3	is	be	AUX
fpq-14292	12	4	an	an	DET
fpq-14292	12	5	abstract	abstract	ADJ
fpq-14292	12	6	representation	representation	NOUN
fpq-14292	12	7	of	of	ADP
fpq-14292	12	8	a	a	DET
fpq-14292	12	9	phenomenon	phenomenon	NOUN
fpq-14292	12	10	,	,	PUNCT
fpq-14292	12	11	such	such	ADJ
fpq-14292	12	12	as	as	ADP
fpq-14292	12	13	how	how	SCONJ
fpq-14292	12	14	words	word	NOUN
fpq-14292	12	15	relate	relate	VERB
fpq-14292	12	16	in	in	ADP
fpq-14292	12	17	a	a	DET
fpq-14292	12	18	natural	natural	ADJ
fpq-14292	12	19	language	language	NOUN
fpq-14292	12	20	.	.	PUNCT
fpq-14292	13	1	we	we	PRON
fpq-14292	13	2	prefer	prefer	VERB
fpq-14292	13	3	the	the	DET
fpq-14292	13	4	term	term	NOUN
fpq-14292	13	5	“	"	PUNCT
fpq-14292	13	6	model	model	NOUN
fpq-14292	13	7	”	"	PUNCT
fpq-14292	13	8	to	to	ADP
fpq-14292	13	9	“	"	PUNCT
fpq-14292	13	10	algorithm	algorithm	NOUN
fpq-14292	13	11	,	,	PUNCT
fpq-14292	13	12	”	"	PUNCT
fpq-14292	13	13	since	since	SCONJ
fpq-14292	13	14	strictly	strictly	ADV
fpq-14292	13	15	speaking	speak	VERB
fpq-14292	13	16	“	"	PUNCT
fpq-14292	13	17	algorithm	algorithm	NOUN
fpq-14292	13	18	”	"	PUNCT
fpq-14292	13	19	refers	refer	VERB
fpq-14292	13	20	only	only	ADV
fpq-14292	13	21	to	to	ADP
fpq-14292	13	22	the	the	DET
fpq-14292	13	23	actual	actual	ADJ
fpq-14292	13	24	sequence	sequence	NOUN
fpq-14292	13	25	of	of	ADP
fpq-14292	13	26	steps	step	NOUN
fpq-14292	13	27	carried	carry	VERB
fpq-14292	13	28	out	out	ADP
fpq-14292	13	29	during	during	ADP
fpq-14292	13	30	the	the	DET
fpq-14292	13	31	training	training	NOUN
fpq-14292	13	32	of	of	ADP
fpq-14292	13	33	a	a	DET
fpq-14292	13	34	model	model	NOUN
fpq-14292	13	35	or	or	CCONJ
fpq-14292	13	36	at	at	ADP
fpq-14292	13	37	the	the	DET
fpq-14292	13	38	time	time	NOUN
fpq-14292	13	39	it	it	PRON
fpq-14292	13	40	issues	issue	VERB
fpq-14292	13	41	predictions	prediction	NOUN
fpq-14292	13	42	.	.	PUNCT
fpq-14292	14	1	however	however	ADV
fpq-14292	14	2	,	,	PUNCT
fpq-14292	14	3	everyday	everyday	ADJ
fpq-14292	14	4	use	use	NOUN
fpq-14292	14	5	of	of	ADP
fpq-14292	14	6	“	"	PUNCT
fpq-14292	14	7	algorithm	algorithm	NOUN
fpq-14292	14	8	”	"	PUNCT
fpq-14292	14	9	generally	generally	ADV
fpq-14292	14	10	picks	pick	VERB
fpq-14292	14	11	out	out	ADP
fpq-14292	14	12	the	the	DET
fpq-14292	14	13	already	already	ADV
fpq-14292	14	14	-	-	PUNCT
fpq-14292	14	15	trained	train	VERB
fpq-14292	14	16	model	model	NOUN
fpq-14292	14	17	,	,	PUNCT
fpq-14292	14	18	which	which	PRON
fpq-14292	14	19	implicitly	implicitly	ADV
fpq-14292	14	20	includes	include	VERB
fpq-14292	14	21	the	the	DET
fpq-14292	14	22	data	datum	NOUN
fpq-14292	14	23	on	on	ADP
fpq-14292	14	24	which	which	PRON
fpq-14292	14	25	that	that	DET
fpq-14292	14	26	model	model	NOUN
fpq-14292	14	27	was	be	AUX
fpq-14292	14	28	trained	train	VERB
fpq-14292	14	29	.	.	PUNCT
fpq-14292	15	1	if	if	SCONJ
fpq-14292	15	2	the	the	DET
fpq-14292	15	3	data	datum	NOUN
fpq-14292	15	4	are	be	AUX
fpq-14292	15	5	biased	bias	VERB
fpq-14292	15	6	,	,	PUNCT
fpq-14292	15	7	then	then	ADV
fpq-14292	15	8	there	there	PRON
fpq-14292	15	9	are	be	VERB
fpq-14292	15	10	biased	bias	VERB
fpq-14292	15	11	algorithms.1	algorithms.1	ADP
fpq-14292	15	12	yet	yet	ADV
fpq-14292	15	13	for	for	ADP
fpq-14292	15	14	simplicity	simplicity	NOUN
fpq-14292	15	15	’s	’s	PART
fpq-14292	15	16	sake	sake	NOUN
fpq-14292	15	17	,	,	PUNCT
fpq-14292	15	18	we	we	PRON
fpq-14292	15	19	will	will	AUX
fpq-14292	15	20	mostly	mostly	ADV
fpq-14292	15	21	stick	stick	VERB
fpq-14292	15	22	with	with	ADP
fpq-14292	15	23	“	"	PUNCT
fpq-14292	15	24	model	model	NOUN
fpq-14292	15	25	.	.	PUNCT
fpq-14292	15	26	”	"	PUNCT
fpq-14292	16	1	reliance	reliance	NOUN
fpq-14292	16	2	on	on	ADP
fpq-14292	16	3	biased	biased	ADJ
fpq-14292	16	4	models	model	NOUN
fpq-14292	16	5	can	can	AUX
fpq-14292	16	6	exacerbate	exacerbate	VERB
fpq-14292	16	7	marginalization	marginalization	NOUN
fpq-14292	16	8	of	of	ADP
fpq-14292	16	9	vulnerable	vulnerable	ADJ
fpq-14292	16	10	groups	group	NOUN
fpq-14292	16	11	.	.	PUNCT
fpq-14292	17	1	for	for	ADP
fpq-14292	17	2	instance	instance	NOUN
fpq-14292	17	3	,	,	PUNCT
fpq-14292	17	4	the	the	DET
fpq-14292	17	5	ethical	ethical	ADJ
fpq-14292	17	6	issues	issue	NOUN
fpq-14292	17	7	related	relate	VERB
fpq-14292	17	8	to	to	ADP
fpq-14292	17	9	statistical	statistical	ADJ
fpq-14292	17	10	models	model	NOUN
fpq-14292	17	11	in	in	ADP
fpq-14292	17	12	hiring	hiring	NOUN
fpq-14292	17	13	have	have	AUX
fpq-14292	17	14	been	be	AUX
fpq-14292	17	15	widely	widely	ADV
fpq-14292	17	16	discussed	discuss	VERB
fpq-14292	17	17	(	(	PUNCT
fpq-14292	17	18	e.g.	e.g.	ADV
fpq-14292	17	19	,	,	PUNCT
fpq-14292	17	20	hu	hu	PROPN
fpq-14292	17	21	and	and	CCONJ
fpq-14292	17	22	chen	chen	PROPN
fpq-14292	17	23	2017	2017	NUM
fpq-14292	17	24	)	)	PUNCT
fpq-14292	17	25	.	.	PUNCT
fpq-14292	18	1	problems	problem	NOUN
fpq-14292	18	2	can	can	AUX
fpq-14292	18	3	arise	arise	VERB
fpq-14292	18	4	even	even	ADV
fpq-14292	18	5	when	when	SCONJ
fpq-14292	18	6	models	model	NOUN
fpq-14292	18	7	are	be	AUX
fpq-14292	18	8	deployed	deploy	VERB
fpq-14292	18	9	to	to	PART
fpq-14292	18	10	reduce	reduce	VERB
fpq-14292	18	11	the	the	DET
fpq-14292	18	12	influence	influence	NOUN
fpq-14292	18	13	of	of	ADP
fpq-14292	18	14	bias	bias	NOUN
fpq-14292	18	15	.	.	PUNCT
fpq-14292	19	1	thus	thus	ADV
fpq-14292	19	2	,	,	PUNCT
fpq-14292	19	3	recidivism	recidivism	NOUN
fpq-14292	19	4	models	model	NOUN
fpq-14292	19	5	aim	aim	VERB
fpq-14292	19	6	to	to	PART
fpq-14292	19	7	reduce	reduce	VERB
fpq-14292	19	8	the	the	DET
fpq-14292	19	9	*	*	PUNCT
fpq-14292	19	10	oisín	oisín	PROPN
fpq-14292	19	11	deery	deery	NOUN
fpq-14292	19	12	,	,	PUNCT
fpq-14292	19	13	departments	department	NOUN
fpq-14292	19	14	of	of	ADP
fpq-14292	19	15	philosophy	philosophy	NOUN
fpq-14292	19	16	at	at	ADP
fpq-14292	19	17	macquarie	macquarie	PROPN
fpq-14292	19	18	university	university	PROPN
fpq-14292	19	19	,	,	PUNCT
fpq-14292	19	20	sydney	sydney	PROPN
fpq-14292	19	21	,	,	PUNCT
fpq-14292	19	22	australia	australia	PROPN
fpq-14292	19	23	&	&	CCONJ
fpq-14292	19	24	york	york	PROPN
fpq-14292	19	25	university	university	PROPN
fpq-14292	19	26	,	,	PUNCT
fpq-14292	19	27	toronto	toronto	PROPN
fpq-14292	19	28	,	,	PUNCT
fpq-14292	19	29	canada	canada	PROPN
fpq-14292	19	30	(	(	PUNCT
fpq-14292	19	31	oisin@oisindeery.com	oisin@oisindeery.com	PROPN
fpq-14292	19	32	)	)	PUNCT
fpq-14292	19	33	katherine	katherine	PROPN
fpq-14292	19	34	bailey	bailey	PROPN
fpq-14292	19	35	,	,	PUNCT
fpq-14292	19	36	shopify	shopify	PROPN
fpq-14292	19	37	inc	inc	PROPN
fpq-14292	19	38	.	.	PROPN
fpq-14292	19	39	,	,	PUNCT
fpq-14292	19	40	toronto	toronto	PROPN
fpq-14292	19	41	,	,	PUNCT
fpq-14292	19	42	canada	canada	PROPN
fpq-14292	19	43	(	(	PUNCT
fpq-14292	19	44	katherine@katbailey.net	katherine@katbailey.net	PROPN
fpq-14292	19	45	)	)	PUNCT
fpq-14292	19	46	1	1	NUM
fpq-14292	19	47	algorithms	algorithm	NOUN
fpq-14292	19	48	can	can	AUX
fpq-14292	19	49	be	be	AUX
fpq-14292	19	50	biased	bias	VERB
fpq-14292	19	51	for	for	ADP
fpq-14292	19	52	many	many	ADJ
fpq-14292	19	53	reasons	reason	NOUN
fpq-14292	19	54	(	(	PUNCT
fpq-14292	19	55	see	see	VERB
fpq-14292	19	56	sections	section	NOUN
fpq-14292	19	57	6.3	6.3	NUM
fpq-14292	19	58	and	and	CCONJ
fpq-14292	19	59	8)	8)	NUM
fpq-14292	19	60	.	.	PUNCT
fpq-14292	20	1	we	we	PRON
fpq-14292	20	2	focus	focus	VERB
fpq-14292	20	3	on	on	ADP
fpq-14292	20	4	bias	bias	NOUN
fpq-14292	20	5	deriving	derive	VERB
fpq-14292	20	6	from	from	ADP
fpq-14292	20	7	data	data	PROPN
fpq-14292	20	8	.	.	PUNCT
fpq-14292	21	1	feminist	feminist	PROPN
fpq-14292	21	2	philosophy	philosophy	PROPN
fpq-14292	21	3	quarterly	quarterly	ADV
fpq-14292	21	4	,	,	PUNCT
fpq-14292	21	5	2022	2022	NUM
fpq-14292	21	6	,	,	PUNCT
fpq-14292	21	7	vol	vol	NOUN
fpq-14292	21	8	.	.	NOUN
fpq-14292	21	9	8	8	NUM
fpq-14292	21	10	,	,	PUNCT
fpq-14292	21	11	iss	iss	PROPN
fpq-14292	21	12	.	.	PUNCT
fpq-14292	21	13	3/4	3/4	NUM
fpq-14292	21	14	,	,	PUNCT
fpq-14292	21	15	article	article	NOUN
fpq-14292	21	16	1	1	NUM
fpq-14292	21	17	published	publish	VERB
fpq-14292	21	18	by	by	ADP
fpq-14292	21	19	scholarship@western	scholarship@western	PROPN
fpq-14292	21	20	,	,	PUNCT
fpq-14292	21	21	2022	2022	NUM
fpq-14292	21	22	2	2	NUM
fpq-14292	21	23	influence	influence	NOUN
fpq-14292	21	24	of	of	ADP
fpq-14292	21	25	bias	bias	NOUN
fpq-14292	21	26	in	in	ADP
fpq-14292	21	27	decision	decision	NOUN
fpq-14292	21	28	-	-	PUNCT
fpq-14292	21	29	making	making	NOUN
fpq-14292	21	30	when	when	SCONJ
fpq-14292	21	31	sentencing	sentencing	NOUN
fpq-14292	21	32	offenders	offender	NOUN
fpq-14292	21	33	yet	yet	ADV
fpq-14292	21	34	often	often	ADV
fpq-14292	21	35	serve	serve	VERB
fpq-14292	21	36	only	only	ADV
fpq-14292	21	37	to	to	PART
fpq-14292	21	38	reproduce	reproduce	VERB
fpq-14292	21	39	or	or	CCONJ
fpq-14292	21	40	disguise	disguise	ADJ
fpq-14292	21	41	bias	bias	NOUN
fpq-14292	21	42	(	(	PUNCT
fpq-14292	21	43	kehl	kehl	PROPN
fpq-14292	21	44	,	,	PUNCT
fpq-14292	21	45	guo	guo	PROPN
fpq-14292	21	46	,	,	PUNCT
fpq-14292	21	47	and	and	CCONJ
fpq-14292	21	48	kessler	kessler	PROPN
fpq-14292	22	1	2017).2	2017).2	NUM
fpq-14292	22	2	more	more	ADV
fpq-14292	22	3	generally	generally	ADV
fpq-14292	22	4	,	,	PUNCT
fpq-14292	22	5	numerous	numerous	ADJ
fpq-14292	22	6	ethical	ethical	ADJ
fpq-14292	22	7	questions	question	NOUN
fpq-14292	22	8	arise	arise	VERB
fpq-14292	22	9	in	in	ADP
fpq-14292	22	10	relation	relation	NOUN
fpq-14292	22	11	to	to	ADP
fpq-14292	22	12	machine	machine	NOUN
fpq-14292	22	13	learning	learning	NOUN
fpq-14292	22	14	(	(	PUNCT
fpq-14292	22	15	ml	ml	NOUN
fpq-14292	22	16	)	)	PUNCT
fpq-14292	22	17	and	and	CCONJ
fpq-14292	22	18	artificial	artificial	ADJ
fpq-14292	22	19	intelligence	intelligence	NOUN
fpq-14292	22	20	(	(	PUNCT
fpq-14292	22	21	ai	ai	NOUN
fpq-14292	22	22	)	)	PUNCT
fpq-14292	22	23	,	,	PUNCT
fpq-14292	22	24	most	most	ADJ
fpq-14292	22	25	of	of	ADP
fpq-14292	22	26	which	which	PRON
fpq-14292	22	27	we	we	PRON
fpq-14292	22	28	leave	leave	VERB
fpq-14292	22	29	aside	aside	ADV
fpq-14292	22	30	.	.	PUNCT
fpq-14292	23	1	instead	instead	ADV
fpq-14292	23	2	,	,	PUNCT
fpq-14292	23	3	our	our	PRON
fpq-14292	23	4	focus	focus	NOUN
fpq-14292	23	5	is	be	AUX
fpq-14292	23	6	on	on	ADP
fpq-14292	23	7	bias	bias	NOUN
fpq-14292	23	8	in	in	ADP
fpq-14292	23	9	the	the	DET
fpq-14292	23	10	data	datum	NOUN
fpq-14292	23	11	on	on	ADP
fpq-14292	23	12	which	which	PRON
fpq-14292	23	13	ml	ml	NOUN
fpq-14292	23	14	models	model	NOUN
fpq-14292	23	15	are	be	AUX
fpq-14292	23	16	trained	train	VERB
fpq-14292	23	17	,	,	PUNCT
fpq-14292	23	18	and	and	CCONJ
fpq-14292	23	19	especially	especially	ADV
fpq-14292	23	20	language	language	NOUN
fpq-14292	23	21	models	model	NOUN
fpq-14292	23	22	.	.	PUNCT
fpq-14292	24	1	the	the	DET
fpq-14292	24	2	central	central	ADJ
fpq-14292	24	3	issue	issue	NOUN
fpq-14292	24	4	here	here	ADV
fpq-14292	24	5	is	be	AUX
fpq-14292	24	6	that	that	SCONJ
fpq-14292	24	7	the	the	DET
fpq-14292	24	8	corpora	corpora	NOUN
fpq-14292	24	9	on	on	ADP
fpq-14292	24	10	which	which	PRON
fpq-14292	24	11	such	such	ADJ
fpq-14292	24	12	models	model	NOUN
fpq-14292	24	13	are	be	AUX
fpq-14292	24	14	trained	train	VERB
fpq-14292	24	15	typically	typically	ADV
fpq-14292	24	16	contain	contain	VERB
fpq-14292	24	17	biases	bias	NOUN
fpq-14292	24	18	—	—	PUNCT
fpq-14292	24	19	for	for	ADP
fpq-14292	24	20	example	example	NOUN
fpq-14292	24	21	,	,	PUNCT
fpq-14292	24	22	sexist	sexist	ADJ
fpq-14292	24	23	or	or	CCONJ
fpq-14292	24	24	racist	racist	ADJ
fpq-14292	24	25	biases	bias	NOUN
fpq-14292	24	26	.	.	PUNCT
fpq-14292	25	1	that	that	PRON
fpq-14292	25	2	is	be	AUX
fpq-14292	25	3	because	because	SCONJ
fpq-14292	25	4	the	the	DET
fpq-14292	25	5	corpora	corpora	NOUN
fpq-14292	25	6	comprise	comprise	NOUN
fpq-14292	25	7	examples	example	NOUN
fpq-14292	25	8	of	of	ADP
fpq-14292	25	9	how	how	SCONJ
fpq-14292	25	10	people	people	NOUN
fpq-14292	25	11	actually	actually	ADV
fpq-14292	25	12	use	use	VERB
fpq-14292	25	13	language	language	NOUN
fpq-14292	25	14	,	,	PUNCT
fpq-14292	25	15	and	and	CCONJ
fpq-14292	25	16	people	people	NOUN
fpq-14292	25	17	use	use	VERB
fpq-14292	25	18	language	language	NOUN
fpq-14292	25	19	in	in	ADP
fpq-14292	25	20	sexist	sexist	ADJ
fpq-14292	25	21	and	and	CCONJ
fpq-14292	25	22	racist	racist	ADJ
fpq-14292	25	23	ways.3	ways.3	PROPN
fpq-14292	25	24	thus	thus	ADV
fpq-14292	25	25	,	,	PUNCT
fpq-14292	25	26	the	the	DET
fpq-14292	25	27	outputs	output	NOUN
fpq-14292	25	28	of	of	ADP
fpq-14292	25	29	models	model	NOUN
fpq-14292	25	30	trained	train	VERB
fpq-14292	25	31	on	on	ADP
fpq-14292	25	32	such	such	ADJ
fpq-14292	25	33	data	datum	NOUN
fpq-14292	25	34	reflect	reflect	VERB
fpq-14292	25	35	biases	bias	NOUN
fpq-14292	25	36	(	(	PUNCT
fpq-14292	25	37	see	see	VERB
fpq-14292	25	38	,	,	PUNCT
fpq-14292	25	39	e.g.	e.g.	ADV
fpq-14292	25	40	,	,	PUNCT
fpq-14292	25	41	the	the	DET
fpq-14292	25	42	papers	paper	NOUN
fpq-14292	25	43	surveyed	survey	VERB
fpq-14292	25	44	by	by	ADP
fpq-14292	25	45	blodgett	blodgett	PROPN
fpq-14292	25	46	et	et	PROPN
fpq-14292	25	47	al	al	PROPN
fpq-14292	25	48	.	.	PROPN
fpq-14292	25	49	2020	2020	NUM
fpq-14292	25	50	)	)	PUNCT
fpq-14292	25	51	.	.	PUNCT
fpq-14292	26	1	we	we	PRON
fpq-14292	26	2	must	must	AUX
fpq-14292	26	3	ask	ask	VERB
fpq-14292	26	4	whether	whether	SCONJ
fpq-14292	26	5	(	(	PUNCT
fpq-14292	26	6	1	1	X
fpq-14292	26	7	)	)	PUNCT
fpq-14292	26	8	it	it	PRON
fpq-14292	26	9	is	be	AUX
fpq-14292	26	10	practically	practically	ADV
fpq-14292	26	11	speaking	speak	VERB
fpq-14292	26	12	possible	possible	ADJ
fpq-14292	26	13	to	to	ADP
fpq-14292	26	14	debias	debias	PROPN
fpq-14292	26	15	language	language	NOUN
fpq-14292	26	16	models	model	NOUN
fpq-14292	26	17	and	and	CCONJ
fpq-14292	26	18	(	(	PUNCT
fpq-14292	26	19	2	2	X
fpq-14292	26	20	)	)	PUNCT
fpq-14292	26	21	whether	whether	SCONJ
fpq-14292	26	22	we	we	PRON
fpq-14292	26	23	should	should	AUX
fpq-14292	26	24	do	do	VERB
fpq-14292	26	25	so	so	ADV
fpq-14292	26	26	,	,	PUNCT
fpq-14292	26	27	if	if	SCONJ
fpq-14292	26	28	possible	possible	ADJ
fpq-14292	26	29	.	.	PUNCT
fpq-14292	27	1	in	in	ADP
fpq-14292	27	2	relation	relation	NOUN
fpq-14292	27	3	to	to	ADP
fpq-14292	27	4	(	(	PUNCT
fpq-14292	27	5	2	2	NUM
fpq-14292	27	6	)	)	PUNCT
fpq-14292	27	7	,	,	PUNCT
fpq-14292	27	8	a	a	DET
fpq-14292	27	9	central	central	ADJ
fpq-14292	27	10	issue	issue	NOUN
fpq-14292	27	11	is	be	AUX
fpq-14292	27	12	whether	whether	SCONJ
fpq-14292	27	13	there	there	PRON
fpq-14292	27	14	might	might	AUX
fpq-14292	27	15	be	be	AUX
fpq-14292	27	16	practical	practical	ADJ
fpq-14292	27	17	or	or	CCONJ
fpq-14292	27	18	ethical	ethical	ADJ
fpq-14292	27	19	value	value	NOUN
fpq-14292	27	20	to	to	ADP
fpq-14292	27	21	not	not	PART
fpq-14292	27	22	debiasing	debiase	VERB
fpq-14292	27	23	models	model	NOUN
fpq-14292	27	24	.	.	PUNCT
fpq-14292	28	1	some	some	DET
fpq-14292	28	2	maintain	maintain	VERB
fpq-14292	28	3	—	—	PUNCT
fpq-14292	28	4	as	as	SCONJ
fpq-14292	28	5	we	we	PRON
fpq-14292	28	6	do	do	VERB
fpq-14292	28	7	—	—	PUNCT
fpq-14292	28	8	that	that	SCONJ
fpq-14292	28	9	not	not	PART
fpq-14292	28	10	debiasing	debiase	VERB
fpq-14292	28	11	can	can	AUX
fpq-14292	28	12	be	be	AUX
fpq-14292	28	13	valuable	valuable	ADJ
fpq-14292	28	14	,	,	PUNCT
fpq-14292	28	15	since	since	SCONJ
fpq-14292	28	16	models	model	NOUN
fpq-14292	28	17	often	often	ADV
fpq-14292	28	18	reflect	reflect	VERB
fpq-14292	28	19	descriptively	descriptively	ADV
fpq-14292	28	20	accurate	accurate	ADJ
fpq-14292	28	21	information	information	NOUN
fpq-14292	28	22	about	about	ADP
fpq-14292	28	23	how	how	SCONJ
fpq-14292	28	24	language	language	NOUN
fpq-14292	28	25	is	be	AUX
fpq-14292	28	26	actually	actually	ADV
fpq-14292	28	27	used	use	VERB
fpq-14292	28	28	(	(	PUNCT
fpq-14292	28	29	see	see	VERB
fpq-14292	28	30	goldberg	goldberg	PROPN
fpq-14292	28	31	2021	2021	NUM
fpq-14292	28	32	)	)	PUNCT
fpq-14292	28	33	.	.	PUNCT
fpq-14292	29	1	by	by	ADP
fpq-14292	29	2	contrast	contrast	NOUN
fpq-14292	29	3	,	,	PUNCT
fpq-14292	29	4	others	other	NOUN
fpq-14292	29	5	claim	claim	VERB
fpq-14292	29	6	that	that	SCONJ
fpq-14292	29	7	not	not	PART
fpq-14292	29	8	debiasing	debiase	VERB
fpq-14292	29	9	runs	run	VERB
fpq-14292	29	10	the	the	DET
fpq-14292	29	11	risk	risk	NOUN
fpq-14292	29	12	of	of	ADP
fpq-14292	29	13	perpetuating	perpetuate	VERB
fpq-14292	29	14	and	and	CCONJ
fpq-14292	29	15	amplifying	amplifying	ADJ
fpq-14292	29	16	biases	bias	NOUN
fpq-14292	29	17	(	(	PUNCT
fpq-14292	29	18	bolukbasi	bolukbasi	NOUN
fpq-14292	29	19	et	et	PROPN
fpq-14292	29	20	al	al	PROPN
fpq-14292	29	21	.	.	PROPN
fpq-14292	29	22	2016	2016	NUM
fpq-14292	29	23	)	)	PUNCT
fpq-14292	29	24	.	.	PUNCT
fpq-14292	30	1	some	some	PRON
fpq-14292	30	2	argue	argue	VERB
fpq-14292	30	3	that	that	SCONJ
fpq-14292	30	4	existing	exist	VERB
fpq-14292	30	5	models	model	NOUN
fpq-14292	30	6	do	do	AUX
fpq-14292	30	7	not	not	PART
fpq-14292	30	8	actually	actually	ADV
fpq-14292	30	9	provide	provide	VERB
fpq-14292	30	10	descriptively	descriptively	ADV
fpq-14292	30	11	accurate	accurate	ADJ
fpq-14292	30	12	information	information	NOUN
fpq-14292	30	13	at	at	ADV
fpq-14292	30	14	all	all	ADV
fpq-14292	30	15	,	,	PUNCT
fpq-14292	30	16	since	since	SCONJ
fpq-14292	30	17	the	the	DET
fpq-14292	30	18	corpora	corpora	NOUN
fpq-14292	30	19	on	on	ADP
fpq-14292	30	20	which	which	PRON
fpq-14292	30	21	the	the	DET
fpq-14292	30	22	models	model	NOUN
fpq-14292	30	23	are	be	AUX
fpq-14292	30	24	trained	train	VERB
fpq-14292	30	25	do	do	AUX
fpq-14292	30	26	not	not	PART
fpq-14292	30	27	reflect	reflect	VERB
fpq-14292	30	28	marginalized	marginalized	ADJ
fpq-14292	30	29	voices	voice	NOUN
fpq-14292	30	30	(	(	PUNCT
fpq-14292	30	31	bender	bender	NOUN
fpq-14292	30	32	,	,	PUNCT
fpq-14292	30	33	gebru	gebru	VERB
fpq-14292	30	34	et	et	PROPN
fpq-14292	30	35	al	al	PROPN
fpq-14292	30	36	.	.	PROPN
fpq-14292	30	37	2021	2021	NUM
fpq-14292	30	38	)	)	PUNCT
fpq-14292	30	39	.	.	PUNCT
fpq-14292	31	1	these	these	DET
fpq-14292	31	2	debates	debate	NOUN
fpq-14292	31	3	do	do	AUX
fpq-14292	31	4	not	not	PART
fpq-14292	31	5	deploy	deploy	VERB
fpq-14292	31	6	the	the	DET
fpq-14292	31	7	descriptive	descriptive	ADJ
fpq-14292	31	8	-	-	ADJ
fpq-14292	31	9	normative	normative	ADJ
fpq-14292	31	10	distinction	distinction	NOUN
fpq-14292	31	11	that	that	SCONJ
fpq-14292	31	12	we	we	PRON
fpq-14292	31	13	will	will	AUX
fpq-14292	31	14	outline	outline	VERB
fpq-14292	31	15	,	,	PUNCT
fpq-14292	31	16	yet	yet	CCONJ
fpq-14292	31	17	this	this	DET
fpq-14292	31	18	distinction	distinction	NOUN
fpq-14292	31	19	helps	help	VERB
fpq-14292	31	20	to	to	PART
fpq-14292	31	21	clarify	clarify	VERB
fpq-14292	31	22	the	the	DET
fpq-14292	31	23	points	point	NOUN
fpq-14292	31	24	of	of	ADP
fpq-14292	31	25	contention	contention	NOUN
fpq-14292	31	26	in	in	ADP
fpq-14292	31	27	such	such	ADJ
fpq-14292	31	28	debates	debate	NOUN
fpq-14292	31	29	.	.	PUNCT
fpq-14292	32	1	furthermore	furthermore	ADV
fpq-14292	32	2	,	,	PUNCT
fpq-14292	32	3	even	even	ADV
fpq-14292	32	4	if	if	SCONJ
fpq-14292	32	5	we	we	PRON
fpq-14292	32	6	had	have	VERB
fpq-14292	32	7	descriptively	descriptively	ADV
fpq-14292	32	8	accurate	accurate	ADJ
fpq-14292	32	9	models	model	NOUN
fpq-14292	32	10	—	—	PUNCT
fpq-14292	32	11	for	for	ADP
fpq-14292	32	12	example	example	NOUN
fpq-14292	32	13	,	,	PUNCT
fpq-14292	32	14	models	model	NOUN
fpq-14292	32	15	that	that	PRON
fpq-14292	32	16	do	do	AUX
fpq-14292	32	17	not	not	PART
fpq-14292	32	18	exclude	exclude	VERB
fpq-14292	32	19	marginalized	marginalized	ADJ
fpq-14292	32	20	voices	voice	NOUN
fpq-14292	32	21	—	—	PUNCT
fpq-14292	32	22	there	there	PRON
fpq-14292	32	23	might	might	AUX
fpq-14292	32	24	be	be	AUX
fpq-14292	32	25	practical	practical	ADJ
fpq-14292	32	26	and	and	CCONJ
fpq-14292	32	27	ethical	ethical	ADJ
fpq-14292	32	28	value	value	NOUN
fpq-14292	32	29	to	to	AUX
fpq-14292	32	30	not	not	PART
fpq-14292	32	31	debiasing	debiase	VERB
fpq-14292	32	32	,	,	PUNCT
fpq-14292	32	33	especially	especially	ADV
fpq-14292	32	34	when	when	SCONJ
fpq-14292	32	35	we	we	PRON
fpq-14292	32	36	can	can	AUX
fpq-14292	32	37	control	control	VERB
fpq-14292	32	38	for	for	ADP
fpq-14292	32	39	the	the	DET
fpq-14292	32	40	negative	negative	ADJ
fpq-14292	32	41	effects	effect	NOUN
fpq-14292	32	42	of	of	ADP
fpq-14292	32	43	bias	bias	NOUN
fpq-14292	32	44	at	at	ADP
fpq-14292	32	45	the	the	DET
fpq-14292	32	46	output	output	NOUN
fpq-14292	32	47	stage	stage	NOUN
fpq-14292	32	48	rather	rather	ADV
fpq-14292	32	49	than	than	ADP
fpq-14292	32	50	by	by	ADP
fpq-14292	32	51	debiasing	debiase	VERB
fpq-14292	32	52	the	the	DET
fpq-14292	32	53	model	model	NOUN
fpq-14292	32	54	itself	itself	PRON
fpq-14292	32	55	.	.	PUNCT
fpq-14292	33	1	once	once	SCONJ
fpq-14292	33	2	we	we	PRON
fpq-14292	33	3	recognize	recognize	VERB
fpq-14292	33	4	that	that	SCONJ
fpq-14292	33	5	there	there	PRON
fpq-14292	33	6	might	might	AUX
fpq-14292	33	7	be	be	AUX
fpq-14292	33	8	an	an	DET
fpq-14292	33	9	ethical	ethical	ADJ
fpq-14292	33	10	value	value	NOUN
fpq-14292	33	11	to	to	ADP
fpq-14292	33	12	not	not	PART
fpq-14292	33	13	debiasing	debiase	VERB
fpq-14292	33	14	,	,	PUNCT
fpq-14292	33	15	we	we	PRON
fpq-14292	33	16	encounter	encounter	VERB
fpq-14292	33	17	an	an	DET
fpq-14292	33	18	ethical	ethical	ADJ
fpq-14292	33	19	dilemma	dilemma	NOUN
fpq-14292	33	20	.	.	PUNCT
fpq-14292	34	1	either	either	CCONJ
fpq-14292	34	2	we	we	PRON
fpq-14292	34	3	do	do	AUX
fpq-14292	34	4	not	not	PART
fpq-14292	34	5	debias	debia	NOUN
fpq-14292	34	6	,	,	PUNCT
fpq-14292	34	7	in	in	ADP
fpq-14292	34	8	which	which	DET
fpq-14292	34	9	case	case	NOUN
fpq-14292	34	10	we	we	PRON
fpq-14292	34	11	retain	retain	VERB
fpq-14292	34	12	2	2	NUM
fpq-14292	34	13	statistical	statistical	ADJ
fpq-14292	34	14	recidivism	recidivism	NOUN
fpq-14292	34	15	models	model	NOUN
fpq-14292	34	16	aim	aim	VERB
fpq-14292	34	17	at	at	ADP
fpq-14292	34	18	reducing	reduce	VERB
fpq-14292	34	19	the	the	DET
fpq-14292	34	20	influence	influence	NOUN
fpq-14292	34	21	of	of	ADP
fpq-14292	34	22	biases	bias	NOUN
fpq-14292	34	23	(	(	PUNCT
fpq-14292	34	24	explicit	explicit	ADJ
fpq-14292	34	25	or	or	CCONJ
fpq-14292	34	26	implicit	implicit	ADJ
fpq-14292	34	27	)	)	PUNCT
fpq-14292	34	28	of	of	ADP
fpq-14292	34	29	judges	judge	NOUN
fpq-14292	34	30	in	in	ADP
fpq-14292	34	31	sentencing	sentencing	NOUN
fpq-14292	34	32	.	.	PUNCT
fpq-14292	35	1	yet	yet	CCONJ
fpq-14292	35	2	the	the	DET
fpq-14292	35	3	data	datum	NOUN
fpq-14292	35	4	these	these	DET
fpq-14292	35	5	models	model	NOUN
fpq-14292	35	6	use	use	VERB
fpq-14292	35	7	are	be	AUX
fpq-14292	35	8	typically	typically	ADV
fpq-14292	35	9	obtained	obtain	VERB
fpq-14292	35	10	from	from	ADP
fpq-14292	35	11	questionnaires	questionnaire	NOUN
fpq-14292	35	12	that	that	SCONJ
fpq-14292	35	13	criminals	criminal	NOUN
fpq-14292	35	14	are	be	AUX
fpq-14292	35	15	asked	ask	VERB
fpq-14292	35	16	to	to	PART
fpq-14292	35	17	complete	complete	VERB
fpq-14292	35	18	.	.	PUNCT
fpq-14292	36	1	these	these	DET
fpq-14292	36	2	data	datum	NOUN
fpq-14292	36	3	often	often	ADV
fpq-14292	36	4	include	include	VERB
fpq-14292	36	5	details	detail	NOUN
fpq-14292	36	6	about	about	ADP
fpq-14292	36	7	a	a	DET
fpq-14292	36	8	criminal	criminal	NOUN
fpq-14292	36	9	’s	’s	NOUN
fpq-14292	36	10	upbringing	upbringing	ADJ
fpq-14292	36	11	,	,	PUNCT
fpq-14292	36	12	family	family	NOUN
fpq-14292	36	13	,	,	PUNCT
fpq-14292	36	14	and	and	CCONJ
fpq-14292	36	15	social	social	ADJ
fpq-14292	36	16	connections	connection	NOUN
fpq-14292	36	17	.	.	PUNCT
fpq-14292	37	1	such	such	ADJ
fpq-14292	37	2	data	datum	NOUN
fpq-14292	37	3	ought	ought	AUX
fpq-14292	37	4	to	to	PART
fpq-14292	37	5	be	be	AUX
fpq-14292	37	6	irrelevant	irrelevant	ADJ
fpq-14292	37	7	to	to	ADP
fpq-14292	37	8	a	a	DET
fpq-14292	37	9	criminal	criminal	NOUN
fpq-14292	37	10	’s	’s	PART
fpq-14292	37	11	sentencing	sentencing	NOUN
fpq-14292	37	12	.	.	PUNCT
fpq-14292	38	1	nevertheless	nevertheless	ADV
fpq-14292	38	2	,	,	PUNCT
fpq-14292	38	3	since	since	SCONJ
fpq-14292	38	4	the	the	DET
fpq-14292	38	5	data	datum	NOUN
fpq-14292	38	6	are	be	AUX
fpq-14292	38	7	used	use	VERB
fpq-14292	38	8	to	to	PART
fpq-14292	38	9	generate	generate	VERB
fpq-14292	38	10	a	a	DET
fpq-14292	38	11	“	"	PUNCT
fpq-14292	38	12	recidivism	recidivism	NOUN
fpq-14292	38	13	score	score	NOUN
fpq-14292	38	14	”	"	PUNCT
fpq-14292	38	15	for	for	ADP
fpq-14292	38	16	criminals	criminal	NOUN
fpq-14292	38	17	,	,	PUNCT
fpq-14292	38	18	which	which	PRON
fpq-14292	38	19	is	be	AUX
fpq-14292	38	20	used	use	VERB
fpq-14292	38	21	in	in	ADP
fpq-14292	38	22	sentencing	sentencing	NOUN
fpq-14292	38	23	,	,	PUNCT
fpq-14292	38	24	the	the	DET
fpq-14292	38	25	data	data	NOUN
fpq-14292	38	26	influence	influence	NOUN
fpq-14292	38	27	sentencing	sentencing	NOUN
fpq-14292	38	28	in	in	ADP
fpq-14292	38	29	a	a	DET
fpq-14292	38	30	way	way	NOUN
fpq-14292	38	31	that	that	PRON
fpq-14292	38	32	it	it	PRON
fpq-14292	38	33	should	should	AUX
fpq-14292	38	34	not	not	PART
fpq-14292	38	35	.	.	PUNCT
fpq-14292	39	1	for	for	ADP
fpq-14292	39	2	example	example	NOUN
fpq-14292	39	3	,	,	PUNCT
fpq-14292	39	4	higher	high	ADJ
fpq-14292	39	5	scores	score	NOUN
fpq-14292	39	6	often	often	ADV
fpq-14292	39	7	result	result	VERB
fpq-14292	39	8	in	in	ADP
fpq-14292	39	9	longer	long	ADJ
fpq-14292	39	10	sentences	sentence	NOUN
fpq-14292	39	11	.	.	PUNCT
fpq-14292	40	1	more	more	ADV
fpq-14292	40	2	perniciously	perniciously	ADV
fpq-14292	40	3	,	,	PUNCT
fpq-14292	40	4	a	a	DET
fpq-14292	40	5	longer	long	ADJ
fpq-14292	40	6	sentence	sentence	NOUN
fpq-14292	40	7	will	will	AUX
fpq-14292	40	8	often	often	ADV
fpq-14292	40	9	subsequently	subsequently	ADV
fpq-14292	40	10	result	result	VERB
fpq-14292	40	11	in	in	ADP
fpq-14292	40	12	a	a	DET
fpq-14292	40	13	higher	high	ADJ
fpq-14292	40	14	score	score	NOUN
fpq-14292	40	15	on	on	ADP
fpq-14292	40	16	the	the	DET
fpq-14292	40	17	recidivism	recidivism	NOUN
fpq-14292	40	18	scale	scale	NOUN
fpq-14292	40	19	.	.	PUNCT
fpq-14292	41	1	3	3	NUM
fpq-14292	41	2	data	datum	NOUN
fpq-14292	41	3	can	can	AUX
fpq-14292	41	4	also	also	ADV
fpq-14292	41	5	be	be	AUX
fpq-14292	41	6	biased	bias	VERB
fpq-14292	41	7	due	due	ADJ
fpq-14292	41	8	to	to	ADP
fpq-14292	41	9	bad	bad	ADJ
fpq-14292	41	10	sampling	sampling	NOUN
fpq-14292	41	11	.	.	PUNCT
fpq-14292	42	1	we	we	PRON
fpq-14292	42	2	do	do	AUX
fpq-14292	42	3	not	not	PART
fpq-14292	42	4	focus	focus	VERB
fpq-14292	42	5	on	on	ADP
fpq-14292	42	6	such	such	ADJ
fpq-14292	42	7	bias	bias	NOUN
fpq-14292	42	8	(	(	PUNCT
fpq-14292	42	9	but	but	CCONJ
fpq-14292	42	10	see	see	VERB
fpq-14292	42	11	sections	section	NOUN
fpq-14292	42	12	6.3	6.3	NUM
fpq-14292	42	13	and	and	CCONJ
fpq-14292	42	14	8)	8)	NUM
fpq-14292	42	15	.	.	PUNCT
fpq-14292	43	1	deery	deery	NOUN
fpq-14292	43	2	and	and	CCONJ
fpq-14292	43	3	bailey	bailey	PROPN
fpq-14292	43	4	–	–	PUNCT
fpq-14292	43	5	the	the	DET
fpq-14292	43	6	bias	bias	NOUN
fpq-14292	43	7	dilemma	dilemma	NOUN
fpq-14292	43	8	:	:	PUNCT
fpq-14292	43	9	the	the	DET
fpq-14292	43	10	ethics	ethic	NOUN
fpq-14292	43	11	of	of	ADP
fpq-14292	43	12	algorithmic	algorithmic	ADJ
fpq-14292	43	13	bias	bias	NOUN
fpq-14292	43	14	in	in	ADP
fpq-14292	43	15	natural	natural	ADJ
fpq-14292	43	16	-	-	PUNCT
fpq-14292	43	17	language	language	NOUN
fpq-14292	43	18	processing	processing	NOUN
fpq-14292	43	19	published	publish	VERB
fpq-14292	43	20	by	by	ADP
fpq-14292	43	21	scholarship@western	scholarship@western	PROPN
fpq-14292	43	22	,	,	PUNCT
fpq-14292	43	23	2022	2022	NUM
fpq-14292	43	24	3	3	NUM
fpq-14292	43	25	descriptive	descriptive	ADJ
fpq-14292	43	26	accuracy	accuracy	NOUN
fpq-14292	43	27	and	and	CCONJ
fpq-14292	43	28	potentially	potentially	ADV
fpq-14292	43	29	ethically	ethically	ADV
fpq-14292	43	30	useful	useful	ADJ
fpq-14292	43	31	information	information	NOUN
fpq-14292	43	32	yet	yet	ADV
fpq-14292	43	33	run	run	VERB
fpq-14292	43	34	the	the	DET
fpq-14292	43	35	risk	risk	NOUN
fpq-14292	43	36	of	of	ADP
fpq-14292	43	37	perpetuating	perpetuate	VERB
fpq-14292	43	38	or	or	CCONJ
fpq-14292	43	39	amplifying	amplifying	ADJ
fpq-14292	43	40	bias	bias	NOUN
fpq-14292	43	41	;	;	PUNCT
fpq-14292	43	42	or	or	CCONJ
fpq-14292	43	43	instead	instead	ADV
fpq-14292	43	44	we	we	PRON
fpq-14292	43	45	debias	debia	NOUN
fpq-14292	43	46	,	,	PUNCT
fpq-14292	43	47	thereby	thereby	ADV
fpq-14292	43	48	avoiding	avoid	VERB
fpq-14292	43	49	perpetuating	perpetuate	VERB
fpq-14292	43	50	bias	bias	NOUN
fpq-14292	43	51	yet	yet	ADV
fpq-14292	43	52	losing	lose	VERB
fpq-14292	43	53	descriptive	descriptive	ADJ
fpq-14292	43	54	accuracy	accuracy	NOUN
fpq-14292	43	55	and	and	CCONJ
fpq-14292	43	56	ethically	ethically	ADV
fpq-14292	43	57	useful	useful	ADJ
fpq-14292	43	58	information	information	NOUN
fpq-14292	43	59	.	.	PUNCT
fpq-14292	44	1	our	our	PRON
fpq-14292	44	2	aim	aim	NOUN
fpq-14292	44	3	is	be	AUX
fpq-14292	44	4	to	to	PART
fpq-14292	44	5	draw	draw	VERB
fpq-14292	44	6	attention	attention	NOUN
fpq-14292	44	7	to	to	ADP
fpq-14292	44	8	this	this	DET
fpq-14292	44	9	dilemma	dilemma	NOUN
fpq-14292	44	10	and	and	CCONJ
fpq-14292	44	11	to	to	PART
fpq-14292	44	12	show	show	VERB
fpq-14292	44	13	how	how	SCONJ
fpq-14292	44	14	it	it	PRON
fpq-14292	44	15	can	can	AUX
fpq-14292	44	16	help	help	VERB
fpq-14292	44	17	in	in	ADP
fpq-14292	44	18	rethinking	rethink	VERB
fpq-14292	44	19	the	the	DET
fpq-14292	44	20	ethics	ethic	NOUN
fpq-14292	44	21	of	of	ADP
fpq-14292	44	22	algorithmic	algorithmic	ADJ
fpq-14292	44	23	bias	bias	NOUN
fpq-14292	44	24	in	in	ADP
fpq-14292	44	25	nlp	nlp	NOUN
fpq-14292	44	26	systems	system	NOUN
fpq-14292	44	27	.	.	PUNCT
fpq-14292	45	1	in	in	ADP
fpq-14292	45	2	section	section	NOUN
fpq-14292	45	3	2	2	NUM
fpq-14292	45	4	,	,	PUNCT
fpq-14292	45	5	we	we	PRON
fpq-14292	45	6	specify	specify	VERB
fpq-14292	45	7	the	the	DET
fpq-14292	45	8	type	type	NOUN
fpq-14292	45	9	of	of	ADP
fpq-14292	45	10	bias	bias	NOUN
fpq-14292	45	11	we	we	PRON
fpq-14292	45	12	will	will	AUX
fpq-14292	45	13	focus	focus	VERB
fpq-14292	45	14	on	on	ADP
fpq-14292	45	15	.	.	PUNCT
fpq-14292	46	1	in	in	ADP
fpq-14292	46	2	section	section	NOUN
fpq-14292	46	3	3	3	NUM
fpq-14292	46	4	,	,	PUNCT
fpq-14292	46	5	we	we	PRON
fpq-14292	46	6	outline	outline	VERB
fpq-14292	46	7	the	the	DET
fpq-14292	46	8	distinction	distinction	NOUN
fpq-14292	46	9	between	between	ADP
fpq-14292	46	10	descriptive	descriptive	ADJ
fpq-14292	46	11	and	and	CCONJ
fpq-14292	46	12	normative	normative	ADJ
fpq-14292	46	13	correctness	correctness	NOUN
fpq-14292	46	14	as	as	ADP
fpq-14292	46	15	often	often	ADV
fpq-14292	46	16	competing	compete	VERB
fpq-14292	46	17	virtues	virtue	NOUN
fpq-14292	46	18	of	of	ADP
fpq-14292	46	19	nlp	nlp	NOUN
fpq-14292	46	20	models	model	NOUN
fpq-14292	46	21	,	,	PUNCT
fpq-14292	46	22	by	by	ADP
fpq-14292	46	23	analogy	analogy	NOUN
fpq-14292	46	24	with	with	ADP
fpq-14292	46	25	how	how	SCONJ
fpq-14292	46	26	biases	bias	NOUN
fpq-14292	46	27	rooted	root	VERB
fpq-14292	46	28	in	in	ADP
fpq-14292	46	29	statistical	statistical	ADJ
fpq-14292	46	30	regularities	regularity	NOUN
fpq-14292	46	31	are	be	AUX
fpq-14292	46	32	often	often	ADV
fpq-14292	46	33	thought	think	VERB
fpq-14292	46	34	to	to	PART
fpq-14292	46	35	offer	offer	VERB
fpq-14292	46	36	support	support	NOUN
fpq-14292	46	37	for	for	ADP
fpq-14292	46	38	beliefs	belief	NOUN
fpq-14292	46	39	even	even	ADV
fpq-14292	46	40	when	when	SCONJ
fpq-14292	46	41	these	these	DET
fpq-14292	46	42	beliefs	belief	NOUN
fpq-14292	46	43	are	be	AUX
fpq-14292	46	44	morally	morally	ADV
fpq-14292	46	45	problematic	problematic	ADJ
fpq-14292	46	46	.	.	PUNCT
fpq-14292	47	1	in	in	ADP
fpq-14292	47	2	section	section	NOUN
fpq-14292	47	3	4	4	NUM
fpq-14292	47	4	,	,	PUNCT
fpq-14292	47	5	we	we	PRON
fpq-14292	47	6	illustrate	illustrate	VERB
fpq-14292	47	7	our	our	PRON
fpq-14292	47	8	distinction	distinction	NOUN
fpq-14292	47	9	by	by	ADP
fpq-14292	47	10	focusing	focus	VERB
fpq-14292	47	11	on	on	ADP
fpq-14292	47	12	a	a	DET
fpq-14292	47	13	particular	particular	ADJ
fpq-14292	47	14	variety	variety	NOUN
fpq-14292	47	15	of	of	ADP
fpq-14292	47	16	nlp	nlp	NOUN
fpq-14292	47	17	model	model	NOUN
fpq-14292	47	18	called	call	VERB
fpq-14292	47	19	word	word	NOUN
fpq-14292	47	20	-	-	PUNCT
fpq-14292	47	21	embeddings	embedding	NOUN
fpq-14292	47	22	.	.	PUNCT
fpq-14292	48	1	in	in	ADP
fpq-14292	48	2	section	section	NOUN
fpq-14292	48	3	5	5	NUM
fpq-14292	48	4	,	,	PUNCT
fpq-14292	48	5	we	we	PRON
fpq-14292	48	6	discuss	discuss	VERB
fpq-14292	48	7	practical	practical	ADJ
fpq-14292	48	8	problems	problem	NOUN
fpq-14292	48	9	with	with	ADP
fpq-14292	48	10	debiasing	debiase	VERB
fpq-14292	48	11	nlp	nlp	NOUN
fpq-14292	48	12	models	model	NOUN
fpq-14292	48	13	.	.	PUNCT
fpq-14292	49	1	these	these	DET
fpq-14292	49	2	problems	problem	NOUN
fpq-14292	49	3	lend	lend	VERB
fpq-14292	49	4	support	support	NOUN
fpq-14292	49	5	to	to	ADP
fpq-14292	49	6	our	our	PRON
fpq-14292	49	7	claim	claim	NOUN
fpq-14292	49	8	that	that	SCONJ
fpq-14292	49	9	it	it	PRON
fpq-14292	49	10	is	be	AUX
fpq-14292	49	11	often	often	ADV
fpq-14292	49	12	preferable	preferable	ADJ
fpq-14292	49	13	to	to	PART
fpq-14292	49	14	control	control	VERB
fpq-14292	49	15	for	for	ADP
fpq-14292	49	16	negative	negative	ADJ
fpq-14292	49	17	effects	effect	NOUN
fpq-14292	49	18	at	at	ADP
fpq-14292	49	19	the	the	DET
fpq-14292	49	20	output	output	NOUN
fpq-14292	49	21	stage	stage	NOUN
fpq-14292	49	22	rather	rather	ADV
fpq-14292	49	23	than	than	ADP
fpq-14292	49	24	try	try	VERB
fpq-14292	49	25	to	to	PART
fpq-14292	49	26	debias	debias	VERB
fpq-14292	49	27	the	the	DET
fpq-14292	49	28	models	model	NOUN
fpq-14292	49	29	themselves	themselves	PRON
fpq-14292	49	30	.	.	PUNCT
fpq-14292	50	1	in	in	ADP
fpq-14292	50	2	section	section	NOUN
fpq-14292	50	3	6	6	NUM
fpq-14292	50	4	,	,	PUNCT
fpq-14292	50	5	we	we	PRON
fpq-14292	50	6	consider	consider	VERB
fpq-14292	50	7	varying	vary	VERB
fpq-14292	50	8	degrees	degree	NOUN
fpq-14292	50	9	of	of	ADP
fpq-14292	50	10	scope	scope	NOUN
fpq-14292	50	11	for	for	ADP
fpq-14292	50	12	debiasing	debiase	VERB
fpq-14292	50	13	nlp	nlp	ADJ
fpq-14292	50	14	systems	system	NOUN
fpq-14292	50	15	’	'	PUNCT
fpq-14292	50	16	outputs	output	NOUN
fpq-14292	50	17	.	.	PUNCT
fpq-14292	51	1	in	in	ADP
fpq-14292	51	2	section	section	NOUN
fpq-14292	51	3	7	7	NUM
fpq-14292	51	4	,	,	PUNCT
fpq-14292	51	5	we	we	PRON
fpq-14292	51	6	explain	explain	VERB
fpq-14292	51	7	the	the	DET
fpq-14292	51	8	ethical	ethical	ADJ
fpq-14292	51	9	dilemma	dilemma	NOUN
fpq-14292	51	10	.	.	PUNCT
fpq-14292	52	1	in	in	ADP
fpq-14292	52	2	section	section	NOUN
fpq-14292	52	3	8	8	NUM
fpq-14292	52	4	,	,	PUNCT
fpq-14292	52	5	we	we	PRON
fpq-14292	52	6	discuss	discuss	VERB
fpq-14292	52	7	how	how	SCONJ
fpq-14292	52	8	our	our	PRON
fpq-14292	52	9	view	view	NOUN
fpq-14292	52	10	helps	help	VERB
fpq-14292	52	11	to	to	PART
fpq-14292	52	12	rethink	rethink	VERB
fpq-14292	52	13	the	the	DET
fpq-14292	52	14	ethics	ethic	NOUN
fpq-14292	52	15	of	of	ADP
fpq-14292	52	16	algorithmic	algorithmic	ADJ
fpq-14292	52	17	bias	bias	NOUN
fpq-14292	52	18	in	in	ADP
fpq-14292	52	19	nlp	nlp	NOUN
fpq-14292	52	20	.	.	PROPN
fpq-14292	53	1	2	2	X
fpq-14292	53	2	.	.	X
fpq-14292	54	1	how	how	SCONJ
fpq-14292	54	2	(	(	PUNCT
fpq-14292	54	3	and	and	CCONJ
fpq-14292	54	4	how	how	SCONJ
fpq-14292	54	5	not	not	PART
fpq-14292	54	6	)	)	PUNCT
fpq-14292	54	7	to	to	PART
fpq-14292	54	8	control	control	VERB
fpq-14292	54	9	for	for	ADP
fpq-14292	54	10	morally	morally	ADV
fpq-14292	54	11	relevant	relevant	ADJ
fpq-14292	54	12	bias	bias	NOUN
fpq-14292	54	13	both	both	CCONJ
fpq-14292	54	14	human	human	ADJ
fpq-14292	54	15	cognition	cognition	NOUN
fpq-14292	54	16	and	and	CCONJ
fpq-14292	54	17	artificial	artificial	ADJ
fpq-14292	54	18	systems	system	NOUN
fpq-14292	54	19	can	can	AUX
fpq-14292	54	20	exhibit	exhibit	VERB
fpq-14292	54	21	morally	morally	ADV
fpq-14292	54	22	relevant	relevant	ADJ
fpq-14292	54	23	bias	bias	NOUN
fpq-14292	54	24	.	.	PUNCT
fpq-14292	55	1	in	in	ADP
fpq-14292	55	2	humans	human	NOUN
fpq-14292	55	3	,	,	PUNCT
fpq-14292	55	4	the	the	DET
fpq-14292	55	5	relevant	relevant	ADJ
fpq-14292	55	6	bias	bias	NOUN
fpq-14292	55	7	is	be	AUX
fpq-14292	55	8	typically	typically	ADV
fpq-14292	55	9	a	a	DET
fpq-14292	55	10	negative	negative	ADJ
fpq-14292	55	11	evaluative	evaluative	ADJ
fpq-14292	55	12	tendency	tendency	NOUN
fpq-14292	55	13	regarding	regard	VERB
fpq-14292	55	14	other	other	ADJ
fpq-14292	55	15	people	people	NOUN
fpq-14292	55	16	based	base	VERB
fpq-14292	55	17	on	on	ADP
fpq-14292	55	18	their	their	PRON
fpq-14292	55	19	apparent	apparent	ADJ
fpq-14292	55	20	membership	membership	NOUN
fpq-14292	55	21	in	in	ADP
fpq-14292	55	22	a	a	DET
fpq-14292	55	23	socially	socially	ADV
fpq-14292	55	24	salient	salient	ADJ
fpq-14292	55	25	category	category	NOUN
fpq-14292	55	26	or	or	CCONJ
fpq-14292	55	27	group	group	NOUN
fpq-14292	55	28	(	(	PUNCT
fpq-14292	55	29	brownstein	brownstein	PROPN
fpq-14292	55	30	2018	2018	NUM
fpq-14292	55	31	,	,	PUNCT
fpq-14292	55	32	43	43	NUM
fpq-14292	55	33	,	,	PUNCT
fpq-14292	55	34	126	126	NUM
fpq-14292	55	35	,	,	PUNCT
fpq-14292	55	36	172	172	NUM
fpq-14292	55	37	;	;	PUNCT
fpq-14292	55	38	brownstein	brownstein	PROPN
fpq-14292	55	39	and	and	CCONJ
fpq-14292	55	40	saul	saul	PROPN
fpq-14292	55	41	2016a	2016a	NUM
fpq-14292	55	42	,	,	PUNCT
fpq-14292	55	43	2016b	2016b	NUM
fpq-14292	55	44	)	)	PUNCT
fpq-14292	55	45	.	.	PUNCT
fpq-14292	56	1	such	such	ADJ
fpq-14292	56	2	biases	bias	NOUN
fpq-14292	56	3	are	be	AUX
fpq-14292	56	4	often	often	ADV
fpq-14292	56	5	explicit	explicit	ADJ
fpq-14292	56	6	,	,	PUNCT
fpq-14292	56	7	consisting	consist	VERB
fpq-14292	56	8	of	of	ADP
fpq-14292	56	9	mental	mental	ADJ
fpq-14292	56	10	attitudes	attitude	NOUN
fpq-14292	56	11	(	(	PUNCT
fpq-14292	56	12	like	like	INTJ
fpq-14292	56	13	belief	belief	NOUN
fpq-14292	56	14	)	)	PUNCT
fpq-14292	56	15	that	that	SCONJ
fpq-14292	56	16	an	an	DET
fpq-14292	56	17	agent	agent	NOUN
fpq-14292	56	18	has	have	AUX
fpq-14292	56	19	and	and	CCONJ
fpq-14292	56	20	consciously	consciously	ADV
fpq-14292	56	21	endorses	endorse	VERB
fpq-14292	56	22	.	.	PUNCT
fpq-14292	57	1	for	for	ADP
fpq-14292	57	2	example	example	NOUN
fpq-14292	57	3	,	,	PUNCT
fpq-14292	57	4	someone	someone	PRON
fpq-14292	57	5	is	be	AUX
fpq-14292	57	6	explicitly	explicitly	ADV
fpq-14292	57	7	sexist	sexist	ADJ
fpq-14292	57	8	if	if	SCONJ
fpq-14292	57	9	they	they	PRON
fpq-14292	57	10	openly	openly	ADV
fpq-14292	57	11	admit	admit	VERB
fpq-14292	57	12	to	to	ADP
fpq-14292	57	13	being	be	AUX
fpq-14292	57	14	sexist	sexist	ADJ
fpq-14292	57	15	and	and	CCONJ
fpq-14292	57	16	approve	approve	VERB
fpq-14292	57	17	of	of	ADP
fpq-14292	57	18	these	these	DET
fpq-14292	57	19	attitudes	attitude	NOUN
fpq-14292	57	20	.	.	PUNCT
fpq-14292	58	1	by	by	ADP
fpq-14292	58	2	contrast	contrast	NOUN
fpq-14292	58	3	,	,	PUNCT
fpq-14292	58	4	implicit	implicit	ADJ
fpq-14292	58	5	biases	bias	NOUN
fpq-14292	58	6	are	be	AUX
fpq-14292	58	7	unconscious	unconscious	ADJ
fpq-14292	58	8	attitudes	attitude	NOUN
fpq-14292	58	9	that	that	PRON
fpq-14292	58	10	can	can	AUX
fpq-14292	58	11	be	be	AUX
fpq-14292	58	12	difficult	difficult	ADJ
fpq-14292	58	13	to	to	PART
fpq-14292	58	14	inhibit	inhibit	VERB
fpq-14292	58	15	,	,	PUNCT
fpq-14292	58	16	seem	seem	VERB
fpq-14292	58	17	insensitive	insensitive	ADJ
fpq-14292	58	18	to	to	ADP
fpq-14292	58	19	an	an	DET
fpq-14292	58	20	agent	agent	NOUN
fpq-14292	58	21	’s	’s	PART
fpq-14292	58	22	explicit	explicit	ADJ
fpq-14292	58	23	attitudes	attitude	NOUN
fpq-14292	58	24	,	,	PUNCT
fpq-14292	58	25	and	and	CCONJ
fpq-14292	58	26	are	be	AUX
fpq-14292	58	27	often	often	ADV
fpq-14292	58	28	in	in	ADP
fpq-14292	58	29	conflict	conflict	NOUN
fpq-14292	58	30	with	with	ADP
fpq-14292	58	31	an	an	DET
fpq-14292	58	32	agent	agent	NOUN
fpq-14292	58	33	’s	’s	PART
fpq-14292	58	34	explicitly	explicitly	ADV
fpq-14292	58	35	held	hold	VERB
fpq-14292	58	36	attitudes	attitude	NOUN
fpq-14292	58	37	(	(	PUNCT
fpq-14292	58	38	see	see	VERB
fpq-14292	58	39	levy	levy	NOUN
fpq-14292	58	40	2017	2017	NUM
fpq-14292	58	41	for	for	ADP
fpq-14292	58	42	a	a	DET
fpq-14292	58	43	review	review	NOUN
fpq-14292	58	44	of	of	ADP
fpq-14292	58	45	the	the	DET
fpq-14292	58	46	empirical	empirical	ADJ
fpq-14292	58	47	literature	literature	NOUN
fpq-14292	58	48	)	)	PUNCT
fpq-14292	58	49	.	.	PUNCT
fpq-14292	59	1	for	for	ADP
fpq-14292	59	2	example	example	NOUN
fpq-14292	59	3	,	,	PUNCT
fpq-14292	59	4	even	even	ADV
fpq-14292	59	5	someone	someone	PRON
fpq-14292	59	6	who	who	PRON
fpq-14292	59	7	explicitly	explicitly	ADV
fpq-14292	59	8	condemns	condemn	VERB
fpq-14292	59	9	sexism	sexism	NOUN
fpq-14292	59	10	might	might	AUX
fpq-14292	59	11	still	still	ADV
fpq-14292	59	12	have	have	VERB
fpq-14292	59	13	implicit	implicit	ADJ
fpq-14292	59	14	sexist	sexist	ADJ
fpq-14292	59	15	attitudes	attitude	NOUN
fpq-14292	59	16	that	that	PRON
fpq-14292	59	17	they	they	PRON
fpq-14292	59	18	are	be	AUX
fpq-14292	59	19	unaware	unaware	ADJ
fpq-14292	59	20	of	of	ADP
fpq-14292	59	21	and	and	CCONJ
fpq-14292	59	22	that	that	PRON
fpq-14292	59	23	are	be	AUX
fpq-14292	59	24	difficult	difficult	ADJ
fpq-14292	59	25	to	to	PART
fpq-14292	59	26	inhibit	inhibit	VERB
fpq-14292	59	27	.	.	PUNCT
fpq-14292	60	1	when	when	SCONJ
fpq-14292	60	2	people	people	NOUN
fpq-14292	60	3	act	act	VERB
fpq-14292	60	4	on	on	ADP
fpq-14292	60	5	explicit	explicit	ADJ
fpq-14292	60	6	biases	bias	NOUN
fpq-14292	60	7	,	,	PUNCT
fpq-14292	60	8	we	we	PRON
fpq-14292	60	9	hold	hold	VERB
fpq-14292	60	10	them	they	PRON
fpq-14292	60	11	morally	morally	ADV
fpq-14292	60	12	responsible	responsible	ADJ
fpq-14292	60	13	for	for	ADP
fpq-14292	60	14	doing	do	VERB
fpq-14292	60	15	so	so	ADV
fpq-14292	60	16	,	,	PUNCT
fpq-14292	60	17	and	and	CCONJ
fpq-14292	60	18	as	as	ADP
fpq-14292	60	19	a	a	DET
fpq-14292	60	20	result	result	NOUN
fpq-14292	60	21	we	we	PRON
fpq-14292	60	22	blame	blame	VERB
fpq-14292	60	23	them	they	PRON
fpq-14292	60	24	for	for	ADP
fpq-14292	60	25	their	their	PRON
fpq-14292	60	26	behavior	behavior	NOUN
fpq-14292	60	27	(	(	PUNCT
fpq-14292	60	28	see	see	VERB
fpq-14292	60	29	,	,	PUNCT
fpq-14292	60	30	e.g.	e.g.	ADV
fpq-14292	60	31	,	,	PUNCT
fpq-14292	60	32	brownstein	brownstein	PROPN
fpq-14292	60	33	and	and	CCONJ
fpq-14292	60	34	saul	saul	PROPN
fpq-14292	60	35	2016b	2016b	NUM
fpq-14292	60	36	)	)	PUNCT
fpq-14292	60	37	.	.	PUNCT
fpq-14292	61	1	it	it	PRON
fpq-14292	61	2	is	be	AUX
fpq-14292	61	3	more	more	ADV
fpq-14292	61	4	difficult	difficult	ADJ
fpq-14292	61	5	to	to	PART
fpq-14292	61	6	know	know	VERB
fpq-14292	61	7	whether	whether	SCONJ
fpq-14292	61	8	we	we	PRON
fpq-14292	61	9	can	can	AUX
fpq-14292	61	10	blame	blame	VERB
fpq-14292	61	11	or	or	CCONJ
fpq-14292	61	12	hold	hold	VERB
fpq-14292	61	13	people	people	NOUN
fpq-14292	61	14	responsible	responsible	ADJ
fpq-14292	61	15	for	for	ADP
fpq-14292	61	16	actions	action	NOUN
fpq-14292	61	17	caused	cause	VERB
fpq-14292	61	18	by	by	ADP
fpq-14292	61	19	implicit	implicit	ADJ
fpq-14292	61	20	biases	bias	NOUN
fpq-14292	61	21	,	,	PUNCT
fpq-14292	61	22	since	since	SCONJ
fpq-14292	61	23	it	it	PRON
fpq-14292	61	24	is	be	AUX
fpq-14292	61	25	unclear	unclear	ADJ
fpq-14292	61	26	whether	whether	SCONJ
fpq-14292	61	27	people	people	NOUN
fpq-14292	61	28	have	have	VERB
fpq-14292	61	29	sufficient	sufficient	ADJ
fpq-14292	61	30	control	control	NOUN
fpq-14292	61	31	over	over	ADP
fpq-14292	61	32	how	how	SCONJ
fpq-14292	61	33	their	their	PRON
fpq-14292	61	34	biases	bias	NOUN
fpq-14292	61	35	influence	influence	VERB
fpq-14292	61	36	behavior	behavior	NOUN
fpq-14292	61	37	for	for	SCONJ
fpq-14292	61	38	them	they	PRON
fpq-14292	61	39	to	to	PART
fpq-14292	61	40	be	be	AUX
fpq-14292	61	41	responsible	responsible	ADJ
fpq-14292	61	42	for	for	ADP
fpq-14292	61	43	these	these	DET
fpq-14292	61	44	behaviors	behavior	NOUN
fpq-14292	61	45	(	(	PUNCT
fpq-14292	61	46	levy	levy	NOUN
fpq-14292	61	47	2017).4	2017).4	NOUN
fpq-14292	61	48	even	even	ADV
fpq-14292	61	49	so	so	ADV
fpq-14292	61	50	,	,	PUNCT
fpq-14292	61	51	it	it	PRON
fpq-14292	61	52	is	be	AUX
fpq-14292	61	53	plausible	plausible	ADJ
fpq-14292	61	54	to	to	PART
fpq-14292	61	55	think	think	VERB
fpq-14292	61	56	that	that	SCONJ
fpq-14292	61	57	we	we	PRON
fpq-14292	61	58	4	4	NUM
fpq-14292	61	59	the	the	DET
fpq-14292	61	60	view	view	NOUN
fpq-14292	61	61	that	that	SCONJ
fpq-14292	61	62	control	control	NOUN
fpq-14292	61	63	is	be	AUX
fpq-14292	61	64	not	not	PART
fpq-14292	61	65	needed	need	VERB
fpq-14292	61	66	for	for	ADP
fpq-14292	61	67	responsibility	responsibility	NOUN
fpq-14292	61	68	is	be	AUX
fpq-14292	61	69	a	a	DET
fpq-14292	61	70	minority	minority	NOUN
fpq-14292	61	71	position	position	NOUN
fpq-14292	61	72	(	(	PUNCT
fpq-14292	61	73	e.g.	e.g.	ADV
fpq-14292	61	74	,	,	PUNCT
fpq-14292	61	75	smith	smith	NOUN
fpq-14292	61	76	2008	2008	NUM
fpq-14292	61	77	)	)	PUNCT
fpq-14292	61	78	.	.	PUNCT
fpq-14292	62	1	it	it	PRON
fpq-14292	62	2	is	be	AUX
fpq-14292	62	3	more	more	ADV
fpq-14292	62	4	widely	widely	ADV
fpq-14292	62	5	held	hold	VERB
fpq-14292	62	6	that	that	DET
fpq-14292	62	7	responsibility	responsibility	NOUN
fpq-14292	62	8	of	of	ADP
fpq-14292	62	9	the	the	DET
fpq-14292	62	10	sort	sort	NOUN
fpq-14292	62	11	that	that	PRON
fpq-14292	62	12	might	might	AUX
fpq-14292	62	13	justify	justify	VERB
fpq-14292	62	14	feminist	feminist	ADJ
fpq-14292	62	15	philosophy	philosophy	NOUN
fpq-14292	62	16	quarterly	quarterly	ADV
fpq-14292	62	17	,	,	PUNCT
fpq-14292	62	18	2022	2022	NUM
fpq-14292	62	19	,	,	PUNCT
fpq-14292	62	20	vol	vol	NOUN
fpq-14292	62	21	.	.	NOUN
fpq-14292	62	22	8	8	NUM
fpq-14292	62	23	,	,	PUNCT
fpq-14292	62	24	iss	iss	PROPN
fpq-14292	62	25	.	.	PUNCT
fpq-14292	62	26	3/4	3/4	NUM
fpq-14292	62	27	,	,	PUNCT
fpq-14292	62	28	article	article	NOUN
fpq-14292	62	29	1	1	NUM
fpq-14292	62	30	published	publish	VERB
fpq-14292	62	31	by	by	ADP
fpq-14292	62	32	scholarship@western	scholarship@western	PROPN
fpq-14292	62	33	,	,	PUNCT
fpq-14292	62	34	2022	2022	NUM
fpq-14292	62	35	4	4	NUM
fpq-14292	62	36	need	need	NOUN
fpq-14292	62	37	,	,	PUNCT
fpq-14292	62	38	at	at	ADP
fpq-14292	62	39	least	least	ADJ
fpq-14292	62	40	,	,	PUNCT
fpq-14292	62	41	to	to	PART
fpq-14292	62	42	control	control	VERB
fpq-14292	62	43	for	for	ADP
fpq-14292	62	44	the	the	DET
fpq-14292	62	45	negative	negative	ADJ
fpq-14292	62	46	influences	influence	NOUN
fpq-14292	62	47	of	of	ADP
fpq-14292	62	48	implicit	implicit	ADJ
fpq-14292	62	49	bias	bias	NOUN
fpq-14292	62	50	when	when	SCONJ
fpq-14292	62	51	expressed	express	VERB
fpq-14292	62	52	in	in	ADP
fpq-14292	62	53	behavior.5	behavior.5	ADP
fpq-14292	62	54	there	there	PRON
fpq-14292	62	55	can	can	AUX
fpq-14292	62	56	be	be	AUX
fpq-14292	62	57	structural	structural	ADJ
fpq-14292	62	58	ways	way	NOUN
fpq-14292	62	59	of	of	ADP
fpq-14292	62	60	doing	do	VERB
fpq-14292	62	61	so	so	ADV
fpq-14292	62	62	.	.	PUNCT
fpq-14292	63	1	for	for	ADP
fpq-14292	63	2	example	example	NOUN
fpq-14292	63	3	,	,	PUNCT
fpq-14292	63	4	the	the	DET
fpq-14292	63	5	biased	biased	ADJ
fpq-14292	63	6	hiring	hiring	NOUN
fpq-14292	63	7	of	of	ADP
fpq-14292	63	8	male	male	ADJ
fpq-14292	63	9	lead	lead	NOUN
fpq-14292	63	10	violinists	violinist	NOUN
fpq-14292	63	11	in	in	ADP
fpq-14292	63	12	orchestras	orchestra	NOUN
fpq-14292	63	13	(	(	PUNCT
fpq-14292	63	14	whether	whether	SCONJ
fpq-14292	63	15	explicit	explicit	ADJ
fpq-14292	63	16	or	or	CCONJ
fpq-14292	63	17	implicit	implicit	ADJ
fpq-14292	63	18	)	)	PUNCT
fpq-14292	63	19	can	can	AUX
fpq-14292	63	20	be	be	AUX
fpq-14292	63	21	controlled	control	VERB
fpq-14292	63	22	for	for	ADP
fpq-14292	63	23	by	by	ADP
fpq-14292	63	24	having	have	VERB
fpq-14292	63	25	musicians	musician	NOUN
fpq-14292	63	26	audition	audition	VERB
fpq-14292	63	27	behind	behind	ADP
fpq-14292	63	28	a	a	DET
fpq-14292	63	29	screen	screen	NOUN
fpq-14292	63	30	so	so	SCONJ
fpq-14292	63	31	that	that	SCONJ
fpq-14292	63	32	the	the	DET
fpq-14292	63	33	hiring	hire	VERB
fpq-14292	63	34	panel	panel	NOUN
fpq-14292	63	35	can	can	AUX
fpq-14292	63	36	not	not	PART
fpq-14292	63	37	see	see	VERB
fpq-14292	63	38	the	the	DET
fpq-14292	63	39	violinist	violinist	NOUN
fpq-14292	63	40	’s	’s	PART
fpq-14292	63	41	gender	gender	NOUN
fpq-14292	63	42	(	(	PUNCT
fpq-14292	63	43	see	see	VERB
fpq-14292	63	44	,	,	PUNCT
fpq-14292	63	45	e.g.	e.g.	ADV
fpq-14292	63	46	,	,	PUNCT
fpq-14292	63	47	goldin	goldin	VERB
fpq-14292	63	48	and	and	CCONJ
fpq-14292	63	49	rouse	rouse	VERB
fpq-14292	63	50	2000	2000	NUM
fpq-14292	63	51	)	)	PUNCT
fpq-14292	63	52	.	.	PUNCT
fpq-14292	64	1	bias	bias	NOUN
fpq-14292	64	2	in	in	ADP
fpq-14292	64	3	the	the	DET
fpq-14292	64	4	relevant	relevant	ADJ
fpq-14292	64	5	sense	sense	NOUN
fpq-14292	64	6	is	be	AUX
fpq-14292	64	7	something	something	PRON
fpq-14292	64	8	that	that	PRON
fpq-14292	64	9	it	it	PRON
fpq-14292	64	10	can	can	AUX
fpq-14292	64	11	be	be	AUX
fpq-14292	64	12	ethically	ethically	ADV
fpq-14292	64	13	problematic	problematic	ADJ
fpq-14292	64	14	for	for	ADP
fpq-14292	64	15	a	a	DET
fpq-14292	64	16	system	system	NOUN
fpq-14292	64	17	to	to	PART
fpq-14292	64	18	have	have	VERB
fpq-14292	64	19	,	,	PUNCT
fpq-14292	64	20	and	and	CCONJ
fpq-14292	64	21	it	it	PRON
fpq-14292	64	22	is	be	AUX
fpq-14292	64	23	especially	especially	ADV
fpq-14292	64	24	bad	bad	ADJ
fpq-14292	64	25	when	when	SCONJ
fpq-14292	64	26	a	a	DET
fpq-14292	64	27	system	system	NOUN
fpq-14292	64	28	’s	’s	PART
fpq-14292	64	29	outputs	output	NOUN
fpq-14292	64	30	have	have	VERB
fpq-14292	64	31	negative	negative	ADJ
fpq-14292	64	32	ethical	ethical	ADJ
fpq-14292	64	33	effects	effect	NOUN
fpq-14292	64	34	by	by	ADP
fpq-14292	64	35	causing	cause	VERB
fpq-14292	64	36	harms	harm	NOUN
fpq-14292	64	37	.	.	PUNCT
fpq-14292	65	1	human	human	ADJ
fpq-14292	65	2	cognition	cognition	NOUN
fpq-14292	65	3	exhibits	exhibit	VERB
fpq-14292	65	4	implicit	implicit	ADJ
fpq-14292	65	5	bias	bias	NOUN
fpq-14292	65	6	,	,	PUNCT
fpq-14292	65	7	and	and	CCONJ
fpq-14292	65	8	various	various	ADJ
fpq-14292	65	9	ml	ml	NOUN
fpq-14292	65	10	systems	system	NOUN
fpq-14292	65	11	,	,	PUNCT
fpq-14292	65	12	including	include	VERB
fpq-14292	65	13	nlp	nlp	NOUN
fpq-14292	65	14	systems	system	NOUN
fpq-14292	65	15	,	,	PUNCT
fpq-14292	65	16	exhibit	exhibit	VERB
fpq-14292	65	17	a	a	DET
fpq-14292	65	18	close	close	ADJ
fpq-14292	65	19	analog	analog	NOUN
fpq-14292	65	20	of	of	ADP
fpq-14292	65	21	such	such	ADJ
fpq-14292	65	22	bias.6	bias.6	PROPN
fpq-14292	65	23	in	in	ADP
fpq-14292	65	24	each	each	DET
fpq-14292	65	25	case	case	NOUN
fpq-14292	65	26	,	,	PUNCT
fpq-14292	65	27	the	the	DET
fpq-14292	65	28	relevant	relevant	ADJ
fpq-14292	65	29	system	system	NOUN
fpq-14292	65	30	learns	learn	VERB
fpq-14292	65	31	or	or	CCONJ
fpq-14292	65	32	acquires	acquire	VERB
fpq-14292	65	33	its	its	PRON
fpq-14292	65	34	biases	bias	NOUN
fpq-14292	65	35	from	from	ADP
fpq-14292	65	36	statistical	statistical	ADJ
fpq-14292	65	37	regularities	regularity	NOUN
fpq-14292	65	38	it	it	PRON
fpq-14292	65	39	identifies	identify	VERB
fpq-14292	65	40	in	in	ADP
fpq-14292	65	41	the	the	DET
fpq-14292	65	42	world.7	world.7	NOUN
fpq-14292	65	43	we	we	PRON
fpq-14292	65	44	humans	human	NOUN
fpq-14292	65	45	learn	learn	VERB
fpq-14292	65	46	biases	bias	NOUN
fpq-14292	65	47	from	from	ADP
fpq-14292	65	48	other	other	ADJ
fpq-14292	65	49	members	member	NOUN
fpq-14292	65	50	of	of	ADP
fpq-14292	65	51	our	our	PRON
fpq-14292	65	52	communities.8	communities.8	PROPN
fpq-14292	65	53	for	for	ADP
fpq-14292	65	54	ml	ml	PROPN
fpq-14292	65	55	systems	system	NOUN
fpq-14292	65	56	,	,	PUNCT
fpq-14292	65	57	biases	bias	NOUN
fpq-14292	65	58	are	be	AUX
fpq-14292	65	59	often	often	ADV
fpq-14292	65	60	learned	learn	VERB
fpq-14292	65	61	from	from	ADP
fpq-14292	65	62	training	train	VERB
fpq-14292	65	63	data	datum	NOUN
fpq-14292	65	64	,	,	PUNCT
fpq-14292	65	65	which	which	PRON
fpq-14292	65	66	for	for	ADP
fpq-14292	65	67	nlp	nlp	NOUN
fpq-14292	65	68	systems	system	NOUN
fpq-14292	65	69	are	be	AUX
fpq-14292	65	70	language	language	NOUN
fpq-14292	65	71	corpora	corpora	NOUN
fpq-14292	65	72	that	that	SCONJ
fpq-14292	65	73	we	we	PRON
fpq-14292	65	74	produce	produce	VERB
fpq-14292	65	75	.	.	PUNCT
fpq-14292	66	1	since	since	SCONJ
fpq-14292	66	2	we	we	PRON
fpq-14292	66	3	produce	produce	VERB
fpq-14292	66	4	the	the	DET
fpq-14292	66	5	data	datum	NOUN
fpq-14292	66	6	containing	contain	VERB
fpq-14292	66	7	the	the	DET
fpq-14292	66	8	biases	bias	NOUN
fpq-14292	66	9	,	,	PUNCT
fpq-14292	66	10	the	the	DET
fpq-14292	66	11	biases	bias	NOUN
fpq-14292	66	12	the	the	DET
fpq-14292	66	13	nlp	nlp	NOUN
fpq-14292	66	14	systems	system	NOUN
fpq-14292	66	15	acquire	acquire	VERB
fpq-14292	66	16	they	they	PRON
fpq-14292	66	17	learn	learn	VERB
fpq-14292	66	18	from	from	ADP
fpq-14292	66	19	us.9	us.9	PROPN
fpq-14292	66	20	to	to	PART
fpq-14292	66	21	preview	preview	VERB
fpq-14292	66	22	one	one	NUM
fpq-14292	66	23	of	of	ADP
fpq-14292	66	24	our	our	PRON
fpq-14292	66	25	central	central	ADJ
fpq-14292	66	26	claims	claim	NOUN
fpq-14292	66	27	:	:	PUNCT
fpq-14292	66	28	it	it	PRON
fpq-14292	66	29	can	can	AUX
fpq-14292	66	30	be	be	AUX
fpq-14292	66	31	difficult	difficult	ADJ
fpq-14292	66	32	to	to	AUX
fpq-14292	66	33	debias	debias	PROPN
fpq-14292	66	34	human	human	ADJ
fpq-14292	66	35	cognition	cognition	NOUN
fpq-14292	66	36	,	,	PUNCT
fpq-14292	66	37	especially	especially	ADV
fpq-14292	66	38	for	for	ADP
fpq-14292	66	39	implicit	implicit	ADJ
fpq-14292	66	40	biases	bias	NOUN
fpq-14292	66	41	,	,	PUNCT
fpq-14292	66	42	and	and	CCONJ
fpq-14292	66	43	we	we	PRON
fpq-14292	66	44	think	think	VERB
fpq-14292	66	45	it	it	PRON
fpq-14292	66	46	is	be	AUX
fpq-14292	66	47	also	also	ADV
fpq-14292	66	48	difficult	difficult	ADJ
fpq-14292	66	49	to	to	ADP
fpq-14292	66	50	debias	debias	X
fpq-14292	66	51	ml	ml	ADP
fpq-14292	66	52	blaming	blame	VERB
fpq-14292	66	53	or	or	CCONJ
fpq-14292	66	54	praising	praise	VERB
fpq-14292	66	55	does	do	AUX
fpq-14292	66	56	require	require	VERB
fpq-14292	66	57	control	control	NOUN
fpq-14292	66	58	.	.	PUNCT
fpq-14292	67	1	the	the	DET
fpq-14292	67	2	empirical	empirical	ADJ
fpq-14292	67	3	literature	literature	NOUN
fpq-14292	67	4	suggests	suggest	VERB
fpq-14292	67	5	that	that	SCONJ
fpq-14292	67	6	we	we	PRON
fpq-14292	67	7	lack	lack	VERB
fpq-14292	67	8	such	such	ADJ
fpq-14292	67	9	control	control	NOUN
fpq-14292	67	10	over	over	ADP
fpq-14292	67	11	how	how	SCONJ
fpq-14292	67	12	implicit	implicit	ADJ
fpq-14292	67	13	biases	bias	NOUN
fpq-14292	67	14	affect	affect	VERB
fpq-14292	67	15	our	our	PRON
fpq-14292	67	16	behavior	behavior	NOUN
fpq-14292	67	17	(	(	PUNCT
fpq-14292	67	18	levy	levy	NOUN
fpq-14292	67	19	2017	2017	NUM
fpq-14292	67	20	)	)	PUNCT
fpq-14292	67	21	.	.	PUNCT
fpq-14292	68	1	5	5	NUM
fpq-14292	68	2	we	we	PRON
fpq-14292	68	3	make	make	VERB
fpq-14292	68	4	no	no	DET
fpq-14292	68	5	claim	claim	NOUN
fpq-14292	68	6	about	about	ADP
fpq-14292	68	7	whether	whether	SCONJ
fpq-14292	68	8	agents	agent	NOUN
fpq-14292	68	9	can	can	AUX
fpq-14292	68	10	be	be	AUX
fpq-14292	68	11	responsible	responsible	ADJ
fpq-14292	68	12	for	for	ADP
fpq-14292	68	13	actions	action	NOUN
fpq-14292	68	14	caused	cause	VERB
fpq-14292	68	15	by	by	ADP
fpq-14292	68	16	their	their	PRON
fpq-14292	68	17	implicit	implicit	ADJ
fpq-14292	68	18	attitudes	attitude	NOUN
fpq-14292	68	19	—	—	PUNCT
fpq-14292	68	20	although	although	SCONJ
fpq-14292	68	21	one	one	NUM
fpq-14292	68	22	of	of	ADP
fpq-14292	68	23	us	we	PRON
fpq-14292	68	24	defends	defend	VERB
fpq-14292	68	25	this	this	DET
fpq-14292	68	26	view	view	NOUN
fpq-14292	68	27	in	in	ADP
fpq-14292	68	28	print	print	NOUN
fpq-14292	68	29	(	(	PUNCT
fpq-14292	68	30	deery	deery	NOUN
fpq-14292	68	31	2021	2021	NUM
fpq-14292	68	32	)	)	PUNCT
fpq-14292	68	33	.	.	PUNCT
fpq-14292	69	1	here	here	ADV
fpq-14292	69	2	,	,	PUNCT
fpq-14292	69	3	we	we	PRON
fpq-14292	69	4	claim	claim	VERB
fpq-14292	69	5	only	only	ADV
fpq-14292	69	6	that	that	SCONJ
fpq-14292	69	7	even	even	ADV
fpq-14292	69	8	when	when	SCONJ
fpq-14292	69	9	behaviors	behavior	NOUN
fpq-14292	69	10	are	be	AUX
fpq-14292	69	11	caused	cause	VERB
fpq-14292	69	12	by	by	ADP
fpq-14292	69	13	implicit	implicit	ADJ
fpq-14292	69	14	attitudes	attitude	NOUN
fpq-14292	69	15	,	,	PUNCT
fpq-14292	69	16	we	we	PRON
fpq-14292	69	17	need	need	VERB
fpq-14292	69	18	to	to	PART
fpq-14292	69	19	do	do	VERB
fpq-14292	69	20	something	something	PRON
fpq-14292	69	21	to	to	PART
fpq-14292	69	22	control	control	VERB
fpq-14292	69	23	for	for	ADP
fpq-14292	69	24	their	their	PRON
fpq-14292	69	25	negative	negative	ADJ
fpq-14292	69	26	effects	effect	NOUN
fpq-14292	69	27	.	.	PUNCT
fpq-14292	70	1	6	6	NUM
fpq-14292	70	2	the	the	DET
fpq-14292	70	3	relevant	relevant	ADJ
fpq-14292	70	4	biases	bias	NOUN
fpq-14292	70	5	for	for	ADP
fpq-14292	70	6	ml	ml	NOUN
fpq-14292	70	7	systems	system	NOUN
fpq-14292	70	8	are	be	AUX
fpq-14292	70	9	analogous	analogous	ADJ
fpq-14292	70	10	to	to	ADP
fpq-14292	70	11	implicit	implicit	ADJ
fpq-14292	70	12	biases	bias	NOUN
fpq-14292	70	13	because	because	SCONJ
fpq-14292	70	14	such	such	ADJ
fpq-14292	70	15	biases	bias	NOUN
fpq-14292	70	16	are	be	AUX
fpq-14292	70	17	not	not	PART
fpq-14292	70	18	explicitly	explicitly	ADV
fpq-14292	70	19	represented	represent	VERB
fpq-14292	70	20	in	in	ADP
fpq-14292	70	21	models	model	NOUN
fpq-14292	70	22	but	but	CCONJ
fpq-14292	70	23	instead	instead	ADV
fpq-14292	70	24	take	take	VERB
fpq-14292	70	25	the	the	DET
fpq-14292	70	26	form	form	NOUN
fpq-14292	70	27	of	of	ADP
fpq-14292	70	28	implicit	implicit	ADJ
fpq-14292	70	29	distributed	distribute	VERB
fpq-14292	70	30	patterns	pattern	NOUN
fpq-14292	70	31	—	—	PUNCT
fpq-14292	70	32	e.g.	e.g.	ADV
fpq-14292	70	33	,	,	PUNCT
fpq-14292	70	34	between	between	ADP
fpq-14292	70	35	vectors	vector	NOUN
fpq-14292	70	36	in	in	ADP
fpq-14292	70	37	a	a	DET
fpq-14292	70	38	model	model	NOUN
fpq-14292	70	39	.	.	PUNCT
fpq-14292	71	1	we	we	PRON
fpq-14292	71	2	expand	expand	VERB
fpq-14292	71	3	on	on	ADP
fpq-14292	71	4	this	this	DET
fpq-14292	71	5	analogy	analogy	NOUN
fpq-14292	71	6	between	between	ADP
fpq-14292	71	7	human	human	ADJ
fpq-14292	71	8	cognition	cognition	NOUN
fpq-14292	71	9	and	and	CCONJ
fpq-14292	71	10	ml	ml	NOUN
fpq-14292	71	11	systems	system	NOUN
fpq-14292	71	12	in	in	ADP
fpq-14292	71	13	section	section	NOUN
fpq-14292	71	14	3	3	NUM
fpq-14292	71	15	.	.	NOUN
fpq-14292	71	16	7	7	NUM
fpq-14292	71	17	some	some	DET
fpq-14292	71	18	human	human	ADJ
fpq-14292	71	19	cognitive	cognitive	ADJ
fpq-14292	71	20	or	or	CCONJ
fpq-14292	71	21	perceptual	perceptual	ADJ
fpq-14292	71	22	biases	bias	NOUN
fpq-14292	71	23	may	may	AUX
fpq-14292	71	24	be	be	AUX
fpq-14292	71	25	innate	innate	ADJ
fpq-14292	71	26	.	.	PUNCT
fpq-14292	72	1	they	they	PRON
fpq-14292	72	2	are	be	AUX
fpq-14292	72	3	not	not	PART
fpq-14292	72	4	our	our	PRON
fpq-14292	72	5	focus	focus	NOUN
fpq-14292	72	6	here	here	ADV
fpq-14292	72	7	.	.	PUNCT
fpq-14292	73	1	8	8	NUM
fpq-14292	74	1	in	in	ADP
fpq-14292	74	2	cases	case	NOUN
fpq-14292	74	3	where	where	SCONJ
fpq-14292	74	4	a	a	DET
fpq-14292	74	5	population	population	NOUN
fpq-14292	74	6	uses	use	VERB
fpq-14292	74	7	words	word	NOUN
fpq-14292	74	8	in	in	ADP
fpq-14292	74	9	certain	certain	ADJ
fpq-14292	74	10	ways	way	NOUN
fpq-14292	74	11	and	and	CCONJ
fpq-14292	74	12	people	people	NOUN
fpq-14292	74	13	acquire	acquire	VERB
fpq-14292	74	14	biases	bias	NOUN
fpq-14292	74	15	from	from	ADP
fpq-14292	74	16	exposure	exposure	NOUN
fpq-14292	74	17	to	to	ADP
fpq-14292	74	18	that	that	DET
fpq-14292	74	19	usage	usage	NOUN
fpq-14292	74	20	,	,	PUNCT
fpq-14292	74	21	the	the	DET
fpq-14292	74	22	exposure	exposure	NOUN
fpq-14292	74	23	might	might	AUX
fpq-14292	74	24	be	be	AUX
fpq-14292	74	25	mediated	mediate	VERB
fpq-14292	74	26	by	by	ADP
fpq-14292	74	27	how	how	SCONJ
fpq-14292	74	28	the	the	DET
fpq-14292	74	29	media	medium	NOUN
fpq-14292	74	30	portrays	portray	VERB
fpq-14292	74	31	such	such	ADJ
fpq-14292	74	32	language	language	NOUN
fpq-14292	74	33	use	use	NOUN
fpq-14292	74	34	.	.	PUNCT
fpq-14292	75	1	even	even	ADV
fpq-14292	75	2	so	so	ADV
fpq-14292	75	3	,	,	PUNCT
fpq-14292	75	4	we	we	PRON
fpq-14292	75	5	acquire	acquire	VERB
fpq-14292	75	6	biases	bias	NOUN
fpq-14292	75	7	from	from	ADP
fpq-14292	75	8	what	what	PRON
fpq-14292	75	9	we	we	PRON
fpq-14292	75	10	are	be	AUX
fpq-14292	75	11	exposed	expose	VERB
fpq-14292	75	12	to	to	ADP
fpq-14292	75	13	(	(	PUNCT
fpq-14292	75	14	the	the	DET
fpq-14292	75	15	world	world	NOUN
fpq-14292	75	16	)	)	PUNCT
fpq-14292	75	17	,	,	PUNCT
fpq-14292	75	18	which	which	PRON
fpq-14292	75	19	here	here	ADV
fpq-14292	75	20	would	would	AUX
fpq-14292	75	21	include	include	VERB
fpq-14292	75	22	how	how	SCONJ
fpq-14292	75	23	the	the	DET
fpq-14292	75	24	media	medium	NOUN
fpq-14292	75	25	portrays	portray	VERB
fpq-14292	75	26	language	language	NOUN
fpq-14292	75	27	use	use	NOUN
fpq-14292	75	28	.	.	PUNCT
fpq-14292	76	1	9	9	NUM
fpq-14292	76	2	might	might	AUX
fpq-14292	76	3	we	we	PRON
fpq-14292	76	4	have	have	VERB
fpq-14292	76	5	more	more	ADJ
fpq-14292	76	6	control	control	NOUN
fpq-14292	76	7	over	over	ADP
fpq-14292	76	8	the	the	DET
fpq-14292	76	9	inputs	input	NOUN
fpq-14292	76	10	to	to	AUX
fpq-14292	76	11	ml	ml	NOUN
fpq-14292	76	12	systems	system	NOUN
fpq-14292	76	13	than	than	SCONJ
fpq-14292	76	14	we	we	PRON
fpq-14292	76	15	do	do	AUX
fpq-14292	76	16	for	for	ADP
fpq-14292	76	17	ourselves	ourselves	PRON
fpq-14292	76	18	?	?	PUNCT
fpq-14292	77	1	if	if	SCONJ
fpq-14292	77	2	so	so	ADV
fpq-14292	77	3	,	,	PUNCT
fpq-14292	77	4	we	we	PRON
fpq-14292	77	5	could	could	AUX
fpq-14292	77	6	control	control	VERB
fpq-14292	77	7	for	for	ADP
fpq-14292	77	8	biases	bias	NOUN
fpq-14292	77	9	at	at	ADP
fpq-14292	77	10	the	the	DET
fpq-14292	77	11	input	input	NOUN
fpq-14292	77	12	stage	stage	NOUN
fpq-14292	77	13	.	.	PUNCT
fpq-14292	78	1	yet	yet	CCONJ
fpq-14292	78	2	the	the	DET
fpq-14292	78	3	corpora	corpora	NOUN
fpq-14292	78	4	on	on	ADP
fpq-14292	78	5	which	which	PRON
fpq-14292	78	6	ml	ml	NOUN
fpq-14292	78	7	systems	system	NOUN
fpq-14292	78	8	are	be	AUX
fpq-14292	78	9	trained	train	VERB
fpq-14292	78	10	are	be	AUX
fpq-14292	78	11	vast	vast	ADJ
fpq-14292	78	12	,	,	PUNCT
fpq-14292	78	13	and	and	CCONJ
fpq-14292	78	14	to	to	PART
fpq-14292	78	15	eradicate	eradicate	VERB
fpq-14292	78	16	biases	bias	NOUN
fpq-14292	78	17	from	from	ADP
fpq-14292	78	18	them	they	PRON
fpq-14292	78	19	would	would	AUX
fpq-14292	78	20	be	be	AUX
fpq-14292	78	21	an	an	DET
fpq-14292	78	22	impractical	impractical	ADJ
fpq-14292	78	23	task	task	NOUN
fpq-14292	78	24	.	.	PUNCT
fpq-14292	79	1	that	that	PRON
fpq-14292	79	2	is	be	AUX
fpq-14292	79	3	why	why	SCONJ
fpq-14292	79	4	those	those	PRON
fpq-14292	79	5	who	who	PRON
fpq-14292	79	6	seek	seek	VERB
fpq-14292	79	7	to	to	PART
fpq-14292	79	8	eliminate	eliminate	VERB
fpq-14292	79	9	bias	bias	NOUN
fpq-14292	79	10	from	from	ADP
fpq-14292	79	11	models	model	NOUN
fpq-14292	79	12	typically	typically	ADV
fpq-14292	79	13	focus	focus	VERB
fpq-14292	79	14	on	on	ADP
fpq-14292	79	15	the	the	DET
fpq-14292	79	16	already	already	ADV
fpq-14292	79	17	-	-	PUNCT
fpq-14292	79	18	trained	train	VERB
fpq-14292	79	19	models	model	NOUN
fpq-14292	79	20	(	(	PUNCT
fpq-14292	79	21	see	see	VERB
fpq-14292	79	22	section	section	NOUN
fpq-14292	79	23	5	5	NUM
fpq-14292	79	24	for	for	ADP
fpq-14292	79	25	details	detail	NOUN
fpq-14292	79	26	)	)	PUNCT
fpq-14292	79	27	.	.	PUNCT
fpq-14292	80	1	exceptions	exception	NOUN
fpq-14292	80	2	include	include	VERB
fpq-14292	80	3	badly	badly	ADV
fpq-14292	80	4	sampled	sample	VERB
fpq-14292	80	5	data	datum	NOUN
fpq-14292	80	6	and	and	CCONJ
fpq-14292	80	7	algorithms	algorithm	NOUN
fpq-14292	80	8	that	that	PRON
fpq-14292	80	9	are	be	AUX
fpq-14292	80	10	biased	bias	VERB
fpq-14292	80	11	for	for	ADP
fpq-14292	80	12	other	other	ADJ
fpq-14292	80	13	reasons	reason	NOUN
fpq-14292	80	14	(	(	PUNCT
fpq-14292	80	15	see	see	VERB
fpq-14292	80	16	sections	section	NOUN
fpq-14292	80	17	6.3	6.3	NUM
fpq-14292	80	18	and	and	CCONJ
fpq-14292	80	19	8	8	NUM
fpq-14292	80	20	below	below	ADV
fpq-14292	80	21	)	)	PUNCT
fpq-14292	80	22	.	.	PUNCT
fpq-14292	81	1	deery	deery	NOUN
fpq-14292	81	2	and	and	CCONJ
fpq-14292	81	3	bailey	bailey	PROPN
fpq-14292	81	4	–	–	PUNCT
fpq-14292	81	5	the	the	DET
fpq-14292	81	6	bias	bias	NOUN
fpq-14292	81	7	dilemma	dilemma	NOUN
fpq-14292	81	8	:	:	PUNCT
fpq-14292	81	9	the	the	DET
fpq-14292	81	10	ethics	ethic	NOUN
fpq-14292	81	11	of	of	ADP
fpq-14292	81	12	algorithmic	algorithmic	ADJ
fpq-14292	81	13	bias	bias	NOUN
fpq-14292	81	14	in	in	ADP
fpq-14292	81	15	natural	natural	ADJ
fpq-14292	81	16	-	-	PUNCT
fpq-14292	81	17	language	language	NOUN
fpq-14292	81	18	processing	processing	NOUN
fpq-14292	81	19	published	publish	VERB
fpq-14292	81	20	by	by	ADP
fpq-14292	81	21	scholarship@western	scholarship@western	PROPN
fpq-14292	81	22	,	,	PUNCT
fpq-14292	81	23	2022	2022	NUM
fpq-14292	81	24	5	5	NUM
fpq-14292	81	25	systems	system	NOUN
fpq-14292	81	26	,	,	PUNCT
fpq-14292	81	27	including	include	VERB
fpq-14292	81	28	nlp	nlp	NOUN
fpq-14292	81	29	systems	system	NOUN
fpq-14292	81	30	.	.	PUNCT
fpq-14292	82	1	for	for	ADP
fpq-14292	82	2	humans	human	NOUN
fpq-14292	82	3	,	,	PUNCT
fpq-14292	82	4	it	it	PRON
fpq-14292	82	5	is	be	AUX
fpq-14292	82	6	often	often	ADV
fpq-14292	82	7	easiest	easy	ADJ
fpq-14292	82	8	to	to	PART
fpq-14292	82	9	control	control	VERB
fpq-14292	82	10	for	for	ADP
fpq-14292	82	11	the	the	DET
fpq-14292	82	12	negative	negative	ADJ
fpq-14292	82	13	effects	effect	NOUN
fpq-14292	82	14	of	of	ADP
fpq-14292	82	15	bias	bias	NOUN
fpq-14292	82	16	at	at	ADP
fpq-14292	82	17	the	the	DET
fpq-14292	82	18	output	output	NOUN
fpq-14292	82	19	stage	stage	NOUN
fpq-14292	82	20	.	.	PUNCT
fpq-14292	83	1	as	as	SCONJ
fpq-14292	83	2	noted	note	VERB
fpq-14292	83	3	already	already	ADV
fpq-14292	83	4	,	,	PUNCT
fpq-14292	83	5	auditioning	audition	VERB
fpq-14292	83	6	musicians	musician	NOUN
fpq-14292	83	7	can	can	AUX
fpq-14292	83	8	perform	perform	VERB
fpq-14292	83	9	behind	behind	ADP
fpq-14292	83	10	a	a	DET
fpq-14292	83	11	screen	screen	NOUN
fpq-14292	83	12	,	,	PUNCT
fpq-14292	83	13	thereby	thereby	ADV
fpq-14292	83	14	defusing	defuse	VERB
fpq-14292	83	15	the	the	DET
fpq-14292	83	16	possibility	possibility	NOUN
fpq-14292	83	17	of	of	ADP
fpq-14292	83	18	implicit	implicit	ADJ
fpq-14292	83	19	biases	bias	NOUN
fpq-14292	83	20	’	'	PUNCT
fpq-14292	83	21	influencing	influence	VERB
fpq-14292	83	22	hiring	hiring	NOUN
fpq-14292	83	23	.	.	PUNCT
fpq-14292	84	1	as	as	ADP
fpq-14292	84	2	a	a	DET
fpq-14292	84	3	general	general	ADJ
fpq-14292	84	4	matter	matter	NOUN
fpq-14292	84	5	,	,	PUNCT
fpq-14292	84	6	it	it	PRON
fpq-14292	84	7	seems	seem	VERB
fpq-14292	84	8	that	that	SCONJ
fpq-14292	84	9	the	the	DET
fpq-14292	84	10	negative	negative	ADJ
fpq-14292	84	11	effects	effect	NOUN
fpq-14292	84	12	of	of	ADP
fpq-14292	84	13	bias	bias	NOUN
fpq-14292	84	14	are	be	AUX
fpq-14292	84	15	often	often	ADV
fpq-14292	84	16	best	well	ADV
fpq-14292	84	17	controlled	control	VERB
fpq-14292	84	18	for	for	ADP
fpq-14292	84	19	when	when	SCONJ
fpq-14292	84	20	we	we	PRON
fpq-14292	84	21	acknowledge	acknowledge	VERB
fpq-14292	84	22	bias	bias	NOUN
fpq-14292	84	23	in	in	ADP
fpq-14292	84	24	a	a	DET
fpq-14292	84	25	system	system	NOUN
fpq-14292	84	26	and	and	CCONJ
fpq-14292	84	27	control	control	NOUN
fpq-14292	84	28	for	for	ADP
fpq-14292	84	29	its	its	PRON
fpq-14292	84	30	negative	negative	ADJ
fpq-14292	84	31	influences	influence	NOUN
fpq-14292	84	32	at	at	ADP
fpq-14292	84	33	the	the	DET
fpq-14292	84	34	output	output	NOUN
fpq-14292	84	35	stage	stage	NOUN
fpq-14292	84	36	rather	rather	ADV
fpq-14292	84	37	than	than	ADP
fpq-14292	84	38	by	by	ADP
fpq-14292	84	39	debiasing	debiase	VERB
fpq-14292	84	40	the	the	DET
fpq-14292	84	41	system	system	NOUN
fpq-14292	84	42	itself.10	itself.10	ADP
fpq-14292	84	43	the	the	DET
fpq-14292	84	44	possibility	possibility	NOUN
fpq-14292	84	45	we	we	PRON
fpq-14292	84	46	explore	explore	VERB
fpq-14292	84	47	in	in	ADP
fpq-14292	84	48	this	this	DET
fpq-14292	84	49	paper	paper	NOUN
fpq-14292	84	50	is	be	AUX
fpq-14292	84	51	whether	whether	SCONJ
fpq-14292	84	52	we	we	PRON
fpq-14292	84	53	might	might	AUX
fpq-14292	84	54	do	do	VERB
fpq-14292	84	55	something	something	PRON
fpq-14292	84	56	similar	similar	ADJ
fpq-14292	84	57	for	for	ADP
fpq-14292	84	58	nlp	nlp	NOUN
fpq-14292	84	59	systems	system	NOUN
fpq-14292	84	60	.	.	PUNCT
fpq-14292	85	1	moreover	moreover	ADV
fpq-14292	85	2	,	,	PUNCT
fpq-14292	85	3	we	we	PRON
fpq-14292	85	4	claim	claim	VERB
fpq-14292	85	5	that	that	SCONJ
fpq-14292	85	6	there	there	PRON
fpq-14292	85	7	can	can	AUX
fpq-14292	85	8	be	be	AUX
fpq-14292	85	9	an	an	DET
fpq-14292	85	10	ethical	ethical	ADJ
fpq-14292	85	11	value	value	NOUN
fpq-14292	85	12	to	to	ADP
fpq-14292	85	13	not	not	PART
fpq-14292	85	14	debiasing	debiase	VERB
fpq-14292	85	15	ml	ml	NOUN
fpq-14292	85	16	systems	system	NOUN
fpq-14292	85	17	or	or	CCONJ
fpq-14292	85	18	their	their	PRON
fpq-14292	85	19	outputs	output	NOUN
fpq-14292	85	20	,	,	PUNCT
fpq-14292	85	21	and	and	CCONJ
fpq-14292	85	22	we	we	PRON
fpq-14292	85	23	maintain	maintain	VERB
fpq-14292	85	24	that	that	SCONJ
fpq-14292	85	25	the	the	DET
fpq-14292	85	26	question	question	NOUN
fpq-14292	85	27	of	of	ADP
fpq-14292	85	28	whether	whether	SCONJ
fpq-14292	85	29	we	we	PRON
fpq-14292	85	30	should	should	AUX
fpq-14292	85	31	debias	debias	PROPN
fpq-14292	85	32	raises	raise	VERB
fpq-14292	85	33	an	an	DET
fpq-14292	85	34	ethical	ethical	ADJ
fpq-14292	85	35	dilemma	dilemma	NOUN
fpq-14292	85	36	.	.	PUNCT
fpq-14292	86	1	3	3	X
fpq-14292	86	2	.	.	X
fpq-14292	86	3	descriptive	descriptive	ADJ
fpq-14292	86	4	vs.	vs.	ADP
fpq-14292	86	5	normative	normative	ADJ
fpq-14292	86	6	accuracy	accuracy	NOUN
fpq-14292	86	7	to	to	PART
fpq-14292	86	8	explain	explain	VERB
fpq-14292	86	9	our	our	PRON
fpq-14292	86	10	claim	claim	NOUN
fpq-14292	86	11	that	that	SCONJ
fpq-14292	86	12	there	there	PRON
fpq-14292	86	13	can	can	AUX
fpq-14292	86	14	be	be	AUX
fpq-14292	86	15	an	an	DET
fpq-14292	86	16	ethical	ethical	ADJ
fpq-14292	86	17	value	value	NOUN
fpq-14292	86	18	to	to	ADP
fpq-14292	86	19	not	not	PART
fpq-14292	86	20	debiasing	debiase	VERB
fpq-14292	86	21	models	model	NOUN
fpq-14292	86	22	or	or	CCONJ
fpq-14292	86	23	outputs	output	NOUN
fpq-14292	86	24	,	,	PUNCT
fpq-14292	86	25	we	we	PRON
fpq-14292	86	26	rely	rely	VERB
fpq-14292	86	27	on	on	ADP
fpq-14292	86	28	a	a	DET
fpq-14292	86	29	particular	particular	ADJ
fpq-14292	86	30	way	way	NOUN
fpq-14292	86	31	of	of	ADP
fpq-14292	86	32	distinguishing	distinguish	VERB
fpq-14292	86	33	descriptive	descriptive	ADJ
fpq-14292	86	34	accuracy	accuracy	NOUN
fpq-14292	86	35	from	from	ADP
fpq-14292	86	36	normative	normative	ADJ
fpq-14292	86	37	correctness.11	correctness.11	ADV
fpq-14292	86	38	we	we	PRON
fpq-14292	86	39	explain	explain	VERB
fpq-14292	86	40	our	our	PRON
fpq-14292	86	41	distinction	distinction	NOUN
fpq-14292	86	42	by	by	ADP
fpq-14292	86	43	analogy	analogy	NOUN
fpq-14292	86	44	with	with	ADP
fpq-14292	86	45	how	how	SCONJ
fpq-14292	86	46	implicit	implicit	ADJ
fpq-14292	86	47	biases	bias	NOUN
fpq-14292	86	48	in	in	ADP
fpq-14292	86	49	human	human	ADJ
fpq-14292	86	50	cognition	cognition	NOUN
fpq-14292	86	51	can	can	AUX
fpq-14292	86	52	offer	offer	VERB
fpq-14292	86	53	support	support	NOUN
fpq-14292	86	54	for	for	ADP
fpq-14292	86	55	beliefs	belief	NOUN
fpq-14292	86	56	even	even	ADV
fpq-14292	86	57	when	when	SCONJ
fpq-14292	86	58	these	these	DET
fpq-14292	86	59	beliefs	belief	NOUN
fpq-14292	86	60	are	be	AUX
fpq-14292	86	61	morally	morally	ADV
fpq-14292	86	62	problematic	problematic	ADJ
fpq-14292	86	63	.	.	PUNCT
fpq-14292	87	1	in	in	ADP
fpq-14292	87	2	this	this	DET
fpq-14292	87	3	way	way	NOUN
fpq-14292	87	4	,	,	PUNCT
fpq-14292	87	5	people	people	NOUN
fpq-14292	87	6	’s	’s	PART
fpq-14292	87	7	epistemic	epistemic	ADJ
fpq-14292	87	8	reasons	reason	NOUN
fpq-14292	87	9	can	can	AUX
fpq-14292	87	10	support	support	VERB
fpq-14292	87	11	biased	biased	ADJ
fpq-14292	87	12	beliefs	belief	NOUN
fpq-14292	87	13	even	even	ADV
fpq-14292	87	14	when	when	SCONJ
fpq-14292	87	15	their	their	PRON
fpq-14292	87	16	ethical	ethical	ADJ
fpq-14292	87	17	reasons	reason	NOUN
fpq-14292	87	18	cut	cut	VERB
fpq-14292	87	19	against	against	ADP
fpq-14292	87	20	these	these	DET
fpq-14292	87	21	beliefs	belief	NOUN
fpq-14292	87	22	.	.	PUNCT
fpq-14292	88	1	in	in	ADP
fpq-14292	88	2	relation	relation	NOUN
fpq-14292	88	3	to	to	ADP
fpq-14292	88	4	racist	racist	ADJ
fpq-14292	88	5	bias	bias	NOUN
fpq-14292	88	6	,	,	PUNCT
fpq-14292	88	7	rima	rima	PROPN
fpq-14292	88	8	basu	basu	PROPN
fpq-14292	88	9	outlines	outline	VERB
fpq-14292	88	10	this	this	DET
fpq-14292	88	11	epistemic	epistemic	ADJ
fpq-14292	88	12	/	/	SYM
fpq-14292	88	13	ethical	ethical	ADJ
fpq-14292	88	14	conflict	conflict	NOUN
fpq-14292	88	15	as	as	SCONJ
fpq-14292	88	16	follows	follow	VERB
fpq-14292	88	17	:	:	PUNCT
fpq-14292	88	18	we	we	PRON
fpq-14292	88	19	may	may	AUX
fpq-14292	88	20	find	find	VERB
fpq-14292	88	21	ourselves	ourselves	PRON
fpq-14292	88	22	facing	face	VERB
fpq-14292	88	23	the	the	DET
fpq-14292	88	24	following	follow	VERB
fpq-14292	88	25	conflict	conflict	NOUN
fpq-14292	88	26	:	:	PUNCT
fpq-14292	88	27	what	what	PRON
fpq-14292	88	28	if	if	SCONJ
fpq-14292	88	29	the	the	DET
fpq-14292	88	30	evidence	evidence	NOUN
fpq-14292	88	31	.	.	PUNCT
fpq-14292	88	32	.	.	PUNCT
fpq-14292	89	1	.	.	PUNCT
fpq-14292	89	2	supports	support	VERB
fpq-14292	89	3	something	something	PRON
fpq-14292	89	4	we	we	PRON
fpq-14292	89	5	morally	morally	ADV
fpq-14292	89	6	should	should	AUX
fpq-14292	89	7	n’t	not	PART
fpq-14292	89	8	believe	believe	VERB
fpq-14292	89	9	?	?	PUNCT
fpq-14292	90	1	for	for	ADP
fpq-14292	90	2	example	example	NOUN
fpq-14292	90	3	,	,	PUNCT
fpq-14292	90	4	it	it	PRON
fpq-14292	90	5	is	be	AUX
fpq-14292	90	6	morally	morally	ADV
fpq-14292	90	7	wrong	wrong	ADJ
fpq-14292	90	8	to	to	PART
fpq-14292	90	9	assume	assume	VERB
fpq-14292	90	10	,	,	PUNCT
fpq-14292	90	11	solely	solely	ADV
fpq-14292	90	12	on	on	ADP
fpq-14292	90	13	the	the	DET
fpq-14292	90	14	basis	basis	NOUN
fpq-14292	90	15	of	of	ADP
fpq-14292	90	16	someone	someone	PRON
fpq-14292	90	17	’s	’s	PART
fpq-14292	90	18	skin	skin	NOUN
fpq-14292	90	19	color	color	NOUN
fpq-14292	90	20	,	,	PUNCT
fpq-14292	90	21	that	that	SCONJ
fpq-14292	90	22	they	they	PRON
fpq-14292	90	23	’re	’re	VERB
fpq-14292	90	24	a	a	DET
fpq-14292	90	25	staff	staff	NOUN
fpq-14292	90	26	member	member	NOUN
fpq-14292	90	27	.	.	PUNCT
fpq-14292	91	1	but	but	CCONJ
fpq-14292	91	2	,	,	PUNCT
fpq-14292	91	3	what	what	PRON
fpq-14292	91	4	if	if	SCONJ
fpq-14292	91	5	you	you	PRON
fpq-14292	91	6	’re	’re	VERB
fpq-14292	91	7	in	in	ADP
fpq-14292	91	8	a	a	DET
fpq-14292	91	9	context	context	NOUN
fpq-14292	91	10	where	where	SCONJ
fpq-14292	91	11	,	,	PUNCT
fpq-14292	91	12	because	because	SCONJ
fpq-14292	91	13	of	of	ADP
fpq-14292	91	14	historical	historical	ADJ
fpq-14292	91	15	patterns	pattern	NOUN
fpq-14292	91	16	of	of	ADP
fpq-14292	91	17	discrimination	discrimination	NOUN
fpq-14292	91	18	,	,	PUNCT
fpq-14292	91	19	someone	someone	PRON
fpq-14292	91	20	’s	’s	PART
fpq-14292	91	21	skin	skin	NOUN
fpq-14292	91	22	color	color	NOUN
fpq-14292	91	23	is	be	AUX
fpq-14292	91	24	a	a	DET
fpq-14292	91	25	very	very	ADV
fpq-14292	91	26	good	good	ADJ
fpq-14292	91	27	indicator	indicator	NOUN
fpq-14292	91	28	that	that	SCONJ
fpq-14292	91	29	they	they	PRON
fpq-14292	91	30	’re	’re	VERB
fpq-14292	91	31	a	a	DET
fpq-14292	91	32	staff	staff	NOUN
fpq-14292	91	33	member	member	NOUN
fpq-14292	91	34	?	?	PUNCT
fpq-14292	92	1	when	when	SCONJ
fpq-14292	92	2	this	this	DET
fpq-14292	92	3	sort	sort	NOUN
fpq-14292	92	4	of	of	ADP
fpq-14292	92	5	normative	normative	ADJ
fpq-14292	92	6	conflict	conflict	NOUN
fpq-14292	92	7	looms	loom	VERB
fpq-14292	92	8	,	,	PUNCT
fpq-14292	92	9	a	a	DET
fpq-14292	92	10	conflict	conflict	NOUN
fpq-14292	92	11	between	between	ADP
fpq-14292	92	12	moral	moral	ADJ
fpq-14292	92	13	considerations	consideration	NOUN
fpq-14292	92	14	on	on	ADP
fpq-14292	92	15	the	the	DET
fpq-14292	92	16	one	one	NUM
fpq-14292	92	17	hand	hand	NOUN
fpq-14292	92	18	and	and	CCONJ
fpq-14292	92	19	what	what	PRON
fpq-14292	92	20	you	you	PRON
fpq-14292	92	21	epistemically	epistemically	ADV
fpq-14292	92	22	ought	ought	AUX
fpq-14292	92	23	to	to	PART
fpq-14292	92	24	believe	believe	VERB
fpq-14292	92	25	given	give	VERB
fpq-14292	92	26	the	the	DET
fpq-14292	92	27	evidence	evidence	NOUN
fpq-14292	92	28	on	on	ADP
fpq-14292	92	29	the	the	DET
fpq-14292	92	30	other	other	ADJ
fpq-14292	92	31	,	,	PUNCT
fpq-14292	92	32	what	what	PRON
fpq-14292	92	33	should	should	AUX
fpq-14292	92	34	we	we	PRON
fpq-14292	92	35	do	do	VERB
fpq-14292	92	36	?	?	PUNCT
fpq-14292	93	1	it	it	PRON
fpq-14292	93	2	might	might	AUX
fpq-14292	93	3	be	be	AUX
fpq-14292	93	4	unfair	unfair	ADJ
fpq-14292	93	5	to	to	PART
fpq-14292	93	6	assume	assume	VERB
fpq-14292	93	7	that	that	SCONJ
fpq-14292	93	8	they	they	PRON
fpq-14292	93	9	’re	’re	VERB
fpq-14292	93	10	a	a	DET
fpq-14292	93	11	staff	staff	NOUN
fpq-14292	93	12	member	member	NOUN
fpq-14292	93	13	,	,	PUNCT
fpq-14292	93	14	but	but	CCONJ
fpq-14292	93	15	to	to	PART
fpq-14292	93	16	ignore	ignore	VERB
fpq-14292	93	17	the	the	DET
fpq-14292	93	18	evidence	evidence	NOUN
fpq-14292	93	19	would	would	AUX
fpq-14292	93	20	mean	mean	VERB
fpq-14292	93	21	risking	risk	VERB
fpq-14292	93	22	inaccurate	inaccurate	ADJ
fpq-14292	93	23	beliefs	belief	NOUN
fpq-14292	93	24	.	.	PUNCT
fpq-14292	94	1	some	some	PRON
fpq-14292	94	2	.	.	PUNCT
fpq-14292	94	3	.	.	PUNCT
fpq-14292	94	4	.	.	PUNCT
fpq-14292	95	1	have	have	AUX
fpq-14292	95	2	suggested	suggest	VERB
fpq-14292	95	3	that	that	SCONJ
fpq-14292	95	4	we	we	PRON
fpq-14292	95	5	simply	simply	ADV
fpq-14292	95	6	face	face	VERB
fpq-14292	95	7	a	a	DET
fpq-14292	95	8	tragic	tragic	ADJ
fpq-14292	95	9	irresolvable	irresolvable	ADJ
fpq-14292	95	10	dilemma	dilemma	NOUN
fpq-14292	95	11	.	.	PUNCT
fpq-14292	96	1	(	(	PUNCT
fpq-14292	96	2	basu	basu	PROPN
fpq-14292	96	3	2020	2020	NUM
fpq-14292	96	4	,	,	PUNCT
fpq-14292	96	5	191	191	NUM
fpq-14292	96	6	;	;	PUNCT
fpq-14292	96	7	cf	cf	NOUN
fpq-14292	96	8	.	.	PUNCT
fpq-14292	97	1	gendler	gendler	PROPN
fpq-14292	97	2	2011	2011	NUM
fpq-14292	97	3	;	;	PUNCT
fpq-14292	97	4	puddifoot	puddifoot	NOUN
fpq-14292	97	5	2017	2017	NUM
fpq-14292	97	6	)	)	PUNCT
fpq-14292	97	7	10	10	NUM
fpq-14292	97	8	if	if	SCONJ
fpq-14292	97	9	the	the	DET
fpq-14292	97	10	bias	bias	NOUN
fpq-14292	97	11	comes	come	VERB
fpq-14292	97	12	from	from	ADP
fpq-14292	97	13	badly	badly	ADV
fpq-14292	97	14	sampled	sample	VERB
fpq-14292	97	15	data	datum	NOUN
fpq-14292	97	16	,	,	PUNCT
fpq-14292	97	17	that	that	PRON
fpq-14292	97	18	is	be	AUX
fpq-14292	97	19	a	a	DET
fpq-14292	97	20	different	different	ADJ
fpq-14292	97	21	matter	matter	NOUN
fpq-14292	97	22	.	.	PUNCT
fpq-14292	98	1	11	11	NUM
fpq-14292	98	2	we	we	PRON
fpq-14292	98	3	introduce	introduce	VERB
fpq-14292	98	4	the	the	DET
fpq-14292	98	5	normative	normative	ADJ
fpq-14292	98	6	/	/	SYM
fpq-14292	98	7	descriptive	descriptive	ADJ
fpq-14292	98	8	distinction	distinction	NOUN
fpq-14292	98	9	at	at	ADP
fpq-14292	98	10	some	some	DET
fpq-14292	98	11	length	length	NOUN
fpq-14292	98	12	since	since	SCONJ
fpq-14292	98	13	we	we	PRON
fpq-14292	98	14	have	have	AUX
fpq-14292	98	15	found	find	VERB
fpq-14292	98	16	it	it	PRON
fpq-14292	98	17	is	be	AUX
fpq-14292	98	18	unfamiliar	unfamiliar	ADJ
fpq-14292	98	19	to	to	ADP
fpq-14292	98	20	nonphilosophers	nonphilosopher	NOUN
fpq-14292	98	21	working	work	VERB
fpq-14292	98	22	in	in	ADP
fpq-14292	98	23	ai	ai	PROPN
fpq-14292	98	24	and	and	CCONJ
fpq-14292	98	25	ml	ml	VERB
fpq-14292	98	26	.	.	PUNCT
fpq-14292	99	1	yet	yet	ADV
fpq-14292	99	2	these	these	DET
fpq-14292	99	3	researchers	researcher	NOUN
fpq-14292	99	4	find	find	VERB
fpq-14292	99	5	it	it	PRON
fpq-14292	99	6	useful	useful	ADJ
fpq-14292	99	7	when	when	SCONJ
fpq-14292	99	8	explained	explain	VERB
fpq-14292	99	9	.	.	PUNCT
fpq-14292	100	1	feminist	feminist	PROPN
fpq-14292	100	2	philosophy	philosophy	PROPN
fpq-14292	100	3	quarterly	quarterly	ADV
fpq-14292	100	4	,	,	PUNCT
fpq-14292	100	5	2022	2022	NUM
fpq-14292	100	6	,	,	PUNCT
fpq-14292	100	7	vol	vol	NOUN
fpq-14292	100	8	.	.	NOUN
fpq-14292	100	9	8	8	NUM
fpq-14292	100	10	,	,	PUNCT
fpq-14292	100	11	iss	iss	PROPN
fpq-14292	100	12	.	.	PUNCT
fpq-14292	100	13	3/4	3/4	NUM
fpq-14292	100	14	,	,	PUNCT
fpq-14292	100	15	article	article	NOUN
fpq-14292	100	16	1	1	NUM
fpq-14292	100	17	published	publish	VERB
fpq-14292	100	18	by	by	ADP
fpq-14292	100	19	scholarship@western	scholarship@western	PROPN
fpq-14292	100	20	,	,	PUNCT
fpq-14292	100	21	2022	2022	NUM
fpq-14292	100	22	6	6	NUM
fpq-14292	100	23	in	in	ADP
fpq-14292	100	24	other	other	ADJ
fpq-14292	100	25	words	word	NOUN
fpq-14292	100	26	,	,	PUNCT
fpq-14292	100	27	descriptively	descriptively	ADV
fpq-14292	100	28	accurate	accurate	ADJ
fpq-14292	100	29	information	information	NOUN
fpq-14292	100	30	about	about	ADP
fpq-14292	100	31	statistical	statistical	ADJ
fpq-14292	100	32	regularities	regularity	NOUN
fpq-14292	100	33	can	can	AUX
fpq-14292	100	34	support	support	VERB
fpq-14292	100	35	beliefs	belief	NOUN
fpq-14292	100	36	that	that	PRON
fpq-14292	100	37	are	be	AUX
fpq-14292	100	38	morally	morally	ADV
fpq-14292	100	39	problematic	problematic	ADJ
fpq-14292	100	40	or	or	CCONJ
fpq-14292	100	41	normatively	normatively	ADV
fpq-14292	100	42	incorrect	incorrect	ADJ
fpq-14292	100	43	,	,	PUNCT
fpq-14292	100	44	in	in	ADP
fpq-14292	100	45	which	which	DET
fpq-14292	100	46	case	case	NOUN
fpq-14292	100	47	we	we	PRON
fpq-14292	100	48	face	face	VERB
fpq-14292	100	49	a	a	DET
fpq-14292	100	50	choice	choice	NOUN
fpq-14292	100	51	about	about	ADP
fpq-14292	100	52	whether	whether	SCONJ
fpq-14292	100	53	to	to	PART
fpq-14292	100	54	prioritize	prioritize	VERB
fpq-14292	100	55	the	the	DET
fpq-14292	100	56	descriptive	descriptive	ADJ
fpq-14292	100	57	or	or	CCONJ
fpq-14292	100	58	instead	instead	ADV
fpq-14292	100	59	the	the	DET
fpq-14292	100	60	normative	normative	NOUN
fpq-14292	100	61	.	.	PUNCT
fpq-14292	101	1	likewise	likewise	ADV
fpq-14292	101	2	,	,	PUNCT
fpq-14292	101	3	we	we	PRON
fpq-14292	101	4	think	think	VERB
fpq-14292	101	5	that	that	SCONJ
fpq-14292	101	6	there	there	PRON
fpq-14292	101	7	are	be	VERB
fpq-14292	101	8	two	two	NUM
fpq-14292	101	9	general	general	ADJ
fpq-14292	101	10	ways	way	NOUN
fpq-14292	101	11	in	in	ADP
fpq-14292	101	12	which	which	PRON
fpq-14292	101	13	ml	ml	ADP
fpq-14292	101	14	systems	system	NOUN
fpq-14292	101	15	or	or	CCONJ
fpq-14292	101	16	humans	human	NOUN
fpq-14292	101	17	can	can	AUX
fpq-14292	101	18	be	be	AUX
fpq-14292	101	19	right	right	ADJ
fpq-14292	101	20	or	or	CCONJ
fpq-14292	101	21	wrong	wrong	ADJ
fpq-14292	101	22	—	—	PUNCT
fpq-14292	101	23	descriptively	descriptively	ADV
fpq-14292	101	24	and	and	CCONJ
fpq-14292	101	25	normatively	normatively	ADV
fpq-14292	101	26	—	—	PUNCT
fpq-14292	101	27	and	and	CCONJ
fpq-14292	101	28	that	that	SCONJ
fpq-14292	101	29	a	a	DET
fpq-14292	101	30	dilemma	dilemma	NOUN
fpq-14292	101	31	similar	similar	ADJ
fpq-14292	101	32	to	to	ADP
fpq-14292	101	33	the	the	DET
fpq-14292	101	34	one	one	NUM
fpq-14292	101	35	basu	basu	PROPN
fpq-14292	101	36	describes	describe	VERB
fpq-14292	101	37	can	can	AUX
fpq-14292	101	38	arise	arise	VERB
fpq-14292	101	39	even	even	ADV
fpq-14292	101	40	for	for	ADP
fpq-14292	101	41	ml	ml	NOUN
fpq-14292	101	42	systems	system	NOUN
fpq-14292	101	43	.	.	PUNCT
fpq-14292	102	1	to	to	PART
fpq-14292	102	2	explain	explain	VERB
fpq-14292	102	3	,	,	PUNCT
fpq-14292	102	4	a	a	DET
fpq-14292	102	5	visual	visual	ADJ
fpq-14292	102	6	representation	representation	NOUN
fpq-14292	102	7	of	of	ADP
fpq-14292	102	8	a	a	DET
fpq-14292	102	9	scene	scene	NOUN
fpq-14292	102	10	gets	get	VERB
fpq-14292	102	11	things	thing	NOUN
fpq-14292	102	12	descriptively	descriptively	ADV
fpq-14292	102	13	right	right	ADJ
fpq-14292	102	14	to	to	ADP
fpq-14292	102	15	the	the	DET
fpq-14292	102	16	extent	extent	NOUN
fpq-14292	102	17	that	that	SCONJ
fpq-14292	102	18	it	it	PRON
fpq-14292	102	19	accurately	accurately	ADV
fpq-14292	102	20	depicts	depict	VERB
fpq-14292	102	21	visual	visual	ADJ
fpq-14292	102	22	information	information	NOUN
fpq-14292	102	23	about	about	ADP
fpq-14292	102	24	a	a	DET
fpq-14292	102	25	scene	scene	NOUN
fpq-14292	102	26	as	as	SCONJ
fpq-14292	102	27	presented	present	VERB
fpq-14292	102	28	and	and	CCONJ
fpq-14292	102	29	wrong	wrong	ADJ
fpq-14292	102	30	to	to	ADP
fpq-14292	102	31	the	the	DET
fpq-14292	102	32	extent	extent	NOUN
fpq-14292	102	33	that	that	SCONJ
fpq-14292	102	34	it	it	PRON
fpq-14292	102	35	does	do	VERB
fpq-14292	102	36	not	not	PART
fpq-14292	102	37	.	.	PUNCT
fpq-14292	103	1	for	for	ADP
fpq-14292	103	2	example	example	NOUN
fpq-14292	103	3	,	,	PUNCT
fpq-14292	103	4	if	if	SCONJ
fpq-14292	103	5	you	you	PRON
fpq-14292	103	6	photograph	photograph	VERB
fpq-14292	103	7	me	i	PRON
fpq-14292	103	8	robbing	rob	VERB
fpq-14292	103	9	a	a	DET
fpq-14292	103	10	store	store	NOUN
fpq-14292	103	11	,	,	PUNCT
fpq-14292	103	12	your	your	PRON
fpq-14292	103	13	camera	camera	NOUN
fpq-14292	103	14	gets	get	VERB
fpq-14292	103	15	things	thing	NOUN
fpq-14292	103	16	descriptively	descriptively	ADV
fpq-14292	103	17	right	right	ADV
fpq-14292	103	18	insofar	insofar	ADV
fpq-14292	103	19	as	as	SCONJ
fpq-14292	103	20	it	it	PRON
fpq-14292	103	21	accurately	accurately	ADV
fpq-14292	103	22	depicts	depict	VERB
fpq-14292	103	23	my	my	PRON
fpq-14292	103	24	face	face	NOUN
fpq-14292	103	25	and	and	CCONJ
fpq-14292	103	26	wrong	wrong	NOUN
fpq-14292	103	27	insofar	insofar	ADV
fpq-14292	103	28	as	as	SCONJ
fpq-14292	103	29	it	it	PRON
fpq-14292	103	30	does	do	AUX
fpq-14292	103	31	not	not	PART
fpq-14292	103	32	—	—	PUNCT
fpq-14292	103	33	for	for	ADP
fpq-14292	103	34	example	example	NOUN
fpq-14292	103	35	,	,	PUNCT
fpq-14292	103	36	because	because	SCONJ
fpq-14292	103	37	its	its	PRON
fpq-14292	103	38	exposure	exposure	NOUN
fpq-14292	103	39	settings	setting	NOUN
fpq-14292	103	40	are	be	AUX
fpq-14292	103	41	off	off	ADP
fpq-14292	103	42	.	.	PUNCT
fpq-14292	104	1	most	most	ADJ
fpq-14292	104	2	tasks	task	NOUN
fpq-14292	104	3	a	a	DET
fpq-14292	104	4	camera	camera	NOUN
fpq-14292	104	5	is	be	AUX
fpq-14292	104	6	designed	design	VERB
fpq-14292	104	7	to	to	PART
fpq-14292	104	8	perform	perform	VERB
fpq-14292	104	9	require	require	NOUN
fpq-14292	104	10	for	for	ADP
fpq-14292	104	11	success	success	NOUN
fpq-14292	104	12	that	that	PRON
fpq-14292	104	13	at	at	ADV
fpq-14292	104	14	least	least	ADJ
fpq-14292	104	15	some	some	PRON
fpq-14292	104	16	threshold	threshold	NOUN
fpq-14292	104	17	of	of	ADP
fpq-14292	104	18	descriptive	descriptive	ADJ
fpq-14292	104	19	accuracy	accuracy	NOUN
fpq-14292	104	20	(	(	PUNCT
fpq-14292	104	21	in	in	ADP
fpq-14292	104	22	our	our	PRON
fpq-14292	104	23	sense	sense	NOUN
fpq-14292	104	24	)	)	PUNCT
fpq-14292	104	25	be	be	AUX
fpq-14292	104	26	achieved	achieve	VERB
fpq-14292	104	27	in	in	ADP
fpq-14292	104	28	the	the	DET
fpq-14292	104	29	camera	camera	NOUN
fpq-14292	104	30	’s	’s	PART
fpq-14292	104	31	outputs	output	NOUN
fpq-14292	104	32	.	.	PUNCT
fpq-14292	105	1	likewise	likewise	ADV
fpq-14292	105	2	,	,	PUNCT
fpq-14292	105	3	an	an	DET
fpq-14292	105	4	audio	audio	ADJ
fpq-14292	105	5	recording	recording	NOUN
fpq-14292	105	6	of	of	ADP
fpq-14292	105	7	my	my	PRON
fpq-14292	105	8	boss	boss	NOUN
fpq-14292	105	9	making	make	VERB
fpq-14292	105	10	sexist	sexist	ADJ
fpq-14292	105	11	or	or	CCONJ
fpq-14292	105	12	racist	racist	ADJ
fpq-14292	105	13	remarks	remark	NOUN
fpq-14292	105	14	is	be	AUX
fpq-14292	105	15	descriptively	descriptively	ADV
fpq-14292	105	16	accurate	accurate	ADJ
fpq-14292	105	17	to	to	ADP
fpq-14292	105	18	the	the	DET
fpq-14292	105	19	extent	extent	NOUN
fpq-14292	105	20	that	that	SCONJ
fpq-14292	105	21	it	it	PRON
fpq-14292	105	22	records	record	VERB
fpq-14292	105	23	what	what	PRON
fpq-14292	105	24	he	he	PRON
fpq-14292	105	25	said	say	VERB
fpq-14292	105	26	and	and	CCONJ
fpq-14292	105	27	descriptively	descriptively	ADV
fpq-14292	105	28	wrong	wrong	ADJ
fpq-14292	105	29	to	to	ADP
fpq-14292	105	30	the	the	DET
fpq-14292	105	31	extent	extent	NOUN
fpq-14292	105	32	that	that	SCONJ
fpq-14292	105	33	it	it	PRON
fpq-14292	105	34	does	do	VERB
fpq-14292	105	35	not	not	PART
fpq-14292	105	36	,	,	PUNCT
fpq-14292	105	37	perhaps	perhaps	ADV
fpq-14292	105	38	due	due	ADP
fpq-14292	105	39	to	to	ADP
fpq-14292	105	40	electromagnetic	electromagnetic	ADJ
fpq-14292	105	41	interference	interference	NOUN
fpq-14292	105	42	from	from	ADP
fpq-14292	105	43	another	another	DET
fpq-14292	105	44	device	device	NOUN
fpq-14292	105	45	.	.	PUNCT
fpq-14292	106	1	by	by	ADP
fpq-14292	106	2	contrast	contrast	NOUN
fpq-14292	106	3	,	,	PUNCT
fpq-14292	106	4	we	we	PRON
fpq-14292	106	5	expect	expect	VERB
fpq-14292	106	6	people	people	NOUN
fpq-14292	106	7	—	—	PUNCT
fpq-14292	106	8	as	as	ADP
fpq-14292	106	9	moral	moral	ADJ
fpq-14292	106	10	agents	agent	NOUN
fpq-14292	106	11	—	—	PUNCT
fpq-14292	106	12	to	to	PART
fpq-14292	106	13	get	get	VERB
fpq-14292	106	14	things	thing	NOUN
fpq-14292	106	15	normatively	normatively	ADV
fpq-14292	106	16	right	right	ADV
fpq-14292	106	17	in	in	ADP
fpq-14292	106	18	a	a	DET
fpq-14292	106	19	moral	moral	ADJ
fpq-14292	106	20	sense	sense	NOUN
fpq-14292	106	21	.	.	PUNCT
fpq-14292	107	1	it	it	PRON
fpq-14292	107	2	is	be	AUX
fpq-14292	107	3	normatively	normatively	ADV
fpq-14292	107	4	correct	correct	ADJ
fpq-14292	107	5	for	for	SCONJ
fpq-14292	107	6	me	i	PRON
fpq-14292	107	7	to	to	PART
fpq-14292	107	8	pay	pay	VERB
fpq-14292	107	9	for	for	ADP
fpq-14292	107	10	my	my	PRON
fpq-14292	107	11	groceries	grocery	NOUN
fpq-14292	107	12	instead	instead	ADV
fpq-14292	107	13	of	of	ADP
fpq-14292	107	14	stealing	steal	VERB
fpq-14292	107	15	them	they	PRON
fpq-14292	107	16	(	(	PUNCT
fpq-14292	107	17	all	all	DET
fpq-14292	107	18	else	else	ADV
fpq-14292	107	19	being	be	AUX
fpq-14292	107	20	equal	equal	ADJ
fpq-14292	107	21	)	)	PUNCT
fpq-14292	107	22	.	.	PUNCT
fpq-14292	108	1	when	when	SCONJ
fpq-14292	108	2	i	i	PRON
fpq-14292	108	3	rob	rob	VERB
fpq-14292	108	4	a	a	DET
fpq-14292	108	5	store	store	NOUN
fpq-14292	108	6	,	,	PUNCT
fpq-14292	108	7	by	by	ADP
fpq-14292	108	8	contrast	contrast	NOUN
fpq-14292	108	9	,	,	PUNCT
fpq-14292	108	10	i	i	PRON
fpq-14292	108	11	get	get	VERB
fpq-14292	108	12	things	thing	NOUN
fpq-14292	108	13	normatively	normatively	ADV
fpq-14292	108	14	incorrect	incorrect	ADJ
fpq-14292	108	15	by	by	ADP
fpq-14292	108	16	failing	fail	VERB
fpq-14292	108	17	to	to	PART
fpq-14292	108	18	do	do	VERB
fpq-14292	108	19	what	what	PRON
fpq-14292	108	20	i	i	PRON
fpq-14292	108	21	ought	ought	AUX
fpq-14292	108	22	to	to	PART
fpq-14292	108	23	do	do	VERB
fpq-14292	108	24	(	(	PUNCT
fpq-14292	108	25	pay	pay	VERB
fpq-14292	108	26	for	for	ADP
fpq-14292	108	27	my	my	PRON
fpq-14292	108	28	groceries	grocery	NOUN
fpq-14292	108	29	)	)	PUNCT
fpq-14292	108	30	and	and	CCONJ
fpq-14292	108	31	by	by	ADP
fpq-14292	108	32	doing	do	VERB
fpq-14292	108	33	what	what	PRON
fpq-14292	108	34	i	i	PRON
fpq-14292	108	35	should	should	AUX
fpq-14292	108	36	not	not	PART
fpq-14292	108	37	(	(	PUNCT
fpq-14292	108	38	stealing	steal	VERB
fpq-14292	108	39	them	they	PRON
fpq-14292	108	40	)	)	PUNCT
fpq-14292	108	41	.	.	PUNCT
fpq-14292	109	1	similarly	similarly	ADV
fpq-14292	109	2	,	,	PUNCT
fpq-14292	109	3	my	my	PRON
fpq-14292	109	4	boss	boss	NOUN
fpq-14292	109	5	behaves	behave	VERB
fpq-14292	109	6	normatively	normatively	ADV
fpq-14292	109	7	correctly	correctly	ADV
fpq-14292	109	8	by	by	ADP
fpq-14292	109	9	not	not	PART
fpq-14292	109	10	making	make	VERB
fpq-14292	109	11	sexist	sexist	ADJ
fpq-14292	109	12	or	or	CCONJ
fpq-14292	109	13	racist	racist	ADJ
fpq-14292	109	14	remarks	remark	NOUN
fpq-14292	109	15	.	.	PUNCT
fpq-14292	110	1	we	we	PRON
fpq-14292	110	2	do	do	AUX
fpq-14292	110	3	not	not	PART
fpq-14292	110	4	expect	expect	VERB
fpq-14292	110	5	systems	system	NOUN
fpq-14292	110	6	like	like	ADP
fpq-14292	110	7	cameras	camera	NOUN
fpq-14292	110	8	or	or	CCONJ
fpq-14292	110	9	audio	audio	NOUN
fpq-14292	110	10	devices	device	NOUN
fpq-14292	110	11	to	to	PART
fpq-14292	110	12	get	get	VERB
fpq-14292	110	13	things	thing	NOUN
fpq-14292	110	14	normatively	normatively	ADV
fpq-14292	110	15	right	right	ADJ
fpq-14292	110	16	.	.	PUNCT
fpq-14292	111	1	consider	consider	VERB
fpq-14292	111	2	an	an	DET
fpq-14292	111	3	imaginary	imaginary	ADJ
fpq-14292	111	4	camera	camera	NOUN
fpq-14292	111	5	,	,	PUNCT
fpq-14292	111	6	the	the	DET
fpq-14292	111	7	iblur	iblur	NOUN
fpq-14292	111	8	,	,	PUNCT
fpq-14292	111	9	which	which	PRON
fpq-14292	111	10	blurs	blur	VERB
fpq-14292	111	11	out	out	ADP
fpq-14292	111	12	the	the	DET
fpq-14292	111	13	morally	morally	ADV
fpq-14292	111	14	objectionable	objectionable	ADJ
fpq-14292	111	15	parts	part	NOUN
fpq-14292	111	16	of	of	ADP
fpq-14292	111	17	a	a	DET
fpq-14292	111	18	visual	visual	ADJ
fpq-14292	111	19	scene.12	scene.12	NOUN
fpq-14292	111	20	as	as	ADP
fpq-14292	111	21	a	a	DET
fpq-14292	111	22	result	result	NOUN
fpq-14292	111	23	,	,	PUNCT
fpq-14292	111	24	the	the	DET
fpq-14292	111	25	iblur	iblur	NOUN
fpq-14292	111	26	blurs	blur	VERB
fpq-14292	111	27	out	out	ADP
fpq-14292	111	28	the	the	DET
fpq-14292	111	29	image	image	NOUN
fpq-14292	111	30	of	of	ADP
fpq-14292	111	31	me	i	PRON
fpq-14292	111	32	robbing	rob	VERB
fpq-14292	111	33	the	the	DET
fpq-14292	111	34	store	store	NOUN
fpq-14292	111	35	,	,	PUNCT
fpq-14292	111	36	since	since	SCONJ
fpq-14292	111	37	robbing	rob	VERB
fpq-14292	111	38	a	a	DET
fpq-14292	111	39	store	store	NOUN
fpq-14292	111	40	is	be	AUX
fpq-14292	111	41	a	a	DET
fpq-14292	111	42	morally	morally	ADV
fpq-14292	111	43	objectionable	objectionable	ADJ
fpq-14292	111	44	action	action	NOUN
fpq-14292	111	45	(	(	PUNCT
fpq-14292	111	46	all	all	PRON
fpq-14292	111	47	else	else	ADV
fpq-14292	111	48	being	be	AUX
fpq-14292	111	49	equal	equal	ADJ
fpq-14292	111	50	)	)	PUNCT
fpq-14292	111	51	.	.	PUNCT
fpq-14292	112	1	not	not	PART
fpq-14292	112	2	only	only	ADV
fpq-14292	112	3	do	do	AUX
fpq-14292	112	4	we	we	PRON
fpq-14292	112	5	not	not	PART
fpq-14292	112	6	expect	expect	VERB
fpq-14292	112	7	a	a	DET
fpq-14292	112	8	camera	camera	NOUN
fpq-14292	112	9	to	to	PART
fpq-14292	112	10	work	work	VERB
fpq-14292	112	11	like	like	ADP
fpq-14292	112	12	this	this	PRON
fpq-14292	112	13	but	but	CCONJ
fpq-14292	112	14	we	we	PRON
fpq-14292	112	15	expect	expect	VERB
fpq-14292	112	16	it	it	PRON
fpq-14292	112	17	not	not	PART
fpq-14292	112	18	to	to	PART
fpq-14292	112	19	work	work	VERB
fpq-14292	112	20	this	this	DET
fpq-14292	112	21	way	way	NOUN
fpq-14292	112	22	,	,	PUNCT
fpq-14292	112	23	since	since	SCONJ
fpq-14292	112	24	such	such	DET
fpq-14292	112	25	an	an	DET
fpq-14292	112	26	imposition	imposition	NOUN
fpq-14292	112	27	of	of	ADP
fpq-14292	112	28	normative	normative	ADJ
fpq-14292	112	29	correctness	correctness	NOUN
fpq-14292	112	30	undermines	undermine	VERB
fpq-14292	112	31	descriptive	descriptive	ADJ
fpq-14292	112	32	accuracy	accuracy	NOUN
fpq-14292	112	33	,	,	PUNCT
fpq-14292	112	34	which	which	PRON
fpq-14292	112	35	the	the	DET
fpq-14292	112	36	functions	function	NOUN
fpq-14292	112	37	of	of	ADP
fpq-14292	112	38	a	a	DET
fpq-14292	112	39	camera	camera	NOUN
fpq-14292	112	40	require.13	require.13	NOUN
fpq-14292	112	41	additionally	additionally	ADV
fpq-14292	112	42	,	,	PUNCT
fpq-14292	112	43	the	the	DET
fpq-14292	112	44	iblur	iblur	NOUN
fpq-14292	112	45	would	would	AUX
fpq-14292	112	46	fail	fail	VERB
fpq-14292	112	47	to	to	PART
fpq-14292	112	48	provide	provide	VERB
fpq-14292	112	49	us	we	PRON
fpq-14292	112	50	with	with	ADP
fpq-14292	112	51	ethically	ethically	ADV
fpq-14292	112	52	useful	useful	ADJ
fpq-14292	112	53	information	information	NOUN
fpq-14292	112	54	,	,	PUNCT
fpq-14292	112	55	since	since	SCONJ
fpq-14292	112	56	12	12	NUM
fpq-14292	112	57	we	we	PRON
fpq-14292	112	58	model	model	VERB
fpq-14292	112	59	our	our	PRON
fpq-14292	112	60	iblur	iblur	NOUN
fpq-14292	112	61	after	after	ADP
fpq-14292	112	62	the	the	DET
fpq-14292	112	63	arkangel	arkangel	NOUN
fpq-14292	112	64	system	system	NOUN
fpq-14292	112	65	in	in	ADP
fpq-14292	112	66	the	the	DET
fpq-14292	112	67	tv	tv	NOUN
fpq-14292	112	68	show	show	NOUN
fpq-14292	112	69	black	black	ADJ
fpq-14292	112	70	mirror	mirror	NOUN
fpq-14292	112	71	.	.	PROPN
fpq-14292	113	1	13	13	NUM
fpq-14292	113	2	cameras	camera	NOUN
fpq-14292	113	3	serve	serve	VERB
fpq-14292	113	4	various	various	ADJ
fpq-14292	113	5	functions	function	NOUN
fpq-14292	113	6	,	,	PUNCT
fpq-14292	113	7	including	include	VERB
fpq-14292	113	8	the	the	DET
fpq-14292	113	9	attainment	attainment	NOUN
fpq-14292	113	10	of	of	ADP
fpq-14292	113	11	artistic	artistic	ADJ
fpq-14292	113	12	or	or	CCONJ
fpq-14292	113	13	imaginative	imaginative	ADJ
fpq-14292	113	14	goals	goal	NOUN
fpq-14292	113	15	.	.	PUNCT
fpq-14292	114	1	likewise	likewise	ADV
fpq-14292	114	2	,	,	PUNCT
fpq-14292	114	3	security	security	NOUN
fpq-14292	114	4	cameras	camera	NOUN
fpq-14292	114	5	might	might	AUX
fpq-14292	114	6	be	be	AUX
fpq-14292	114	7	used	use	VERB
fpq-14292	114	8	to	to	PART
fpq-14292	114	9	intimidate	intimidate	VERB
fpq-14292	114	10	people	people	NOUN
fpq-14292	114	11	or	or	CCONJ
fpq-14292	114	12	to	to	PART
fpq-14292	114	13	shed	shed	VERB
fpq-14292	114	14	light	light	NOUN
fpq-14292	114	15	on	on	ADP
fpq-14292	114	16	the	the	DET
fpq-14292	114	17	activities	activity	NOUN
fpq-14292	114	18	of	of	ADP
fpq-14292	114	19	select	select	ADJ
fpq-14292	114	20	people	people	NOUN
fpq-14292	114	21	in	in	ADP
fpq-14292	114	22	select	select	ADJ
fpq-14292	114	23	places	place	NOUN
fpq-14292	114	24	in	in	ADP
fpq-14292	114	25	a	a	DET
fpq-14292	114	26	way	way	NOUN
fpq-14292	114	27	that	that	PRON
fpq-14292	114	28	is	be	AUX
fpq-14292	114	29	morally	morally	ADV
fpq-14292	114	30	problematic	problematic	ADJ
fpq-14292	114	31	.	.	PUNCT
fpq-14292	115	1	yet	yet	CCONJ
fpq-14292	115	2	the	the	DET
fpq-14292	115	3	tasks	task	NOUN
fpq-14292	115	4	cameras	camera	NOUN
fpq-14292	115	5	perform	perform	VERB
fpq-14292	115	6	still	still	ADV
fpq-14292	115	7	require	require	VERB
fpq-14292	115	8	that	that	SCONJ
fpq-14292	115	9	cameras	camera	NOUN
fpq-14292	115	10	accurately	accurately	ADV
fpq-14292	115	11	depict	depict	VERB
fpq-14292	115	12	what	what	PRON
fpq-14292	115	13	is	be	AUX
fpq-14292	115	14	in	in	ADP
fpq-14292	115	15	front	front	NOUN
fpq-14292	115	16	of	of	ADP
fpq-14292	115	17	them	they	PRON
fpq-14292	115	18	,	,	PUNCT
fpq-14292	115	19	and	and	CCONJ
fpq-14292	115	20	artistic	artistic	ADJ
fpq-14292	115	21	or	or	CCONJ
fpq-14292	115	22	nefarious	nefarious	ADJ
fpq-14292	115	23	goals	goal	NOUN
fpq-14292	115	24	are	be	AUX
fpq-14292	115	25	served	serve	VERB
fpq-14292	115	26	via	via	ADP
fpq-14292	115	27	such	such	ADJ
fpq-14292	115	28	accuracy	accuracy	NOUN
fpq-14292	115	29	.	.	PUNCT
fpq-14292	116	1	deery	deery	NOUN
fpq-14292	116	2	and	and	CCONJ
fpq-14292	116	3	bailey	bailey	PROPN
fpq-14292	116	4	–	–	PUNCT
fpq-14292	116	5	the	the	DET
fpq-14292	116	6	bias	bias	NOUN
fpq-14292	116	7	dilemma	dilemma	NOUN
fpq-14292	116	8	:	:	PUNCT
fpq-14292	116	9	the	the	DET
fpq-14292	116	10	ethics	ethic	NOUN
fpq-14292	116	11	of	of	ADP
fpq-14292	116	12	algorithmic	algorithmic	ADJ
fpq-14292	116	13	bias	bias	NOUN
fpq-14292	116	14	in	in	ADP
fpq-14292	116	15	natural	natural	ADJ
fpq-14292	116	16	-	-	PUNCT
fpq-14292	116	17	language	language	NOUN
fpq-14292	116	18	processing	processing	NOUN
fpq-14292	116	19	published	publish	VERB
fpq-14292	116	20	by	by	ADP
fpq-14292	116	21	scholarship@western	scholarship@western	PROPN
fpq-14292	116	22	,	,	PUNCT
fpq-14292	116	23	2022	2022	NUM
fpq-14292	116	24	7	7	NUM
fpq-14292	116	25	the	the	DET
fpq-14292	116	26	authorities	authority	NOUN
fpq-14292	116	27	will	will	AUX
fpq-14292	116	28	need	need	VERB
fpq-14292	116	29	an	an	DET
fpq-14292	116	30	accurate	accurate	ADJ
fpq-14292	116	31	depiction	depiction	NOUN
fpq-14292	116	32	of	of	ADP
fpq-14292	116	33	my	my	PRON
fpq-14292	116	34	face	face	NOUN
fpq-14292	116	35	to	to	PART
fpq-14292	116	36	hold	hold	VERB
fpq-14292	116	37	me	i	PRON
fpq-14292	116	38	accountable.14	accountable.14	PROPN
fpq-14292	116	39	and	and	CCONJ
fpq-14292	116	40	this	this	PRON
fpq-14292	116	41	is	be	AUX
fpq-14292	116	42	likewise	likewise	ADV
fpq-14292	116	43	the	the	DET
fpq-14292	116	44	case	case	NOUN
fpq-14292	116	45	for	for	ADP
fpq-14292	116	46	an	an	DET
fpq-14292	116	47	audio	audio	ADJ
fpq-14292	116	48	recording	recording	NOUN
fpq-14292	116	49	of	of	ADP
fpq-14292	116	50	my	my	PRON
fpq-14292	116	51	boss	boss	NOUN
fpq-14292	116	52	making	make	VERB
fpq-14292	116	53	sexist	sexist	ADJ
fpq-14292	116	54	or	or	CCONJ
fpq-14292	116	55	racist	racist	ADJ
fpq-14292	116	56	remarks	remark	NOUN
fpq-14292	116	57	.	.	PUNCT
fpq-14292	117	1	regarding	regard	VERB
fpq-14292	117	2	descriptive	descriptive	ADJ
fpq-14292	117	3	accuracy	accuracy	NOUN
fpq-14292	117	4	,	,	PUNCT
fpq-14292	117	5	one	one	PRON
fpq-14292	117	6	might	might	AUX
fpq-14292	117	7	worry	worry	VERB
fpq-14292	117	8	whether	whether	SCONJ
fpq-14292	117	9	devices	device	NOUN
fpq-14292	117	10	such	such	ADJ
fpq-14292	117	11	as	as	ADP
fpq-14292	117	12	cameras	camera	NOUN
fpq-14292	117	13	really	really	ADV
fpq-14292	117	14	do	do	AUX
fpq-14292	117	15	get	get	VERB
fpq-14292	117	16	things	thing	NOUN
fpq-14292	117	17	descriptively	descriptively	ADV
fpq-14292	117	18	right	right	ADJ
fpq-14292	117	19	(	(	PUNCT
fpq-14292	117	20	even	even	ADV
fpq-14292	117	21	if	if	SCONJ
fpq-14292	117	22	we	we	PRON
fpq-14292	117	23	grant	grant	VERB
fpq-14292	117	24	that	that	SCONJ
fpq-14292	117	25	there	there	PRON
fpq-14292	117	26	is	be	VERB
fpq-14292	117	27	less	less	ADJ
fpq-14292	117	28	of	of	ADP
fpq-14292	117	29	a	a	DET
fpq-14292	117	30	worry	worry	NOUN
fpq-14292	117	31	in	in	ADP
fpq-14292	117	32	this	this	DET
fpq-14292	117	33	regard	regard	NOUN
fpq-14292	117	34	for	for	ADP
fpq-14292	117	35	audio	audio	ADJ
fpq-14292	117	36	-	-	PUNCT
fpq-14292	117	37	recording	record	VERB
fpq-14292	117	38	devices	device	NOUN
fpq-14292	117	39	)	)	PUNCT
fpq-14292	117	40	.	.	PUNCT
fpq-14292	118	1	for	for	ADP
fpq-14292	118	2	instance	instance	NOUN
fpq-14292	118	3	,	,	PUNCT
fpq-14292	118	4	shen	shen	PROPN
fpq-14292	118	5	-	-	PUNCT
fpq-14292	118	6	yi	yi	PROPN
fpq-14292	118	7	liao	liao	PROPN
fpq-14292	118	8	and	and	CCONJ
fpq-14292	118	9	bryce	bryce	PROPN
fpq-14292	118	10	huebner	huebner	PROPN
fpq-14292	118	11	(	(	PUNCT
fpq-14292	118	12	2021	2021	NUM
fpq-14292	118	13	,	,	PUNCT
fpq-14292	118	14	94	94	NUM
fpq-14292	118	15	)	)	PUNCT
fpq-14292	118	16	have	have	AUX
fpq-14292	118	17	argued	argue	VERB
fpq-14292	118	18	that	that	SCONJ
fpq-14292	118	19	cameras	camera	NOUN
fpq-14292	118	20	have	have	VERB
fpq-14292	118	21	a	a	DET
fpq-14292	118	22	“	"	PUNCT
fpq-14292	118	23	light	light	ADJ
fpq-14292	118	24	-	-	PUNCT
fpq-14292	118	25	skin	skin	NOUN
fpq-14292	118	26	bias	bias	NOUN
fpq-14292	118	27	.	.	PUNCT
fpq-14292	118	28	”	"	PUNCT
fpq-14292	119	1	the	the	DET
fpq-14292	119	2	early	early	ADJ
fpq-14292	119	3	development	development	NOUN
fpq-14292	119	4	of	of	ADP
fpq-14292	119	5	kodak	kodak	PROPN
fpq-14292	119	6	film	film	NOUN
fpq-14292	119	7	printing	printing	NOUN
fpq-14292	119	8	required	require	VERB
fpq-14292	119	9	that	that	SCONJ
fpq-14292	119	10	skin	skin	NOUN
fpq-14292	119	11	tones	tone	NOUN
fpq-14292	119	12	be	be	AUX
fpq-14292	119	13	matched	match	VERB
fpq-14292	119	14	to	to	ADP
fpq-14292	119	15	an	an	DET
fpq-14292	119	16	image	image	NOUN
fpq-14292	119	17	of	of	ADP
fpq-14292	119	18	a	a	DET
fpq-14292	119	19	white	white	ADJ
fpq-14292	119	20	model	model	NOUN
fpq-14292	119	21	,	,	PUNCT
fpq-14292	119	22	with	with	ADP
fpq-14292	119	23	the	the	DET
fpq-14292	119	24	result	result	NOUN
fpq-14292	119	25	that	that	SCONJ
fpq-14292	119	26	“	"	PUNCT
fpq-14292	119	27	darker	dark	ADJ
fpq-14292	119	28	skin	skin	NOUN
fpq-14292	119	29	tones	tone	NOUN
fpq-14292	119	30	.	.	PUNCT
fpq-14292	119	31	.	.	PUNCT
fpq-14292	119	32	.	.	PUNCT
fpq-14292	120	1	[	[	X
fpq-14292	120	2	were	be	AUX
fpq-14292	120	3	]	]	PUNCT
fpq-14292	120	4	.	.	PUNCT
fpq-14292	120	5	.	.	PUNCT
fpq-14292	120	6	.	.	PUNCT
fpq-14292	121	1	over	over	ADP
fpq-14292	121	2	saturated	saturate	VERB
fpq-14292	121	3	,	,	PUNCT
fpq-14292	121	4	or	or	CCONJ
fpq-14292	121	5	under	under	ADV
fpq-14292	121	6	-	-	PUNCT
fpq-14292	121	7	lit	light	VERB
fpq-14292	121	8	,	,	PUNCT
fpq-14292	121	9	so	so	CCONJ
fpq-14292	121	10	the	the	DET
fpq-14292	121	11	only	only	ADJ
fpq-14292	121	12	images	image	NOUN
fpq-14292	121	13	that	that	PRON
fpq-14292	121	14	.	.	PUNCT
fpq-14292	121	15	.	.	PUNCT
fpq-14292	121	16	.	.	PUNCT
fpq-14292	122	1	looked	look	VERB
fpq-14292	122	2	‘	'	PUNCT
fpq-14292	122	3	right	right	ADJ
fpq-14292	122	4	’	'	PUNCT
fpq-14292	122	5	were	be	AUX
fpq-14292	122	6	images	image	NOUN
fpq-14292	122	7	of	of	ADP
fpq-14292	122	8	light	light	ADJ
fpq-14292	122	9	-	-	PUNCT
fpq-14292	122	10	skinned	skinned	ADJ
fpq-14292	122	11	people	people	NOUN
fpq-14292	122	12	”	"	PUNCT
fpq-14292	122	13	(	(	PUNCT
fpq-14292	122	14	95	95	NUM
fpq-14292	122	15	)	)	PUNCT
fpq-14292	122	16	.	.	PUNCT
fpq-14292	123	1	moreover	moreover	ADV
fpq-14292	123	2	,	,	PUNCT
fpq-14292	123	3	“	"	PUNCT
fpq-14292	123	4	light	light	ADJ
fpq-14292	123	5	-	-	PUNCT
fpq-14292	123	6	skin	skin	NOUN
fpq-14292	123	7	bias	bias	NOUN
fpq-14292	123	8	.	.	PUNCT
fpq-14292	123	9	.	.	PUNCT
fpq-14292	124	1	.	.	PUNCT
fpq-14292	125	1	is	be	AUX
fpq-14292	125	2	not	not	PART
fpq-14292	125	3	primarily	primarily	ADV
fpq-14292	125	4	a	a	DET
fpq-14292	125	5	technical	technical	ADJ
fpq-14292	125	6	issue	issue	NOUN
fpq-14292	125	7	.	.	PUNCT
fpq-14292	125	8	.	.	PUNCT
fpq-14292	125	9	.	.	PUNCT
fpq-14292	125	10	.	.	PUNCT
fpq-14292	126	1	film	film	NOUN
fpq-14292	126	2	emulsions	emulsion	NOUN
fpq-14292	126	3	could	could	AUX
fpq-14292	126	4	have	have	AUX
fpq-14292	126	5	been	be	AUX
fpq-14292	126	6	designed	design	VERB
fpq-14292	126	7	that	that	PRON
fpq-14292	126	8	were	be	AUX
fpq-14292	126	9	more	more	ADV
fpq-14292	126	10	sensitive	sensitive	ADJ
fpq-14292	126	11	to	to	ADP
fpq-14292	126	12	a	a	DET
fpq-14292	126	13	wider	wide	ADJ
fpq-14292	126	14	range	range	NOUN
fpq-14292	126	15	of	of	ADP
fpq-14292	126	16	skin	skin	NOUN
fpq-14292	126	17	tones	tone	NOUN
fpq-14292	126	18	,	,	PUNCT
fpq-14292	126	19	”	"	PUNCT
fpq-14292	126	20	(	(	PUNCT
fpq-14292	126	21	95	95	NUM
fpq-14292	126	22	)	)	PUNCT
fpq-14292	126	23	but	but	CCONJ
fpq-14292	126	24	they	they	PRON
fpq-14292	126	25	were	be	AUX
fpq-14292	126	26	not	not	PART
fpq-14292	126	27	.	.	PUNCT
fpq-14292	127	1	why	why	SCONJ
fpq-14292	127	2	?	?	PUNCT
fpq-14292	128	1	because	because	SCONJ
fpq-14292	128	2	“	"	PUNCT
fpq-14292	128	3	deeply	deeply	ADV
fpq-14292	128	4	entrenched	entrenched	ADJ
fpq-14292	128	5	forms	form	NOUN
fpq-14292	128	6	of	of	ADP
fpq-14292	128	7	racial	racial	ADJ
fpq-14292	128	8	ignorance	ignorance	NOUN
fpq-14292	128	9	and	and	CCONJ
fpq-14292	128	10	racial	racial	ADJ
fpq-14292	128	11	biases	bias	NOUN
fpq-14292	128	12	led	lead	VERB
fpq-14292	128	13	the	the	DET
fpq-14292	128	14	people	people	NOUN
fpq-14292	128	15	who	who	PRON
fpq-14292	128	16	were	be	AUX
fpq-14292	128	17	developing	develop	VERB
fpq-14292	128	18	emulsion	emulsion	NOUN
fpq-14292	128	19	technologies	technology	NOUN
fpq-14292	128	20	to	to	PART
fpq-14292	128	21	ignore	ignore	VERB
fpq-14292	128	22	variations	variation	NOUN
fpq-14292	128	23	in	in	ADP
fpq-14292	128	24	skin	skin	NOUN
fpq-14292	128	25	tone	tone	NOUN
fpq-14292	128	26	,	,	PUNCT
fpq-14292	128	27	or	or	CCONJ
fpq-14292	128	28	assume	assume	VERB
fpq-14292	128	29	that	that	SCONJ
fpq-14292	128	30	racialized	racialize	VERB
fpq-14292	128	31	differences	difference	NOUN
fpq-14292	128	32	were	be	AUX
fpq-14292	128	33	irrelevant	irrelevant	ADJ
fpq-14292	128	34	to	to	ADP
fpq-14292	128	35	the	the	DET
fpq-14292	128	36	design	design	NOUN
fpq-14292	128	37	of	of	ADP
fpq-14292	128	38	this	this	DET
fpq-14292	128	39	technology	technology	NOUN
fpq-14292	128	40	”	"	PUNCT
fpq-14292	128	41	(	(	PUNCT
fpq-14292	128	42	95	95	NUM
fpq-14292	128	43	;	;	PUNCT
fpq-14292	128	44	see	see	VERB
fpq-14292	128	45	also	also	ADV
fpq-14292	128	46	roth	roth	PROPN
fpq-14292	128	47	2009	2009	NUM
fpq-14292	128	48	;	;	PUNCT
fpq-14292	128	49	smith	smith	PROPN
fpq-14292	128	50	2013	2013	NUM
fpq-14292	128	51	;	;	PUNCT
fpq-14292	128	52	benjamin	benjamin	PROPN
fpq-14292	128	53	2019	2019	NUM
fpq-14292	128	54	)	)	PUNCT
fpq-14292	128	55	.	.	PUNCT
fpq-14292	129	1	however	however	ADV
fpq-14292	129	2	,	,	PUNCT
fpq-14292	129	3	when	when	SCONJ
fpq-14292	129	4	we	we	PRON
fpq-14292	129	5	use	use	VERB
fpq-14292	129	6	cameras	camera	NOUN
fpq-14292	129	7	as	as	ADP
fpq-14292	129	8	a	a	DET
fpq-14292	129	9	paradigm	paradigm	NOUN
fpq-14292	129	10	case	case	NOUN
fpq-14292	129	11	of	of	ADP
fpq-14292	129	12	descriptive	descriptive	ADJ
fpq-14292	129	13	accuracy	accuracy	NOUN
fpq-14292	129	14	,	,	PUNCT
fpq-14292	129	15	we	we	PRON
fpq-14292	129	16	are	be	AUX
fpq-14292	129	17	considering	consider	VERB
fpq-14292	129	18	an	an	DET
fpq-14292	129	19	ideal	ideal	NOUN
fpq-14292	129	20	—	—	PUNCT
fpq-14292	129	21	a	a	DET
fpq-14292	129	22	system	system	NOUN
fpq-14292	129	23	that	that	PRON
fpq-14292	129	24	does	do	AUX
fpq-14292	129	25	get	get	VERB
fpq-14292	129	26	things	thing	NOUN
fpq-14292	129	27	descriptively	descriptively	ADV
fpq-14292	129	28	right	right	ADJ
fpq-14292	129	29	(	(	PUNCT
fpq-14292	129	30	in	in	ADP
fpq-14292	129	31	a	a	DET
fpq-14292	129	32	relevant	relevant	ADJ
fpq-14292	129	33	regard	regard	NOUN
fpq-14292	129	34	)	)	PUNCT
fpq-14292	129	35	.	.	PUNCT
fpq-14292	130	1	if	if	SCONJ
fpq-14292	130	2	a	a	DET
fpq-14292	130	3	camera	camera	NOUN
fpq-14292	130	4	using	use	VERB
fpq-14292	130	5	kodak	kodak	PROPN
fpq-14292	130	6	film	film	NOUN
fpq-14292	130	7	fails	fail	VERB
fpq-14292	130	8	to	to	PART
fpq-14292	130	9	accurately	accurately	ADV
fpq-14292	130	10	depict	depict	VERB
fpq-14292	130	11	darker	dark	ADJ
fpq-14292	130	12	skin	skin	NOUN
fpq-14292	130	13	tones	tone	NOUN
fpq-14292	130	14	,	,	PUNCT
fpq-14292	130	15	then	then	ADV
fpq-14292	130	16	in	in	ADP
fpq-14292	130	17	that	that	DET
fpq-14292	130	18	regard	regard	NOUN
fpq-14292	130	19	it	it	PRON
fpq-14292	130	20	will	will	AUX
fpq-14292	130	21	be	be	AUX
fpq-14292	130	22	descriptively	descriptively	ADV
fpq-14292	130	23	inaccurate	inaccurate	ADJ
fpq-14292	130	24	,	,	PUNCT
fpq-14292	130	25	notwithstanding	notwithstanding	ADP
fpq-14292	130	26	its	its	PRON
fpq-14292	130	27	perhaps	perhaps	ADV
fpq-14292	130	28	achieving	achieve	VERB
fpq-14292	130	29	descriptive	descriptive	ADJ
fpq-14292	130	30	accuracy	accuracy	NOUN
fpq-14292	130	31	in	in	ADP
fpq-14292	130	32	other	other	ADJ
fpq-14292	130	33	regards	regard	NOUN
fpq-14292	130	34	.	.	PUNCT
fpq-14292	131	1	still	still	ADV
fpq-14292	131	2	,	,	PUNCT
fpq-14292	131	3	there	there	PRON
fpq-14292	131	4	are	be	VERB
fpq-14292	131	5	two	two	NUM
fpq-14292	131	6	important	important	ADJ
fpq-14292	131	7	lessons	lesson	NOUN
fpq-14292	131	8	.	.	PUNCT
fpq-14292	132	1	first	first	ADV
fpq-14292	132	2	,	,	PUNCT
fpq-14292	132	3	if	if	SCONJ
fpq-14292	132	4	a	a	DET
fpq-14292	132	5	lack	lack	NOUN
fpq-14292	132	6	of	of	ADP
fpq-14292	132	7	descriptive	descriptive	ADJ
fpq-14292	132	8	accuracy	accuracy	NOUN
fpq-14292	132	9	in	in	ADP
fpq-14292	132	10	any	any	DET
fpq-14292	132	11	particular	particular	ADJ
fpq-14292	132	12	respect	respect	NOUN
fpq-14292	132	13	is	be	AUX
fpq-14292	132	14	due	due	ADJ
fpq-14292	132	15	to	to	PART
fpq-14292	132	16	bias	bias	VERB
fpq-14292	132	17	,	,	PUNCT
fpq-14292	132	18	as	as	ADP
fpq-14292	132	19	in	in	ADP
fpq-14292	132	20	the	the	DET
fpq-14292	132	21	kodak	kodak	PROPN
fpq-14292	132	22	film	film	NOUN
fpq-14292	132	23	case	case	NOUN
fpq-14292	132	24	,	,	PUNCT
fpq-14292	132	25	we	we	PRON
fpq-14292	132	26	must	must	AUX
fpq-14292	132	27	be	be	AUX
fpq-14292	132	28	alive	alive	ADJ
fpq-14292	132	29	to	to	ADP
fpq-14292	132	30	the	the	DET
fpq-14292	132	31	possibility	possibility	NOUN
fpq-14292	132	32	that	that	SCONJ
fpq-14292	132	33	for	for	ADP
fpq-14292	132	34	language	language	NOUN
fpq-14292	132	35	models	model	NOUN
fpq-14292	132	36	,	,	PUNCT
fpq-14292	132	37	even	even	ADV
fpq-14292	132	38	when	when	SCONJ
fpq-14292	132	39	we	we	PRON
fpq-14292	132	40	appear	appear	VERB
fpq-14292	132	41	to	to	PART
fpq-14292	132	42	have	have	VERB
fpq-14292	132	43	descriptive	descriptive	ADJ
fpq-14292	132	44	accuracy	accuracy	NOUN
fpq-14292	132	45	about	about	ADP
fpq-14292	132	46	how	how	SCONJ
fpq-14292	132	47	people	people	NOUN
fpq-14292	132	48	use	use	VERB
fpq-14292	132	49	language	language	NOUN
fpq-14292	132	50	,	,	PUNCT
fpq-14292	132	51	we	we	PRON
fpq-14292	132	52	might	might	AUX
fpq-14292	132	53	not	not	PART
fpq-14292	132	54	,	,	PUNCT
fpq-14292	132	55	at	at	ADP
fpq-14292	132	56	least	least	ADJ
fpq-14292	132	57	in	in	ADP
fpq-14292	132	58	some	some	DET
fpq-14292	132	59	relevant	relevant	ADJ
fpq-14292	132	60	respect	respect	NOUN
fpq-14292	132	61	,	,	PUNCT
fpq-14292	132	62	perhaps	perhaps	ADV
fpq-14292	132	63	due	due	ADJ
fpq-14292	132	64	to	to	PART
fpq-14292	132	65	bias	bias	VERB
fpq-14292	132	66	in	in	ADP
fpq-14292	132	67	the	the	DET
fpq-14292	132	68	training	training	NOUN
fpq-14292	132	69	data.15	data.15	PROPN
fpq-14292	132	70	second	second	ADV
fpq-14292	132	71	,	,	PUNCT
fpq-14292	132	72	there	there	PRON
fpq-14292	132	73	are	be	VERB
fpq-14292	132	74	various	various	ADJ
fpq-14292	132	75	respects	respect	NOUN
fpq-14292	132	76	in	in	ADP
fpq-14292	132	77	which	which	PRON
fpq-14292	132	78	a	a	DET
fpq-14292	132	79	system	system	NOUN
fpq-14292	132	80	or	or	CCONJ
fpq-14292	132	81	model	model	NOUN
fpq-14292	132	82	might	might	AUX
fpq-14292	132	83	get	get	VERB
fpq-14292	132	84	matters	matter	NOUN
fpq-14292	132	85	descriptively	descriptively	ADV
fpq-14292	132	86	right	right	ADJ
fpq-14292	132	87	about	about	ADP
fpq-14292	132	88	how	how	SCONJ
fpq-14292	132	89	people	people	NOUN
fpq-14292	132	90	use	use	VERB
fpq-14292	132	91	language	language	NOUN
fpq-14292	132	92	,	,	PUNCT
fpq-14292	132	93	and	and	CCONJ
fpq-14292	132	94	different	different	ADJ
fpq-14292	132	95	systems	system	NOUN
fpq-14292	132	96	might	might	AUX
fpq-14292	132	97	have	have	VERB
fpq-14292	132	98	differing	differ	VERB
fpq-14292	132	99	aims	aim	NOUN
fpq-14292	132	100	in	in	ADP
fpq-14292	132	101	this	this	DET
fpq-14292	132	102	regard	regard	NOUN
fpq-14292	132	103	—	—	PUNCT
fpq-14292	132	104	for	for	ADP
fpq-14292	132	105	14	14	NUM
fpq-14292	132	106	we	we	PRON
fpq-14292	132	107	will	will	AUX
fpq-14292	132	108	say	say	VERB
fpq-14292	132	109	more	more	ADJ
fpq-14292	132	110	about	about	ADP
fpq-14292	132	111	why	why	SCONJ
fpq-14292	132	112	descriptively	descriptively	ADV
fpq-14292	132	113	accurate	accurate	ADJ
fpq-14292	132	114	information	information	NOUN
fpq-14292	132	115	can	can	AUX
fpq-14292	132	116	be	be	AUX
fpq-14292	132	117	ethically	ethically	ADV
fpq-14292	132	118	useful	useful	ADJ
fpq-14292	132	119	in	in	ADP
fpq-14292	132	120	section	section	NOUN
fpq-14292	132	121	6	6	NUM
fpq-14292	132	122	.	.	PUNCT
fpq-14292	133	1	but	but	CCONJ
fpq-14292	133	2	note	note	VERB
fpq-14292	133	3	that	that	SCONJ
fpq-14292	133	4	even	even	ADV
fpq-14292	133	5	if	if	SCONJ
fpq-14292	133	6	knowing	know	VERB
fpq-14292	133	7	who	who	PRON
fpq-14292	133	8	perpetrated	perpetrate	VERB
fpq-14292	133	9	a	a	DET
fpq-14292	133	10	wrongdoing	wrongdoing	NOUN
fpq-14292	133	11	is	be	AUX
fpq-14292	133	12	ethically	ethically	ADV
fpq-14292	133	13	useful	useful	ADJ
fpq-14292	133	14	information	information	NOUN
fpq-14292	133	15	,	,	PUNCT
fpq-14292	133	16	we	we	PRON
fpq-14292	133	17	agree	agree	VERB
fpq-14292	133	18	with	with	ADP
fpq-14292	133	19	benjamin	benjamin	PROPN
fpq-14292	133	20	(	(	PUNCT
fpq-14292	133	21	2019	2019	NUM
fpq-14292	133	22	)	)	PUNCT
fpq-14292	133	23	that	that	PRON
fpq-14292	133	24	often	often	ADV
fpq-14292	133	25	we	we	PRON
fpq-14292	133	26	have	have	VERB
fpq-14292	133	27	to	to	PART
fpq-14292	133	28	ask	ask	VERB
fpq-14292	133	29	prior	prior	ADJ
fpq-14292	133	30	questions	question	NOUN
fpq-14292	133	31	about	about	ADP
fpq-14292	133	32	whose	whose	DET
fpq-14292	133	33	faces	face	NOUN
fpq-14292	133	34	are	be	AUX
fpq-14292	133	35	being	be	AUX
fpq-14292	133	36	disproportionally	disproportionally	ADV
fpq-14292	133	37	recorded	record	VERB
fpq-14292	133	38	by	by	ADP
fpq-14292	133	39	cameras	camera	NOUN
fpq-14292	133	40	(	(	PUNCT
fpq-14292	133	41	e.g.	e.g.	ADV
fpq-14292	133	42	,	,	PUNCT
fpq-14292	133	43	members	member	NOUN
fpq-14292	133	44	of	of	ADP
fpq-14292	133	45	overpoliced	overpoliced	ADJ
fpq-14292	133	46	or	or	CCONJ
fpq-14292	133	47	oversurveilled	oversurveille	VERB
fpq-14292	133	48	communities	community	NOUN
fpq-14292	133	49	)	)	PUNCT
fpq-14292	133	50	,	,	PUNCT
fpq-14292	133	51	and	and	CCONJ
fpq-14292	133	52	in	in	ADP
fpq-14292	133	53	particular	particular	ADJ
fpq-14292	133	54	by	by	ADP
fpq-14292	133	55	devices	device	NOUN
fpq-14292	133	56	such	such	ADJ
fpq-14292	133	57	as	as	ADP
fpq-14292	133	58	cctv	cctv	NOUN
fpq-14292	133	59	.	.	PUNCT
fpq-14292	134	1	these	these	DET
fpq-14292	134	2	considerations	consideration	NOUN
fpq-14292	134	3	are	be	AUX
fpq-14292	134	4	consistent	consistent	ADJ
fpq-14292	134	5	with	with	ADP
fpq-14292	134	6	our	our	PRON
fpq-14292	134	7	claims	claim	NOUN
fpq-14292	134	8	.	.	PUNCT
fpq-14292	135	1	15	15	NUM
fpq-14292	135	2	this	this	DET
fpq-14292	135	3	issue	issue	NOUN
fpq-14292	135	4	will	will	AUX
fpq-14292	135	5	be	be	AUX
fpq-14292	135	6	important	important	ADJ
fpq-14292	135	7	later	later	ADV
fpq-14292	135	8	,	,	PUNCT
fpq-14292	135	9	when	when	SCONJ
fpq-14292	135	10	we	we	PRON
fpq-14292	135	11	discuss	discuss	VERB
fpq-14292	135	12	large	large	ADJ
fpq-14292	135	13	language	language	NOUN
fpq-14292	135	14	models	model	NOUN
fpq-14292	135	15	.	.	PUNCT
fpq-14292	136	1	some	some	PRON
fpq-14292	136	2	argue	argue	VERB
fpq-14292	136	3	that	that	SCONJ
fpq-14292	136	4	models	model	NOUN
fpq-14292	136	5	like	like	ADP
fpq-14292	136	6	bert	bert	NOUN
fpq-14292	136	7	or	or	CCONJ
fpq-14292	136	8	gtp-2/3	gtp-2/3	ADJ
fpq-14292	136	9	not	not	PART
fpq-14292	136	10	only	only	ADV
fpq-14292	136	11	reproduce	reproduce	VERB
fpq-14292	136	12	bias	bias	NOUN
fpq-14292	136	13	(	(	PUNCT
fpq-14292	136	14	and	and	CCONJ
fpq-14292	136	15	are	be	AUX
fpq-14292	136	16	thereby	thereby	ADV
fpq-14292	136	17	normatively	normatively	ADV
fpq-14292	136	18	incorrect	incorrect	ADJ
fpq-14292	136	19	)	)	PUNCT
fpq-14292	136	20	but	but	CCONJ
fpq-14292	136	21	are	be	AUX
fpq-14292	136	22	descriptively	descriptively	ADV
fpq-14292	136	23	inaccurate	inaccurate	ADJ
fpq-14292	136	24	since	since	SCONJ
fpq-14292	136	25	the	the	DET
fpq-14292	136	26	training	training	NOUN
fpq-14292	136	27	data	datum	NOUN
fpq-14292	136	28	do	do	AUX
fpq-14292	136	29	not	not	PART
fpq-14292	136	30	reflect	reflect	VERB
fpq-14292	136	31	marginalized	marginalized	ADJ
fpq-14292	136	32	voices	voice	NOUN
fpq-14292	136	33	(	(	PUNCT
fpq-14292	136	34	bender	bender	NOUN
fpq-14292	136	35	,	,	PUNCT
fpq-14292	136	36	gebru	gebru	VERB
fpq-14292	136	37	et	et	PROPN
fpq-14292	136	38	al	al	PROPN
fpq-14292	136	39	.	.	PROPN
fpq-14292	136	40	2021	2021	NUM
fpq-14292	136	41	)	)	PUNCT
fpq-14292	136	42	.	.	PUNCT
fpq-14292	137	1	feminist	feminist	PROPN
fpq-14292	137	2	philosophy	philosophy	PROPN
fpq-14292	137	3	quarterly	quarterly	ADV
fpq-14292	137	4	,	,	PUNCT
fpq-14292	137	5	2022	2022	NUM
fpq-14292	137	6	,	,	PUNCT
fpq-14292	137	7	vol	vol	NOUN
fpq-14292	137	8	.	.	NOUN
fpq-14292	137	9	8	8	NUM
fpq-14292	137	10	,	,	PUNCT
fpq-14292	137	11	iss	iss	PROPN
fpq-14292	137	12	.	.	PUNCT
fpq-14292	137	13	3/4	3/4	NUM
fpq-14292	137	14	,	,	PUNCT
fpq-14292	137	15	article	article	NOUN
fpq-14292	137	16	1	1	NUM
fpq-14292	137	17	published	publish	VERB
fpq-14292	137	18	by	by	ADP
fpq-14292	137	19	scholarship@western	scholarship@western	PROPN
fpq-14292	137	20	,	,	PUNCT
fpq-14292	137	21	2022	2022	NUM
fpq-14292	137	22	8	8	NUM
fpq-14292	137	23	example	example	NOUN
fpq-14292	137	24	,	,	PUNCT
fpq-14292	137	25	a	a	DET
fpq-14292	137	26	model	model	NOUN
fpq-14292	137	27	might	might	AUX
fpq-14292	137	28	aim	aim	VERB
fpq-14292	137	29	for	for	ADP
fpq-14292	137	30	descriptive	descriptive	ADJ
fpq-14292	137	31	accuracy	accuracy	NOUN
fpq-14292	137	32	in	in	ADP
fpq-14292	137	33	one	one	NUM
fpq-14292	137	34	respect	respect	NOUN
fpq-14292	137	35	yet	yet	ADV
fpq-14292	137	36	not	not	PART
fpq-14292	137	37	in	in	ADP
fpq-14292	137	38	another	another	PRON
fpq-14292	137	39	.	.	PUNCT
fpq-14292	138	1	for	for	ADP
fpq-14292	138	2	example	example	NOUN
fpq-14292	138	3	,	,	PUNCT
fpq-14292	138	4	word	word	NOUN
fpq-14292	138	5	-	-	PUNCT
fpq-14292	138	6	embeddings	embedding	NOUN
fpq-14292	138	7	models	model	NOUN
fpq-14292	138	8	(	(	PUNCT
fpq-14292	138	9	which	which	PRON
fpq-14292	138	10	we	we	PRON
fpq-14292	138	11	will	will	AUX
fpq-14292	138	12	discuss	discuss	VERB
fpq-14292	138	13	in	in	ADP
fpq-14292	138	14	detail	detail	NOUN
fpq-14292	138	15	in	in	ADP
fpq-14292	138	16	section	section	NOUN
fpq-14292	138	17	4	4	NUM
fpq-14292	138	18	)	)	PUNCT
fpq-14292	138	19	aim	aim	NOUN
fpq-14292	138	20	at	at	ADP
fpq-14292	138	21	capturing	capture	VERB
fpq-14292	138	22	how	how	SCONJ
fpq-14292	138	23	words	word	NOUN
fpq-14292	138	24	cluster	cluster	NOUN
fpq-14292	138	25	in	in	ADP
fpq-14292	138	26	a	a	DET
fpq-14292	138	27	vector	vector	NOUN
fpq-14292	138	28	space	space	NOUN
fpq-14292	138	29	based	base	VERB
fpq-14292	138	30	on	on	ADP
fpq-14292	138	31	how	how	SCONJ
fpq-14292	138	32	people	people	NOUN
fpq-14292	138	33	actually	actually	ADV
fpq-14292	138	34	use	use	VERB
fpq-14292	138	35	language	language	NOUN
fpq-14292	138	36	,	,	PUNCT
fpq-14292	138	37	and	and	CCONJ
fpq-14292	138	38	these	these	DET
fpq-14292	138	39	models	model	NOUN
fpq-14292	138	40	might	might	AUX
fpq-14292	138	41	be	be	AUX
fpq-14292	138	42	assessed	assess	VERB
fpq-14292	138	43	for	for	ADP
fpq-14292	138	44	descriptive	descriptive	ADJ
fpq-14292	138	45	accuracy	accuracy	NOUN
fpq-14292	138	46	in	in	ADP
fpq-14292	138	47	this	this	DET
fpq-14292	138	48	respect	respect	NOUN
fpq-14292	138	49	.	.	PUNCT
fpq-14292	139	1	as	as	SCONJ
fpq-14292	139	2	we	we	PRON
fpq-14292	139	3	shall	shall	AUX
fpq-14292	139	4	see	see	VERB
fpq-14292	139	5	,	,	PUNCT
fpq-14292	139	6	“	"	PUNCT
fpq-14292	139	7	woman	woman	NOUN
fpq-14292	139	8	”	"	PUNCT
fpq-14292	139	9	and	and	CCONJ
fpq-14292	139	10	“	"	PUNCT
fpq-14292	139	11	homemaker	homemaker	NOUN
fpq-14292	139	12	”	"	PUNCT
fpq-14292	139	13	do	do	AUX
fpq-14292	139	14	cluster	cluster	NOUN
fpq-14292	139	15	together	together	ADV
fpq-14292	139	16	,	,	PUNCT
fpq-14292	139	17	whereas	whereas	SCONJ
fpq-14292	139	18	“	"	PUNCT
fpq-14292	139	19	woman	woman	NOUN
fpq-14292	139	20	”	"	PUNCT
fpq-14292	139	21	and	and	CCONJ
fpq-14292	139	22	“	"	PUNCT
fpq-14292	139	23	computer	computer	NOUN
fpq-14292	139	24	programmer	programmer	NOUN
fpq-14292	139	25	”	"	PUNCT
fpq-14292	139	26	do	do	VERB
fpq-14292	139	27	not	not	PART
fpq-14292	139	28	.	.	PUNCT
fpq-14292	140	1	even	even	ADV
fpq-14292	140	2	so	so	ADV
fpq-14292	140	3	,	,	PUNCT
fpq-14292	140	4	such	such	DET
fpq-14292	140	5	a	a	DET
fpq-14292	140	6	model	model	NOUN
fpq-14292	140	7	is	be	AUX
fpq-14292	140	8	clearly	clearly	ADV
fpq-14292	140	9	descriptively	descriptively	ADV
fpq-14292	140	10	inaccurate	inaccurate	ADJ
fpq-14292	140	11	in	in	ADP
fpq-14292	140	12	another	another	DET
fpq-14292	140	13	sense	sense	NOUN
fpq-14292	140	14	,	,	PUNCT
fpq-14292	140	15	since	since	SCONJ
fpq-14292	140	16	these	these	DET
fpq-14292	140	17	clusterings	clustering	NOUN
fpq-14292	140	18	are	be	AUX
fpq-14292	140	19	descriptively	descriptively	ADV
fpq-14292	140	20	inaccurate	inaccurate	ADJ
fpq-14292	140	21	when	when	SCONJ
fpq-14292	140	22	it	it	PRON
fpq-14292	140	23	comes	come	VERB
fpq-14292	140	24	to	to	ADP
fpq-14292	140	25	these	these	DET
fpq-14292	140	26	words	word	NOUN
fpq-14292	140	27	’	'	PUNCT
fpq-14292	140	28	semantics	semantic	NOUN
fpq-14292	140	29	—	—	PUNCT
fpq-14292	140	30	there	there	PRON
fpq-14292	140	31	is	be	VERB
fpq-14292	140	32	nothing	nothing	PRON
fpq-14292	140	33	about	about	ADP
fpq-14292	140	34	“	"	PUNCT
fpq-14292	140	35	woman	woman	NOUN
fpq-14292	140	36	,	,	PUNCT
fpq-14292	140	37	”	"	PUNCT
fpq-14292	140	38	semantically	semantically	ADV
fpq-14292	140	39	,	,	PUNCT
fpq-14292	140	40	in	in	ADP
fpq-14292	140	41	virtue	virtue	NOUN
fpq-14292	140	42	of	of	ADP
fpq-14292	140	43	which	which	PRON
fpq-14292	140	44	it	it	PRON
fpq-14292	140	45	clusters	cluster	VERB
fpq-14292	140	46	with	with	ADP
fpq-14292	140	47	“	"	PUNCT
fpq-14292	140	48	homemaker	homemaker	NOUN
fpq-14292	140	49	”	"	PUNCT
fpq-14292	140	50	(	(	PUNCT
fpq-14292	140	51	by	by	ADP
fpq-14292	140	52	contrast	contrast	NOUN
fpq-14292	140	53	with	with	ADP
fpq-14292	140	54	“	"	PUNCT
fpq-14292	140	55	bachelor	bachelor	NOUN
fpq-14292	140	56	,	,	PUNCT
fpq-14292	140	57	”	"	PUNCT
fpq-14292	140	58	which	which	PRON
fpq-14292	140	59	semantically	semantically	ADV
fpq-14292	140	60	clusters	cluster	VERB
fpq-14292	140	61	with	with	ADP
fpq-14292	140	62	the	the	DET
fpq-14292	140	63	word	word	NOUN
fpq-14292	140	64	“	"	PUNCT
fpq-14292	140	65	man	man	NOUN
fpq-14292	140	66	”	"	PUNCT
fpq-14292	140	67	)	)	PUNCT
fpq-14292	140	68	.	.	PUNCT
fpq-14292	141	1	one	one	PRON
fpq-14292	141	2	might	might	AUX
fpq-14292	141	3	think	think	VERB
fpq-14292	141	4	that	that	SCONJ
fpq-14292	141	5	whether	whether	SCONJ
fpq-14292	141	6	a	a	DET
fpq-14292	141	7	representation	representation	NOUN
fpq-14292	141	8	is	be	AUX
fpq-14292	141	9	descriptively	descriptively	ADV
fpq-14292	141	10	accurate	accurate	ADJ
fpq-14292	141	11	is	be	AUX
fpq-14292	141	12	a	a	DET
fpq-14292	141	13	function	function	NOUN
fpq-14292	141	14	of	of	ADP
fpq-14292	141	15	who	who	PRON
fpq-14292	141	16	we	we	PRON
fpq-14292	141	17	are	be	AUX
fpq-14292	141	18	and	and	CCONJ
fpq-14292	141	19	what	what	PRON
fpq-14292	141	20	we	we	PRON
fpq-14292	141	21	want	want	VERB
fpq-14292	141	22	to	to	PART
fpq-14292	141	23	do	do	VERB
fpq-14292	141	24	with	with	ADP
fpq-14292	141	25	it	it	PRON
fpq-14292	141	26	—	—	PUNCT
fpq-14292	141	27	in	in	ADP
fpq-14292	141	28	which	which	DET
fpq-14292	141	29	case	case	NOUN
fpq-14292	141	30	,	,	PUNCT
fpq-14292	141	31	descriptive	descriptive	ADJ
fpq-14292	141	32	accuracy	accuracy	NOUN
fpq-14292	141	33	seems	seem	VERB
fpq-14292	141	34	context	context	NOUN
fpq-14292	141	35	-	-	PUNCT
fpq-14292	141	36	dependent	dependent	ADJ
fpq-14292	141	37	.	.	PUNCT
fpq-14292	142	1	but	but	CCONJ
fpq-14292	142	2	this	this	DET
fpq-14292	142	3	issue	issue	NOUN
fpq-14292	142	4	is	be	AUX
fpq-14292	142	5	more	more	ADJ
fpq-14292	142	6	one	one	NUM
fpq-14292	142	7	of	of	ADP
fpq-14292	142	8	salience	salience	NOUN
fpq-14292	142	9	,	,	PUNCT
fpq-14292	142	10	not	not	PART
fpq-14292	142	11	accuracy	accuracy	NOUN
fpq-14292	142	12	.	.	PUNCT
fpq-14292	143	1	many	many	ADJ
fpq-14292	143	2	representations	representation	NOUN
fpq-14292	143	3	are	be	AUX
fpq-14292	143	4	accurate	accurate	ADJ
fpq-14292	143	5	even	even	ADV
fpq-14292	143	6	if	if	SCONJ
fpq-14292	143	7	some	some	DET
fpq-14292	143	8	information	information	NOUN
fpq-14292	143	9	they	they	PRON
fpq-14292	143	10	convey	convey	VERB
fpq-14292	143	11	is	be	AUX
fpq-14292	143	12	not	not	PART
fpq-14292	143	13	salient	salient	ADJ
fpq-14292	143	14	,	,	PUNCT
fpq-14292	143	15	and	and	CCONJ
fpq-14292	143	16	salience	salience	NOUN
fpq-14292	143	17	depends	depend	VERB
fpq-14292	143	18	on	on	ADP
fpq-14292	143	19	the	the	DET
fpq-14292	143	20	context	context	NOUN
fpq-14292	143	21	.	.	PUNCT
fpq-14292	144	1	the	the	DET
fpq-14292	144	2	family	family	NOUN
fpq-14292	144	3	photograph	photograph	NOUN
fpq-14292	144	4	i	i	PRON
fpq-14292	144	5	took	take	VERB
fpq-14292	144	6	in	in	ADP
fpq-14292	144	7	my	my	PRON
fpq-14292	144	8	front	front	ADJ
fpq-14292	144	9	garden	garden	NOUN
fpq-14292	144	10	is	be	AUX
fpq-14292	144	11	descriptively	descriptively	ADV
fpq-14292	144	12	accurate	accurate	ADJ
fpq-14292	144	13	in	in	ADP
fpq-14292	144	14	depicting	depict	VERB
fpq-14292	144	15	us	we	PRON
fpq-14292	144	16	smiling	smile	VERB
fpq-14292	144	17	.	.	PUNCT
fpq-14292	145	1	in	in	ADP
fpq-14292	145	2	the	the	DET
fpq-14292	145	3	context	context	NOUN
fpq-14292	145	4	of	of	ADP
fpq-14292	145	5	sending	send	VERB
fpq-14292	145	6	a	a	DET
fpq-14292	145	7	holiday	holiday	NOUN
fpq-14292	145	8	card	card	NOUN
fpq-14292	145	9	to	to	ADP
fpq-14292	145	10	friends	friend	NOUN
fpq-14292	145	11	,	,	PUNCT
fpq-14292	145	12	this	this	DET
fpq-14292	145	13	information	information	NOUN
fpq-14292	145	14	may	may	AUX
fpq-14292	145	15	be	be	AUX
fpq-14292	145	16	most	most	ADV
fpq-14292	145	17	salient	salient	NOUN
fpq-14292	145	18	.	.	PUNCT
fpq-14292	146	1	yet	yet	CCONJ
fpq-14292	146	2	the	the	DET
fpq-14292	146	3	photograph	photograph	NOUN
fpq-14292	146	4	might	might	AUX
fpq-14292	146	5	also	also	ADV
fpq-14292	146	6	accurately	accurately	ADV
fpq-14292	146	7	depict	depict	VERB
fpq-14292	146	8	the	the	DET
fpq-14292	146	9	hole	hole	NOUN
fpq-14292	146	10	in	in	ADP
fpq-14292	146	11	my	my	PRON
fpq-14292	146	12	eavestrough	eavestrough	NOUN
fpq-14292	146	13	,	,	PUNCT
fpq-14292	146	14	and	and	CCONJ
fpq-14292	146	15	in	in	ADP
fpq-14292	146	16	a	a	DET
fpq-14292	146	17	different	different	ADJ
fpq-14292	146	18	context	context	NOUN
fpq-14292	146	19	—	—	PUNCT
fpq-14292	146	20	such	such	ADJ
fpq-14292	146	21	as	as	ADP
fpq-14292	146	22	getting	get	VERB
fpq-14292	146	23	a	a	DET
fpq-14292	146	24	quote	quote	NOUN
fpq-14292	146	25	for	for	ADP
fpq-14292	146	26	fixing	fix	VERB
fpq-14292	146	27	the	the	DET
fpq-14292	146	28	hole	hole	NOUN
fpq-14292	146	29	—	—	PUNCT
fpq-14292	146	30	this	this	DET
fpq-14292	146	31	descriptively	descriptively	ADV
fpq-14292	146	32	accurate	accurate	ADJ
fpq-14292	146	33	information	information	NOUN
fpq-14292	146	34	will	will	AUX
fpq-14292	146	35	be	be	AUX
fpq-14292	146	36	of	of	ADP
fpq-14292	146	37	primary	primary	ADJ
fpq-14292	146	38	salience	salience	NOUN
fpq-14292	146	39	rather	rather	ADV
fpq-14292	146	40	than	than	ADP
fpq-14292	146	41	the	the	DET
fpq-14292	146	42	fact	fact	NOUN
fpq-14292	146	43	that	that	SCONJ
fpq-14292	146	44	we	we	PRON
fpq-14292	146	45	are	be	AUX
fpq-14292	146	46	smiling.16	smiling.16	PROPN
fpq-14292	146	47	finally	finally	ADV
fpq-14292	146	48	,	,	PUNCT
fpq-14292	146	49	note	note	VERB
fpq-14292	146	50	that	that	SCONJ
fpq-14292	146	51	descriptive	descriptive	ADJ
fpq-14292	146	52	accuracy	accuracy	NOUN
fpq-14292	146	53	does	do	AUX
fpq-14292	146	54	not	not	PART
fpq-14292	146	55	require	require	VERB
fpq-14292	146	56	perfect	perfect	ADJ
fpq-14292	146	57	descriptive	descriptive	ADJ
fpq-14292	146	58	accuracy	accuracy	NOUN
fpq-14292	146	59	.	.	PUNCT
fpq-14292	147	1	many	many	ADJ
fpq-14292	147	2	representations	representation	NOUN
fpq-14292	147	3	of	of	ADP
fpq-14292	147	4	aspects	aspect	NOUN
fpq-14292	147	5	of	of	ADP
fpq-14292	147	6	the	the	DET
fpq-14292	147	7	world	world	NOUN
fpq-14292	147	8	are	be	AUX
fpq-14292	147	9	descriptively	descriptively	ADV
fpq-14292	147	10	accurate	accurate	ADJ
fpq-14292	147	11	even	even	ADV
fpq-14292	147	12	when	when	SCONJ
fpq-14292	147	13	they	they	PRON
fpq-14292	147	14	are	be	AUX
fpq-14292	147	15	less	less	ADJ
fpq-14292	147	16	than	than	ADP
fpq-14292	147	17	perfectly	perfectly	ADV
fpq-14292	147	18	accurate	accurate	ADJ
fpq-14292	147	19	.	.	PUNCT
fpq-14292	148	1	that	that	PRON
fpq-14292	148	2	is	be	AUX
fpq-14292	148	3	the	the	DET
fpq-14292	148	4	case	case	NOUN
fpq-14292	148	5	even	even	ADV
fpq-14292	148	6	for	for	ADP
fpq-14292	148	7	scientific	scientific	ADJ
fpq-14292	148	8	theories	theory	NOUN
fpq-14292	148	9	,	,	PUNCT
fpq-14292	148	10	including	include	VERB
fpq-14292	148	11	theories	theory	NOUN
fpq-14292	148	12	of	of	ADP
fpq-14292	148	13	physics	physics	NOUN
fpq-14292	148	14	(	(	PUNCT
fpq-14292	148	15	see	see	VERB
fpq-14292	148	16	,	,	PUNCT
fpq-14292	148	17	e.g.	e.g.	ADV
fpq-14292	148	18	,	,	PUNCT
fpq-14292	148	19	hitchcock	hitchcock	PROPN
fpq-14292	148	20	2004	2004	NUM
fpq-14292	148	21	,	,	PUNCT
fpq-14292	148	22	10	10	NUM
fpq-14292	148	23	)	)	PUNCT
fpq-14292	148	24	.	.	PUNCT
fpq-14292	149	1	moreover	moreover	ADV
fpq-14292	149	2	,	,	PUNCT
fpq-14292	149	3	a	a	DET
fpq-14292	149	4	visual	visual	ADJ
fpq-14292	149	5	representation	representation	NOUN
fpq-14292	149	6	can	can	AUX
fpq-14292	149	7	be	be	AUX
fpq-14292	149	8	accurate	accurate	ADJ
fpq-14292	149	9	without	without	ADP
fpq-14292	149	10	being	be	AUX
fpq-14292	149	11	perfectly	perfectly	ADV
fpq-14292	149	12	accurate	accurate	ADJ
fpq-14292	149	13	.	.	PUNCT
fpq-14292	150	1	for	for	ADP
fpq-14292	150	2	instance	instance	NOUN
fpq-14292	150	3	,	,	PUNCT
fpq-14292	150	4	canine	canine	PROPN
fpq-14292	150	5	visual	visual	ADJ
fpq-14292	150	6	systems	system	NOUN
fpq-14292	150	7	are	be	AUX
fpq-14292	150	8	optimized	optimize	VERB
fpq-14292	150	9	more	more	ADV
fpq-14292	150	10	for	for	ADP
fpq-14292	150	11	low	low	ADJ
fpq-14292	150	12	-	-	PUNCT
fpq-14292	150	13	light	light	NOUN
fpq-14292	150	14	conditions	condition	NOUN
fpq-14292	150	15	than	than	ADP
fpq-14292	150	16	color	color	NOUN
fpq-14292	150	17	,	,	PUNCT
fpq-14292	150	18	by	by	ADP
fpq-14292	150	19	contrast	contrast	NOUN
fpq-14292	150	20	with	with	ADP
fpq-14292	150	21	our	our	PRON
fpq-14292	150	22	own	own	ADJ
fpq-14292	150	23	vision	vision	NOUN
fpq-14292	150	24	(	(	PUNCT
fpq-14292	150	25	barber	barber	PROPN
fpq-14292	150	26	et	et	PROPN
fpq-14292	150	27	al	al	PROPN
fpq-14292	150	28	.	.	PROPN
fpq-14292	150	29	2020	2020	NUM
fpq-14292	150	30	)	)	PUNCT
fpq-14292	150	31	.	.	PUNCT
fpq-14292	151	1	even	even	ADV
fpq-14292	151	2	so	so	ADV
fpq-14292	151	3	,	,	PUNCT
fpq-14292	151	4	dogs	dog	NOUN
fpq-14292	151	5	’	'	PUNCT
fpq-14292	151	6	vision	vision	NOUN
fpq-14292	151	7	is	be	AUX
fpq-14292	151	8	descriptively	descriptively	ADV
fpq-14292	151	9	accurate	accurate	ADJ
fpq-14292	151	10	in	in	ADP
fpq-14292	151	11	many	many	ADJ
fpq-14292	151	12	respects	respect	NOUN
fpq-14292	151	13	and	and	CCONJ
fpq-14292	151	14	is	be	AUX
fpq-14292	151	15	often	often	ADV
fpq-14292	151	16	more	more	ADV
fpq-14292	151	17	descriptively	descriptively	ADV
fpq-14292	151	18	accurate	accurate	ADJ
fpq-14292	151	19	than	than	ADP
fpq-14292	151	20	ours	our	NOUN
fpq-14292	151	21	.	.	PUNCT
fpq-14292	152	1	finally	finally	ADV
fpq-14292	152	2	,	,	PUNCT
fpq-14292	152	3	it	it	PRON
fpq-14292	152	4	is	be	AUX
fpq-14292	152	5	a	a	DET
fpq-14292	152	6	truism	truism	NOUN
fpq-14292	152	7	that	that	SCONJ
fpq-14292	152	8	all	all	DET
fpq-14292	152	9	models	model	NOUN
fpq-14292	152	10	are	be	AUX
fpq-14292	152	11	inaccurate	inaccurate	ADJ
fpq-14292	152	12	to	to	ADP
fpq-14292	152	13	some	some	DET
fpq-14292	152	14	degree	degree	NOUN
fpq-14292	152	15	(	(	PUNCT
fpq-14292	152	16	see	see	VERB
fpq-14292	152	17	,	,	PUNCT
fpq-14292	152	18	e.g.	e.g.	ADV
fpq-14292	152	19	,	,	PUNCT
fpq-14292	152	20	box	box	NOUN
fpq-14292	152	21	1976	1976	NUM
fpq-14292	152	22	,	,	PUNCT
fpq-14292	152	23	792	792	NUM
fpq-14292	152	24	)	)	PUNCT
fpq-14292	152	25	while	while	SCONJ
fpq-14292	152	26	still	still	ADV
fpq-14292	152	27	being	be	AUX
fpq-14292	152	28	usefully	usefully	ADV
fpq-14292	152	29	descriptively	descriptively	ADV
fpq-14292	152	30	accurate	accurate	ADJ
fpq-14292	152	31	(	(	PUNCT
fpq-14292	152	32	and	and	CCONJ
fpq-14292	152	33	thus	thus	ADV
fpq-14292	152	34	also	also	ADV
fpq-14292	152	35	predictively	predictively	ADV
fpq-14292	152	36	or	or	CCONJ
fpq-14292	152	37	explanatorily	explanatorily	ADV
fpq-14292	152	38	useful	useful	ADJ
fpq-14292	152	39	)	)	PUNCT
fpq-14292	152	40	.	.	PUNCT
fpq-14292	153	1	leaving	leave	VERB
fpq-14292	153	2	aside	aside	ADV
fpq-14292	153	3	descriptive	descriptive	ADJ
fpq-14292	153	4	accuracy	accuracy	NOUN
fpq-14292	153	5	,	,	PUNCT
fpq-14292	153	6	a	a	DET
fpq-14292	153	7	system	system	NOUN
fpq-14292	153	8	or	or	CCONJ
fpq-14292	153	9	its	its	PRON
fpq-14292	153	10	outputs	output	NOUN
fpq-14292	153	11	are	be	AUX
fpq-14292	153	12	normatively	normatively	ADV
fpq-14292	153	13	correct	correct	ADJ
fpq-14292	153	14	if	if	SCONJ
fpq-14292	153	15	they	they	PRON
fpq-14292	153	16	reflect	reflect	VERB
fpq-14292	153	17	the	the	DET
fpq-14292	153	18	world	world	NOUN
fpq-14292	153	19	as	as	SCONJ
fpq-14292	153	20	it	it	PRON
fpq-14292	153	21	ought	ought	AUX
fpq-14292	153	22	to	to	PART
fpq-14292	153	23	be	be	AUX
fpq-14292	153	24	—	—	PUNCT
fpq-14292	153	25	in	in	ADP
fpq-14292	153	26	some	some	DET
fpq-14292	153	27	respect	respect	NOUN
fpq-14292	153	28	and	and	CCONJ
fpq-14292	153	29	on	on	ADP
fpq-14292	153	30	some	some	DET
fpq-14292	153	31	conception	conception	NOUN
fpq-14292	153	32	of	of	ADP
fpq-14292	153	33	how	how	SCONJ
fpq-14292	153	34	the	the	DET
fpq-14292	153	35	world	world	NOUN
fpq-14292	153	36	should	should	AUX
fpq-14292	153	37	be	be	AUX
fpq-14292	153	38	,	,	PUNCT
fpq-14292	153	39	even	even	ADV
fpq-14292	153	40	if	if	SCONJ
fpq-14292	153	41	that	that	PRON
fpq-14292	153	42	is	be	AUX
fpq-14292	153	43	not	not	PART
fpq-14292	153	44	how	how	SCONJ
fpq-14292	153	45	it	it	PRON
fpq-14292	153	46	is	be	AUX
fpq-14292	153	47	.	.	PUNCT
fpq-14292	154	1	language	language	NOUN
fpq-14292	154	2	models	model	NOUN
fpq-14292	154	3	are	be	AUX
fpq-14292	154	4	normatively	normatively	ADV
fpq-14292	154	5	correct	correct	ADJ
fpq-14292	154	6	if	if	SCONJ
fpq-14292	154	7	they	they	PRON
fpq-14292	154	8	reflect	reflect	VERB
fpq-14292	154	9	language	language	NOUN
fpq-14292	154	10	use	use	NOUN
fpq-14292	154	11	as	as	SCONJ
fpq-14292	154	12	it	it	PRON
fpq-14292	154	13	should	should	AUX
fpq-14292	154	14	be	be	AUX
fpq-14292	154	15	.	.	PUNCT
fpq-14292	155	1	for	for	ADP
fpq-14292	155	2	example	example	NOUN
fpq-14292	155	3	,	,	PUNCT
fpq-14292	155	4	a	a	DET
fpq-14292	155	5	model	model	NOUN
fpq-14292	155	6	might	might	AUX
fpq-14292	155	7	be	be	AUX
fpq-14292	155	8	normatively	normatively	ADV
fpq-14292	155	9	correct	correct	ADJ
fpq-14292	155	10	in	in	ADP
fpq-14292	155	11	not	not	PART
fpq-14292	155	12	giving	give	VERB
fpq-14292	155	13	sexist	sexist	ADJ
fpq-14292	155	14	outputs	output	NOUN
fpq-14292	155	15	.	.	PUNCT
fpq-14292	156	1	such	such	DET
fpq-14292	156	2	a	a	DET
fpq-14292	156	3	system	system	NOUN
fpq-14292	156	4	would	would	AUX
fpq-14292	156	5	not	not	PART
fpq-14292	156	6	resolve	resolve	VERB
fpq-14292	156	7	the	the	DET
fpq-14292	156	8	analogy	analogy	NOUN
fpq-14292	156	9	task	task	NOUN
fpq-14292	156	10	“	"	PUNCT
fpq-14292	156	11	man	man	NOUN
fpq-14292	156	12	is	be	AUX
fpq-14292	156	13	to	to	PART
fpq-14292	156	14	computer	computer	NOUN
fpq-14292	156	15	programmer	programmer	NOUN
fpq-14292	156	16	as	as	SCONJ
fpq-14292	156	17	woman	woman	NOUN
fpq-14292	156	18	is	be	AUX
fpq-14292	156	19	to	to	ADP
fpq-14292	156	20	x	x	PRON
fpq-14292	156	21	”	"	PUNCT
fpq-14292	156	22	by	by	ADP
fpq-14292	156	23	16	16	NUM
fpq-14292	156	24	we	we	PRON
fpq-14292	156	25	thank	thank	VERB
fpq-14292	156	26	shelley	shelley	PROPN
fpq-14292	156	27	park	park	PROPN
fpq-14292	156	28	(	(	PUNCT
fpq-14292	156	29	at	at	ADP
fpq-14292	156	30	the	the	DET
fpq-14292	156	31	2021	2021	NUM
fpq-14292	156	32	fsjai	fsjai	NOUN
fpq-14292	156	33	workshop	workshop	NOUN
fpq-14292	156	34	)	)	PUNCT
fpq-14292	156	35	for	for	ADP
fpq-14292	156	36	prompting	prompt	VERB
fpq-14292	156	37	this	this	DET
fpq-14292	156	38	clarification	clarification	NOUN
fpq-14292	156	39	.	.	PUNCT
fpq-14292	157	1	deery	deery	NOUN
fpq-14292	157	2	and	and	CCONJ
fpq-14292	157	3	bailey	bailey	PROPN
fpq-14292	157	4	–	–	PUNCT
fpq-14292	157	5	the	the	DET
fpq-14292	157	6	bias	bias	NOUN
fpq-14292	157	7	dilemma	dilemma	NOUN
fpq-14292	157	8	:	:	PUNCT
fpq-14292	157	9	the	the	DET
fpq-14292	157	10	ethics	ethic	NOUN
fpq-14292	157	11	of	of	ADP
fpq-14292	157	12	algorithmic	algorithmic	ADJ
fpq-14292	157	13	bias	bias	NOUN
fpq-14292	157	14	in	in	ADP
fpq-14292	157	15	natural	natural	ADJ
fpq-14292	157	16	-	-	PUNCT
fpq-14292	157	17	language	language	NOUN
fpq-14292	157	18	processing	processing	NOUN
fpq-14292	157	19	published	publish	VERB
fpq-14292	157	20	by	by	ADP
fpq-14292	157	21	scholarship@western	scholarship@western	PROPN
fpq-14292	157	22	,	,	PUNCT
fpq-14292	157	23	2022	2022	NUM
fpq-14292	157	24	9	9	NUM
fpq-14292	157	25	filling	fill	VERB
fpq-14292	157	26	in	in	ADP
fpq-14292	157	27	“	"	PUNCT
fpq-14292	157	28	homemaker	homemaker	NOUN
fpq-14292	157	29	”	"	PUNCT
fpq-14292	157	30	for	for	ADP
fpq-14292	157	31	x	x	SYM
fpq-14292	157	32	(	(	PUNCT
fpq-14292	157	33	we	we	PRON
fpq-14292	157	34	will	will	AUX
fpq-14292	157	35	discuss	discuss	VERB
fpq-14292	157	36	this	this	DET
fpq-14292	157	37	example	example	NOUN
fpq-14292	157	38	in	in	ADP
fpq-14292	157	39	greater	great	ADJ
fpq-14292	157	40	detail	detail	NOUN
fpq-14292	157	41	in	in	ADP
fpq-14292	157	42	section	section	NOUN
fpq-14292	157	43	4	4	NUM
fpq-14292	157	44	)	)	PUNCT
fpq-14292	157	45	.	.	PUNCT
fpq-14292	158	1	by	by	ADP
fpq-14292	158	2	contrast	contrast	NOUN
fpq-14292	158	3	,	,	PUNCT
fpq-14292	158	4	normatively	normatively	ADV
fpq-14292	158	5	incorrect	incorrect	ADJ
fpq-14292	158	6	models	model	NOUN
fpq-14292	158	7	or	or	CCONJ
fpq-14292	158	8	outputs	output	NOUN
fpq-14292	158	9	reflect	reflect	VERB
fpq-14292	158	10	language	language	NOUN
fpq-14292	158	11	use	use	NOUN
fpq-14292	158	12	as	as	SCONJ
fpq-14292	158	13	it	it	PRON
fpq-14292	158	14	ought	ought	AUX
fpq-14292	158	15	not	not	PART
fpq-14292	158	16	to	to	PART
fpq-14292	158	17	be	be	AUX
fpq-14292	158	18	;	;	PUNCT
fpq-14292	158	19	such	such	DET
fpq-14292	158	20	a	a	DET
fpq-14292	158	21	system	system	NOUN
fpq-14292	158	22	might	might	AUX
fpq-14292	158	23	fill	fill	VERB
fpq-14292	158	24	in	in	ADP
fpq-14292	158	25	“	"	PUNCT
fpq-14292	158	26	homemaker	homemaker	NOUN
fpq-14292	158	27	”	"	PUNCT
fpq-14292	158	28	for	for	ADP
fpq-14292	158	29	x	x	SYM
fpq-14292	158	30	in	in	ADP
fpq-14292	158	31	our	our	PRON
fpq-14292	158	32	example.17	example.17	PROPN
fpq-14292	158	33	for	for	ADP
fpq-14292	158	34	our	our	PRON
fpq-14292	158	35	purposes	purpose	NOUN
fpq-14292	158	36	,	,	PUNCT
fpq-14292	158	37	morally	morally	ADV
fpq-14292	158	38	relevant	relevant	ADJ
fpq-14292	158	39	bias	bias	NOUN
fpq-14292	158	40	in	in	ADP
fpq-14292	158	41	a	a	DET
fpq-14292	158	42	language	language	NOUN
fpq-14292	158	43	model	model	NOUN
fpq-14292	158	44	is	be	AUX
fpq-14292	158	45	sufficient	sufficient	ADJ
fpq-14292	158	46	for	for	SCONJ
fpq-14292	158	47	that	that	DET
fpq-14292	158	48	model	model	NOUN
fpq-14292	158	49	to	to	PART
fpq-14292	158	50	be	be	AUX
fpq-14292	158	51	normatively	normatively	ADV
fpq-14292	158	52	incorrect	incorrect	ADJ
fpq-14292	158	53	,	,	PUNCT
fpq-14292	158	54	and	and	CCONJ
fpq-14292	158	55	in	in	ADP
fpq-14292	158	56	such	such	ADJ
fpq-14292	158	57	cases	case	NOUN
fpq-14292	158	58	these	these	DET
fpq-14292	158	59	models	model	NOUN
fpq-14292	158	60	reflect	reflect	VERB
fpq-14292	158	61	bias	bias	NOUN
fpq-14292	158	62	,	,	PUNCT
fpq-14292	158	63	as	as	SCONJ
fpq-14292	158	64	outlined	outline	VERB
fpq-14292	158	65	in	in	ADP
fpq-14292	158	66	section	section	NOUN
fpq-14292	158	67	2	2	NUM
fpq-14292	158	68	,	,	PUNCT
fpq-14292	158	69	whereas	whereas	SCONJ
fpq-14292	158	70	normatively	normatively	ADV
fpq-14292	158	71	correct	correct	ADJ
fpq-14292	158	72	models	model	NOUN
fpq-14292	158	73	or	or	CCONJ
fpq-14292	158	74	outputs	output	NOUN
fpq-14292	158	75	do	do	VERB
fpq-14292	158	76	not	not	PART
fpq-14292	158	77	(	(	PUNCT
fpq-14292	158	78	nor	nor	CCONJ
fpq-14292	158	79	,	,	PUNCT
fpq-14292	158	80	consequently	consequently	ADV
fpq-14292	158	81	,	,	PUNCT
fpq-14292	158	82	do	do	AUX
fpq-14292	158	83	they	they	PRON
fpq-14292	158	84	perpetuate	perpetuate	VERB
fpq-14292	158	85	bias	bias	NOUN
fpq-14292	158	86	)	)	PUNCT
fpq-14292	158	87	.	.	PUNCT
fpq-14292	159	1	a	a	DET
fpq-14292	159	2	system	system	NOUN
fpq-14292	159	3	might	might	AUX
fpq-14292	159	4	also	also	ADV
fpq-14292	159	5	be	be	AUX
fpq-14292	159	6	normatively	normatively	ADV
fpq-14292	159	7	correct	correct	ADJ
fpq-14292	159	8	in	in	ADP
fpq-14292	159	9	one	one	NUM
fpq-14292	159	10	regard	regard	NOUN
fpq-14292	159	11	(	(	PUNCT
fpq-14292	159	12	e.g.	e.g.	ADV
fpq-14292	159	13	,	,	PUNCT
fpq-14292	159	14	in	in	ADP
fpq-14292	159	15	not	not	PART
fpq-14292	159	16	giving	give	VERB
fpq-14292	159	17	sexist	sexist	ADJ
fpq-14292	159	18	outputs	output	NOUN
fpq-14292	159	19	)	)	PUNCT
fpq-14292	159	20	yet	yet	ADV
fpq-14292	159	21	incorrect	incorrect	ADJ
fpq-14292	159	22	in	in	ADP
fpq-14292	159	23	another	another	PRON
fpq-14292	159	24	(	(	PUNCT
fpq-14292	159	25	e.g.	e.g.	ADV
fpq-14292	159	26	,	,	PUNCT
fpq-14292	159	27	in	in	ADP
fpq-14292	159	28	giving	give	VERB
fpq-14292	159	29	racist	racist	NOUN
fpq-14292	159	30	outputs).18	outputs).18	PROPN
fpq-14292	159	31	additionally	additionally	ADV
fpq-14292	159	32	,	,	PUNCT
fpq-14292	159	33	normative	normative	ADJ
fpq-14292	159	34	correctness	correctness	NOUN
fpq-14292	159	35	can	can	AUX
fpq-14292	159	36	have	have	VERB
fpq-14292	159	37	various	various	ADJ
fpq-14292	159	38	goals	goal	NOUN
fpq-14292	159	39	.	.	PUNCT
fpq-14292	160	1	not	not	PART
fpq-14292	160	2	reflecting	reflect	VERB
fpq-14292	160	3	ethically	ethically	ADV
fpq-14292	160	4	problematic	problematic	ADJ
fpq-14292	160	5	biases	bias	NOUN
fpq-14292	160	6	might	might	AUX
fpq-14292	160	7	be	be	AUX
fpq-14292	160	8	one	one	NUM
fpq-14292	160	9	normatively	normatively	ADV
fpq-14292	160	10	correct	correct	ADJ
fpq-14292	160	11	goal	goal	NOUN
fpq-14292	160	12	,	,	PUNCT
fpq-14292	160	13	whereas	whereas	SCONJ
fpq-14292	160	14	other	other	ADJ
fpq-14292	160	15	goals	goal	NOUN
fpq-14292	160	16	might	might	AUX
fpq-14292	160	17	include	include	VERB
fpq-14292	160	18	promoting	promote	VERB
fpq-14292	160	19	outcomes	outcome	NOUN
fpq-14292	160	20	related	relate	VERB
fpq-14292	160	21	to	to	ADP
fpq-14292	160	22	social	social	ADJ
fpq-14292	160	23	justice	justice	NOUN
fpq-14292	160	24	.	.	PUNCT
fpq-14292	161	1	we	we	PRON
fpq-14292	161	2	acknowledge	acknowledge	VERB
fpq-14292	161	3	these	these	DET
fpq-14292	161	4	various	various	ADJ
fpq-14292	161	5	goals	goal	NOUN
fpq-14292	161	6	as	as	ADP
fpq-14292	161	7	being	be	AUX
fpq-14292	161	8	normative	normative	ADJ
fpq-14292	161	9	in	in	ADP
fpq-14292	161	10	a	a	DET
fpq-14292	161	11	relevant	relevant	ADJ
fpq-14292	161	12	sense	sense	NOUN
fpq-14292	161	13	.	.	PUNCT
fpq-14292	162	1	even	even	ADV
fpq-14292	162	2	so	so	ADV
fpq-14292	162	3	,	,	PUNCT
fpq-14292	162	4	our	our	PRON
fpq-14292	162	5	primary	primary	ADJ
fpq-14292	162	6	focus	focus	NOUN
fpq-14292	162	7	will	will	AUX
fpq-14292	162	8	be	be	AUX
fpq-14292	162	9	on	on	ADP
fpq-14292	162	10	the	the	DET
fpq-14292	162	11	aim	aim	NOUN
fpq-14292	162	12	of	of	ADP
fpq-14292	162	13	not	not	PART
fpq-14292	162	14	reflecting	reflect	VERB
fpq-14292	162	15	bias	bias	NOUN
fpq-14292	162	16	.	.	PUNCT
fpq-14292	163	1	to	to	PART
fpq-14292	163	2	see	see	VERB
fpq-14292	163	3	more	more	ADV
fpq-14292	163	4	clearly	clearly	ADV
fpq-14292	163	5	how	how	SCONJ
fpq-14292	163	6	descriptive	descriptive	ADJ
fpq-14292	163	7	and	and	CCONJ
fpq-14292	163	8	normative	normative	ADJ
fpq-14292	163	9	correctness	correctness	NOUN
fpq-14292	163	10	relate	relate	VERB
fpq-14292	163	11	to	to	PART
fpq-14292	163	12	bias	bias	VERB
fpq-14292	163	13	in	in	ADP
fpq-14292	163	14	nlp	nlp	NOUN
fpq-14292	163	15	systems	system	NOUN
fpq-14292	163	16	,	,	PUNCT
fpq-14292	163	17	let	let	VERB
fpq-14292	163	18	us	we	PRON
fpq-14292	163	19	consider	consider	VERB
fpq-14292	163	20	the	the	DET
fpq-14292	163	21	widespread	widespread	ADJ
fpq-14292	163	22	nlp	nlp	NOUN
fpq-14292	163	23	method	method	NOUN
fpq-14292	163	24	of	of	ADP
fpq-14292	163	25	word	word	NOUN
fpq-14292	163	26	-	-	PUNCT
fpq-14292	163	27	embeddings	embedding	NOUN
fpq-14292	163	28	,	,	PUNCT
fpq-14292	163	29	which	which	PRON
fpq-14292	163	30	we	we	PRON
fpq-14292	163	31	will	will	AUX
fpq-14292	163	32	now	now	ADV
fpq-14292	163	33	describe	describe	VERB
fpq-14292	163	34	.	.	PUNCT
fpq-14292	164	1	4	4	X
fpq-14292	164	2	.	.	X
fpq-14292	164	3	word	word	NOUN
fpq-14292	164	4	-	-	PUNCT
fpq-14292	164	5	embeddings	embedding	NOUN
fpq-14292	164	6	:	:	PUNCT
fpq-14292	164	7	descriptively	descriptively	ADV
fpq-14292	164	8	right	right	ADJ
fpq-14292	164	9	,	,	PUNCT
fpq-14292	164	10	normatively	normatively	ADV
fpq-14292	164	11	wrong	wrong	ADJ
fpq-14292	164	12	later	later	ADV
fpq-14292	164	13	,	,	PUNCT
fpq-14292	164	14	we	we	PRON
fpq-14292	164	15	will	will	AUX
fpq-14292	164	16	consider	consider	VERB
fpq-14292	164	17	applications	application	NOUN
fpq-14292	164	18	that	that	PRON
fpq-14292	164	19	rely	rely	VERB
fpq-14292	164	20	on	on	ADP
fpq-14292	164	21	large	large	ADJ
fpq-14292	164	22	language	language	NOUN
fpq-14292	164	23	models	model	NOUN
fpq-14292	164	24	.	.	PUNCT
fpq-14292	165	1	yet	yet	ADV
fpq-14292	165	2	,	,	PUNCT
fpq-14292	165	3	for	for	ADP
fpq-14292	165	4	now	now	ADV
fpq-14292	165	5	,	,	PUNCT
fpq-14292	165	6	in	in	ADP
fpq-14292	165	7	order	order	NOUN
fpq-14292	165	8	to	to	PART
fpq-14292	165	9	introduce	introduce	VERB
fpq-14292	165	10	the	the	DET
fpq-14292	165	11	idea	idea	NOUN
fpq-14292	165	12	of	of	ADP
fpq-14292	165	13	bias	bias	NOUN
fpq-14292	165	14	in	in	ADP
fpq-14292	165	15	nlp	nlp	NOUN
fpq-14292	165	16	models	model	NOUN
fpq-14292	165	17	,	,	PUNCT
fpq-14292	165	18	we	we	PRON
fpq-14292	165	19	will	will	AUX
fpq-14292	165	20	outline	outline	VERB
fpq-14292	165	21	the	the	DET
fpq-14292	165	22	simpler	simple	ADJ
fpq-14292	165	23	case	case	NOUN
fpq-14292	165	24	of	of	ADP
fpq-14292	165	25	word	word	NOUN
fpq-14292	165	26	-	-	PUNCT
fpq-14292	165	27	embeddings	embedding	NOUN
fpq-14292	165	28	.	.	PUNCT
fpq-14292	166	1	word	word	NOUN
fpq-14292	166	2	-	-	PUNCT
fpq-14292	166	3	embeddings	embedding	NOUN
fpq-14292	166	4	are	be	AUX
fpq-14292	166	5	learned	learn	VERB
fpq-14292	166	6	representations	representation	NOUN
fpq-14292	166	7	of	of	ADP
fpq-14292	166	8	words	word	NOUN
fpq-14292	166	9	that	that	PRON
fpq-14292	166	10	aim	aim	VERB
fpq-14292	166	11	to	to	PART
fpq-14292	166	12	approximate	approximate	VERB
fpq-14292	166	13	the	the	DET
fpq-14292	166	14	semantic	semantic	ADJ
fpq-14292	166	15	relationships	relationship	NOUN
fpq-14292	166	16	between	between	ADP
fpq-14292	166	17	words	word	NOUN
fpq-14292	166	18	through	through	ADP
fpq-14292	166	19	mathematical	mathematical	ADJ
fpq-14292	166	20	relationships	relationship	NOUN
fpq-14292	166	21	between	between	ADP
fpq-14292	166	22	vectors	vector	NOUN
fpq-14292	166	23	(	(	PUNCT
fpq-14292	166	24	mikolov	mikolov	NOUN
fpq-14292	166	25	et	et	PROPN
fpq-14292	166	26	al	al	PROPN
fpq-14292	166	27	.	.	PROPN
fpq-14292	166	28	2013	2013	NUM
fpq-14292	166	29	)	)	PUNCT
fpq-14292	166	30	.	.	PUNCT
fpq-14292	167	1	a	a	DET
fpq-14292	167	2	word	word	NOUN
fpq-14292	167	3	-	-	PUNCT
fpq-14292	167	4	embeddings	embedding	NOUN
fpq-14292	167	5	model	model	NOUN
fpq-14292	167	6	learns	learn	VERB
fpq-14292	167	7	from	from	ADP
fpq-14292	167	8	vast	vast	ADJ
fpq-14292	167	9	language	language	NOUN
fpq-14292	167	10	corpora	corpora	NOUN
fpq-14292	167	11	produced	produce	VERB
fpq-14292	167	12	by	by	ADP
fpq-14292	167	13	us	we	PRON
fpq-14292	167	14	,	,	PUNCT
fpq-14292	167	15	which	which	PRON
fpq-14292	167	16	are	be	AUX
fpq-14292	167	17	therefore	therefore	ADV
fpq-14292	167	18	examples	example	NOUN
fpq-14292	167	19	of	of	ADP
fpq-14292	167	20	how	how	SCONJ
fpq-14292	167	21	we	we	PRON
fpq-14292	167	22	use	use	VERB
fpq-14292	167	23	language	language	NOUN
fpq-14292	167	24	.	.	PUNCT
fpq-14292	168	1	by	by	ADP
fpq-14292	168	2	definition	definition	NOUN
fpq-14292	168	3	,	,	PUNCT
fpq-14292	168	4	a	a	DET
fpq-14292	168	5	good	good	ADJ
fpq-14292	168	6	set	set	NOUN
fpq-14292	168	7	of	of	ADP
fpq-14292	168	8	representations	representation	NOUN
fpq-14292	168	9	will	will	AUX
fpq-14292	168	10	accurately	accurately	ADV
fpq-14292	168	11	reflect	reflect	VERB
fpq-14292	168	12	how	how	SCONJ
fpq-14292	168	13	language	language	NOUN
fpq-14292	168	14	is	be	AUX
fpq-14292	168	15	,	,	PUNCT
fpq-14292	168	16	as	as	ADP
fpq-14292	168	17	a	a	DET
fpq-14292	168	18	matter	matter	NOUN
fpq-14292	168	19	of	of	ADP
fpq-14292	168	20	fact	fact	NOUN
fpq-14292	168	21	,	,	PUNCT
fpq-14292	168	22	used	use	VERB
fpq-14292	168	23	by	by	ADP
fpq-14292	168	24	the	the	DET
fpq-14292	168	25	sampled	sample	VERB
fpq-14292	168	26	people	people	NOUN
fpq-14292	168	27	.	.	PUNCT
fpq-14292	169	1	in	in	ADP
fpq-14292	169	2	this	this	DET
fpq-14292	169	3	sense	sense	NOUN
fpq-14292	169	4	,	,	PUNCT
fpq-14292	169	5	word	word	NOUN
fpq-14292	169	6	-	-	PUNCT
fpq-14292	169	7	embeddings	embedding	NOUN
fpq-14292	169	8	do	do	VERB
fpq-14292	169	9	a	a	DET
fpq-14292	169	10	remarkably	remarkably	ADV
fpq-14292	169	11	good	good	ADJ
fpq-14292	169	12	job	job	NOUN
fpq-14292	169	13	of	of	ADP
fpq-14292	169	14	getting	get	VERB
fpq-14292	169	15	things	thing	NOUN
fpq-14292	169	16	descriptively	descriptively	ADV
fpq-14292	169	17	right	right	ADJ
fpq-14292	169	18	,	,	PUNCT
fpq-14292	169	19	since	since	SCONJ
fpq-14292	169	20	17	17	NUM
fpq-14292	169	21	normatively	normatively	ADV
fpq-14292	169	22	incorrect	incorrect	ADJ
fpq-14292	169	23	outputs	output	NOUN
fpq-14292	169	24	need	need	AUX
fpq-14292	169	25	not	not	PART
fpq-14292	169	26	be	be	AUX
fpq-14292	169	27	descriptively	descriptively	ADV
fpq-14292	169	28	accurate	accurate	ADJ
fpq-14292	169	29	,	,	PUNCT
fpq-14292	169	30	although	although	SCONJ
fpq-14292	169	31	often	often	ADV
fpq-14292	169	32	they	they	PRON
fpq-14292	169	33	are	be	AUX
fpq-14292	169	34	(	(	PUNCT
fpq-14292	169	35	see	see	VERB
fpq-14292	169	36	section	section	NOUN
fpq-14292	169	37	6.3	6.3	NUM
fpq-14292	169	38	)	)	PUNCT
fpq-14292	169	39	.	.	PUNCT
fpq-14292	170	1	18	18	NUM
fpq-14292	170	2	arguably	arguably	ADV
fpq-14292	170	3	,	,	PUNCT
fpq-14292	170	4	the	the	DET
fpq-14292	170	5	best	good	ADJ
fpq-14292	170	6	way	way	NOUN
fpq-14292	170	7	to	to	PART
fpq-14292	170	8	ensure	ensure	VERB
fpq-14292	170	9	that	that	SCONJ
fpq-14292	170	10	language	language	NOUN
fpq-14292	170	11	use	use	NOUN
fpq-14292	170	12	is	be	AUX
fpq-14292	170	13	as	as	SCONJ
fpq-14292	170	14	it	it	PRON
fpq-14292	170	15	should	should	AUX
fpq-14292	170	16	be	be	AUX
fpq-14292	170	17	and	and	CCONJ
fpq-14292	170	18	that	that	DET
fpq-14292	170	19	language	language	NOUN
fpq-14292	170	20	models	model	NOUN
fpq-14292	170	21	trained	train	VERB
fpq-14292	170	22	on	on	ADP
fpq-14292	170	23	such	such	ADJ
fpq-14292	170	24	language	language	NOUN
fpq-14292	170	25	use	use	NOUN
fpq-14292	170	26	are	be	AUX
fpq-14292	170	27	normatively	normatively	ADV
fpq-14292	170	28	correct	correct	ADJ
fpq-14292	170	29	would	would	AUX
fpq-14292	170	30	be	be	AUX
fpq-14292	170	31	to	to	PART
fpq-14292	170	32	have	have	VERB
fpq-14292	170	33	the	the	DET
fpq-14292	170	34	nonlinguistic	nonlinguistic	ADJ
fpq-14292	170	35	part	part	NOUN
fpq-14292	170	36	of	of	ADP
fpq-14292	170	37	the	the	DET
fpq-14292	170	38	world	world	NOUN
fpq-14292	170	39	that	that	SCONJ
fpq-14292	170	40	the	the	DET
fpq-14292	170	41	language	language	NOUN
fpq-14292	170	42	use	use	NOUN
fpq-14292	170	43	aims	aim	VERB
fpq-14292	170	44	to	to	PART
fpq-14292	170	45	describe	describe	VERB
fpq-14292	170	46	—	—	PUNCT
fpq-14292	170	47	e.g.	e.g.	ADV
fpq-14292	170	48	,	,	PUNCT
fpq-14292	170	49	the	the	DET
fpq-14292	170	50	proportion	proportion	NOUN
fpq-14292	170	51	of	of	ADP
fpq-14292	170	52	women	woman	NOUN
fpq-14292	170	53	and	and	CCONJ
fpq-14292	170	54	men	man	NOUN
fpq-14292	170	55	who	who	PRON
fpq-14292	170	56	are	be	AUX
fpq-14292	170	57	computer	computer	NOUN
fpq-14292	170	58	programmers	programmer	NOUN
fpq-14292	170	59	—	—	PUNCT
fpq-14292	170	60	be	be	VERB
fpq-14292	170	61	as	as	SCONJ
fpq-14292	170	62	it	it	PRON
fpq-14292	170	63	should	should	AUX
fpq-14292	170	64	be	be	AUX
fpq-14292	170	65	.	.	PUNCT
fpq-14292	171	1	after	after	ADV
fpq-14292	171	2	all	all	ADV
fpq-14292	171	3	,	,	PUNCT
fpq-14292	171	4	people	people	NOUN
fpq-14292	171	5	associate	associate	VERB
fpq-14292	171	6	“	"	PUNCT
fpq-14292	171	7	he	he	PRON
fpq-14292	171	8	”	"	PUNCT
fpq-14292	171	9	with	with	ADP
fpq-14292	171	10	programmers	programmer	NOUN
fpq-14292	171	11	more	more	ADJ
fpq-14292	171	12	than	than	ADP
fpq-14292	171	13	“	"	PUNCT
fpq-14292	171	14	she	she	PRON
fpq-14292	171	15	”	"	PUNCT
fpq-14292	171	16	because	because	SCONJ
fpq-14292	171	17	traditionally	traditionally	ADV
fpq-14292	171	18	most	most	ADJ
fpq-14292	171	19	programmers	programmer	NOUN
fpq-14292	171	20	have	have	AUX
fpq-14292	171	21	tended	tend	VERB
fpq-14292	171	22	to	to	PART
fpq-14292	171	23	be	be	AUX
fpq-14292	171	24	men	man	NOUN
fpq-14292	171	25	.	.	PUNCT
fpq-14292	172	1	yet	yet	ADV
fpq-14292	172	2	over	over	ADP
fpq-14292	172	3	time	time	NOUN
fpq-14292	172	4	,	,	PUNCT
fpq-14292	172	5	if	if	SCONJ
fpq-14292	172	6	more	more	ADJ
fpq-14292	172	7	programmers	programmer	NOUN
fpq-14292	172	8	were	be	AUX
fpq-14292	172	9	women	woman	NOUN
fpq-14292	172	10	,	,	PUNCT
fpq-14292	172	11	that	that	PRON
fpq-14292	172	12	would	would	AUX
fpq-14292	172	13	be	be	AUX
fpq-14292	172	14	less	less	ADV
fpq-14292	172	15	true	true	ADJ
fpq-14292	172	16	.	.	PUNCT
fpq-14292	173	1	normative	normative	ADJ
fpq-14292	173	2	correctness	correctness	NOUN
fpq-14292	173	3	would	would	AUX
fpq-14292	173	4	be	be	AUX
fpq-14292	173	5	descriptively	descriptively	ADV
fpq-14292	173	6	accurate	accurate	ADJ
fpq-14292	173	7	.	.	PUNCT
fpq-14292	174	1	feminist	feminist	PROPN
fpq-14292	174	2	philosophy	philosophy	PROPN
fpq-14292	174	3	quarterly	quarterly	ADV
fpq-14292	174	4	,	,	PUNCT
fpq-14292	174	5	2022	2022	NUM
fpq-14292	174	6	,	,	PUNCT
fpq-14292	174	7	vol	vol	NOUN
fpq-14292	174	8	.	.	NOUN
fpq-14292	174	9	8	8	NUM
fpq-14292	174	10	,	,	PUNCT
fpq-14292	174	11	iss	iss	PROPN
fpq-14292	174	12	.	.	PUNCT
fpq-14292	174	13	3/4	3/4	NUM
fpq-14292	174	14	,	,	PUNCT
fpq-14292	174	15	article	article	NOUN
fpq-14292	174	16	1	1	NUM
fpq-14292	174	17	published	publish	VERB
fpq-14292	174	18	by	by	ADP
fpq-14292	174	19	scholarship@western	scholarship@western	PROPN
fpq-14292	174	20	,	,	PUNCT
fpq-14292	174	21	2022	2022	NUM
fpq-14292	174	22	10	10	NUM
fpq-14292	174	23	they	they	PRON
fpq-14292	174	24	are	be	AUX
fpq-14292	174	25	usefully	usefully	ADV
fpq-14292	174	26	deployed	deploy	VERB
fpq-14292	174	27	for	for	ADP
fpq-14292	174	28	many	many	ADJ
fpq-14292	174	29	purposes	purpose	NOUN
fpq-14292	174	30	—	—	PUNCT
fpq-14292	174	31	for	for	ADP
fpq-14292	174	32	example	example	NOUN
fpq-14292	174	33	,	,	PUNCT
fpq-14292	174	34	to	to	PART
fpq-14292	174	35	successfully	successfully	ADV
fpq-14292	174	36	resolve	resolve	VERB
fpq-14292	174	37	analogy	analogy	NOUN
fpq-14292	174	38	tasks	task	NOUN
fpq-14292	174	39	.	.	PUNCT
fpq-14292	175	1	yet	yet	ADV
fpq-14292	175	2	there	there	PRON
fpq-14292	175	3	is	be	VERB
fpq-14292	175	4	a	a	DET
fpq-14292	175	5	sense	sense	NOUN
fpq-14292	175	6	in	in	ADP
fpq-14292	175	7	which	which	PRON
fpq-14292	175	8	word	word	NOUN
fpq-14292	175	9	-	-	PUNCT
fpq-14292	175	10	embeddings	embedding	NOUN
fpq-14292	175	11	get	get	VERB
fpq-14292	175	12	things	thing	NOUN
fpq-14292	175	13	normatively	normatively	ADV
fpq-14292	175	14	wrong	wrong	ADJ
fpq-14292	175	15	.	.	PUNCT
fpq-14292	176	1	in	in	ADP
fpq-14292	176	2	their	their	PRON
fpq-14292	176	3	paper	paper	NOUN
fpq-14292	176	4	,	,	PUNCT
fpq-14292	176	5	“	"	PUNCT
fpq-14292	176	6	man	man	NOUN
fpq-14292	176	7	is	be	AUX
fpq-14292	176	8	to	to	PART
fpq-14292	176	9	computer	computer	NOUN
fpq-14292	176	10	programmer	programmer	NOUN
fpq-14292	176	11	as	as	SCONJ
fpq-14292	176	12	woman	woman	NOUN
fpq-14292	176	13	is	be	AUX
fpq-14292	176	14	to	to	ADP
fpq-14292	176	15	homemaker	homemaker	NOUN
fpq-14292	176	16	,	,	PUNCT
fpq-14292	176	17	”	"	PUNCT
fpq-14292	176	18	researchers	researcher	NOUN
fpq-14292	176	19	at	at	ADP
fpq-14292	176	20	boston	boston	PROPN
fpq-14292	176	21	university	university	PROPN
fpq-14292	176	22	and	and	CCONJ
fpq-14292	176	23	microsoft	microsoft	PROPN
fpq-14292	176	24	research	research	PROPN
fpq-14292	176	25	(	(	PUNCT
fpq-14292	176	26	bolukbasi	bolukbasi	NOUN
fpq-14292	176	27	et	et	PROPN
fpq-14292	176	28	al	al	PROPN
fpq-14292	176	29	.	.	PROPN
fpq-14292	176	30	2016	2016	NUM
fpq-14292	176	31	)	)	PUNCT
fpq-14292	176	32	demonstrated	demonstrate	VERB
fpq-14292	176	33	that	that	SCONJ
fpq-14292	176	34	pretrained	pretraine	VERB
fpq-14292	176	35	models	model	NOUN
fpq-14292	176	36	such	such	ADJ
fpq-14292	176	37	as	as	ADP
fpq-14292	176	38	word2vec	word2vec	PRON
fpq-14292	176	39	seem	seem	VERB
fpq-14292	176	40	sexist	sexist	ADJ
fpq-14292	176	41	in	in	ADP
fpq-14292	176	42	certain	certain	ADJ
fpq-14292	176	43	ways	way	NOUN
fpq-14292	176	44	.	.	PUNCT
fpq-14292	177	1	just	just	ADV
fpq-14292	177	2	as	as	SCONJ
fpq-14292	177	3	the	the	DET
fpq-14292	177	4	analogy	analogy	NOUN
fpq-14292	177	5	relationships	relationship	NOUN
fpq-14292	177	6	“	"	PUNCT
fpq-14292	177	7	man	man	NOUN
fpq-14292	177	8	is	be	AUX
fpq-14292	177	9	to	to	ADP
fpq-14292	177	10	woman	woman	NOUN
fpq-14292	177	11	as	as	ADP
fpq-14292	177	12	king	king	NOUN
fpq-14292	177	13	is	be	AUX
fpq-14292	177	14	to	to	ADP
fpq-14292	177	15	queen	queen	PROPN
fpq-14292	177	16	”	"	PUNCT
fpq-14292	177	17	and	and	CCONJ
fpq-14292	177	18	“	"	PUNCT
fpq-14292	177	19	sister	sister	NOUN
fpq-14292	177	20	is	be	AUX
fpq-14292	177	21	to	to	ADP
fpq-14292	177	22	woman	woman	NOUN
fpq-14292	177	23	as	as	SCONJ
fpq-14292	177	24	brother	brother	NOUN
fpq-14292	177	25	is	be	AUX
fpq-14292	177	26	to	to	ADP
fpq-14292	177	27	man	man	NOUN
fpq-14292	177	28	”	"	PUNCT
fpq-14292	177	29	were	be	AUX
fpq-14292	177	30	captured	capture	VERB
fpq-14292	177	31	by	by	ADP
fpq-14292	177	32	the	the	DET
fpq-14292	177	33	models	model	NOUN
fpq-14292	177	34	,	,	PUNCT
fpq-14292	177	35	so	so	ADV
fpq-14292	177	36	too	too	ADV
fpq-14292	177	37	were	be	AUX
fpq-14292	177	38	the	the	DET
fpq-14292	177	39	sexist	sexist	ADJ
fpq-14292	177	40	relationships	relationship	NOUN
fpq-14292	177	41	“	"	PUNCT
fpq-14292	177	42	man	man	NOUN
fpq-14292	177	43	is	be	AUX
fpq-14292	177	44	to	to	PART
fpq-14292	177	45	computer	computer	NOUN
fpq-14292	177	46	programmer	programmer	NOUN
fpq-14292	177	47	as	as	SCONJ
fpq-14292	177	48	woman	woman	NOUN
fpq-14292	177	49	is	be	AUX
fpq-14292	177	50	to	to	ADP
fpq-14292	177	51	homemaker	homemaker	NOUN
fpq-14292	177	52	”	"	PUNCT
fpq-14292	177	53	and	and	CCONJ
fpq-14292	177	54	“	"	PUNCT
fpq-14292	177	55	father	father	NOUN
fpq-14292	177	56	is	be	AUX
fpq-14292	177	57	to	to	PART
fpq-14292	177	58	doctor	doctor	VERB
fpq-14292	177	59	as	as	SCONJ
fpq-14292	177	60	mother	mother	NOUN
fpq-14292	177	61	is	be	AUX
fpq-14292	177	62	to	to	PART
fpq-14292	177	63	nurse	nurse	VERB
fpq-14292	177	64	.	.	PUNCT
fpq-14292	178	1	”19	”19	VERB
fpq-14292	178	2	the	the	DET
fpq-14292	178	3	difficulty	difficulty	NOUN
fpq-14292	178	4	is	be	AUX
fpq-14292	178	5	that	that	PRON
fpq-14292	178	6	to	to	PART
fpq-14292	178	7	work	work	VERB
fpq-14292	178	8	successfully	successfully	ADV
fpq-14292	178	9	,	,	PUNCT
fpq-14292	178	10	nlp	nlp	NOUN
fpq-14292	178	11	systems	system	NOUN
fpq-14292	178	12	relying	rely	VERB
fpq-14292	178	13	on	on	ADP
fpq-14292	178	14	wordembeddings	wordembedding	NOUN
fpq-14292	178	15	often	often	ADV
fpq-14292	178	16	have	have	VERB
fpq-14292	178	17	to	to	PART
fpq-14292	178	18	learn	learn	VERB
fpq-14292	178	19	the	the	DET
fpq-14292	178	20	biases	bias	NOUN
fpq-14292	178	21	that	that	PRON
fpq-14292	178	22	exist	exist	VERB
fpq-14292	178	23	in	in	ADP
fpq-14292	178	24	the	the	DET
fpq-14292	178	25	corpora	corpora	NOUN
fpq-14292	178	26	on	on	ADP
fpq-14292	178	27	which	which	PRON
fpq-14292	178	28	they	they	PRON
fpq-14292	178	29	are	be	AUX
fpq-14292	178	30	trained	train	VERB
fpq-14292	178	31	(	(	PUNCT
fpq-14292	178	32	caliskan	caliskan	PROPN
fpq-14292	178	33	,	,	PUNCT
fpq-14292	178	34	bryson	bryson	PROPN
fpq-14292	178	35	,	,	PUNCT
fpq-14292	178	36	and	and	CCONJ
fpq-14292	178	37	narayanan	narayanan	PROPN
fpq-14292	178	38	2017	2017	NUM
fpq-14292	178	39	,	,	PUNCT
fpq-14292	178	40	186	186	NUM
fpq-14292	178	41	)	)	PUNCT
fpq-14292	178	42	.	.	PUNCT
fpq-14292	179	1	these	these	DET
fpq-14292	179	2	biases	bias	NOUN
fpq-14292	179	3	are	be	AUX
fpq-14292	179	4	expressed	express	VERB
fpq-14292	179	5	by	by	ADP
fpq-14292	179	6	us	we	PRON
fpq-14292	179	7	in	in	ADP
fpq-14292	179	8	the	the	DET
fpq-14292	179	9	corpora	corpora	NOUN
fpq-14292	179	10	,	,	PUNCT
fpq-14292	179	11	and	and	CCONJ
fpq-14292	179	12	to	to	PART
fpq-14292	179	13	learn	learn	VERB
fpq-14292	179	14	the	the	DET
fpq-14292	179	15	relationships	relationship	NOUN
fpq-14292	179	16	that	that	PRON
fpq-14292	179	17	actually	actually	ADV
fpq-14292	179	18	exist	exist	VERB
fpq-14292	179	19	between	between	ADP
fpq-14292	179	20	words	word	NOUN
fpq-14292	179	21	in	in	ADP
fpq-14292	179	22	our	our	PRON
fpq-14292	179	23	uses	use	NOUN
fpq-14292	179	24	of	of	ADP
fpq-14292	179	25	language	language	NOUN
fpq-14292	179	26	,	,	PUNCT
fpq-14292	179	27	the	the	DET
fpq-14292	179	28	models	model	NOUN
fpq-14292	179	29	must	must	AUX
fpq-14292	179	30	learn	learn	VERB
fpq-14292	179	31	biased	biased	ADJ
fpq-14292	179	32	relationships	relationship	NOUN
fpq-14292	179	33	—	—	PUNCT
fpq-14292	179	34	including	include	VERB
fpq-14292	179	35	sexist	sexist	ADJ
fpq-14292	179	36	relationships	relationship	NOUN
fpq-14292	179	37	.	.	PUNCT
fpq-14292	180	1	biases	bias	NOUN
fpq-14292	180	2	in	in	ADP
fpq-14292	180	3	the	the	DET
fpq-14292	180	4	corpora	corpora	NOUN
fpq-14292	180	5	on	on	ADP
fpq-14292	180	6	which	which	PRON
fpq-14292	180	7	the	the	DET
fpq-14292	180	8	models	model	NOUN
fpq-14292	180	9	are	be	AUX
fpq-14292	180	10	trained	train	VERB
fpq-14292	180	11	will	will	AUX
fpq-14292	180	12	thus	thus	ADV
fpq-14292	180	13	naturally	naturally	ADV
fpq-14292	180	14	be	be	AUX
fpq-14292	180	15	captured	capture	VERB
fpq-14292	180	16	in	in	ADP
fpq-14292	180	17	the	the	DET
fpq-14292	180	18	geometry	geometry	NOUN
fpq-14292	180	19	of	of	ADP
fpq-14292	180	20	the	the	DET
fpq-14292	180	21	word	word	NOUN
fpq-14292	180	22	-	-	PUNCT
fpq-14292	180	23	embeddings	embedding	NOUN
fpq-14292	180	24	vector	vector	NOUN
fpq-14292	180	25	space	space	NOUN
fpq-14292	180	26	.	.	PUNCT
fpq-14292	181	1	worse	bad	ADJ
fpq-14292	181	2	,	,	PUNCT
fpq-14292	181	3	according	accord	VERB
fpq-14292	181	4	to	to	ADP
fpq-14292	181	5	bolukbasi	bolukbasi	NOUN
fpq-14292	181	6	and	and	CCONJ
fpq-14292	181	7	colleagues	colleague	NOUN
fpq-14292	181	8	’	'	PUNCT
fpq-14292	181	9	assessment	assessment	NOUN
fpq-14292	181	10	,	,	PUNCT
fpq-14292	181	11	“	"	PUNCT
fpq-14292	181	12	the	the	DET
fpq-14292	181	13	blind	blind	ADJ
fpq-14292	181	14	application	application	NOUN
fpq-14292	181	15	of	of	ADP
fpq-14292	181	16	machine	machine	NOUN
fpq-14292	181	17	learning	learning	NOUN
fpq-14292	181	18	runs	run	VERB
fpq-14292	181	19	the	the	DET
fpq-14292	181	20	risk	risk	NOUN
fpq-14292	181	21	of	of	ADP
fpq-14292	181	22	amplifying	amplify	VERB
fpq-14292	181	23	[	[	X
fpq-14292	181	24	emphasis	emphasis	NOUN
fpq-14292	181	25	added	add	VERB
fpq-14292	181	26	]	]	PUNCT
fpq-14292	181	27	biases	bias	NOUN
fpq-14292	181	28	present	present	ADJ
fpq-14292	181	29	in	in	ADP
fpq-14292	181	30	data	datum	NOUN
fpq-14292	181	31	”	"	PUNCT
fpq-14292	181	32	(	(	PUNCT
fpq-14292	181	33	bolukbasi	bolukbasi	NOUN
fpq-14292	181	34	et	et	PROPN
fpq-14292	181	35	al	al	PROPN
fpq-14292	181	36	.	.	PUNCT
fpq-14292	182	1	2016).20	2016).20	NUM
fpq-14292	182	2	still	still	ADV
fpq-14292	182	3	worse	bad	ADJ
fpq-14292	182	4	,	,	PUNCT
fpq-14292	182	5	word	word	NOUN
fpq-14292	182	6	-	-	PUNCT
fpq-14292	182	7	embeddings	embedding	NOUN
fpq-14292	182	8	models	model	NOUN
fpq-14292	182	9	often	often	ADV
fpq-14292	182	10	get	get	VERB
fpq-14292	182	11	things	thing	NOUN
fpq-14292	182	12	normatively	normatively	ADV
fpq-14292	182	13	wrong	wrong	ADJ
fpq-14292	182	14	precisely	precisely	ADV
fpq-14292	182	15	because	because	SCONJ
fpq-14292	182	16	—	—	PUNCT
fpq-14292	182	17	and	and	CCONJ
fpq-14292	182	18	to	to	ADP
fpq-14292	182	19	the	the	DET
fpq-14292	182	20	extent	extent	NOUN
fpq-14292	182	21	to	to	PART
fpq-14292	182	22	which	which	PRON
fpq-14292	182	23	—	—	PUNCT
fpq-14292	182	24	they	they	PRON
fpq-14292	182	25	get	get	VERB
fpq-14292	182	26	things	thing	NOUN
fpq-14292	182	27	descriptively	descriptively	ADV
fpq-14292	182	28	right	right	ADJ
fpq-14292	182	29	about	about	ADP
fpq-14292	182	30	people	people	NOUN
fpq-14292	182	31	’s	’s	PART
fpq-14292	182	32	language	language	NOUN
fpq-14292	182	33	use	use	NOUN
fpq-14292	182	34	.	.	PUNCT
fpq-14292	183	1	that	that	PRON
fpq-14292	183	2	is	is	ADV
fpq-14292	183	3	,	,	PUNCT
fpq-14292	183	4	the	the	DET
fpq-14292	183	5	success	success	NOUN
fpq-14292	183	6	of	of	ADP
fpq-14292	183	7	systems	system	NOUN
fpq-14292	183	8	using	use	VERB
fpq-14292	183	9	word	word	NOUN
fpq-14292	183	10	-	-	PUNCT
fpq-14292	183	11	embeddings	embedding	NOUN
fpq-14292	183	12	can	can	AUX
fpq-14292	183	13	require	require	VERB
fpq-14292	183	14	reproducing	reproduce	VERB
fpq-14292	183	15	bias	bias	NOUN
fpq-14292	183	16	,	,	PUNCT
fpq-14292	183	17	as	as	ADP
fpq-14292	183	18	when	when	SCONJ
fpq-14292	183	19	they	they	PRON
fpq-14292	183	20	solve	solve	VERB
fpq-14292	183	21	analogy	analogy	NOUN
fpq-14292	183	22	tasks	task	NOUN
fpq-14292	183	23	like	like	ADP
fpq-14292	183	24	“	"	PUNCT
fpq-14292	183	25	man	man	NOUN
fpq-14292	183	26	is	be	AUX
fpq-14292	183	27	to	to	ADP
fpq-14292	183	28	19	19	NUM
fpq-14292	183	29	for	for	ADP
fpq-14292	183	30	criticism	criticism	NOUN
fpq-14292	183	31	of	of	ADP
fpq-14292	183	32	bolukbasi	bolukbasi	NOUN
fpq-14292	183	33	et	et	PROPN
fpq-14292	183	34	al	al	PROPN
fpq-14292	183	35	.	.	PUNCT
fpq-14292	184	1	(	(	PUNCT
fpq-14292	184	2	2016	2016	NUM
fpq-14292	184	3	)	)	PUNCT
fpq-14292	184	4	,	,	PUNCT
fpq-14292	184	5	see	see	VERB
fpq-14292	184	6	schluter	schluter	PROPN
fpq-14292	184	7	(	(	PUNCT
fpq-14292	184	8	2018	2018	NUM
fpq-14292	184	9	)	)	PUNCT
fpq-14292	184	10	,	,	PUNCT
fpq-14292	184	11	nissim	nissim	PROPN
fpq-14292	184	12	,	,	PUNCT
fpq-14292	184	13	van	van	PROPN
fpq-14292	184	14	noord	noord	PROPN
fpq-14292	184	15	,	,	PUNCT
fpq-14292	184	16	and	and	CCONJ
fpq-14292	184	17	van	van	PROPN
fpq-14292	184	18	der	der	PROPN
fpq-14292	184	19	goot	goot	NOUN
fpq-14292	184	20	(	(	PUNCT
fpq-14292	184	21	2019	2019	NUM
fpq-14292	184	22	)	)	PUNCT
fpq-14292	184	23	,	,	PUNCT
fpq-14292	184	24	and	and	CCONJ
fpq-14292	184	25	footnote	footnote	VERB
fpq-14292	184	26	24	24	NUM
fpq-14292	184	27	below	below	ADV
fpq-14292	184	28	.	.	PUNCT
fpq-14292	185	1	20	20	NUM
fpq-14292	185	2	bias	bias	NOUN
fpq-14292	185	3	is	be	AUX
fpq-14292	185	4	amplified	amplify	VERB
fpq-14292	185	5	in	in	ADP
fpq-14292	185	6	two	two	NUM
fpq-14292	185	7	ways	way	NOUN
fpq-14292	185	8	.	.	PUNCT
fpq-14292	186	1	first	first	ADJ
fpq-14292	186	2	,	,	PUNCT
fpq-14292	186	3	sexist	sexist	ADJ
fpq-14292	186	4	bias	bias	NOUN
fpq-14292	186	5	(	(	PUNCT
fpq-14292	186	6	for	for	ADP
fpq-14292	186	7	example	example	NOUN
fpq-14292	186	8	)	)	PUNCT
fpq-14292	186	9	in	in	ADP
fpq-14292	186	10	a	a	DET
fpq-14292	186	11	language	language	NOUN
fpq-14292	186	12	model	model	NOUN
fpq-14292	186	13	is	be	AUX
fpq-14292	186	14	acquired	acquire	VERB
fpq-14292	186	15	from	from	ADP
fpq-14292	186	16	training	train	VERB
fpq-14292	186	17	data	datum	NOUN
fpq-14292	186	18	that	that	PRON
fpq-14292	186	19	include	include	VERB
fpq-14292	186	20	expressions	expression	NOUN
fpq-14292	186	21	of	of	ADP
fpq-14292	186	22	sexism	sexism	NOUN
fpq-14292	186	23	from	from	ADP
fpq-14292	186	24	(	(	PUNCT
fpq-14292	186	25	p	p	ADJ
fpq-14292	186	26	number	number	NOUN
fpq-14292	186	27	of	of	ADP
fpq-14292	186	28	)	)	PUNCT
fpq-14292	186	29	individuals	individual	NOUN
fpq-14292	186	30	expressing	express	VERB
fpq-14292	186	31	the	the	DET
fpq-14292	186	32	bias	bias	NOUN
fpq-14292	186	33	.	.	PUNCT
fpq-14292	187	1	this	this	DET
fpq-14292	187	2	model	model	NOUN
fpq-14292	187	3	may	may	AUX
fpq-14292	187	4	then	then	ADV
fpq-14292	187	5	be	be	AUX
fpq-14292	187	6	used	use	VERB
fpq-14292	187	7	for	for	ADP
fpq-14292	187	8	various	various	ADJ
fpq-14292	187	9	purposes	purpose	NOUN
fpq-14292	187	10	in	in	ADP
fpq-14292	187	11	the	the	DET
fpq-14292	187	12	world	world	NOUN
fpq-14292	187	13	such	such	ADJ
fpq-14292	187	14	that	that	SCONJ
fpq-14292	187	15	its	its	PRON
fpq-14292	187	16	outputs	output	NOUN
fpq-14292	187	17	reach	reach	VERB
fpq-14292	187	18	many	many	ADJ
fpq-14292	187	19	more	more	ADJ
fpq-14292	187	20	people	people	NOUN
fpq-14292	187	21	(	(	PUNCT
fpq-14292	187	22	p	p	NOUN
fpq-14292	187	23	+	+	NOUN
fpq-14292	187	24	n	n	CCONJ
fpq-14292	187	25	)	)	PUNCT
fpq-14292	187	26	than	than	ADP
fpq-14292	187	27	those	those	DET
fpq-14292	187	28	people	people	NOUN
fpq-14292	187	29	(	(	PUNCT
fpq-14292	187	30	p	p	NOUN
fpq-14292	187	31	)	)	PUNCT
fpq-14292	187	32	who	who	PRON
fpq-14292	187	33	expressed	express	VERB
fpq-14292	187	34	the	the	DET
fpq-14292	187	35	biases	bias	NOUN
fpq-14292	187	36	that	that	PRON
fpq-14292	187	37	influenced	influence	VERB
fpq-14292	187	38	the	the	DET
fpq-14292	187	39	training	training	NOUN
fpq-14292	187	40	data	datum	NOUN
fpq-14292	187	41	.	.	PUNCT
fpq-14292	188	1	mere	mere	ADJ
fpq-14292	188	2	repetition	repetition	NOUN
fpq-14292	188	3	is	be	AUX
fpq-14292	188	4	widely	widely	ADV
fpq-14292	188	5	known	know	VERB
fpq-14292	188	6	to	to	PART
fpq-14292	188	7	increase	increase	VERB
fpq-14292	188	8	people	people	NOUN
fpq-14292	188	9	’s	’s	PART
fpq-14292	188	10	tendency	tendency	NOUN
fpq-14292	188	11	to	to	PART
fpq-14292	188	12	believe	believe	VERB
fpq-14292	188	13	claims	claim	NOUN
fpq-14292	188	14	or	or	CCONJ
fpq-14292	188	15	to	to	PART
fpq-14292	188	16	reproduce	reproduce	VERB
fpq-14292	188	17	bias	bias	NOUN
fpq-14292	188	18	,	,	PUNCT
fpq-14292	188	19	including	include	VERB
fpq-14292	188	20	in	in	ADP
fpq-14292	188	21	written	write	VERB
fpq-14292	188	22	language	language	NOUN
fpq-14292	188	23	(	(	PUNCT
fpq-14292	188	24	e.g.	e.g.	ADV
fpq-14292	188	25	,	,	PUNCT
fpq-14292	188	26	lacassagne	lacassagne	NOUN
fpq-14292	188	27	,	,	PUNCT
fpq-14292	188	28	béna	béna	NOUN
fpq-14292	188	29	,	,	PUNCT
fpq-14292	188	30	and	and	CCONJ
fpq-14292	188	31	corneille	corneille	PROPN
fpq-14292	188	32	2022	2022	NUM
fpq-14292	188	33	)	)	PUNCT
fpq-14292	188	34	.	.	PUNCT
fpq-14292	189	1	in	in	ADP
fpq-14292	189	2	this	this	DET
fpq-14292	189	3	way	way	NOUN
fpq-14292	189	4	,	,	PUNCT
fpq-14292	189	5	bias	bias	NOUN
fpq-14292	189	6	gets	get	AUX
fpq-14292	189	7	amplified	amplify	VERB
fpq-14292	189	8	in	in	ADP
fpq-14292	189	9	the	the	DET
fpq-14292	189	10	population	population	NOUN
fpq-14292	189	11	due	due	ADP
fpq-14292	189	12	to	to	ADP
fpq-14292	189	13	biased	biased	ADJ
fpq-14292	189	14	models	model	NOUN
fpq-14292	189	15	.	.	PUNCT
fpq-14292	190	1	second	second	ADJ
fpq-14292	190	2	,	,	PUNCT
fpq-14292	190	3	this	this	DET
fpq-14292	190	4	larger	large	ADJ
fpq-14292	190	5	pool	pool	NOUN
fpq-14292	190	6	of	of	ADP
fpq-14292	190	7	individuals	individual	NOUN
fpq-14292	190	8	(	(	PUNCT
fpq-14292	190	9	p	p	NOUN
fpq-14292	190	10	+	+	NOUN
fpq-14292	190	11	n	n	CCONJ
fpq-14292	190	12	)	)	PUNCT
fpq-14292	190	13	is	be	AUX
fpq-14292	190	14	therefore	therefore	ADV
fpq-14292	190	15	more	more	ADV
fpq-14292	190	16	likely	likely	ADJ
fpq-14292	190	17	to	to	PART
fpq-14292	190	18	reproduce	reproduce	VERB
fpq-14292	190	19	examples	example	NOUN
fpq-14292	190	20	of	of	ADP
fpq-14292	190	21	language	language	NOUN
fpq-14292	190	22	containing	contain	VERB
fpq-14292	190	23	bias	bias	NOUN
fpq-14292	190	24	than	than	SCONJ
fpq-14292	190	25	it	it	PRON
fpq-14292	190	26	would	would	AUX
fpq-14292	190	27	have	have	AUX
fpq-14292	190	28	been	be	AUX
fpq-14292	190	29	otherwise	otherwise	ADV
fpq-14292	190	30	,	,	PUNCT
fpq-14292	190	31	and	and	CCONJ
fpq-14292	190	32	these	these	DET
fpq-14292	190	33	examples	example	NOUN
fpq-14292	190	34	may	may	AUX
fpq-14292	190	35	be	be	AUX
fpq-14292	190	36	used	use	VERB
fpq-14292	190	37	as	as	ADP
fpq-14292	190	38	input	input	NOUN
fpq-14292	190	39	for	for	ADP
fpq-14292	190	40	training	train	VERB
fpq-14292	190	41	further	further	ADJ
fpq-14292	190	42	models	model	NOUN
fpq-14292	190	43	.	.	PUNCT
fpq-14292	191	1	thus	thus	ADV
fpq-14292	191	2	,	,	PUNCT
fpq-14292	191	3	the	the	DET
fpq-14292	191	4	bias	bias	NOUN
fpq-14292	191	5	gets	get	AUX
fpq-14292	191	6	amplified	amplify	VERB
fpq-14292	191	7	to	to	ADP
fpq-14292	191	8	a	a	DET
fpq-14292	191	9	wider	wide	ADJ
fpq-14292	191	10	range	range	NOUN
fpq-14292	191	11	of	of	ADP
fpq-14292	191	12	models	model	NOUN
fpq-14292	191	13	than	than	SCONJ
fpq-14292	191	14	it	it	PRON
fpq-14292	191	15	would	would	AUX
fpq-14292	191	16	have	have	AUX
fpq-14292	191	17	been	be	AUX
fpq-14292	191	18	otherwise	otherwise	ADV
fpq-14292	191	19	;	;	PUNCT
fpq-14292	191	20	see	see	VERB
fpq-14292	191	21	bolukbasi	bolukbasi	PROPN
fpq-14292	191	22	et	et	PROPN
fpq-14292	191	23	al	al	PROPN
fpq-14292	191	24	.	.	PROPN
fpq-14292	192	1	(	(	PUNCT
fpq-14292	192	2	2016	2016	NUM
fpq-14292	192	3	)	)	PUNCT
fpq-14292	192	4	for	for	ADP
fpq-14292	192	5	a	a	DET
fpq-14292	192	6	similar	similar	ADJ
fpq-14292	192	7	example	example	NOUN
fpq-14292	192	8	.	.	PUNCT
fpq-14292	193	1	deery	deery	NOUN
fpq-14292	193	2	and	and	CCONJ
fpq-14292	193	3	bailey	bailey	PROPN
fpq-14292	193	4	–	–	PUNCT
fpq-14292	193	5	the	the	DET
fpq-14292	193	6	bias	bias	NOUN
fpq-14292	193	7	dilemma	dilemma	NOUN
fpq-14292	193	8	:	:	PUNCT
fpq-14292	193	9	the	the	DET
fpq-14292	193	10	ethics	ethic	NOUN
fpq-14292	193	11	of	of	ADP
fpq-14292	193	12	algorithmic	algorithmic	ADJ
fpq-14292	193	13	bias	bias	NOUN
fpq-14292	193	14	in	in	ADP
fpq-14292	193	15	natural	natural	ADJ
fpq-14292	193	16	-	-	PUNCT
fpq-14292	193	17	language	language	NOUN
fpq-14292	193	18	processing	processing	NOUN
fpq-14292	193	19	published	publish	VERB
fpq-14292	193	20	by	by	ADP
fpq-14292	193	21	scholarship@western	scholarship@western	PROPN
fpq-14292	193	22	,	,	PUNCT
fpq-14292	193	23	2022	2022	NUM
fpq-14292	193	24	11	11	NUM
fpq-14292	193	25	computer	computer	NOUN
fpq-14292	193	26	programmer	programmer	NOUN
fpq-14292	193	27	as	as	SCONJ
fpq-14292	193	28	woman	woman	NOUN
fpq-14292	193	29	is	be	AUX
fpq-14292	193	30	to	to	ADP
fpq-14292	193	31	x	x	PRON
fpq-14292	193	32	”	"	PUNCT
fpq-14292	193	33	by	by	ADP
fpq-14292	193	34	filling	fill	VERB
fpq-14292	193	35	in	in	ADP
fpq-14292	193	36	“	"	PUNCT
fpq-14292	193	37	homemaker	homemaker	NOUN
fpq-14292	193	38	”	"	PUNCT
fpq-14292	193	39	for	for	ADP
fpq-14292	193	40	x.	x.	PROPN
fpq-14292	193	41	a	a	DET
fpq-14292	193	42	model	model	NOUN
fpq-14292	193	43	gets	get	VERB
fpq-14292	193	44	things	thing	NOUN
fpq-14292	193	45	normatively	normatively	ADV
fpq-14292	193	46	wrong	wrong	ADJ
fpq-14292	193	47	in	in	ADP
fpq-14292	193	48	producing	produce	VERB
fpq-14292	193	49	this	this	DET
fpq-14292	193	50	output	output	NOUN
fpq-14292	193	51	,	,	PUNCT
fpq-14292	193	52	despite	despite	SCONJ
fpq-14292	193	53	its	its	PRON
fpq-14292	193	54	being	be	AUX
fpq-14292	193	55	descriptively	descriptively	ADV
fpq-14292	193	56	right	right	ADJ
fpq-14292	193	57	in	in	ADP
fpq-14292	193	58	reflecting	reflect	VERB
fpq-14292	193	59	actual	actual	ADJ
fpq-14292	193	60	language	language	NOUN
fpq-14292	193	61	use	use	NOUN
fpq-14292	193	62	.	.	PUNCT
fpq-14292	194	1	and	and	CCONJ
fpq-14292	194	2	it	it	PRON
fpq-14292	194	3	gets	get	VERB
fpq-14292	194	4	things	thing	NOUN
fpq-14292	194	5	normatively	normatively	ADV
fpq-14292	194	6	wrong	wrong	ADJ
fpq-14292	194	7	because	because	SCONJ
fpq-14292	194	8	and	and	CCONJ
fpq-14292	194	9	to	to	ADP
fpq-14292	194	10	the	the	DET
fpq-14292	194	11	extent	extent	NOUN
fpq-14292	194	12	to	to	PART
fpq-14292	194	13	which	which	PRON
fpq-14292	194	14	it	it	PRON
fpq-14292	194	15	gets	get	VERB
fpq-14292	194	16	things	thing	NOUN
fpq-14292	194	17	descriptively	descriptively	ADV
fpq-14292	194	18	right	right	ADJ
fpq-14292	194	19	in	in	ADP
fpq-14292	194	20	this	this	DET
fpq-14292	194	21	way.21	way.21	NOUN
fpq-14292	194	22	the	the	DET
fpq-14292	194	23	ideal	ideal	ADJ
fpq-14292	194	24	solution	solution	NOUN
fpq-14292	194	25	might	might	AUX
fpq-14292	194	26	be	be	AUX
fpq-14292	194	27	to	to	ADP
fpq-14292	194	28	debias	debias	PROPN
fpq-14292	194	29	ourselves	ourselves	PRON
fpq-14292	194	30	.	.	PUNCT
fpq-14292	195	1	in	in	ADP
fpq-14292	195	2	that	that	DET
fpq-14292	195	3	case	case	NOUN
fpq-14292	195	4	,	,	PUNCT
fpq-14292	195	5	we	we	PRON
fpq-14292	195	6	would	would	AUX
fpq-14292	195	7	produce	produce	VERB
fpq-14292	195	8	language	language	NOUN
fpq-14292	195	9	corpora	corpora	NOUN
fpq-14292	195	10	that	that	PRON
fpq-14292	195	11	do	do	AUX
fpq-14292	195	12	not	not	PART
fpq-14292	195	13	contain	contain	VERB
fpq-14292	195	14	biases	bias	NOUN
fpq-14292	195	15	,	,	PUNCT
fpq-14292	195	16	the	the	DET
fpq-14292	195	17	nlp	nlp	NOUN
fpq-14292	195	18	models	model	NOUN
fpq-14292	195	19	would	would	AUX
fpq-14292	195	20	be	be	AUX
fpq-14292	195	21	trained	train	VERB
fpq-14292	195	22	on	on	ADP
fpq-14292	195	23	such	such	ADJ
fpq-14292	195	24	data	datum	NOUN
fpq-14292	195	25	,	,	PUNCT
fpq-14292	195	26	and	and	CCONJ
fpq-14292	195	27	all	all	PRON
fpq-14292	195	28	would	would	AUX
fpq-14292	195	29	be	be	AUX
fpq-14292	195	30	well	well	ADJ
fpq-14292	195	31	.	.	PUNCT
fpq-14292	196	1	yet	yet	ADV
fpq-14292	196	2	that	that	PRON
fpq-14292	196	3	is	be	AUX
fpq-14292	196	4	not	not	PART
fpq-14292	196	5	a	a	DET
fpq-14292	196	6	practical	practical	ADJ
fpq-14292	196	7	suggestion	suggestion	NOUN
fpq-14292	196	8	,	,	PUNCT
fpq-14292	196	9	much	much	ADV
fpq-14292	196	10	as	as	SCONJ
fpq-14292	196	11	it	it	PRON
fpq-14292	196	12	might	might	AUX
fpq-14292	196	13	be	be	AUX
fpq-14292	196	14	nice	nice	ADJ
fpq-14292	196	15	to	to	PART
fpq-14292	196	16	imagine	imagine	VERB
fpq-14292	196	17	living	live	VERB
fpq-14292	196	18	in	in	ADP
fpq-14292	196	19	such	such	DET
fpq-14292	196	20	a	a	DET
fpq-14292	196	21	world	world	NOUN
fpq-14292	196	22	.	.	PUNCT
fpq-14292	197	1	or	or	CCONJ
fpq-14292	197	2	we	we	PRON
fpq-14292	197	3	might	might	AUX
fpq-14292	197	4	debias	debia	VERB
fpq-14292	197	5	the	the	DET
fpq-14292	197	6	models	model	NOUN
fpq-14292	197	7	that	that	PRON
fpq-14292	197	8	have	have	AUX
fpq-14292	197	9	acquired	acquire	VERB
fpq-14292	197	10	biases	bias	NOUN
fpq-14292	197	11	.	.	PUNCT
fpq-14292	198	1	as	as	SCONJ
fpq-14292	198	2	mentioned	mention	VERB
fpq-14292	198	3	earlier	early	ADV
fpq-14292	198	4	,	,	PUNCT
fpq-14292	198	5	it	it	PRON
fpq-14292	198	6	can	can	AUX
fpq-14292	198	7	be	be	AUX
fpq-14292	198	8	difficult	difficult	ADJ
fpq-14292	198	9	,	,	PUNCT
fpq-14292	198	10	as	as	ADP
fpq-14292	198	11	cognitive	cognitive	ADJ
fpq-14292	198	12	science	science	NOUN
fpq-14292	198	13	shows	show	NOUN
fpq-14292	198	14	,	,	PUNCT
fpq-14292	198	15	to	to	PART
fpq-14292	198	16	eliminate	eliminate	VERB
fpq-14292	198	17	biases	bias	NOUN
fpq-14292	198	18	from	from	ADP
fpq-14292	198	19	human	human	ADJ
fpq-14292	198	20	cognition	cognition	NOUN
fpq-14292	198	21	.	.	PUNCT
fpq-14292	199	1	we	we	PRON
fpq-14292	199	2	think	think	VERB
fpq-14292	199	3	that	that	SCONJ
fpq-14292	199	4	it	it	PRON
fpq-14292	199	5	is	be	AUX
fpq-14292	199	6	no	no	ADV
fpq-14292	199	7	easier	easy	ADJ
fpq-14292	199	8	to	to	PART
fpq-14292	199	9	debias	debias	VERB
fpq-14292	199	10	nlp	nlp	NOUN
fpq-14292	199	11	models	model	NOUN
fpq-14292	199	12	,	,	PUNCT
fpq-14292	199	13	which	which	PRON
fpq-14292	199	14	raises	raise	VERB
fpq-14292	199	15	a	a	DET
fpq-14292	199	16	practical	practical	ADJ
fpq-14292	199	17	problem	problem	NOUN
fpq-14292	199	18	,	,	PUNCT
fpq-14292	199	19	which	which	PRON
fpq-14292	199	20	we	we	PRON
fpq-14292	199	21	outline	outline	VERB
fpq-14292	199	22	in	in	ADP
fpq-14292	199	23	section	section	NOUN
fpq-14292	199	24	5	5	NUM
fpq-14292	199	25	,	,	PUNCT
fpq-14292	199	26	and	and	CCONJ
fpq-14292	199	27	an	an	DET
fpq-14292	199	28	ethical	ethical	ADJ
fpq-14292	199	29	dilemma	dilemma	NOUN
fpq-14292	199	30	,	,	PUNCT
fpq-14292	199	31	which	which	PRON
fpq-14292	199	32	we	we	PRON
fpq-14292	199	33	outline	outline	VERB
fpq-14292	199	34	in	in	ADP
fpq-14292	199	35	section	section	NOUN
fpq-14292	199	36	6	6	NUM
fpq-14292	199	37	.	.	NOUN
fpq-14292	199	38	5	5	NUM
fpq-14292	199	39	.	.	X
fpq-14292	200	1	debiasing	debiase	VERB
fpq-14292	200	2	:	:	PUNCT
fpq-14292	200	3	practical	practical	ADJ
fpq-14292	200	4	problems	problem	NOUN
fpq-14292	200	5	word	word	NOUN
fpq-14292	200	6	-	-	PUNCT
fpq-14292	200	7	embeddings	embedding	NOUN
fpq-14292	200	8	and	and	CCONJ
fpq-14292	200	9	large	large	ADJ
fpq-14292	200	10	language	language	NOUN
fpq-14292	200	11	models	model	NOUN
fpq-14292	200	12	are	be	AUX
fpq-14292	200	13	trained	train	VERB
fpq-14292	200	14	using	use	VERB
fpq-14292	200	15	neural	neural	ADJ
fpq-14292	200	16	networks	network	NOUN
fpq-14292	200	17	whose	whose	DET
fpq-14292	200	18	objective	objective	ADJ
fpq-14292	200	19	function	function	NOUN
fpq-14292	200	20	is	be	AUX
fpq-14292	200	21	to	to	PART
fpq-14292	200	22	minimize	minimize	VERB
fpq-14292	200	23	error	error	NOUN
fpq-14292	200	24	on	on	ADP
fpq-14292	200	25	a	a	DET
fpq-14292	200	26	particular	particular	ADJ
fpq-14292	200	27	task	task	NOUN
fpq-14292	200	28	,	,	PUNCT
fpq-14292	200	29	often	often	ADV
fpq-14292	200	30	the	the	DET
fpq-14292	200	31	ability	ability	NOUN
fpq-14292	200	32	to	to	PART
fpq-14292	200	33	predict	predict	VERB
fpq-14292	200	34	each	each	DET
fpq-14292	200	35	word	word	NOUN
fpq-14292	200	36	in	in	ADP
fpq-14292	200	37	a	a	DET
fpq-14292	200	38	text	text	NOUN
fpq-14292	200	39	given	give	VERB
fpq-14292	200	40	the	the	DET
fpq-14292	200	41	surrounding	surround	VERB
fpq-14292	200	42	words	word	NOUN
fpq-14292	200	43	(	(	PUNCT
fpq-14292	200	44	see	see	VERB
fpq-14292	200	45	,	,	PUNCT
fpq-14292	200	46	e.g.	e.g.	ADV
fpq-14292	200	47	,	,	PUNCT
fpq-14292	200	48	mikolov	mikolov	PROPN
fpq-14292	200	49	et	et	PROPN
fpq-14292	200	50	al	al	PROPN
fpq-14292	200	51	.	.	PROPN
fpq-14292	200	52	2013	2013	NUM
fpq-14292	200	53	)	)	PUNCT
fpq-14292	200	54	.	.	PUNCT
fpq-14292	201	1	that	that	PRON
fpq-14292	201	2	is	be	AUX
fpq-14292	201	3	the	the	DET
fpq-14292	201	4	training	training	NOUN
fpq-14292	201	5	task	task	NOUN
fpq-14292	201	6	.	.	PUNCT
fpq-14292	202	1	when	when	SCONJ
fpq-14292	202	2	we	we	PRON
fpq-14292	202	3	consider	consider	VERB
fpq-14292	202	4	downstream	downstream	ADJ
fpq-14292	202	5	tasks	task	NOUN
fpq-14292	202	6	,	,	PUNCT
fpq-14292	202	7	we	we	PRON
fpq-14292	202	8	are	be	AUX
fpq-14292	202	9	taking	take	VERB
fpq-14292	202	10	a	a	DET
fpq-14292	202	11	model	model	NOUN
fpq-14292	202	12	that	that	PRON
fpq-14292	202	13	has	have	AUX
fpq-14292	202	14	already	already	ADV
fpq-14292	202	15	been	be	AUX
fpq-14292	202	16	trained	train	VERB
fpq-14292	202	17	on	on	ADP
fpq-14292	202	18	a	a	DET
fpq-14292	202	19	training	training	NOUN
fpq-14292	202	20	task	task	NOUN
fpq-14292	202	21	using	use	VERB
fpq-14292	202	22	a	a	DET
fpq-14292	202	23	massive	massive	ADJ
fpq-14292	202	24	language	language	NOUN
fpq-14292	202	25	corpus	corpus	NOUN
fpq-14292	202	26	and	and	CCONJ
fpq-14292	202	27	fine	fine	ADV
fpq-14292	202	28	-	-	PUNCT
fpq-14292	202	29	tuning	tune	VERB
fpq-14292	202	30	it	it	PRON
fpq-14292	202	31	or	or	CCONJ
fpq-14292	202	32	using	use	VERB
fpq-14292	202	33	it	it	PRON
fpq-14292	202	34	as	as	SCONJ
fpq-14292	202	35	is	be	AUX
fpq-14292	202	36	on	on	ADP
fpq-14292	202	37	a	a	DET
fpq-14292	202	38	new	new	ADJ
fpq-14292	202	39	task	task	NOUN
fpq-14292	202	40	.	.	PUNCT
fpq-14292	203	1	the	the	DET
fpq-14292	203	2	new	new	ADJ
fpq-14292	203	3	task	task	NOUN
fpq-14292	203	4	could	could	AUX
fpq-14292	203	5	be	be	AUX
fpq-14292	203	6	text	text	NOUN
fpq-14292	203	7	classification	classification	NOUN
fpq-14292	203	8	,	,	PUNCT
fpq-14292	203	9	which	which	PRON
fpq-14292	203	10	might	might	AUX
fpq-14292	203	11	then	then	ADV
fpq-14292	203	12	be	be	AUX
fpq-14292	203	13	used	use	VERB
fpq-14292	203	14	in	in	ADP
fpq-14292	203	15	a	a	DET
fpq-14292	203	16	further	further	ADJ
fpq-14292	203	17	application	application	NOUN
fpq-14292	203	18	—	—	PUNCT
fpq-14292	203	19	for	for	ADP
fpq-14292	203	20	example	example	NOUN
fpq-14292	203	21	,	,	PUNCT
fpq-14292	203	22	a	a	DET
fpq-14292	203	23	content	content	NOUN
fpq-14292	203	24	-	-	PUNCT
fpq-14292	203	25	management	management	NOUN
fpq-14292	203	26	system	system	NOUN
fpq-14292	203	27	that	that	PRON
fpq-14292	203	28	adds	add	VERB
fpq-14292	203	29	keywords	keyword	NOUN
fpq-14292	203	30	and	and	CCONJ
fpq-14292	203	31	categories	category	NOUN
fpq-14292	203	32	to	to	AUX
fpq-14292	203	33	content	content	VERB
fpq-14292	203	34	.	.	PUNCT
fpq-14292	204	1	this	this	DET
fpq-14292	204	2	case	case	NOUN
fpq-14292	204	3	is	be	AUX
fpq-14292	204	4	an	an	DET
fpq-14292	204	5	example	example	NOUN
fpq-14292	204	6	of	of	ADP
fpq-14292	204	7	transfer	transfer	NOUN
fpq-14292	204	8	learning	learning	NOUN
fpq-14292	204	9	,	,	PUNCT
fpq-14292	204	10	where	where	SCONJ
fpq-14292	204	11	learning	learn	VERB
fpq-14292	204	12	from	from	ADP
fpq-14292	204	13	one	one	NUM
fpq-14292	204	14	task	task	NOUN
fpq-14292	204	15	on	on	ADP
fpq-14292	204	16	a	a	DET
fpq-14292	204	17	massive	massive	ADJ
fpq-14292	204	18	corpus	corpus	NOUN
fpq-14292	204	19	is	be	AUX
fpq-14292	204	20	transferred	transfer	VERB
fpq-14292	204	21	to	to	ADP
fpq-14292	204	22	another	another	DET
fpq-14292	204	23	task	task	NOUN
fpq-14292	204	24	on	on	ADP
fpq-14292	204	25	a	a	DET
fpq-14292	204	26	different	different	ADJ
fpq-14292	204	27	,	,	PUNCT
fpq-14292	204	28	smaller	small	ADJ
fpq-14292	204	29	corpus	corpus	NOUN
fpq-14292	204	30	.	.	PUNCT
fpq-14292	205	1	often	often	ADV
fpq-14292	205	2	,	,	PUNCT
fpq-14292	205	3	the	the	DET
fpq-14292	205	4	so	so	ADV
fpq-14292	205	5	-	-	PUNCT
fpq-14292	205	6	called	call	VERB
fpq-14292	205	7	pretrained	pretraine	VERB
fpq-14292	205	8	model	model	NOUN
fpq-14292	205	9	is	be	AUX
fpq-14292	205	10	made	make	VERB
fpq-14292	205	11	publicly	publicly	ADV
fpq-14292	205	12	available	available	ADJ
fpq-14292	205	13	by	by	ADP
fpq-14292	205	14	those	those	PRON
fpq-14292	205	15	who	who	PRON
fpq-14292	205	16	trained	train	VERB
fpq-14292	205	17	it	it	PRON
fpq-14292	205	18	and	and	CCONJ
fpq-14292	205	19	can	can	AUX
fpq-14292	205	20	end	end	VERB
fpq-14292	205	21	up	up	ADP
fpq-14292	205	22	being	be	AUX
fpq-14292	205	23	used	use	VERB
fpq-14292	205	24	in	in	ADP
fpq-14292	205	25	countless	countless	ADJ
fpq-14292	205	26	downstream	downstream	ADJ
fpq-14292	205	27	applications	application	NOUN
fpq-14292	205	28	that	that	PRON
fpq-14292	205	29	the	the	DET
fpq-14292	205	30	original	original	ADJ
fpq-14292	205	31	developers	developer	NOUN
fpq-14292	205	32	of	of	ADP
fpq-14292	205	33	the	the	DET
fpq-14292	205	34	model	model	NOUN
fpq-14292	205	35	might	might	AUX
fpq-14292	205	36	never	never	ADV
fpq-14292	205	37	know	know	VERB
fpq-14292	205	38	about	about	ADP
fpq-14292	205	39	.	.	PUNCT
fpq-14292	206	1	that	that	PRON
fpq-14292	206	2	is	be	AUX
fpq-14292	206	3	what	what	PRON
fpq-14292	206	4	happened	happen	VERB
fpq-14292	206	5	with	with	ADP
fpq-14292	206	6	word2vec	word2vec	PROPN
fpq-14292	206	7	,	,	PUNCT
fpq-14292	206	8	the	the	DET
fpq-14292	206	9	word	word	NOUN
fpq-14292	206	10	-	-	PUNCT
fpq-14292	206	11	embeddings	embeddings	NOUN
fpq-14292	206	12	model	model	NOUN
fpq-14292	206	13	we	we	PRON
fpq-14292	206	14	mentioned	mention	VERB
fpq-14292	206	15	above	above	ADV
fpq-14292	206	16	(	(	PUNCT
fpq-14292	206	17	mikolov	mikolov	PROPN
fpq-14292	206	18	et	et	PROPN
fpq-14292	206	19	al	al	PROPN
fpq-14292	206	20	.	.	PROPN
fpq-14292	206	21	2013	2013	NUM
fpq-14292	206	22	)	)	PUNCT
fpq-14292	206	23	,	,	PUNCT
fpq-14292	206	24	and	and	CCONJ
fpq-14292	206	25	bert	bert	PROPN
fpq-14292	206	26	,	,	PUNCT
fpq-14292	206	27	one	one	NUM
fpq-14292	206	28	of	of	ADP
fpq-14292	206	29	the	the	DET
fpq-14292	206	30	first	first	ADJ
fpq-14292	206	31	large	large	ADJ
fpq-14292	206	32	language	language	NOUN
fpq-14292	206	33	models	model	NOUN
fpq-14292	206	34	to	to	PART
fpq-14292	206	35	be	be	AUX
fpq-14292	206	36	made	make	VERB
fpq-14292	206	37	publicly	publicly	ADV
fpq-14292	206	38	available	available	ADJ
fpq-14292	206	39	(	(	PUNCT
fpq-14292	206	40	devlin	devlin	PROPN
fpq-14292	206	41	et	et	PROPN
fpq-14292	206	42	al	al	PROPN
fpq-14292	206	43	.	.	PROPN
fpq-14292	206	44	2019	2019	NUM
fpq-14292	206	45	)	)	PUNCT
fpq-14292	206	46	.	.	PUNCT
fpq-14292	207	1	the	the	DET
fpq-14292	207	2	focus	focus	NOUN
fpq-14292	207	3	on	on	ADP
fpq-14292	207	4	debiasing	debiase	VERB
fpq-14292	207	5	has	have	AUX
fpq-14292	207	6	tended	tend	VERB
fpq-14292	207	7	to	to	PART
fpq-14292	207	8	be	be	AUX
fpq-14292	207	9	on	on	ADP
fpq-14292	207	10	this	this	DET
fpq-14292	207	11	initial	initial	ADJ
fpq-14292	207	12	stage	stage	NOUN
fpq-14292	207	13	of	of	ADP
fpq-14292	207	14	the	the	DET
fpq-14292	207	15	training	training	NOUN
fpq-14292	207	16	pipeline.22	pipeline.22	ADP
fpq-14292	207	17	the	the	DET
fpq-14292	207	18	resulting	result	VERB
fpq-14292	207	19	models	model	NOUN
fpq-14292	207	20	are	be	AUX
fpq-14292	207	21	subsequently	subsequently	ADV
fpq-14292	207	22	released	release	VERB
fpq-14292	207	23	for	for	ADP
fpq-14292	207	24	use	use	NOUN
fpq-14292	207	25	in	in	ADP
fpq-14292	207	26	various	various	ADJ
fpq-14292	207	27	applications	application	NOUN
fpq-14292	207	28	;	;	PUNCT
fpq-14292	207	29	if	if	SCONJ
fpq-14292	207	30	we	we	PRON
fpq-14292	207	31	debias	debia	VERB
fpq-14292	207	32	these	these	DET
fpq-14292	207	33	pretrained	pretraine	VERB
fpq-14292	207	34	models	model	NOUN
fpq-14292	207	35	,	,	PUNCT
fpq-14292	207	36	then	then	ADV
fpq-14292	207	37	the	the	DET
fpq-14292	207	38	applications	application	NOUN
fpq-14292	207	39	built	build	VERB
fpq-14292	207	40	21	21	NUM
fpq-14292	207	41	we	we	PRON
fpq-14292	207	42	acknowledged	acknowledge	VERB
fpq-14292	207	43	in	in	ADP
fpq-14292	207	44	section	section	NOUN
fpq-14292	207	45	3	3	NUM
fpq-14292	207	46	that	that	SCONJ
fpq-14292	207	47	such	such	DET
fpq-14292	207	48	a	a	DET
fpq-14292	207	49	model	model	NOUN
fpq-14292	207	50	is	be	AUX
fpq-14292	207	51	descriptively	descriptively	ADV
fpq-14292	207	52	inaccurate	inaccurate	ADJ
fpq-14292	207	53	regarding	regard	VERB
fpq-14292	207	54	these	these	DET
fpq-14292	207	55	words	word	NOUN
fpq-14292	207	56	’	'	PUNCT
fpq-14292	207	57	semantics	semantic	NOUN
fpq-14292	207	58	—	—	PUNCT
fpq-14292	207	59	there	there	PRON
fpq-14292	207	60	is	be	VERB
fpq-14292	207	61	nothing	nothing	PRON
fpq-14292	207	62	about	about	ADP
fpq-14292	207	63	the	the	DET
fpq-14292	207	64	word	word	NOUN
fpq-14292	207	65	“	"	PUNCT
fpq-14292	207	66	woman	woman	NOUN
fpq-14292	207	67	,	,	PUNCT
fpq-14292	207	68	”	"	PUNCT
fpq-14292	207	69	semantically	semantically	ADV
fpq-14292	207	70	,	,	PUNCT
fpq-14292	207	71	in	in	ADP
fpq-14292	207	72	virtue	virtue	NOUN
fpq-14292	207	73	of	of	ADP
fpq-14292	207	74	which	which	PRON
fpq-14292	207	75	it	it	PRON
fpq-14292	207	76	clusters	cluster	VERB
fpq-14292	207	77	with	with	ADP
fpq-14292	207	78	“	"	PUNCT
fpq-14292	207	79	homemaker	homemaker	NOUN
fpq-14292	207	80	.	.	PUNCT
fpq-14292	207	81	”	"	PUNCT
fpq-14292	208	1	22	22	NUM
fpq-14292	208	2	here	here	ADV
fpq-14292	208	3	,	,	PUNCT
fpq-14292	208	4	word	word	NOUN
fpq-14292	208	5	-	-	PUNCT
fpq-14292	208	6	embeddings	embedding	NOUN
fpq-14292	208	7	models	model	NOUN
fpq-14292	208	8	are	be	AUX
fpq-14292	208	9	the	the	DET
fpq-14292	208	10	initial	initial	ADJ
fpq-14292	208	11	stage	stage	NOUN
fpq-14292	208	12	of	of	ADP
fpq-14292	208	13	the	the	DET
fpq-14292	208	14	training	training	NOUN
fpq-14292	208	15	pipeline	pipeline	NOUN
fpq-14292	208	16	and	and	CCONJ
fpq-14292	208	17	can	can	AUX
fpq-14292	208	18	subsequently	subsequently	ADV
fpq-14292	208	19	be	be	AUX
fpq-14292	208	20	used	use	VERB
fpq-14292	208	21	in	in	ADP
fpq-14292	208	22	downstream	downstream	ADJ
fpq-14292	208	23	applications	application	NOUN
fpq-14292	208	24	(	(	PUNCT
fpq-14292	208	25	mikolov	mikolov	X
fpq-14292	208	26	et	et	PROPN
fpq-14292	208	27	al	al	PROPN
fpq-14292	208	28	.	.	PROPN
fpq-14292	208	29	2013	2013	NUM
fpq-14292	208	30	)	)	PUNCT
fpq-14292	208	31	.	.	PUNCT
fpq-14292	209	1	blodgett	blodgett	PROPN
fpq-14292	209	2	et	et	PROPN
fpq-14292	209	3	al	al	PROPN
fpq-14292	209	4	.	.	PROPN
fpq-14292	209	5	(	(	PUNCT
fpq-14292	209	6	2020	2020	NUM
fpq-14292	209	7	)	)	PUNCT
fpq-14292	209	8	surveyed	survey	VERB
fpq-14292	209	9	146	146	NUM
fpq-14292	209	10	papers	paper	NOUN
fpq-14292	209	11	on	on	ADP
fpq-14292	209	12	bias	bias	NOUN
fpq-14292	209	13	in	in	ADP
fpq-14292	209	14	nlp	nlp	NOUN
fpq-14292	209	15	systems	system	NOUN
fpq-14292	209	16	and	and	CCONJ
fpq-14292	209	17	54	54	NUM
fpq-14292	209	18	of	of	ADP
fpq-14292	209	19	these	these	DET
fpq-14292	209	20	papers	paper	NOUN
fpq-14292	209	21	specifically	specifically	ADV
fpq-14292	209	22	dealt	deal	VERB
fpq-14292	209	23	with	with	ADP
fpq-14292	209	24	bias	bias	NOUN
fpq-14292	209	25	in	in	ADP
fpq-14292	209	26	word	word	NOUN
fpq-14292	209	27	-	-	PUNCT
fpq-14292	209	28	embeddings	embedding	NOUN
fpq-14292	209	29	systems	system	NOUN
fpq-14292	209	30	.	.	PUNCT
fpq-14292	210	1	feminist	feminist	ADJ
fpq-14292	210	2	philosophy	philosophy	PROPN
fpq-14292	210	3	quarterly	quarterly	ADV
fpq-14292	210	4	,	,	PUNCT
fpq-14292	210	5	2022	2022	NUM
fpq-14292	210	6	,	,	PUNCT
fpq-14292	210	7	vol	vol	NOUN
fpq-14292	210	8	.	.	NOUN
fpq-14292	210	9	8	8	NUM
fpq-14292	210	10	,	,	PUNCT
fpq-14292	210	11	iss	iss	PROPN
fpq-14292	210	12	.	.	PUNCT
fpq-14292	210	13	3/4	3/4	NUM
fpq-14292	210	14	,	,	PUNCT
fpq-14292	210	15	article	article	NOUN
fpq-14292	210	16	1	1	NUM
fpq-14292	210	17	published	publish	VERB
fpq-14292	210	18	by	by	ADP
fpq-14292	210	19	scholarship@western	scholarship@western	PROPN
fpq-14292	210	20	,	,	PUNCT
fpq-14292	210	21	2022	2022	NUM
fpq-14292	210	22	12	12	NUM
fpq-14292	210	23	downstream	downstream	NOUN
fpq-14292	210	24	of	of	ADP
fpq-14292	210	25	them	they	PRON
fpq-14292	210	26	will	will	AUX
fpq-14292	210	27	also	also	ADV
fpq-14292	210	28	be	be	AUX
fpq-14292	210	29	debiased	debiase	VERB
fpq-14292	210	30	unless	unless	SCONJ
fpq-14292	210	31	bias	bias	NOUN
fpq-14292	210	32	is	be	AUX
fpq-14292	210	33	introduced	introduce	VERB
fpq-14292	210	34	somehow	somehow	ADV
fpq-14292	210	35	later	later	ADV
fpq-14292	210	36	.	.	PUNCT
fpq-14292	211	1	consequently	consequently	ADV
fpq-14292	211	2	,	,	PUNCT
fpq-14292	211	3	debiasing	debiase	VERB
fpq-14292	211	4	the	the	DET
fpq-14292	211	5	models	model	NOUN
fpq-14292	211	6	is	be	AUX
fpq-14292	211	7	seen	see	VERB
fpq-14292	211	8	as	as	ADP
fpq-14292	211	9	an	an	DET
fpq-14292	211	10	attractive	attractive	ADJ
fpq-14292	211	11	solution	solution	NOUN
fpq-14292	211	12	to	to	ADP
fpq-14292	211	13	many	many	ADJ
fpq-14292	211	14	researchers	researcher	NOUN
fpq-14292	211	15	.	.	PUNCT
fpq-14292	212	1	however	however	ADV
fpq-14292	212	2	,	,	PUNCT
fpq-14292	212	3	we	we	PRON
fpq-14292	212	4	think	think	VERB
fpq-14292	212	5	that	that	SCONJ
fpq-14292	212	6	researchers	researcher	NOUN
fpq-14292	212	7	would	would	AUX
fpq-14292	212	8	do	do	VERB
fpq-14292	212	9	better	well	ADV
fpq-14292	212	10	to	to	PART
fpq-14292	212	11	focus	focus	VERB
fpq-14292	212	12	on	on	ADP
fpq-14292	212	13	debiasing	debiase	VERB
fpq-14292	212	14	the	the	DET
fpq-14292	212	15	outputs	output	NOUN
fpq-14292	212	16	of	of	ADP
fpq-14292	212	17	models	model	NOUN
fpq-14292	212	18	rather	rather	ADV
fpq-14292	212	19	the	the	DET
fpq-14292	212	20	models	model	NOUN
fpq-14292	212	21	themselves	themselves	PRON
fpq-14292	212	22	.	.	PUNCT
fpq-14292	213	1	in	in	ADP
fpq-14292	213	2	turn	turn	NOUN
fpq-14292	213	3	,	,	PUNCT
fpq-14292	213	4	we	we	PRON
fpq-14292	213	5	claim	claim	VERB
fpq-14292	213	6	that	that	SCONJ
fpq-14292	213	7	given	give	VERB
fpq-14292	213	8	the	the	DET
fpq-14292	213	9	serious	serious	ADJ
fpq-14292	213	10	practical	practical	ADJ
fpq-14292	213	11	difficulties	difficulty	NOUN
fpq-14292	213	12	with	with	ADP
fpq-14292	213	13	debiasing	debiase	VERB
fpq-14292	213	14	models	model	NOUN
fpq-14292	213	15	,	,	PUNCT
fpq-14292	213	16	it	it	PRON
fpq-14292	213	17	behooves	behoove	VERB
fpq-14292	213	18	us	we	PRON
fpq-14292	213	19	to	to	PART
fpq-14292	213	20	consider	consider	VERB
fpq-14292	213	21	whether	whether	SCONJ
fpq-14292	213	22	the	the	DET
fpq-14292	213	23	biased	biased	ADJ
fpq-14292	213	24	models	model	NOUN
fpq-14292	213	25	we	we	PRON
fpq-14292	213	26	are	be	AUX
fpq-14292	213	27	left	leave	VERB
fpq-14292	213	28	with	with	ADP
fpq-14292	213	29	are	be	AUX
fpq-14292	213	30	always	always	ADV
fpq-14292	213	31	bad	bad	ADJ
fpq-14292	213	32	.	.	PUNCT
fpq-14292	214	1	we	we	PRON
fpq-14292	214	2	will	will	AUX
fpq-14292	214	3	argue	argue	VERB
fpq-14292	214	4	that	that	SCONJ
fpq-14292	214	5	they	they	PRON
fpq-14292	214	6	are	be	AUX
fpq-14292	214	7	not	not	PART
fpq-14292	214	8	(	(	PUNCT
fpq-14292	214	9	in	in	ADP
fpq-14292	214	10	,	,	PUNCT
fpq-14292	214	11	e.g.	e.g.	ADV
fpq-14292	214	12	,	,	PUNCT
fpq-14292	214	13	section	section	NOUN
fpq-14292	214	14	7	7	NUM
fpq-14292	214	15	below	below	ADP
fpq-14292	214	16	)	)	PUNCT
fpq-14292	214	17	.	.	PUNCT
fpq-14292	215	1	in	in	ADP
fpq-14292	215	2	their	their	PRON
fpq-14292	215	3	previously	previously	ADV
fpq-14292	215	4	mentioned	mention	VERB
fpq-14292	215	5	paper	paper	NOUN
fpq-14292	215	6	,	,	PUNCT
fpq-14292	215	7	bolukbasi	bolukbasi	NOUN
fpq-14292	215	8	and	and	CCONJ
fpq-14292	215	9	colleagues	colleague	NOUN
fpq-14292	215	10	(	(	PUNCT
fpq-14292	215	11	bolukbasi	bolukbasi	NOUN
fpq-14292	215	12	et	et	PROPN
fpq-14292	215	13	al	al	PROPN
fpq-14292	215	14	.	.	PROPN
fpq-14292	215	15	2016	2016	NUM
fpq-14292	215	16	)	)	PUNCT
fpq-14292	215	17	introduce	introduce	VERB
fpq-14292	215	18	a	a	DET
fpq-14292	215	19	method	method	NOUN
fpq-14292	215	20	they	they	PRON
fpq-14292	215	21	call	call	VERB
fpq-14292	215	22	“	"	PUNCT
fpq-14292	215	23	hard	hard	ADV
fpq-14292	215	24	-	-	PUNCT
fpq-14292	215	25	debiasing	debiasing	NOUN
fpq-14292	215	26	,	,	PUNCT
fpq-14292	215	27	”	"	PUNCT
fpq-14292	215	28	which	which	PRON
fpq-14292	215	29	aims	aim	VERB
fpq-14292	215	30	to	to	PART
fpq-14292	215	31	reduce	reduce	VERB
fpq-14292	215	32	sexist	sexist	ADJ
fpq-14292	215	33	bias	bias	NOUN
fpq-14292	215	34	within	within	ADP
fpq-14292	215	35	an	an	DET
fpq-14292	215	36	embeddings	embedding	NOUN
fpq-14292	215	37	vector	vector	NOUN
fpq-14292	215	38	space	space	NOUN
fpq-14292	215	39	without	without	ADP
fpq-14292	215	40	compromising	compromise	VERB
fpq-14292	215	41	the	the	DET
fpq-14292	215	42	overall	overall	ADJ
fpq-14292	215	43	structure	structure	NOUN
fpq-14292	215	44	of	of	ADP
fpq-14292	215	45	that	that	DET
fpq-14292	215	46	space	space	NOUN
fpq-14292	215	47	or	or	CCONJ
fpq-14292	215	48	,	,	PUNCT
fpq-14292	215	49	therefore	therefore	ADV
fpq-14292	215	50	,	,	PUNCT
fpq-14292	215	51	its	its	PRON
fpq-14292	215	52	usefulness	usefulness	NOUN
fpq-14292	215	53	.	.	PUNCT
fpq-14292	216	1	they	they	PRON
fpq-14292	216	2	claim	claim	VERB
fpq-14292	216	3	that	that	SCONJ
fpq-14292	216	4	their	their	PRON
fpq-14292	216	5	method	method	NOUN
fpq-14292	216	6	can	can	AUX
fpq-14292	216	7	“	"	PUNCT
fpq-14292	216	8	significantly	significantly	ADV
fpq-14292	216	9	reduce	reduce	VERB
fpq-14292	216	10	gender	gender	NOUN
fpq-14292	216	11	bias	bias	NOUN
fpq-14292	216	12	in	in	ADP
fpq-14292	216	13	embeddings	embedding	NOUN
fpq-14292	216	14	while	while	SCONJ
fpq-14292	216	15	preserving	preserve	VERB
fpq-14292	216	16	[	[	PUNCT
fpq-14292	216	17	their	their	PRON
fpq-14292	216	18	]	]	X
fpq-14292	216	19	useful	useful	ADJ
fpq-14292	216	20	properties	property	NOUN
fpq-14292	216	21	such	such	ADJ
fpq-14292	216	22	as	as	ADP
fpq-14292	216	23	the	the	DET
fpq-14292	216	24	ability	ability	NOUN
fpq-14292	216	25	to	to	PART
fpq-14292	216	26	cluster	cluster	VERB
fpq-14292	216	27	related	relate	VERB
fpq-14292	216	28	concepts	concept	NOUN
fpq-14292	216	29	and	and	CCONJ
fpq-14292	216	30	to	to	PART
fpq-14292	216	31	solve	solve	VERB
fpq-14292	216	32	analogy	analogy	NOUN
fpq-14292	216	33	tasks	task	NOUN
fpq-14292	216	34	”	"	PUNCT
fpq-14292	216	35	(	(	PUNCT
fpq-14292	216	36	bolukbasi	bolukbasi	NOUN
fpq-14292	216	37	et	et	PROPN
fpq-14292	216	38	al	al	PROPN
fpq-14292	216	39	.	.	PROPN
fpq-14292	216	40	2016	2016	NUM
fpq-14292	216	41	,	,	PUNCT
fpq-14292	216	42	1).23	1).23	NUM
fpq-14292	216	43	,	,	PUNCT
fpq-14292	216	44	24	24	NUM
fpq-14292	216	45	we	we	PRON
fpq-14292	216	46	doubt	doubt	VERB
fpq-14292	216	47	whether	whether	SCONJ
fpq-14292	216	48	word	word	NOUN
fpq-14292	216	49	-	-	PUNCT
fpq-14292	216	50	embeddings	embedding	NOUN
fpq-14292	216	51	can	can	AUX
fpq-14292	216	52	be	be	AUX
fpq-14292	216	53	completely	completely	ADV
fpq-14292	216	54	debiased	debiase	VERB
fpq-14292	216	55	using	use	VERB
fpq-14292	216	56	this	this	DET
fpq-14292	216	57	method	method	NOUN
fpq-14292	216	58	.	.	PUNCT
fpq-14292	217	1	what	what	PRON
fpq-14292	217	2	bolukbasi	bolukbasi	NOUN
fpq-14292	217	3	and	and	CCONJ
fpq-14292	217	4	colleagues	colleague	NOUN
fpq-14292	217	5	call	call	VERB
fpq-14292	217	6	hard	hard	ADV
fpq-14292	217	7	-	-	PUNCT
fpq-14292	217	8	debiasing	debiase	VERB
fpq-14292	217	9	works	work	NOUN
fpq-14292	217	10	largely	largely	ADV
fpq-14292	217	11	by	by	ADP
fpq-14292	217	12	maintaining	maintain	VERB
fpq-14292	217	13	humans	human	NOUN
fpq-14292	217	14	in	in	ADP
fpq-14292	217	15	the	the	DET
fpq-14292	217	16	loop	loop	NOUN
fpq-14292	217	17	at	at	ADP
fpq-14292	217	18	multiple	multiple	ADJ
fpq-14292	217	19	steps	step	NOUN
fpq-14292	217	20	to	to	PART
fpq-14292	217	21	identify	identify	VERB
fpq-14292	217	22	and	and	CCONJ
fpq-14292	217	23	compensate	compensate	VERB
fpq-14292	217	24	for	for	ADP
fpq-14292	217	25	biases	bias	NOUN
fpq-14292	217	26	picked	pick	VERB
fpq-14292	217	27	up	up	ADP
fpq-14292	217	28	by	by	ADP
fpq-14292	217	29	the	the	DET
fpq-14292	217	30	model	model	NOUN
fpq-14292	217	31	.	.	PUNCT
fpq-14292	218	1	this	this	DET
fpq-14292	218	2	feature	feature	NOUN
fpq-14292	218	3	is	be	AUX
fpq-14292	218	4	partly	partly	ADV
fpq-14292	218	5	attractive	attractive	ADJ
fpq-14292	218	6	,	,	PUNCT
fpq-14292	218	7	since	since	SCONJ
fpq-14292	218	8	humans	human	NOUN
fpq-14292	218	9	in	in	ADP
fpq-14292	218	10	the	the	DET
fpq-14292	218	11	loop	loop	NOUN
fpq-14292	218	12	are	be	AUX
fpq-14292	218	13	frequently	frequently	ADV
fpq-14292	218	14	practically	practically	ADV
fpq-14292	218	15	and	and	CCONJ
fpq-14292	218	16	ethically	ethically	ADV
fpq-14292	218	17	beneficial	beneficial	ADJ
fpq-14292	218	18	(	(	PUNCT
fpq-14292	218	19	especially	especially	ADV
fpq-14292	218	20	when	when	SCONJ
fpq-14292	218	21	the	the	DET
fpq-14292	218	22	stakes	stake	NOUN
fpq-14292	218	23	are	be	AUX
fpq-14292	218	24	high	high	ADJ
fpq-14292	218	25	,	,	PUNCT
fpq-14292	218	26	as	as	ADP
fpq-14292	218	27	in	in	ADP
fpq-14292	218	28	medical	medical	ADJ
fpq-14292	218	29	applications	application	NOUN
fpq-14292	218	30	,	,	PUNCT
fpq-14292	218	31	since	since	SCONJ
fpq-14292	218	32	the	the	DET
fpq-14292	218	33	instance	instance	NOUN
fpq-14292	218	34	of	of	ADP
fpq-14292	218	35	harmful	harmful	ADJ
fpq-14292	218	36	errors	error	NOUN
fpq-14292	218	37	will	will	AUX
fpq-14292	218	38	thereby	thereby	ADV
fpq-14292	218	39	be	be	AUX
fpq-14292	218	40	lowered	lower	VERB
fpq-14292	218	41	;	;	PUNCT
fpq-14292	218	42	arguably	arguably	ADV
fpq-14292	218	43	,	,	PUNCT
fpq-14292	218	44	humans	human	NOUN
fpq-14292	218	45	must	must	AUX
fpq-14292	218	46	always	always	ADV
fpq-14292	218	47	be	be	AUX
fpq-14292	218	48	kept	keep	VERB
fpq-14292	218	49	in	in	ADP
fpq-14292	218	50	the	the	DET
fpq-14292	218	51	loop	loop	NOUN
fpq-14292	218	52	in	in	ADP
fpq-14292	218	53	such	such	ADJ
fpq-14292	218	54	cases	case	NOUN
fpq-14292	218	55	)	)	PUNCT
fpq-14292	218	56	.	.	PUNCT
fpq-14292	219	1	yet	yet	ADV
fpq-14292	219	2	as	as	SCONJ
fpq-14292	219	3	applied	apply	VERB
fpq-14292	219	4	to	to	ADP
fpq-14292	219	5	word	word	NOUN
fpq-14292	219	6	-	-	PUNCT
fpq-14292	219	7	embeddings	embedding	NOUN
fpq-14292	219	8	,	,	PUNCT
fpq-14292	219	9	the	the	DET
fpq-14292	219	10	end	end	NOUN
fpq-14292	219	11	product	product	NOUN
fpq-14292	219	12	is	be	AUX
fpq-14292	219	13	less	less	ADJ
fpq-14292	219	14	a	a	DET
fpq-14292	219	15	well	well	ADV
fpq-14292	219	16	-	-	PUNCT
fpq-14292	219	17	oiled	oil	VERB
fpq-14292	219	18	machine	machine	NOUN
fpq-14292	219	19	that	that	PRON
fpq-14292	219	20	works	work	VERB
fpq-14292	219	21	by	by	ADP
fpq-14292	219	22	itself	itself	PRON
fpq-14292	219	23	and	and	CCONJ
fpq-14292	219	24	more	more	ADV
fpq-14292	219	25	an	an	DET
fpq-14292	219	26	old	old	ADJ
fpq-14292	219	27	tv	tv	NOUN
fpq-14292	219	28	that	that	PRON
fpq-14292	219	29	must	must	AUX
fpq-14292	219	30	be	be	AUX
fpq-14292	219	31	continually	continually	ADV
fpq-14292	219	32	retuned	retune	VERB
fpq-14292	219	33	for	for	ADP
fpq-14292	219	34	clear	clear	ADJ
fpq-14292	219	35	reception.25	reception.25	VERB
fpq-14292	219	36	the	the	DET
fpq-14292	219	37	retention	retention	NOUN
fpq-14292	219	38	of	of	ADP
fpq-14292	219	39	humans	human	NOUN
fpq-14292	219	40	in	in	ADP
fpq-14292	219	41	the	the	DET
fpq-14292	219	42	loop	loop	NOUN
fpq-14292	219	43	as	as	ADP
fpq-14292	219	44	a	a	DET
fpq-14292	219	45	means	means	NOUN
fpq-14292	219	46	of	of	ADP
fpq-14292	219	47	policing	police	VERB
fpq-14292	219	48	a	a	DET
fpq-14292	219	49	system	system	NOUN
fpq-14292	219	50	’s	’s	PART
fpq-14292	219	51	acquisition	acquisition	NOUN
fpq-14292	219	52	of	of	ADP
fpq-14292	219	53	bias	bias	NOUN
fpq-14292	219	54	also	also	ADV
fpq-14292	219	55	raises	raise	VERB
fpq-14292	219	56	worries	worry	NOUN
fpq-14292	219	57	about	about	ADP
fpq-14292	219	58	implicit	implicit	ADJ
fpq-14292	219	59	bias	bias	NOUN
fpq-14292	219	60	in	in	ADP
fpq-14292	219	61	human	human	ADJ
fpq-14292	219	62	cognition	cognition	NOUN
fpq-14292	219	63	.	.	PUNCT
fpq-14292	220	1	in	in	ADP
fpq-14292	220	2	particular	particular	ADJ
fpq-14292	220	3	,	,	PUNCT
fpq-14292	220	4	people	people	NOUN
fpq-14292	220	5	23	23	NUM
fpq-14292	220	6	there	there	PRON
fpq-14292	220	7	are	be	VERB
fpq-14292	220	8	also	also	ADV
fpq-14292	220	9	technical	technical	ADJ
fpq-14292	220	10	aspects	aspect	NOUN
fpq-14292	220	11	to	to	ADP
fpq-14292	220	12	bolukbasi	bolukbasi	NOUN
fpq-14292	220	13	and	and	CCONJ
fpq-14292	220	14	colleagues	colleague	NOUN
fpq-14292	220	15	’	'	PUNCT
fpq-14292	220	16	hard	hard	ADV
fpq-14292	220	17	-	-	PUNCT
fpq-14292	220	18	debiasing	debiase	VERB
fpq-14292	220	19	method	method	NOUN
fpq-14292	220	20	that	that	PRON
fpq-14292	220	21	we	we	PRON
fpq-14292	220	22	do	do	AUX
fpq-14292	220	23	not	not	PART
fpq-14292	220	24	discuss	discuss	VERB
fpq-14292	220	25	.	.	PUNCT
fpq-14292	221	1	24	24	NUM
fpq-14292	221	2	a	a	DET
fpq-14292	221	3	word	word	NOUN
fpq-14292	221	4	about	about	ADP
fpq-14292	221	5	analogy	analogy	NOUN
fpq-14292	221	6	tasks	task	NOUN
fpq-14292	221	7	:	:	PUNCT
fpq-14292	221	8	nissim	nissim	PROPN
fpq-14292	221	9	,	,	PUNCT
fpq-14292	221	10	van	van	PROPN
fpq-14292	221	11	noord	noord	PROPN
fpq-14292	221	12	,	,	PUNCT
fpq-14292	221	13	and	and	CCONJ
fpq-14292	221	14	van	van	PROPN
fpq-14292	221	15	der	der	PROPN
fpq-14292	221	16	goot	goot	NOUN
fpq-14292	221	17	(	(	PUNCT
fpq-14292	221	18	2019	2019	NUM
fpq-14292	221	19	)	)	PUNCT
fpq-14292	221	20	outline	outline	NOUN
fpq-14292	221	21	various	various	ADJ
fpq-14292	221	22	problems	problem	NOUN
fpq-14292	221	23	with	with	ADP
fpq-14292	221	24	using	use	VERB
fpq-14292	221	25	analogy	analogy	NOUN
fpq-14292	221	26	tasks	task	NOUN
fpq-14292	221	27	for	for	ADP
fpq-14292	221	28	bias	bias	NOUN
fpq-14292	221	29	detection	detection	NOUN
fpq-14292	221	30	.	.	PUNCT
fpq-14292	222	1	at	at	ADP
fpq-14292	222	2	their	their	PRON
fpq-14292	222	3	worst	bad	ADJ
fpq-14292	222	4	,	,	PUNCT
fpq-14292	222	5	some	some	DET
fpq-14292	222	6	algorithms	algorithm	NOUN
fpq-14292	222	7	ignore	ignore	VERB
fpq-14292	222	8	input	input	NOUN
fpq-14292	222	9	vectors	vector	NOUN
fpq-14292	222	10	,	,	PUNCT
fpq-14292	222	11	so	so	SCONJ
fpq-14292	222	12	that	that	SCONJ
fpq-14292	222	13	“	"	PUNCT
fpq-14292	222	14	a	a	PRON
fpq-14292	222	15	is	be	AUX
fpq-14292	222	16	to	to	ADP
fpq-14292	222	17	b	b	NOUN
fpq-14292	222	18	as	as	SCONJ
fpq-14292	222	19	c	c	PROPN
fpq-14292	222	20	is	be	AUX
fpq-14292	222	21	to	to	PART
fpq-14292	222	22	x	x	PRON
fpq-14292	222	23	”	"	PUNCT
fpq-14292	222	24	is	be	AUX
fpq-14292	222	25	often	often	ADV
fpq-14292	222	26	prevented	prevent	VERB
fpq-14292	222	27	from	from	ADP
fpq-14292	222	28	returning	return	VERB
fpq-14292	222	29	a	a	DET
fpq-14292	222	30	,	,	PUNCT
fpq-14292	222	31	b	b	NOUN
fpq-14292	222	32	,	,	PUNCT
fpq-14292	222	33	or	or	CCONJ
fpq-14292	222	34	c	c	NOUN
fpq-14292	222	35	for	for	ADP
fpq-14292	222	36	x	x	PROPN
fpq-14292	222	37	(	(	PUNCT
fpq-14292	222	38	nissim	nissim	PROPN
fpq-14292	222	39	,	,	PUNCT
fpq-14292	222	40	van	van	PROPN
fpq-14292	222	41	noord	noord	PROPN
fpq-14292	222	42	,	,	PUNCT
fpq-14292	222	43	and	and	CCONJ
fpq-14292	222	44	van	van	PROPN
fpq-14292	222	45	der	der	PROPN
fpq-14292	222	46	goot	goot	PROPN
fpq-14292	222	47	2019	2019	NUM
fpq-14292	222	48	,	,	PUNCT
fpq-14292	222	49	491–92	491–92	NOUN
fpq-14292	222	50	;	;	PUNCT
fpq-14292	222	51	cf	cf	X
fpq-14292	222	52	.	.	PUNCT
fpq-14292	223	1	schluter	schluter	PROPN
fpq-14292	223	2	2018	2018	NUM
fpq-14292	223	3	)	)	PUNCT
fpq-14292	223	4	.	.	PUNCT
fpq-14292	224	1	nissim	nissim	PROPN
fpq-14292	224	2	and	and	CCONJ
fpq-14292	224	3	colleagues	colleague	NOUN
fpq-14292	224	4	think	think	VERB
fpq-14292	224	5	using	use	VERB
fpq-14292	224	6	analogy	analogy	NOUN
fpq-14292	224	7	tasks	task	NOUN
fpq-14292	224	8	to	to	PART
fpq-14292	224	9	detect	detect	VERB
fpq-14292	224	10	bias	bias	NOUN
fpq-14292	224	11	in	in	ADP
fpq-14292	224	12	embeddings	embedding	NOUN
fpq-14292	224	13	results	result	NOUN
fpq-14292	224	14	in	in	ADP
fpq-14292	224	15	claims	claim	NOUN
fpq-14292	224	16	about	about	ADP
fpq-14292	224	17	biases	bias	NOUN
fpq-14292	224	18	in	in	ADP
fpq-14292	224	19	models	model	NOUN
fpq-14292	224	20	where	where	SCONJ
fpq-14292	224	21	the	the	DET
fpq-14292	224	22	evidence	evidence	NOUN
fpq-14292	224	23	does	do	AUX
fpq-14292	224	24	not	not	PART
fpq-14292	224	25	warrant	warrant	VERB
fpq-14292	224	26	it	it	PRON
fpq-14292	224	27	and	and	CCONJ
fpq-14292	224	28	may	may	AUX
fpq-14292	224	29	even	even	ADV
fpq-14292	224	30	conceal	conceal	VERB
fpq-14292	224	31	existing	exist	VERB
fpq-14292	224	32	biases	bias	NOUN
fpq-14292	224	33	.	.	PUNCT
fpq-14292	225	1	25	25	NUM
fpq-14292	225	2	a	a	DET
fpq-14292	225	3	solution	solution	NOUN
fpq-14292	225	4	might	might	AUX
fpq-14292	225	5	be	be	AUX
fpq-14292	225	6	not	not	PART
fpq-14292	225	7	to	to	PART
fpq-14292	225	8	use	use	VERB
fpq-14292	225	9	such	such	ADJ
fpq-14292	225	10	models	model	NOUN
fpq-14292	225	11	.	.	PUNCT
fpq-14292	226	1	we	we	PRON
fpq-14292	226	2	are	be	AUX
fpq-14292	226	3	concerned	concerned	ADJ
fpq-14292	226	4	with	with	ADP
fpq-14292	226	5	what	what	PRON
fpq-14292	226	6	to	to	PART
fpq-14292	226	7	do	do	AUX
fpq-14292	226	8	given	give	VERB
fpq-14292	226	9	that	that	SCONJ
fpq-14292	226	10	they	they	PRON
fpq-14292	226	11	are	be	AUX
fpq-14292	226	12	used	use	VERB
fpq-14292	226	13	.	.	PUNCT
fpq-14292	227	1	deery	deery	NOUN
fpq-14292	227	2	and	and	CCONJ
fpq-14292	227	3	bailey	bailey	PROPN
fpq-14292	227	4	–	–	PUNCT
fpq-14292	227	5	the	the	DET
fpq-14292	227	6	bias	bias	NOUN
fpq-14292	227	7	dilemma	dilemma	NOUN
fpq-14292	227	8	:	:	PUNCT
fpq-14292	227	9	the	the	DET
fpq-14292	227	10	ethics	ethic	NOUN
fpq-14292	227	11	of	of	ADP
fpq-14292	227	12	algorithmic	algorithmic	ADJ
fpq-14292	227	13	bias	bias	NOUN
fpq-14292	227	14	in	in	ADP
fpq-14292	227	15	natural	natural	ADJ
fpq-14292	227	16	-	-	PUNCT
fpq-14292	227	17	language	language	NOUN
fpq-14292	227	18	processing	processing	NOUN
fpq-14292	227	19	published	publish	VERB
fpq-14292	227	20	by	by	ADP
fpq-14292	227	21	scholarship@western	scholarship@western	PROPN
fpq-14292	227	22	,	,	PUNCT
fpq-14292	227	23	2022	2022	NUM
fpq-14292	227	24	13	13	NUM
fpq-14292	227	25	may	may	AUX
fpq-14292	227	26	unconsciously	unconsciously	ADV
fpq-14292	227	27	add	add	VERB
fpq-14292	227	28	biases	bias	NOUN
fpq-14292	227	29	even	even	ADV
fpq-14292	227	30	if	if	SCONJ
fpq-14292	227	31	they	they	PRON
fpq-14292	227	32	succeed	succeed	VERB
fpq-14292	227	33	in	in	ADP
fpq-14292	227	34	eliminating	eliminate	VERB
fpq-14292	227	35	others	other	NOUN
fpq-14292	227	36	(	(	PUNCT
fpq-14292	227	37	see	see	VERB
fpq-14292	227	38	caliskan	caliskan	PROPN
fpq-14292	227	39	,	,	PUNCT
fpq-14292	227	40	bryson	bryson	PROPN
fpq-14292	227	41	,	,	PUNCT
fpq-14292	227	42	and	and	CCONJ
fpq-14292	227	43	narayanan	narayanan	PROPN
fpq-14292	227	44	2017).26	2017).26	PROPN
fpq-14292	227	45	finally	finally	ADV
fpq-14292	227	46	,	,	PUNCT
fpq-14292	227	47	it	it	PRON
fpq-14292	227	48	is	be	AUX
fpq-14292	227	49	unclear	unclear	ADJ
fpq-14292	227	50	whether	whether	SCONJ
fpq-14292	227	51	we	we	PRON
fpq-14292	227	52	can	can	AUX
fpq-14292	227	53	realistically	realistically	ADV
fpq-14292	227	54	remove	remove	VERB
fpq-14292	227	55	all	all	DET
fpq-14292	227	56	morally	morally	ADV
fpq-14292	227	57	relevant	relevant	ADJ
fpq-14292	227	58	biases	bias	NOUN
fpq-14292	227	59	from	from	ADP
fpq-14292	227	60	a	a	DET
fpq-14292	227	61	word	word	NOUN
fpq-14292	227	62	-	-	PUNCT
fpq-14292	227	63	embeddings	embeddings	NOUN
fpq-14292	227	64	model	model	NOUN
fpq-14292	227	65	while	while	SCONJ
fpq-14292	227	66	preserving	preserve	VERB
fpq-14292	227	67	its	its	PRON
fpq-14292	227	68	usefulness	usefulness	NOUN
fpq-14292	227	69	.	.	PUNCT
fpq-14292	228	1	any	any	DET
fpq-14292	228	2	tweaks	tweak	NOUN
fpq-14292	228	3	made	make	VERB
fpq-14292	228	4	to	to	ADP
fpq-14292	228	5	a	a	DET
fpq-14292	228	6	pretrained	pretraine	VERB
fpq-14292	228	7	word	word	NOUN
fpq-14292	228	8	-	-	PUNCT
fpq-14292	228	9	embeddings	embeddings	ADJ
fpq-14292	228	10	model	model	NOUN
fpq-14292	228	11	to	to	PART
fpq-14292	228	12	increase	increase	VERB
fpq-14292	228	13	its	its	PRON
fpq-14292	228	14	normative	normative	ADJ
fpq-14292	228	15	correctness	correctness	NOUN
fpq-14292	228	16	in	in	ADP
fpq-14292	228	17	a	a	DET
fpq-14292	228	18	particular	particular	ADJ
fpq-14292	228	19	regard	regard	NOUN
fpq-14292	228	20	will	will	AUX
fpq-14292	228	21	result	result	VERB
fpq-14292	228	22	in	in	ADP
fpq-14292	228	23	a	a	DET
fpq-14292	228	24	less	less	ADV
fpq-14292	228	25	descriptively	descriptively	ADV
fpq-14292	228	26	accurate	accurate	ADJ
fpq-14292	228	27	model	model	NOUN
fpq-14292	228	28	regarding	regard	VERB
fpq-14292	228	29	actual	actual	ADJ
fpq-14292	228	30	language	language	NOUN
fpq-14292	228	31	use	use	NOUN
fpq-14292	228	32	,	,	PUNCT
fpq-14292	228	33	at	at	ADP
fpq-14292	228	34	least	least	ADJ
fpq-14292	228	35	in	in	ADP
fpq-14292	228	36	cases	case	NOUN
fpq-14292	228	37	where	where	SCONJ
fpq-14292	228	38	the	the	DET
fpq-14292	228	39	models	model	NOUN
fpq-14292	228	40	reflect	reflect	VERB
fpq-14292	228	41	bias	bias	NOUN
fpq-14292	228	42	.	.	PUNCT
fpq-14292	229	1	many	many	ADJ
fpq-14292	229	2	applications	application	NOUN
fpq-14292	229	3	will	will	AUX
fpq-14292	229	4	be	be	AUX
fpq-14292	229	5	negatively	negatively	ADV
fpq-14292	229	6	affected	affect	VERB
fpq-14292	229	7	by	by	ADP
fpq-14292	229	8	this	this	DET
fpq-14292	229	9	reduced	reduce	VERB
fpq-14292	229	10	accuracy	accuracy	NOUN
fpq-14292	229	11	.	.	PUNCT
fpq-14292	230	1	for	for	ADP
fpq-14292	230	2	example	example	NOUN
fpq-14292	230	3	,	,	PUNCT
fpq-14292	230	4	when	when	SCONJ
fpq-14292	230	5	pretrained	pretraine	VERB
fpq-14292	230	6	models	model	NOUN
fpq-14292	230	7	are	be	AUX
fpq-14292	230	8	used	use	VERB
fpq-14292	230	9	in	in	ADP
fpq-14292	230	10	sentiment	sentiment	NOUN
fpq-14292	230	11	analysis	analysis	NOUN
fpq-14292	230	12	and	and	CCONJ
fpq-14292	230	13	text	text	NOUN
fpq-14292	230	14	classification	classification	NOUN
fpq-14292	230	15	,	,	PUNCT
fpq-14292	230	16	which	which	PRON
fpq-14292	230	17	rely	rely	VERB
fpq-14292	230	18	on	on	ADP
fpq-14292	230	19	descriptively	descriptively	ADV
fpq-14292	230	20	accurate	accurate	ADJ
fpq-14292	230	21	representations	representation	NOUN
fpq-14292	230	22	,	,	PUNCT
fpq-14292	230	23	these	these	DET
fpq-14292	230	24	models	model	NOUN
fpq-14292	230	25	seem	seem	VERB
fpq-14292	230	26	a	a	DET
fpq-14292	230	27	poor	poor	ADJ
fpq-14292	230	28	fit	fit	NOUN
fpq-14292	230	29	for	for	ADP
fpq-14292	230	30	those	those	DET
fpq-14292	230	31	tasks	task	NOUN
fpq-14292	230	32	.	.	PUNCT
fpq-14292	231	1	such	such	ADJ
fpq-14292	231	2	practical	practical	ADJ
fpq-14292	231	3	problems	problem	NOUN
fpq-14292	231	4	with	with	ADP
fpq-14292	231	5	debiasing	debiase	VERB
fpq-14292	231	6	suggest	suggest	VERB
fpq-14292	231	7	that	that	SCONJ
fpq-14292	231	8	a	a	DET
fpq-14292	231	9	more	more	ADV
fpq-14292	231	10	effective	effective	ADJ
fpq-14292	231	11	way	way	NOUN
fpq-14292	231	12	to	to	PART
fpq-14292	231	13	control	control	VERB
fpq-14292	231	14	for	for	ADP
fpq-14292	231	15	the	the	DET
fpq-14292	231	16	negative	negative	ADJ
fpq-14292	231	17	effects	effect	NOUN
fpq-14292	231	18	of	of	ADP
fpq-14292	231	19	bias	bias	NOUN
fpq-14292	231	20	in	in	ADP
fpq-14292	231	21	word	word	NOUN
fpq-14292	231	22	-	-	PUNCT
fpq-14292	231	23	embeddings	embedding	NOUN
fpq-14292	231	24	might	might	AUX
fpq-14292	231	25	instead	instead	ADV
fpq-14292	231	26	be	be	AUX
fpq-14292	231	27	to	to	PART
fpq-14292	231	28	see	see	VERB
fpq-14292	231	29	how	how	SCONJ
fpq-14292	231	30	bias	bias	NOUN
fpq-14292	231	31	manifests	manifest	VERB
fpq-14292	231	32	itself	itself	PRON
fpq-14292	231	33	in	in	ADP
fpq-14292	231	34	a	a	DET
fpq-14292	231	35	system	system	NOUN
fpq-14292	231	36	’s	’s	PART
fpq-14292	231	37	outputs	output	NOUN
fpq-14292	231	38	and	and	CCONJ
fpq-14292	231	39	control	control	NOUN
fpq-14292	231	40	for	for	ADP
fpq-14292	231	41	bias	bias	NOUN
fpq-14292	231	42	there	there	ADV
fpq-14292	231	43	to	to	PART
fpq-14292	231	44	reduce	reduce	VERB
fpq-14292	231	45	the	the	DET
fpq-14292	231	46	negative	negative	ADJ
fpq-14292	231	47	effects	effect	NOUN
fpq-14292	231	48	.	.	PUNCT
fpq-14292	232	1	6	6	X
fpq-14292	232	2	.	.	PUNCT
fpq-14292	232	3	applications	application	NOUN
fpq-14292	232	4	controlling	control	VERB
fpq-14292	232	5	for	for	ADP
fpq-14292	232	6	bias	bias	NOUN
fpq-14292	232	7	at	at	ADP
fpq-14292	232	8	the	the	DET
fpq-14292	232	9	output	output	NOUN
fpq-14292	232	10	stage	stage	NOUN
fpq-14292	232	11	means	mean	VERB
fpq-14292	232	12	we	we	PRON
fpq-14292	232	13	can	can	AUX
fpq-14292	232	14	retain	retain	VERB
fpq-14292	232	15	descriptive	descriptive	ADJ
fpq-14292	232	16	accuracy	accuracy	NOUN
fpq-14292	232	17	.	.	PUNCT
fpq-14292	233	1	exactly	exactly	ADV
fpq-14292	233	2	how	how	SCONJ
fpq-14292	233	3	we	we	PRON
fpq-14292	233	4	decide	decide	VERB
fpq-14292	233	5	to	to	PART
fpq-14292	233	6	control	control	VERB
fpq-14292	233	7	for	for	ADP
fpq-14292	233	8	bias	bias	NOUN
fpq-14292	233	9	in	in	ADP
fpq-14292	233	10	this	this	DET
fpq-14292	233	11	way	way	NOUN
fpq-14292	233	12	will	will	AUX
fpq-14292	233	13	,	,	PUNCT
fpq-14292	233	14	of	of	ADP
fpq-14292	233	15	course	course	NOUN
fpq-14292	233	16	,	,	PUNCT
fpq-14292	233	17	depend	depend	VERB
fpq-14292	233	18	on	on	ADP
fpq-14292	233	19	the	the	DET
fpq-14292	233	20	application	application	NOUN
fpq-14292	233	21	in	in	ADP
fpq-14292	233	22	question	question	NOUN
fpq-14292	233	23	.	.	PUNCT
fpq-14292	234	1	let	let	VERB
fpq-14292	234	2	us	we	PRON
fpq-14292	234	3	therefore	therefore	ADV
fpq-14292	234	4	consider	consider	VERB
fpq-14292	234	5	some	some	DET
fpq-14292	234	6	applications	application	NOUN
fpq-14292	234	7	.	.	PUNCT
fpq-14292	235	1	6.1	6.1	NUM
fpq-14292	235	2	.	.	PUNCT
fpq-14292	235	3	translation	translation	NOUN
fpq-14292	235	4	machine	machine	NOUN
fpq-14292	235	5	translation	translation	NOUN
fpq-14292	235	6	is	be	AUX
fpq-14292	235	7	the	the	DET
fpq-14292	235	8	oldest	old	ADJ
fpq-14292	235	9	nlp	nlp	ADJ
fpq-14292	235	10	task	task	NOUN
fpq-14292	235	11	and	and	CCONJ
fpq-14292	235	12	is	be	AUX
fpq-14292	235	13	often	often	ADV
fpq-14292	235	14	regarded	regard	VERB
fpq-14292	235	15	as	as	ADP
fpq-14292	235	16	one	one	NUM
fpq-14292	235	17	of	of	ADP
fpq-14292	235	18	the	the	DET
fpq-14292	235	19	field	field	NOUN
fpq-14292	235	20	’s	’s	PART
fpq-14292	235	21	major	major	ADJ
fpq-14292	235	22	successes	success	NOUN
fpq-14292	235	23	.	.	PUNCT
fpq-14292	236	1	deep	deep	ADJ
fpq-14292	236	2	-	-	PUNCT
fpq-14292	236	3	learning	learn	VERB
fpq-14292	236	4	techniques	technique	NOUN
fpq-14292	236	5	similar	similar	ADJ
fpq-14292	236	6	to	to	ADP
fpq-14292	236	7	those	those	PRON
fpq-14292	236	8	employed	employ	VERB
fpq-14292	236	9	in	in	ADP
fpq-14292	236	10	the	the	DET
fpq-14292	236	11	training	training	NOUN
fpq-14292	236	12	of	of	ADP
fpq-14292	236	13	large	large	ADJ
fpq-14292	236	14	language	language	NOUN
fpq-14292	236	15	models	model	NOUN
fpq-14292	236	16	such	such	ADJ
fpq-14292	236	17	as	as	ADP
fpq-14292	236	18	word2vec	word2vec	PRON
fpq-14292	236	19	and	and	CCONJ
fpq-14292	236	20	bert	bert	PROPN
fpq-14292	236	21	are	be	AUX
fpq-14292	236	22	employed	employ	VERB
fpq-14292	236	23	in	in	ADP
fpq-14292	236	24	today	today	NOUN
fpq-14292	236	25	’s	’s	PART
fpq-14292	236	26	most	most	ADV
fpq-14292	236	27	commonly	commonly	ADV
fpq-14292	236	28	used	use	VERB
fpq-14292	236	29	machine	machine	NOUN
fpq-14292	236	30	-	-	PUNCT
fpq-14292	236	31	translation	translation	NOUN
fpq-14292	236	32	systems	system	NOUN
fpq-14292	236	33	,	,	PUNCT
fpq-14292	236	34	such	such	ADJ
fpq-14292	236	35	as	as	ADP
fpq-14292	236	36	google	google	PROPN
fpq-14292	236	37	translate	translate	NOUN
fpq-14292	236	38	.	.	PUNCT
fpq-14292	237	1	one	one	NUM
fpq-14292	237	2	difference	difference	NOUN
fpq-14292	237	3	is	be	AUX
fpq-14292	237	4	that	that	SCONJ
fpq-14292	237	5	these	these	DET
fpq-14292	237	6	systems	system	NOUN
fpq-14292	237	7	are	be	AUX
fpq-14292	237	8	trained	train	VERB
fpq-14292	237	9	end	end	NOUN
fpq-14292	237	10	-	-	PUNCT
fpq-14292	237	11	to	to	ADP
fpq-14292	237	12	-	-	PUNCT
fpq-14292	237	13	end	end	NOUN
fpq-14292	237	14	,	,	PUNCT
fpq-14292	237	15	meaning	mean	VERB
fpq-14292	237	16	that	that	SCONJ
fpq-14292	237	17	there	there	PRON
fpq-14292	237	18	is	be	VERB
fpq-14292	237	19	no	no	DET
fpq-14292	237	20	breaking	breaking	NOUN
fpq-14292	237	21	up	up	ADP
fpq-14292	237	22	of	of	ADP
fpq-14292	237	23	the	the	DET
fpq-14292	237	24	problem	problem	NOUN
fpq-14292	237	25	into	into	ADP
fpq-14292	237	26	separate	separate	ADJ
fpq-14292	237	27	steps	step	NOUN
fpq-14292	237	28	whereby	whereby	SCONJ
fpq-14292	237	29	a	a	DET
fpq-14292	237	30	language	language	NOUN
fpq-14292	237	31	model	model	NOUN
fpq-14292	237	32	is	be	AUX
fpq-14292	237	33	first	first	ADV
fpq-14292	237	34	trained	train	VERB
fpq-14292	237	35	on	on	ADP
fpq-14292	237	36	a	a	DET
fpq-14292	237	37	separate	separate	ADJ
fpq-14292	237	38	training	training	NOUN
fpq-14292	237	39	task	task	NOUN
fpq-14292	237	40	and	and	CCONJ
fpq-14292	237	41	only	only	ADV
fpq-14292	237	42	later	later	ADV
fpq-14292	237	43	fine	fine	ADV
fpq-14292	237	44	-	-	PUNCT
fpq-14292	237	45	tuned	tune	VERB
fpq-14292	237	46	for	for	ADP
fpq-14292	237	47	the	the	DET
fpq-14292	237	48	task	task	NOUN
fpq-14292	237	49	of	of	ADP
fpq-14292	237	50	translation	translation	NOUN
fpq-14292	237	51	.	.	PUNCT
fpq-14292	238	1	instead	instead	ADV
fpq-14292	238	2	,	,	PUNCT
fpq-14292	238	3	the	the	DET
fpq-14292	238	4	translation	translation	NOUN
fpq-14292	238	5	task	task	NOUN
fpq-14292	238	6	is	be	AUX
fpq-14292	238	7	the	the	DET
fpq-14292	238	8	training	training	NOUN
fpq-14292	238	9	task	task	NOUN
fpq-14292	238	10	(	(	PUNCT
fpq-14292	238	11	wu	wu	PROPN
fpq-14292	238	12	et	et	PROPN
fpq-14292	238	13	al	al	PROPN
fpq-14292	238	14	.	.	PROPN
fpq-14292	238	15	2016	2016	NUM
fpq-14292	238	16	)	)	PUNCT
fpq-14292	238	17	.	.	PUNCT
fpq-14292	239	1	in	in	ADP
fpq-14292	239	2	spite	spite	NOUN
fpq-14292	239	3	of	of	ADP
fpq-14292	239	4	their	their	PRON
fpq-14292	239	5	successes	success	NOUN
fpq-14292	239	6	,	,	PUNCT
fpq-14292	239	7	machine	machine	NOUN
fpq-14292	239	8	-	-	PUNCT
fpq-14292	239	9	translation	translation	NOUN
fpq-14292	239	10	systems	system	NOUN
fpq-14292	239	11	still	still	ADV
fpq-14292	239	12	encounter	encounter	VERB
fpq-14292	239	13	trouble	trouble	NOUN
fpq-14292	239	14	with	with	ADP
fpq-14292	239	15	pronouns	pronoun	NOUN
fpq-14292	239	16	.	.	PUNCT
fpq-14292	240	1	douglas	douglas	PROPN
fpq-14292	240	2	hofstadter	hofstadter	PROPN
fpq-14292	240	3	(	(	PUNCT
fpq-14292	240	4	2018	2018	NUM
fpq-14292	240	5	)	)	PUNCT
fpq-14292	240	6	has	have	AUX
fpq-14292	240	7	written	write	VERB
fpq-14292	240	8	about	about	ADP
fpq-14292	240	9	the	the	DET
fpq-14292	240	10	“	"	PUNCT
fpq-14292	240	11	shallowness	shallowness	NOUN
fpq-14292	240	12	”	"	PUNCT
fpq-14292	240	13	of	of	ADP
fpq-14292	240	14	machine	machine	NOUN
fpq-14292	240	15	translation	translation	NOUN
fpq-14292	240	16	,	,	PUNCT
fpq-14292	240	17	pointing	point	VERB
fpq-14292	240	18	to	to	ADP
fpq-14292	240	19	such	such	ADJ
fpq-14292	240	20	systems	system	NOUN
fpq-14292	240	21	’	'	PUNCT
fpq-14292	240	22	inability	inability	NOUN
fpq-14292	240	23	to	to	PART
fpq-14292	240	24	understand	understand	VERB
fpq-14292	240	25	what	what	PRON
fpq-14292	240	26	they	they	PRON
fpq-14292	240	27	are	be	AUX
fpq-14292	240	28	translating	translate	VERB
fpq-14292	240	29	.	.	PUNCT
fpq-14292	241	1	many	many	ADJ
fpq-14292	241	2	of	of	ADP
fpq-14292	241	3	hofstadter	hofstadter	NOUN
fpq-14292	241	4	’s	’s	PART
fpq-14292	241	5	examples	example	NOUN
fpq-14292	241	6	are	be	AUX
fpq-14292	241	7	of	of	ADP
fpq-14292	241	8	mistranslating	mistranslate	VERB
fpq-14292	241	9	pronouns	pronoun	NOUN
fpq-14292	241	10	.	.	PUNCT
fpq-14292	242	1	the	the	DET
fpq-14292	242	2	reason	reason	NOUN
fpq-14292	242	3	pronouns	pronoun	NOUN
fpq-14292	242	4	are	be	AUX
fpq-14292	242	5	so	so	ADV
fpq-14292	242	6	difficult	difficult	ADJ
fpq-14292	242	7	is	be	AUX
fpq-14292	242	8	that	that	PRON
fpq-14292	242	9	to	to	PART
fpq-14292	242	10	translate	translate	VERB
fpq-14292	242	11	a	a	DET
fpq-14292	242	12	pronoun	pronoun	NOUN
fpq-14292	242	13	correctly	correctly	ADV
fpq-14292	242	14	,	,	PUNCT
fpq-14292	242	15	a	a	DET
fpq-14292	242	16	system	system	NOUN
fpq-14292	242	17	must	must	AUX
fpq-14292	242	18	26	26	NUM
fpq-14292	242	19	bolukbasi	bolukbasi	NOUN
fpq-14292	242	20	et	et	PROPN
fpq-14292	242	21	al	al	PROPN
fpq-14292	242	22	.	.	PROPN
fpq-14292	243	1	(	(	PUNCT
fpq-14292	243	2	2016	2016	NUM
fpq-14292	243	3	)	)	PUNCT
fpq-14292	243	4	concede	concede	VERB
fpq-14292	243	5	that	that	SCONJ
fpq-14292	243	6	their	their	PRON
fpq-14292	243	7	method	method	NOUN
fpq-14292	243	8	is	be	AUX
fpq-14292	243	9	limited	limited	ADJ
fpq-14292	243	10	.	.	PUNCT
fpq-14292	244	1	gonen	gonen	PROPN
fpq-14292	244	2	and	and	CCONJ
fpq-14292	244	3	goldberg	goldberg	PROPN
fpq-14292	244	4	(	(	PUNCT
fpq-14292	244	5	2019	2019	NUM
fpq-14292	244	6	)	)	PUNCT
fpq-14292	244	7	argue	argue	VERB
fpq-14292	244	8	that	that	SCONJ
fpq-14292	244	9	even	even	ADV
fpq-14292	244	10	the	the	DET
fpq-14292	244	11	technical	technical	ADJ
fpq-14292	244	12	aspects	aspect	NOUN
fpq-14292	244	13	of	of	ADP
fpq-14292	244	14	bolukbasi	bolukbasi	NOUN
fpq-14292	244	15	and	and	CCONJ
fpq-14292	244	16	colleagues	colleague	NOUN
fpq-14292	244	17	’	'	PUNCT
fpq-14292	244	18	method	method	NOUN
fpq-14292	244	19	serve	serve	VERB
fpq-14292	244	20	primarily	primarily	ADV
fpq-14292	244	21	to	to	PART
fpq-14292	244	22	mask	mask	VERB
fpq-14292	244	23	the	the	DET
fpq-14292	244	24	problem	problem	NOUN
fpq-14292	244	25	of	of	ADP
fpq-14292	244	26	gender	gender	NOUN
fpq-14292	244	27	bias	bias	NOUN
fpq-14292	244	28	without	without	ADP
fpq-14292	244	29	actually	actually	ADV
fpq-14292	244	30	eliminating	eliminate	VERB
fpq-14292	244	31	it	it	PRON
fpq-14292	244	32	.	.	PUNCT
fpq-14292	245	1	they	they	PRON
fpq-14292	245	2	conclude	conclude	VERB
fpq-14292	245	3	that	that	SCONJ
fpq-14292	245	4	,	,	PUNCT
fpq-14292	245	5	“	"	PUNCT
fpq-14292	245	6	while	while	SCONJ
fpq-14292	245	7	the	the	DET
fpq-14292	245	8	bias	bias	NOUN
fpq-14292	245	9	is	be	AUX
fpq-14292	245	10	indeed	indeed	ADV
fpq-14292	245	11	substantially	substantially	ADV
fpq-14292	245	12	reduced	reduce	VERB
fpq-14292	245	13	…	…	PUNCT
fpq-14292	245	14	,	,	PUNCT
fpq-14292	245	15	the	the	DET
fpq-14292	245	16	…	…	PUNCT
fpq-14292	245	17	effect	effect	NOUN
fpq-14292	245	18	is	be	AUX
fpq-14292	245	19	mostly	mostly	ADV
fpq-14292	245	20	hiding	hide	VERB
fpq-14292	245	21	the	the	DET
fpq-14292	245	22	bias	bias	NOUN
fpq-14292	245	23	,	,	PUNCT
fpq-14292	245	24	not	not	PART
fpq-14292	245	25	removing	remove	VERB
fpq-14292	245	26	it	it	PRON
fpq-14292	245	27	”	"	PUNCT
fpq-14292	245	28	(	(	PUNCT
fpq-14292	245	29	gonen	gonen	NOUN
fpq-14292	245	30	and	and	CCONJ
fpq-14292	245	31	goldberg	goldberg	PROPN
fpq-14292	245	32	2019	2019	NUM
fpq-14292	245	33	,	,	PUNCT
fpq-14292	245	34	609	609	NUM
fpq-14292	245	35	)	)	PUNCT
fpq-14292	245	36	.	.	PUNCT
fpq-14292	246	1	feminist	feminist	PROPN
fpq-14292	246	2	philosophy	philosophy	PROPN
fpq-14292	246	3	quarterly	quarterly	ADV
fpq-14292	246	4	,	,	PUNCT
fpq-14292	246	5	2022	2022	NUM
fpq-14292	246	6	,	,	PUNCT
fpq-14292	246	7	vol	vol	NOUN
fpq-14292	246	8	.	.	NOUN
fpq-14292	246	9	8	8	NUM
fpq-14292	246	10	,	,	PUNCT
fpq-14292	246	11	iss	iss	PROPN
fpq-14292	246	12	.	.	PUNCT
fpq-14292	246	13	3/4	3/4	NUM
fpq-14292	246	14	,	,	PUNCT
fpq-14292	246	15	article	article	NOUN
fpq-14292	246	16	1	1	NUM
fpq-14292	246	17	published	publish	VERB
fpq-14292	246	18	by	by	ADP
fpq-14292	246	19	scholarship@western	scholarship@western	PROPN
fpq-14292	246	20	,	,	PUNCT
fpq-14292	246	21	2022	2022	NUM
fpq-14292	246	22	14	14	NUM
fpq-14292	246	23	know	know	VERB
fpq-14292	246	24	what	what	PRON
fpq-14292	246	25	it	it	PRON
fpq-14292	246	26	refers	refer	VERB
fpq-14292	246	27	to	to	ADP
fpq-14292	246	28	.	.	PUNCT
fpq-14292	247	1	with	with	ADP
fpq-14292	247	2	gendered	gendere	VERB
fpq-14292	247	3	pronouns	pronoun	NOUN
fpq-14292	247	4	,	,	PUNCT
fpq-14292	247	5	the	the	DET
fpq-14292	247	6	problem	problem	NOUN
fpq-14292	247	7	is	be	AUX
fpq-14292	247	8	clear	clear	ADJ
fpq-14292	247	9	.	.	PUNCT
fpq-14292	248	1	some	some	DET
fpq-14292	248	2	languages	language	NOUN
fpq-14292	248	3	do	do	AUX
fpq-14292	248	4	not	not	PART
fpq-14292	248	5	have	have	AUX
fpq-14292	248	6	gendered	gendere	VERB
fpq-14292	248	7	pronouns	pronoun	NOUN
fpq-14292	248	8	,	,	PUNCT
fpq-14292	248	9	and	and	CCONJ
fpq-14292	248	10	those	those	PRON
fpq-14292	248	11	that	that	PRON
fpq-14292	248	12	do	do	VERB
fpq-14292	248	13	can	can	AUX
fpq-14292	248	14	behave	behave	VERB
fpq-14292	248	15	very	very	ADV
fpq-14292	248	16	differently	differently	ADV
fpq-14292	248	17	—	—	PUNCT
fpq-14292	248	18	for	for	ADP
fpq-14292	248	19	example	example	NOUN
fpq-14292	248	20	,	,	PUNCT
fpq-14292	248	21	with	with	ADP
fpq-14292	248	22	possessive	possessive	ADJ
fpq-14292	248	23	pronouns	pronoun	NOUN
fpq-14292	248	24	in	in	ADP
fpq-14292	248	25	french	french	PROPN
fpq-14292	248	26	,	,	PUNCT
fpq-14292	248	27	the	the	DET
fpq-14292	248	28	gender	gender	NOUN
fpq-14292	248	29	agrees	agree	VERB
fpq-14292	248	30	with	with	ADP
fpq-14292	248	31	the	the	DET
fpq-14292	248	32	noun	noun	NOUN
fpq-14292	248	33	,	,	PUNCT
fpq-14292	248	34	not	not	PART
fpq-14292	248	35	the	the	DET
fpq-14292	248	36	possessor	possessor	NOUN
fpq-14292	248	37	(	(	PUNCT
fpq-14292	248	38	as	as	ADP
fpq-14292	248	39	in	in	ADP
fpq-14292	248	40	english	english	PROPN
fpq-14292	248	41	)	)	PUNCT
fpq-14292	248	42	.	.	PUNCT
fpq-14292	249	1	hofstadter	hofstadter	NOUN
fpq-14292	249	2	illustrates	illustrate	VERB
fpq-14292	249	3	the	the	DET
fpq-14292	249	4	problem	problem	NOUN
fpq-14292	249	5	as	as	SCONJ
fpq-14292	249	6	follows	follow	VERB
fpq-14292	249	7	:	:	PUNCT
fpq-14292	249	8	i	i	PRON
fpq-14292	249	9	began	begin	VERB
fpq-14292	249	10	my	my	PRON
fpq-14292	249	11	explorations	exploration	NOUN
fpq-14292	249	12	.	.	PUNCT
fpq-14292	249	13	.	.	PUNCT
fpq-14292	250	1	.	.	PUNCT
fpq-14292	251	1	using	use	VERB
fpq-14292	251	2	the	the	DET
fpq-14292	251	3	following	follow	VERB
fpq-14292	251	4	short	short	ADJ
fpq-14292	251	5	remark	remark	NOUN
fpq-14292	251	6	.	.	PUNCT
fpq-14292	251	7	.	.	PUNCT
fpq-14292	252	1	.	.	PUNCT
fpq-14292	253	1	in	in	ADP
fpq-14292	253	2	their	their	PRON
fpq-14292	253	3	house	house	NOUN
fpq-14292	253	4	,	,	PUNCT
fpq-14292	253	5	everything	everything	PRON
fpq-14292	253	6	comes	come	VERB
fpq-14292	253	7	in	in	ADP
fpq-14292	253	8	pairs	pair	NOUN
fpq-14292	253	9	.	.	PUNCT
fpq-14292	254	1	there	there	PRON
fpq-14292	254	2	’s	’	VERB
fpq-14292	254	3	his	his	PRON
fpq-14292	254	4	car	car	NOUN
fpq-14292	254	5	and	and	CCONJ
fpq-14292	254	6	her	her	PRON
fpq-14292	254	7	car	car	NOUN
fpq-14292	254	8	,	,	PUNCT
fpq-14292	254	9	his	his	PRON
fpq-14292	254	10	towels	towel	NOUN
fpq-14292	254	11	and	and	CCONJ
fpq-14292	254	12	her	her	PRON
fpq-14292	254	13	towels	towel	NOUN
fpq-14292	254	14	,	,	PUNCT
fpq-14292	254	15	and	and	CCONJ
fpq-14292	254	16	his	his	PRON
fpq-14292	254	17	library	library	NOUN
fpq-14292	254	18	and	and	CCONJ
fpq-14292	254	19	hers	her	NOUN
fpq-14292	254	20	.	.	PUNCT
fpq-14292	254	21	.	.	PUNCT
fpq-14292	254	22	.	.	PUNCT
fpq-14292	254	23	.	.	PUNCT
fpq-14292	255	1	here	here	ADV
fpq-14292	255	2	’s	’	VERB
fpq-14292	255	3	what	what	PRON
fpq-14292	255	4	google	google	PROPN
fpq-14292	255	5	translate	translate	NOUN
fpq-14292	255	6	gave	give	VERB
fpq-14292	255	7	me	i	PRON
fpq-14292	256	1	[	[	X
fpq-14292	256	2	translating	translate	VERB
fpq-14292	256	3	into	into	ADP
fpq-14292	256	4	french	french	PROPN
fpq-14292	256	5	]	]	PUNCT
fpq-14292	256	6	:	:	PUNCT
fpq-14292	256	7	dans	dans	PROPN
fpq-14292	256	8	leur	leur	PROPN
fpq-14292	256	9	maison	maison	PROPN
fpq-14292	256	10	,	,	PUNCT
fpq-14292	256	11	tout	tout	X
fpq-14292	256	12	vient	vient	PROPN
fpq-14292	256	13	en	en	ADP
fpq-14292	256	14	paires	paire	NOUN
fpq-14292	256	15	.	.	PUNCT
fpq-14292	257	1	il	il	PROPN
fpq-14292	257	2	y	y	PROPN
fpq-14292	257	3	a	a	DET
fpq-14292	257	4	sa	sa	PROPN
fpq-14292	257	5	voiture	voiture	PROPN
fpq-14292	257	6	et	et	PROPN
fpq-14292	257	7	sa	sa	PROPN
fpq-14292	257	8	voiture	voiture	PROPN
fpq-14292	257	9	,	,	PUNCT
fpq-14292	257	10	ses	ses	PROPN
fpq-14292	257	11	serviettes	serviette	NOUN
fpq-14292	257	12	et	et	PROPN
fpq-14292	257	13	ses	ses	PROPN
fpq-14292	257	14	serviettes	serviette	NOUN
fpq-14292	257	15	,	,	PUNCT
fpq-14292	257	16	sa	sa	PROPN
fpq-14292	257	17	bibliothèque	bibliothèque	PROPN
fpq-14292	257	18	et	et	PROPN
fpq-14292	257	19	les	les	PROPN
fpq-14292	257	20	siennes	sienne	NOUN
fpq-14292	257	21	.	.	PUNCT
fpq-14292	258	1	the	the	DET
fpq-14292	258	2	program	program	NOUN
fpq-14292	258	3	fell	fall	VERB
fpq-14292	258	4	into	into	ADP
fpq-14292	258	5	my	my	PRON
fpq-14292	258	6	trap	trap	NOUN
fpq-14292	258	7	,	,	PUNCT
fpq-14292	258	8	not	not	PART
fpq-14292	258	9	realizing	realize	VERB
fpq-14292	258	10	,	,	PUNCT
fpq-14292	258	11	as	as	SCONJ
fpq-14292	258	12	any	any	DET
fpq-14292	258	13	human	human	ADJ
fpq-14292	258	14	reader	reader	NOUN
fpq-14292	258	15	would	would	AUX
fpq-14292	258	16	,	,	PUNCT
fpq-14292	258	17	that	that	SCONJ
fpq-14292	258	18	i	i	PRON
fpq-14292	258	19	was	be	AUX
fpq-14292	258	20	describing	describe	VERB
fpq-14292	258	21	a	a	DET
fpq-14292	258	22	couple	couple	NOUN
fpq-14292	258	23	,	,	PUNCT
fpq-14292	258	24	stressing	stress	VERB
fpq-14292	258	25	that	that	SCONJ
fpq-14292	258	26	for	for	ADP
fpq-14292	258	27	each	each	DET
fpq-14292	258	28	item	item	NOUN
fpq-14292	258	29	he	he	PRON
fpq-14292	258	30	had	have	VERB
fpq-14292	258	31	,	,	PUNCT
fpq-14292	258	32	she	she	PRON
fpq-14292	258	33	had	have	VERB
fpq-14292	258	34	a	a	DET
fpq-14292	258	35	similar	similar	ADJ
fpq-14292	258	36	one	one	NOUN
fpq-14292	258	37	.	.	PUNCT
fpq-14292	258	38	.	.	PUNCT
fpq-14292	258	39	.	.	PUNCT
fpq-14292	258	40	.	.	PUNCT
fpq-14292	259	1	the	the	DET
fpq-14292	259	2	deep	deep	ADV
fpq-14292	259	3	-	-	PUNCT
fpq-14292	259	4	learning	learn	VERB
fpq-14292	259	5	engine	engine	NOUN
fpq-14292	259	6	used	use	VERB
fpq-14292	259	7	the	the	DET
fpq-14292	259	8	word	word	NOUN
fpq-14292	259	9	sa	sa	NOUN
fpq-14292	259	10	for	for	ADP
fpq-14292	259	11	both	both	DET
fpq-14292	259	12	“	"	PUNCT
fpq-14292	259	13	his	his	PRON
fpq-14292	259	14	car	car	NOUN
fpq-14292	259	15	”	"	PUNCT
fpq-14292	259	16	and	and	CCONJ
fpq-14292	259	17	“	"	PUNCT
fpq-14292	259	18	her	her	PRON
fpq-14292	259	19	car	car	NOUN
fpq-14292	259	20	,	,	PUNCT
fpq-14292	259	21	”	"	PUNCT
fpq-14292	259	22	so	so	SCONJ
fpq-14292	259	23	you	you	PRON
fpq-14292	259	24	ca	can	AUX
fpq-14292	259	25	n’t	not	PART
fpq-14292	259	26	tell	tell	VERB
fpq-14292	259	27	anything	anything	PRON
fpq-14292	259	28	about	about	ADP
fpq-14292	259	29	either	either	CCONJ
fpq-14292	259	30	car	car	NOUN
fpq-14292	259	31	owner	owner	NOUN
fpq-14292	259	32	’s	’s	PART
fpq-14292	259	33	gender	gender	NOUN
fpq-14292	259	34	.	.	PUNCT
fpq-14292	260	1	likewise	likewise	ADV
fpq-14292	260	2	,	,	PUNCT
fpq-14292	260	3	it	it	PRON
fpq-14292	260	4	used	use	VERB
fpq-14292	260	5	the	the	DET
fpq-14292	260	6	genderless	genderless	ADJ
fpq-14292	260	7	plural	plural	ADJ
fpq-14292	260	8	ses	se	NOUN
fpq-14292	260	9	both	both	PRON
fpq-14292	260	10	for	for	ADP
fpq-14292	260	11	“	"	PUNCT
fpq-14292	260	12	his	his	PRON
fpq-14292	260	13	towels	towel	NOUN
fpq-14292	260	14	”	"	PUNCT
fpq-14292	260	15	and	and	CCONJ
fpq-14292	260	16	“	"	PUNCT
fpq-14292	260	17	her	her	PRON
fpq-14292	260	18	towels	towel	NOUN
fpq-14292	260	19	,	,	PUNCT
fpq-14292	260	20	”	"	PUNCT
fpq-14292	260	21	and	and	CCONJ
fpq-14292	260	22	in	in	ADP
fpq-14292	260	23	the	the	DET
fpq-14292	260	24	last	last	ADJ
fpq-14292	260	25	case	case	NOUN
fpq-14292	260	26	of	of	ADP
fpq-14292	260	27	the	the	DET
fpq-14292	260	28	two	two	NUM
fpq-14292	260	29	libraries	library	NOUN
fpq-14292	260	30	,	,	PUNCT
fpq-14292	260	31	his	his	PRON
fpq-14292	260	32	and	and	CCONJ
fpq-14292	260	33	hers	her	NOUN
fpq-14292	260	34	,	,	PUNCT
fpq-14292	260	35	it	it	PRON
fpq-14292	260	36	got	get	AUX
fpq-14292	260	37	thrown	throw	VERB
fpq-14292	260	38	by	by	ADP
fpq-14292	260	39	the	the	DET
fpq-14292	260	40	final	final	ADJ
fpq-14292	260	41	s	s	X
fpq-14292	260	42	in	in	ADP
fpq-14292	260	43	“	"	PUNCT
fpq-14292	260	44	hers	hers	NOUN
fpq-14292	260	45	”	"	PUNCT
fpq-14292	260	46	and	and	CCONJ
fpq-14292	260	47	somehow	somehow	ADV
fpq-14292	260	48	decided	decide	VERB
fpq-14292	260	49	that	that	SCONJ
fpq-14292	260	50	that	that	PRON
fpq-14292	260	51	s	s	AUX
fpq-14292	260	52	represented	represent	VERB
fpq-14292	260	53	a	a	DET
fpq-14292	260	54	plural	plural	NOUN
fpq-14292	260	55	(	(	PUNCT
fpq-14292	260	56	“	"	PUNCT
fpq-14292	260	57	les	les	X
fpq-14292	260	58	siennes	sienne	NOUN
fpq-14292	260	59	”	"	PUNCT
fpq-14292	260	60	)	)	PUNCT
fpq-14292	260	61	.	.	PUNCT
fpq-14292	261	1	google	google	PROPN
fpq-14292	261	2	translate	translate	PROPN
fpq-14292	261	3	’s	’s	PART
fpq-14292	261	4	french	french	ADJ
fpq-14292	261	5	sentence	sentence	NOUN
fpq-14292	261	6	missed	miss	VERB
fpq-14292	261	7	the	the	DET
fpq-14292	261	8	whole	whole	ADJ
fpq-14292	261	9	point	point	NOUN
fpq-14292	261	10	.	.	PUNCT
fpq-14292	262	1	(	(	PUNCT
fpq-14292	262	2	hofstadter	hofstadter	NOUN
fpq-14292	262	3	2018	2018	NUM
fpq-14292	262	4	;	;	PUNCT
fpq-14292	262	5	bold	bold	ADJ
fpq-14292	262	6	emphasis	emphasis	NOUN
fpq-14292	262	7	added	add	VERB
fpq-14292	262	8	)	)	PUNCT
fpq-14292	262	9	data	data	NOUN
fpq-14292	262	10	-	-	PUNCT
fpq-14292	262	11	driven	drive	VERB
fpq-14292	262	12	approaches	approach	NOUN
fpq-14292	262	13	seem	seem	VERB
fpq-14292	262	14	unlikely	unlikely	ADJ
fpq-14292	262	15	to	to	PART
fpq-14292	262	16	solve	solve	VERB
fpq-14292	262	17	this	this	DET
fpq-14292	262	18	sort	sort	NOUN
fpq-14292	262	19	of	of	ADP
fpq-14292	262	20	problem	problem	NOUN
fpq-14292	262	21	.	.	PUNCT
fpq-14292	263	1	yet	yet	ADV
fpq-14292	263	2	gendered	gendere	VERB
fpq-14292	263	3	pronouns	pronoun	NOUN
fpq-14292	263	4	are	be	AUX
fpq-14292	263	5	an	an	DET
fpq-14292	263	6	obvious	obvious	ADJ
fpq-14292	263	7	place	place	NOUN
fpq-14292	263	8	where	where	SCONJ
fpq-14292	263	9	bias	bias	NOUN
fpq-14292	263	10	reveals	reveal	VERB
fpq-14292	263	11	itself	itself	PRON
fpq-14292	263	12	in	in	ADP
fpq-14292	263	13	machine	machine	NOUN
fpq-14292	263	14	translation	translation	NOUN
fpq-14292	263	15	.	.	PUNCT
fpq-14292	264	1	for	for	ADP
fpq-14292	264	2	example	example	NOUN
fpq-14292	264	3	,	,	PUNCT
fpq-14292	264	4	turkish	turkish	ADJ
fpq-14292	264	5	does	do	AUX
fpq-14292	264	6	not	not	PART
fpq-14292	264	7	have	have	AUX
fpq-14292	264	8	gendered	gendere	VERB
fpq-14292	264	9	pronouns	pronoun	NOUN
fpq-14292	264	10	yet	yet	CCONJ
fpq-14292	264	11	there	there	PRON
fpq-14292	264	12	are	be	VERB
fpq-14292	264	13	numerous	numerous	ADJ
fpq-14292	264	14	examples	example	NOUN
fpq-14292	264	15	of	of	ADP
fpq-14292	264	16	turkish	turkish	ADJ
fpq-14292	264	17	sentences	sentence	NOUN
fpq-14292	264	18	translated	translate	VERB
fpq-14292	264	19	into	into	ADP
fpq-14292	264	20	english	english	PROPN
fpq-14292	264	21	by	by	ADP
fpq-14292	264	22	google	google	PROPN
fpq-14292	264	23	where	where	SCONJ
fpq-14292	264	24	a	a	DET
fpq-14292	264	25	genderless	genderless	ADJ
fpq-14292	264	26	pronoun	pronoun	NOUN
fpq-14292	264	27	becomes	becomes	AUX
fpq-14292	264	28	gendered	gendere	VERB
fpq-14292	264	29	according	accord	VERB
fpq-14292	264	30	to	to	ADP
fpq-14292	264	31	whether	whether	SCONJ
fpq-14292	264	32	the	the	DET
fpq-14292	264	33	noun	noun	NOUN
fpq-14292	264	34	is	be	AUX
fpq-14292	264	35	a	a	DET
fpq-14292	264	36	word	word	NOUN
fpq-14292	264	37	for	for	ADP
fpq-14292	264	38	a	a	DET
fpq-14292	264	39	stereotypically	stereotypically	ADV
fpq-14292	264	40	male	male	ADJ
fpq-14292	264	41	or	or	CCONJ
fpq-14292	264	42	instead	instead	ADV
fpq-14292	264	43	female	female	ADJ
fpq-14292	264	44	profession	profession	NOUN
fpq-14292	264	45	:	:	PUNCT
fpq-14292	264	46	“	"	PUNCT
fpq-14292	264	47	he	he	PRON
fpq-14292	264	48	”	"	PUNCT
fpq-14292	264	49	for	for	ADP
fpq-14292	264	50	doctor	doctor	NOUN
fpq-14292	264	51	,	,	PUNCT
fpq-14292	264	52	“	"	PUNCT
fpq-14292	264	53	she	she	PRON
fpq-14292	264	54	”	"	PUNCT
fpq-14292	264	55	for	for	ADP
fpq-14292	264	56	nurse	nurse	NOUN
fpq-14292	264	57	,	,	PUNCT
fpq-14292	264	58	and	and	CCONJ
fpq-14292	264	59	so	so	ADV
fpq-14292	264	60	forth	forth	ADV
fpq-14292	264	61	.	.	PUNCT
fpq-14292	265	1	thus	thus	ADV
fpq-14292	265	2	,	,	PUNCT
fpq-14292	265	3	“	"	PUNCT
fpq-14292	265	4	o	o	NOUN
fpq-14292	265	5	bir	bir	NOUN
fpq-14292	265	6	doktor	doktor	NOUN
fpq-14292	265	7	”	"	PUNCT
fpq-14292	265	8	will	will	AUX
fpq-14292	265	9	be	be	AUX
fpq-14292	265	10	translated	translate	VERB
fpq-14292	265	11	as	as	ADP
fpq-14292	265	12	“	"	PUNCT
fpq-14292	265	13	he	he	PRON
fpq-14292	265	14	is	be	AUX
fpq-14292	265	15	a	a	DET
fpq-14292	265	16	doctor	doctor	NOUN
fpq-14292	265	17	,	,	PUNCT
fpq-14292	265	18	”	"	PUNCT
fpq-14292	265	19	even	even	ADV
fpq-14292	265	20	though	though	SCONJ
fpq-14292	265	21	the	the	DET
fpq-14292	265	22	turkish	turkish	ADJ
fpq-14292	265	23	pronoun	pronoun	NOUN
fpq-14292	265	24	“	"	PUNCT
fpq-14292	265	25	o	o	NOUN
fpq-14292	265	26	”	"	PUNCT
fpq-14292	265	27	is	be	AUX
fpq-14292	265	28	ungendered	ungendered	ADJ
fpq-14292	265	29	and	and	CCONJ
fpq-14292	265	30	can	can	AUX
fpq-14292	265	31	be	be	AUX
fpq-14292	265	32	translated	translate	VERB
fpq-14292	265	33	as	as	ADP
fpq-14292	265	34	either	either	CCONJ
fpq-14292	265	35	“	"	PUNCT
fpq-14292	265	36	he	he	PRON
fpq-14292	265	37	”	"	PUNCT
fpq-14292	265	38	or	or	CCONJ
fpq-14292	265	39	“	"	PUNCT
fpq-14292	265	40	she	she	PRON
fpq-14292	265	41	.	.	PUNCT
fpq-14292	265	42	”	"	PUNCT
fpq-14292	266	1	in	in	ADP
fpq-14292	266	2	2018	2018	NUM
fpq-14292	266	3	,	,	PUNCT
fpq-14292	266	4	google	google	PROPN
fpq-14292	266	5	introduced	introduce	VERB
fpq-14292	266	6	gender	gender	NOUN
fpq-14292	266	7	-	-	PUNCT
fpq-14292	266	8	specific	specific	ADJ
fpq-14292	266	9	translations	translation	NOUN
fpq-14292	266	10	in	in	ADP
fpq-14292	266	11	order	order	NOUN
fpq-14292	266	12	to	to	PART
fpq-14292	266	13	avoid	avoid	VERB
fpq-14292	266	14	gender	gender	NOUN
fpq-14292	266	15	bias	bias	NOUN
fpq-14292	266	16	in	in	ADP
fpq-14292	266	17	google	google	PROPN
fpq-14292	266	18	translate	translate	NOUN
fpq-14292	266	19	.	.	PUNCT
fpq-14292	267	1	as	as	ADP
fpq-14292	267	2	a	a	DET
fpq-14292	267	3	result	result	NOUN
fpq-14292	267	4	,	,	PUNCT
fpq-14292	267	5	when	when	SCONJ
fpq-14292	267	6	you	you	PRON
fpq-14292	267	7	enter	enter	VERB
fpq-14292	267	8	“	"	PUNCT
fpq-14292	267	9	o	o	NOUN
fpq-14292	267	10	bir	bir	NOUN
fpq-14292	267	11	doktor	doktor	NOUN
fpq-14292	267	12	,	,	PUNCT
fpq-14292	267	13	”	"	PUNCT
fpq-14292	267	14	the	the	DET
fpq-14292	267	15	system	system	NOUN
fpq-14292	267	16	now	now	ADV
fpq-14292	267	17	offers	offer	VERB
fpq-14292	267	18	a	a	DET
fpq-14292	267	19	deery	deery	NOUN
fpq-14292	267	20	and	and	CCONJ
fpq-14292	267	21	bailey	bailey	PROPN
fpq-14292	267	22	–	–	PUNCT
fpq-14292	267	23	the	the	DET
fpq-14292	267	24	bias	bias	NOUN
fpq-14292	267	25	dilemma	dilemma	NOUN
fpq-14292	267	26	:	:	PUNCT
fpq-14292	267	27	the	the	DET
fpq-14292	267	28	ethics	ethic	NOUN
fpq-14292	267	29	of	of	ADP
fpq-14292	267	30	algorithmic	algorithmic	ADJ
fpq-14292	267	31	bias	bias	NOUN
fpq-14292	267	32	in	in	ADP
fpq-14292	267	33	natural	natural	ADJ
fpq-14292	267	34	-	-	PUNCT
fpq-14292	267	35	language	language	NOUN
fpq-14292	267	36	processing	processing	NOUN
fpq-14292	267	37	published	publish	VERB
fpq-14292	267	38	by	by	ADP
fpq-14292	267	39	scholarship@western	scholarship@western	PROPN
fpq-14292	267	40	,	,	PUNCT
fpq-14292	267	41	2022	2022	NUM
fpq-14292	267	42	15	15	NUM
fpq-14292	267	43	choice	choice	NOUN
fpq-14292	267	44	—	—	PUNCT
fpq-14292	267	45	either	either	CCONJ
fpq-14292	267	46	“	"	PUNCT
fpq-14292	267	47	he	he	PRON
fpq-14292	267	48	”	"	PUNCT
fpq-14292	267	49	or	or	CCONJ
fpq-14292	267	50	“	"	PUNCT
fpq-14292	267	51	she”—and	she”—and	NOUN
fpq-14292	267	52	it	it	PRON
fpq-14292	267	53	lets	let	VERB
fpq-14292	267	54	the	the	DET
fpq-14292	267	55	end	end	NOUN
fpq-14292	267	56	user	user	NOUN
fpq-14292	267	57	decide	decide	VERB
fpq-14292	267	58	on	on	ADP
fpq-14292	267	59	the	the	DET
fpq-14292	267	60	correct	correct	ADJ
fpq-14292	267	61	pronoun.27	pronoun.27	NOUN
fpq-14292	267	62	but	but	CCONJ
fpq-14292	267	63	when	when	SCONJ
fpq-14292	267	64	context	context	NOUN
fpq-14292	267	65	is	be	AUX
fpq-14292	267	66	provided	provide	VERB
fpq-14292	267	67	,	,	PUNCT
fpq-14292	267	68	google	google	PROPN
fpq-14292	267	69	translate	translate	NOUN
fpq-14292	267	70	gets	get	VERB
fpq-14292	267	71	it	it	PRON
fpq-14292	267	72	wrong	wrong	ADJ
fpq-14292	267	73	.	.	PUNCT
fpq-14292	268	1	if	if	SCONJ
fpq-14292	268	2	you	you	PRON
fpq-14292	268	3	are	be	AUX
fpq-14292	268	4	simply	simply	ADV
fpq-14292	268	5	translating	translate	VERB
fpq-14292	268	6	“	"	PUNCT
fpq-14292	268	7	o	o	PROPN
fpq-14292	268	8	bir	bir	NOUN
fpq-14292	268	9	doktor	doktor	NOUN
fpq-14292	268	10	,	,	PUNCT
fpq-14292	268	11	”	"	PUNCT
fpq-14292	268	12	there	there	PRON
fpq-14292	268	13	is	be	VERB
fpq-14292	268	14	no	no	DET
fpq-14292	268	15	wrong	wrong	ADJ
fpq-14292	268	16	answer	answer	NOUN
fpq-14292	268	17	,	,	PUNCT
fpq-14292	268	18	since	since	SCONJ
fpq-14292	268	19	there	there	PRON
fpq-14292	268	20	is	be	VERB
fpq-14292	268	21	no	no	DET
fpq-14292	268	22	context	context	NOUN
fpq-14292	268	23	provided	provide	VERB
fpq-14292	268	24	;	;	PUNCT
fpq-14292	268	25	either	either	CCONJ
fpq-14292	268	26	“	"	PUNCT
fpq-14292	268	27	he	he	PRON
fpq-14292	268	28	”	"	PUNCT
fpq-14292	268	29	or	or	CCONJ
fpq-14292	268	30	“	"	PUNCT
fpq-14292	268	31	she	she	PRON
fpq-14292	268	32	”	"	PUNCT
fpq-14292	268	33	is	be	AUX
fpq-14292	268	34	acceptable	acceptable	ADJ
fpq-14292	268	35	.	.	PUNCT
fpq-14292	269	1	however	however	ADV
fpq-14292	269	2	,	,	PUNCT
fpq-14292	269	3	with	with	ADP
fpq-14292	269	4	a	a	DET
fpq-14292	269	5	real	real	ADJ
fpq-14292	269	6	piece	piece	NOUN
fpq-14292	269	7	of	of	ADP
fpq-14292	269	8	text	text	NOUN
fpq-14292	269	9	,	,	PUNCT
fpq-14292	269	10	google	google	PROPN
fpq-14292	269	11	’s	’s	PART
fpq-14292	269	12	system	system	NOUN
fpq-14292	269	13	reveals	reveal	VERB
fpq-14292	269	14	its	its	PRON
fpq-14292	269	15	bias	bias	NOUN
fpq-14292	269	16	.	.	PUNCT
fpq-14292	270	1	for	for	ADP
fpq-14292	270	2	example	example	NOUN
fpq-14292	270	3	,	,	PUNCT
fpq-14292	270	4	it	it	PRON
fpq-14292	270	5	will	will	AUX
fpq-14292	270	6	provide	provide	VERB
fpq-14292	270	7	the	the	DET
fpq-14292	270	8	following	follow	VERB
fpq-14292	270	9	(	(	PUNCT
fpq-14292	270	10	correct	correct	ADJ
fpq-14292	270	11	)	)	PUNCT
fpq-14292	270	12	translation	translation	NOUN
fpq-14292	270	13	from	from	ADP
fpq-14292	270	14	english	english	PROPN
fpq-14292	270	15	into	into	ADP
fpq-14292	270	16	turkish	turkish	NOUN
fpq-14292	270	17	:	:	PUNCT
fpq-14292	270	18	did	do	AUX
fpq-14292	270	19	you	you	PRON
fpq-14292	270	20	meet	meet	VERB
fpq-14292	270	21	john	john	PROPN
fpq-14292	270	22	’s	’s	PART
fpq-14292	270	23	brother	brother	NOUN
fpq-14292	270	24	?	?	PUNCT
fpq-14292	271	1	he	he	PRON
fpq-14292	271	2	is	be	AUX
fpq-14292	271	3	a	a	DET
fpq-14292	271	4	nurse	nurse	NOUN
fpq-14292	271	5	.	.	PUNCT
fpq-14292	272	1	john’un	john’un	PROPN
fpq-14292	272	2	erkek	erkek	PROPN
fpq-14292	272	3	kardeşiyle	kardeşiyle	PROPN
fpq-14292	272	4	tanıştın	tanıştın	NOUN
fpq-14292	272	5	mı	mı	PROPN
fpq-14292	272	6	?	?	PUNCT
fpq-14292	272	7	o	o	NOUN
fpq-14292	272	8	bir	bir	NOUN
fpq-14292	272	9	hemşire	hemşire	NOUN
fpq-14292	272	10	.	.	PUNCT
fpq-14292	273	1	yet	yet	ADV
fpq-14292	273	2	when	when	SCONJ
fpq-14292	273	3	translating	translate	VERB
fpq-14292	273	4	this	this	DET
fpq-14292	273	5	turkish	turkish	ADJ
fpq-14292	273	6	sentence	sentence	NOUN
fpq-14292	273	7	back	back	ADV
fpq-14292	273	8	into	into	ADP
fpq-14292	273	9	english	english	PROPN
fpq-14292	273	10	,	,	PUNCT
fpq-14292	273	11	the	the	DET
fpq-14292	273	12	output	output	NOUN
fpq-14292	273	13	is	be	AUX
fpq-14292	273	14	:	:	PUNCT
fpq-14292	273	15	did	do	AUX
fpq-14292	273	16	you	you	PRON
fpq-14292	273	17	meet	meet	VERB
fpq-14292	273	18	john	john	PROPN
fpq-14292	273	19	’s	’s	PART
fpq-14292	273	20	brother	brother	NOUN
fpq-14292	273	21	?	?	PUNCT
fpq-14292	274	1	she	she	PRON
fpq-14292	274	2	is	be	AUX
fpq-14292	274	3	a	a	DET
fpq-14292	274	4	nurse	nurse	NOUN
fpq-14292	274	5	.	.	PUNCT
fpq-14292	275	1	this	this	DET
fpq-14292	275	2	result	result	NOUN
fpq-14292	275	3	is	be	AUX
fpq-14292	275	4	a	a	DET
fpq-14292	275	5	double	double	ADJ
fpq-14292	275	6	fail	fail	NOUN
fpq-14292	275	7	:	:	PUNCT
fpq-14292	275	8	the	the	DET
fpq-14292	275	9	translation	translation	NOUN
fpq-14292	275	10	is	be	AUX
fpq-14292	275	11	incorrect	incorrect	ADJ
fpq-14292	275	12	(	(	PUNCT
fpq-14292	275	13	even	even	ADV
fpq-14292	275	14	if	if	SCONJ
fpq-14292	275	15	this	this	DET
fpq-14292	275	16	usage	usage	NOUN
fpq-14292	275	17	of	of	ADP
fpq-14292	275	18	“	"	PUNCT
fpq-14292	275	19	she	she	PRON
fpq-14292	275	20	”	"	PUNCT
fpq-14292	275	21	in	in	ADP
fpq-14292	275	22	relation	relation	NOUN
fpq-14292	275	23	to	to	PART
fpq-14292	275	24	“	"	PUNCT
fpq-14292	275	25	nurse	nurse	NOUN
fpq-14292	275	26	”	"	PUNCT
fpq-14292	275	27	remains	remain	VERB
fpq-14292	275	28	descriptively	descriptively	ADV
fpq-14292	275	29	accurate	accurate	ADJ
fpq-14292	275	30	about	about	ADP
fpq-14292	275	31	people	people	NOUN
fpq-14292	275	32	’s	’s	PART
fpq-14292	275	33	actual	actual	ADJ
fpq-14292	275	34	language	language	NOUN
fpq-14292	275	35	use	use	NOUN
fpq-14292	275	36	)	)	PUNCT
fpq-14292	275	37	,	,	PUNCT
fpq-14292	275	38	and	and	CCONJ
fpq-14292	275	39	the	the	DET
fpq-14292	275	40	output	output	NOUN
fpq-14292	275	41	is	be	AUX
fpq-14292	275	42	biased	bias	VERB
fpq-14292	275	43	.	.	PUNCT
fpq-14292	276	1	moreover	moreover	ADV
fpq-14292	276	2	,	,	PUNCT
fpq-14292	276	3	it	it	PRON
fpq-14292	276	4	reveals	reveal	VERB
fpq-14292	276	5	the	the	DET
fpq-14292	276	6	changes	change	NOUN
fpq-14292	276	7	that	that	PRON
fpq-14292	276	8	google	google	PROPN
fpq-14292	276	9	implemented	implement	VERB
fpq-14292	276	10	as	as	ADP
fpq-14292	276	11	relatively	relatively	ADV
fpq-14292	276	12	superficial	superficial	ADJ
fpq-14292	276	13	and	and	CCONJ
fpq-14292	276	14	insubstantial	insubstantial	ADJ
fpq-14292	276	15	.	.	PUNCT
fpq-14292	277	1	6.2	6.2	NUM
fpq-14292	277	2	.	.	PUNCT
fpq-14292	278	1	assorted	assort	VERB
fpq-14292	278	2	nlg	nlg	PROPN
fpq-14292	278	3	tasks	task	NOUN
fpq-14292	278	4	translation	translation	NOUN
fpq-14292	278	5	is	be	AUX
fpq-14292	278	6	an	an	DET
fpq-14292	278	7	example	example	NOUN
fpq-14292	278	8	of	of	ADP
fpq-14292	278	9	a	a	DET
fpq-14292	278	10	natural	natural	ADJ
fpq-14292	278	11	-	-	PUNCT
fpq-14292	278	12	language	language	NOUN
fpq-14292	278	13	-	-	PUNCT
fpq-14292	278	14	generation	generation	NOUN
fpq-14292	278	15	(	(	PUNCT
fpq-14292	278	16	nlg	nlg	PROPN
fpq-14292	278	17	)	)	PUNCT
fpq-14292	278	18	task	task	NOUN
fpq-14292	278	19	,	,	PUNCT
fpq-14292	278	20	where	where	SCONJ
fpq-14292	278	21	the	the	DET
fpq-14292	278	22	output	output	NOUN
fpq-14292	278	23	is	be	AUX
fpq-14292	278	24	a	a	DET
fpq-14292	278	25	sentence	sentence	NOUN
fpq-14292	278	26	or	or	CCONJ
fpq-14292	278	27	sentences	sentence	NOUN
fpq-14292	278	28	in	in	ADP
fpq-14292	278	29	a	a	DET
fpq-14292	278	30	natural	natural	ADJ
fpq-14292	278	31	language	language	NOUN
fpq-14292	278	32	.	.	PUNCT
fpq-14292	279	1	other	other	ADJ
fpq-14292	279	2	examples	example	NOUN
fpq-14292	279	3	of	of	ADP
fpq-14292	279	4	nlg	nlg	NOUN
fpq-14292	279	5	tasks	task	NOUN
fpq-14292	279	6	include	include	VERB
fpq-14292	279	7	speech	speech	NOUN
fpq-14292	279	8	recognition	recognition	NOUN
fpq-14292	279	9	,	,	PUNCT
fpq-14292	279	10	image	image	NOUN
fpq-14292	279	11	captioning	captioning	NOUN
fpq-14292	279	12	,	,	PUNCT
fpq-14292	279	13	and	and	CCONJ
fpq-14292	279	14	dialogue	dialogue	NOUN
fpq-14292	279	15	systems	system	NOUN
fpq-14292	279	16	.	.	PUNCT
fpq-14292	280	1	one	one	NUM
fpq-14292	280	2	naïve	naïve	ADJ
fpq-14292	280	3	suggestion	suggestion	NOUN
fpq-14292	280	4	for	for	ADP
fpq-14292	280	5	systems	system	NOUN
fpq-14292	280	6	that	that	PRON
fpq-14292	280	7	perform	perform	VERB
fpq-14292	280	8	nlg	nlg	NOUN
fpq-14292	280	9	tasks	task	NOUN
fpq-14292	280	10	might	might	AUX
fpq-14292	280	11	be	be	AUX
fpq-14292	280	12	that	that	SCONJ
fpq-14292	280	13	they	they	PRON
fpq-14292	280	14	should	should	AUX
fpq-14292	280	15	never	never	ADV
fpq-14292	280	16	give	give	VERB
fpq-14292	280	17	biased	biased	ADJ
fpq-14292	280	18	outputs	output	NOUN
fpq-14292	280	19	,	,	PUNCT
fpq-14292	280	20	such	such	ADJ
fpq-14292	280	21	as	as	ADP
fpq-14292	280	22	sexist	sexist	ADJ
fpq-14292	280	23	or	or	CCONJ
fpq-14292	280	24	racist	racist	ADJ
fpq-14292	280	25	outputs	output	NOUN
fpq-14292	280	26	.	.	PUNCT
fpq-14292	281	1	when	when	SCONJ
fpq-14292	281	2	we	we	PRON
fpq-14292	281	3	look	look	VERB
fpq-14292	281	4	at	at	ADP
fpq-14292	281	5	particular	particular	ADJ
fpq-14292	281	6	applications	application	NOUN
fpq-14292	281	7	,	,	PUNCT
fpq-14292	281	8	we	we	PRON
fpq-14292	281	9	can	can	AUX
fpq-14292	281	10	see	see	VERB
fpq-14292	281	11	why	why	SCONJ
fpq-14292	281	12	this	this	DET
fpq-14292	281	13	suggestion	suggestion	NOUN
fpq-14292	281	14	is	be	AUX
fpq-14292	281	15	naïve	naïve	ADJ
fpq-14292	281	16	.	.	PUNCT
fpq-14292	282	1	if	if	SCONJ
fpq-14292	282	2	a	a	DET
fpq-14292	282	3	speech	speech	NOUN
fpq-14292	282	4	-	-	PUNCT
fpq-14292	282	5	recognition	recognition	NOUN
fpq-14292	282	6	system	system	NOUN
fpq-14292	282	7	receives	receive	VERB
fpq-14292	282	8	as	as	ADP
fpq-14292	282	9	audio	audio	ADJ
fpq-14292	282	10	input	input	NOUN
fpq-14292	282	11	the	the	DET
fpq-14292	282	12	sentence	sentence	NOUN
fpq-14292	282	13	“	"	PUNCT
fpq-14292	282	14	a	a	DET
fpq-14292	282	15	proper	proper	ADJ
fpq-14292	282	16	wife	wife	NOUN
fpq-14292	282	17	should	should	AUX
fpq-14292	282	18	be	be	AUX
fpq-14292	282	19	as	as	ADV
fpq-14292	282	20	obedient	obedient	ADJ
fpq-14292	282	21	as	as	ADP
fpq-14292	282	22	a	a	DET
fpq-14292	282	23	slave	slave	NOUN
fpq-14292	282	24	,	,	PUNCT
fpq-14292	282	25	”	"	PUNCT
fpq-14292	282	26	the	the	DET
fpq-14292	282	27	only	only	ADJ
fpq-14292	282	28	appropriate	appropriate	ADJ
fpq-14292	282	29	output	output	NOUN
fpq-14292	282	30	is	be	AUX
fpq-14292	282	31	text	text	NOUN
fpq-14292	282	32	of	of	ADP
fpq-14292	282	33	these	these	DET
fpq-14292	282	34	words.28	words.28	NOUN
fpq-14292	282	35	there	there	PRON
fpq-14292	282	36	is	be	VERB
fpq-14292	282	37	no	no	DET
fpq-14292	282	38	scope	scope	NOUN
fpq-14292	282	39	for	for	ADP
fpq-14292	282	40	debiasing	debiase	VERB
fpq-14292	282	41	without	without	ADP
fpq-14292	282	42	undermining	undermine	VERB
fpq-14292	282	43	the	the	DET
fpq-14292	282	44	task	task	NOUN
fpq-14292	282	45	that	that	PRON
fpq-14292	282	46	the	the	DET
fpq-14292	282	47	system	system	NOUN
fpq-14292	282	48	is	be	AUX
fpq-14292	282	49	designed	design	VERB
fpq-14292	282	50	to	to	PART
fpq-14292	282	51	perform	perform	VERB
fpq-14292	282	52	.	.	PUNCT
fpq-14292	283	1	descriptive	descriptive	ADJ
fpq-14292	283	2	accuracy	accuracy	NOUN
fpq-14292	283	3	trumps	trump	VERB
fpq-14292	283	4	normative	normative	ADJ
fpq-14292	283	5	correctness	correctness	NOUN
fpq-14292	283	6	in	in	ADP
fpq-14292	283	7	this	this	DET
fpq-14292	283	8	case	case	NOUN
fpq-14292	283	9	,	,	PUNCT
fpq-14292	283	10	since	since	SCONJ
fpq-14292	283	11	,	,	PUNCT
fpq-14292	283	12	as	as	ADP
fpq-14292	283	13	with	with	ADP
fpq-14292	283	14	a	a	DET
fpq-14292	283	15	camera	camera	NOUN
fpq-14292	283	16	,	,	PUNCT
fpq-14292	283	17	the	the	DET
fpq-14292	283	18	functions	function	NOUN
fpq-14292	283	19	of	of	ADP
fpq-14292	283	20	a	a	DET
fpq-14292	283	21	speech	speech	NOUN
fpq-14292	283	22	-	-	PUNCT
fpq-14292	283	23	recognition	recognition	NOUN
fpq-14292	283	24	system	system	NOUN
fpq-14292	283	25	require	require	VERB
fpq-14292	283	26	descriptive	descriptive	ADJ
fpq-14292	283	27	accuracy	accuracy	NOUN
fpq-14292	283	28	.	.	PUNCT
fpq-14292	284	1	27	27	NUM
fpq-14292	284	2	we	we	PRON
fpq-14292	284	3	made	make	VERB
fpq-14292	284	4	this	this	DET
fpq-14292	284	5	suggestion	suggestion	NOUN
fpq-14292	284	6	in	in	ADP
fpq-14292	284	7	a	a	DET
fpq-14292	284	8	talk	talk	NOUN
fpq-14292	284	9	we	we	PRON
fpq-14292	284	10	gave	give	VERB
fpq-14292	284	11	at	at	ADP
fpq-14292	284	12	google	google	PROPN
fpq-14292	284	13	in	in	ADP
fpq-14292	284	14	2018	2018	NUM
fpq-14292	284	15	.	.	PUNCT
fpq-14292	285	1	much	much	ADV
fpq-14292	285	2	as	as	SCONJ
fpq-14292	285	3	we	we	PRON
fpq-14292	285	4	would	would	AUX
fpq-14292	285	5	like	like	VERB
fpq-14292	285	6	to	to	PART
fpq-14292	285	7	take	take	VERB
fpq-14292	285	8	credit	credit	NOUN
fpq-14292	285	9	for	for	ADP
fpq-14292	285	10	the	the	DET
fpq-14292	285	11	change	change	NOUN
fpq-14292	285	12	,	,	PUNCT
fpq-14292	285	13	it	it	PRON
fpq-14292	285	14	was	be	AUX
fpq-14292	285	15	no	no	ADV
fpq-14292	285	16	doubt	doubt	ADV
fpq-14292	285	17	something	something	PRON
fpq-14292	285	18	that	that	PRON
fpq-14292	285	19	the	the	DET
fpq-14292	285	20	people	people	NOUN
fpq-14292	285	21	at	at	ADP
fpq-14292	285	22	google	google	PROPN
fpq-14292	285	23	had	have	AUX
fpq-14292	285	24	already	already	ADV
fpq-14292	285	25	been	be	AUX
fpq-14292	285	26	discussing	discuss	VERB
fpq-14292	285	27	long	long	ADV
fpq-14292	285	28	beforehand	beforehand	ADV
fpq-14292	285	29	.	.	PUNCT
fpq-14292	286	1	28	28	NUM
fpq-14292	287	1	this	this	DET
fpq-14292	287	2	phrase	phrase	NOUN
fpq-14292	287	3	,	,	PUNCT
fpq-14292	287	4	“	"	PUNCT
fpq-14292	287	5	a	a	DET
fpq-14292	287	6	proper	proper	ADJ
fpq-14292	287	7	wife	wife	NOUN
fpq-14292	287	8	should	should	AUX
fpq-14292	287	9	be	be	AUX
fpq-14292	287	10	as	as	ADV
fpq-14292	287	11	obedient	obedient	ADJ
fpq-14292	287	12	as	as	ADP
fpq-14292	287	13	a	a	DET
fpq-14292	287	14	slave	slave	NOUN
fpq-14292	287	15	,	,	PUNCT
fpq-14292	287	16	”	"	PUNCT
fpq-14292	287	17	often	often	ADV
fpq-14292	287	18	appears	appear	VERB
fpq-14292	287	19	online	online	ADV
fpq-14292	287	20	as	as	SCONJ
fpq-14292	287	21	attributed	attribute	VERB
fpq-14292	287	22	to	to	ADP
fpq-14292	287	23	aristotle	aristotle	PROPN
fpq-14292	287	24	.	.	PUNCT
fpq-14292	288	1	it	it	PRON
fpq-14292	288	2	seems	seem	VERB
fpq-14292	288	3	to	to	PART
fpq-14292	288	4	be	be	AUX
fpq-14292	288	5	a	a	DET
fpq-14292	288	6	broad	broad	ADJ
fpq-14292	288	7	paraphrase	paraphrase	NOUN
fpq-14292	288	8	of	of	ADP
fpq-14292	288	9	scattered	scatter	VERB
fpq-14292	288	10	parts	part	NOUN
fpq-14292	288	11	of	of	ADP
fpq-14292	288	12	aristotle	aristotle	PROPN
fpq-14292	288	13	’s	’s	PART
fpq-14292	288	14	oeconomica	oeconomica	PROPN
fpq-14292	288	15	,	,	PUNCT
fpq-14292	288	16	book	book	NOUN
fpq-14292	288	17	iii	iii	PROPN
fpq-14292	288	18	,	,	PUNCT
fpq-14292	288	19	section	section	NOUN
fpq-14292	288	20	1	1	NUM
fpq-14292	288	21	,	,	PUNCT
fpq-14292	288	22	paragraph	paragraph	NOUN
fpq-14292	288	23	2	2	NUM
fpq-14292	288	24	.	.	PUNCT
fpq-14292	288	25	feminist	feminist	PROPN
fpq-14292	288	26	philosophy	philosophy	PROPN
fpq-14292	288	27	quarterly	quarterly	ADV
fpq-14292	288	28	,	,	PUNCT
fpq-14292	288	29	2022	2022	NUM
fpq-14292	288	30	,	,	PUNCT
fpq-14292	288	31	vol	vol	NOUN
fpq-14292	288	32	.	.	NOUN
fpq-14292	288	33	8	8	NUM
fpq-14292	288	34	,	,	PUNCT
fpq-14292	288	35	iss	iss	PROPN
fpq-14292	288	36	.	.	PUNCT
fpq-14292	288	37	3/4	3/4	NUM
fpq-14292	288	38	,	,	PUNCT
fpq-14292	288	39	article	article	NOUN
fpq-14292	288	40	1	1	NUM
fpq-14292	288	41	published	publish	VERB
fpq-14292	288	42	by	by	ADP
fpq-14292	288	43	scholarship@western	scholarship@western	PROPN
fpq-14292	288	44	,	,	PUNCT
fpq-14292	288	45	2022	2022	NUM
fpq-14292	288	46	16	16	NUM
fpq-14292	288	47	in	in	ADP
fpq-14292	288	48	translation	translation	NOUN
fpq-14292	288	49	tasks	task	NOUN
fpq-14292	288	50	,	,	PUNCT
fpq-14292	288	51	there	there	PRON
fpq-14292	288	52	may	may	AUX
fpq-14292	288	53	not	not	PART
fpq-14292	288	54	be	be	AUX
fpq-14292	288	55	only	only	ADV
fpq-14292	288	56	one	one	NUM
fpq-14292	288	57	acceptable	acceptable	ADJ
fpq-14292	288	58	output	output	NOUN
fpq-14292	288	59	,	,	PUNCT
fpq-14292	288	60	since	since	SCONJ
fpq-14292	288	61	for	for	ADP
fpq-14292	288	62	any	any	DET
fpq-14292	288	63	phrase	phrase	NOUN
fpq-14292	288	64	we	we	PRON
fpq-14292	288	65	input	input	VERB
fpq-14292	288	66	,	,	PUNCT
fpq-14292	288	67	there	there	PRON
fpq-14292	288	68	may	may	AUX
fpq-14292	288	69	be	be	AUX
fpq-14292	288	70	multiple	multiple	ADJ
fpq-14292	288	71	acceptable	acceptable	ADJ
fpq-14292	288	72	ways	way	NOUN
fpq-14292	288	73	of	of	ADP
fpq-14292	288	74	translating	translate	VERB
fpq-14292	288	75	it	it	PRON
fpq-14292	288	76	into	into	ADP
fpq-14292	288	77	another	another	DET
fpq-14292	288	78	language	language	NOUN
fpq-14292	288	79	.	.	PUNCT
fpq-14292	289	1	even	even	ADV
fpq-14292	289	2	so	so	ADV
fpq-14292	289	3	,	,	PUNCT
fpq-14292	289	4	there	there	PRON
fpq-14292	289	5	appears	appear	VERB
fpq-14292	289	6	to	to	PART
fpq-14292	289	7	be	be	AUX
fpq-14292	289	8	limited	limited	ADJ
fpq-14292	289	9	scope	scope	NOUN
fpq-14292	289	10	for	for	ADP
fpq-14292	289	11	debiasing	debiase	VERB
fpq-14292	289	12	or	or	CCONJ
fpq-14292	289	13	imposing	impose	VERB
fpq-14292	289	14	normative	normative	ADJ
fpq-14292	289	15	correctness	correctness	NOUN
fpq-14292	289	16	.	.	PUNCT
fpq-14292	290	1	the	the	DET
fpq-14292	290	2	output	output	NOUN
fpq-14292	290	3	must	must	AUX
fpq-14292	290	4	be	be	AUX
fpq-14292	290	5	an	an	DET
fpq-14292	290	6	accurate	accurate	ADJ
fpq-14292	290	7	translation	translation	NOUN
fpq-14292	290	8	of	of	ADP
fpq-14292	290	9	the	the	DET
fpq-14292	290	10	sentence	sentence	NOUN
fpq-14292	290	11	—	—	PUNCT
fpq-14292	290	12	for	for	ADP
fpq-14292	290	13	example	example	NOUN
fpq-14292	290	14	,	,	PUNCT
fpq-14292	290	15	“	"	PUNCT
fpq-14292	290	16	a	a	DET
fpq-14292	290	17	proper	proper	ADJ
fpq-14292	290	18	wife	wife	NOUN
fpq-14292	290	19	should	should	AUX
fpq-14292	290	20	be	be	AUX
fpq-14292	290	21	as	as	ADV
fpq-14292	290	22	obedient	obedient	ADJ
fpq-14292	290	23	as	as	ADP
fpq-14292	290	24	a	a	DET
fpq-14292	290	25	slave”—into	slave”—into	NOUN
fpq-14292	290	26	the	the	DET
fpq-14292	290	27	target	target	NOUN
fpq-14292	290	28	language	language	NOUN
fpq-14292	290	29	,	,	PUNCT
fpq-14292	290	30	however	however	ADV
fpq-14292	290	31	sexist	sexist	ADJ
fpq-14292	290	32	or	or	CCONJ
fpq-14292	290	33	morally	morally	ADV
fpq-14292	290	34	objectionable	objectionable	ADJ
fpq-14292	290	35	that	that	SCONJ
fpq-14292	290	36	sentence	sentence	NOUN
fpq-14292	290	37	is	be	AUX
fpq-14292	290	38	.	.	PUNCT
fpq-14292	291	1	again	again	ADV
fpq-14292	291	2	,	,	PUNCT
fpq-14292	291	3	we	we	PRON
fpq-14292	291	4	can	can	AUX
fpq-14292	291	5	not	not	PART
fpq-14292	291	6	impose	impose	VERB
fpq-14292	291	7	normative	normative	ADJ
fpq-14292	291	8	correctness	correctness	NOUN
fpq-14292	291	9	without	without	ADP
fpq-14292	291	10	undermining	undermine	VERB
fpq-14292	291	11	accuracy	accuracy	NOUN
fpq-14292	291	12	.	.	PUNCT
fpq-14292	292	1	with	with	ADP
fpq-14292	292	2	image	image	NOUN
fpq-14292	292	3	captioning	captioning	NOUN
fpq-14292	292	4	,	,	PUNCT
fpq-14292	292	5	we	we	PRON
fpq-14292	292	6	begin	begin	VERB
fpq-14292	292	7	to	to	PART
fpq-14292	292	8	see	see	VERB
fpq-14292	292	9	scope	scope	NOUN
fpq-14292	292	10	for	for	ADP
fpq-14292	292	11	imposing	impose	VERB
fpq-14292	292	12	normative	normative	ADJ
fpq-14292	292	13	correctness	correctness	NOUN
fpq-14292	292	14	.	.	PUNCT
fpq-14292	293	1	clearly	clearly	ADV
fpq-14292	293	2	,	,	PUNCT
fpq-14292	293	3	there	there	PRON
fpq-14292	293	4	are	be	VERB
fpq-14292	293	5	numerous	numerous	ADJ
fpq-14292	293	6	suitable	suitable	ADJ
fpq-14292	293	7	captions	caption	NOUN
fpq-14292	293	8	that	that	PRON
fpq-14292	293	9	a	a	DET
fpq-14292	293	10	system	system	NOUN
fpq-14292	293	11	might	might	AUX
fpq-14292	293	12	suggest	suggest	VERB
fpq-14292	293	13	for	for	ADP
fpq-14292	293	14	any	any	DET
fpq-14292	293	15	image	image	NOUN
fpq-14292	293	16	.	.	PUNCT
fpq-14292	294	1	for	for	ADP
fpq-14292	294	2	a	a	DET
fpq-14292	294	3	photo	photo	NOUN
fpq-14292	294	4	of	of	ADP
fpq-14292	294	5	a	a	DET
fpq-14292	294	6	woman	woman	NOUN
fpq-14292	294	7	kitesurfing	kitesurfe	VERB
fpq-14292	294	8	,	,	PUNCT
fpq-14292	294	9	a	a	DET
fpq-14292	294	10	sexist	sexist	ADJ
fpq-14292	294	11	person	person	NOUN
fpq-14292	294	12	might	might	AUX
fpq-14292	294	13	suggest	suggest	VERB
fpq-14292	294	14	the	the	DET
fpq-14292	294	15	caption	caption	NOUN
fpq-14292	294	16	,	,	PUNCT
fpq-14292	294	17	“	"	PUNCT
fpq-14292	294	18	a	a	DET
fpq-14292	294	19	disobedient	disobedient	ADJ
fpq-14292	294	20	wife	wife	NOUN
fpq-14292	294	21	.	.	PUNCT
fpq-14292	294	22	”	"	PUNCT
fpq-14292	295	1	yet	yet	CCONJ
fpq-14292	295	2	we	we	PRON
fpq-14292	295	3	would	would	AUX
fpq-14292	295	4	hope	hope	VERB
fpq-14292	295	5	that	that	SCONJ
fpq-14292	295	6	no	no	DET
fpq-14292	295	7	imagecaptioning	imagecaptione	VERB
fpq-14292	295	8	system	system	NOUN
fpq-14292	295	9	would	would	AUX
fpq-14292	295	10	ever	ever	ADV
fpq-14292	295	11	do	do	VERB
fpq-14292	295	12	so	so	ADV
fpq-14292	295	13	.	.	PUNCT
fpq-14292	296	1	thus	thus	ADV
fpq-14292	296	2	,	,	PUNCT
fpq-14292	296	3	there	there	PRON
fpq-14292	296	4	seems	seem	VERB
fpq-14292	296	5	to	to	PART
fpq-14292	296	6	be	be	AUX
fpq-14292	296	7	a	a	DET
fpq-14292	296	8	normative	normative	ADJ
fpq-14292	296	9	constraint	constraint	NOUN
fpq-14292	296	10	on	on	ADP
fpq-14292	296	11	the	the	DET
fpq-14292	296	12	output	output	NOUN
fpq-14292	296	13	.	.	PUNCT
fpq-14292	297	1	obviously	obviously	ADV
fpq-14292	297	2	,	,	PUNCT
fpq-14292	297	3	we	we	PRON
fpq-14292	297	4	do	do	AUX
fpq-14292	297	5	not	not	PART
fpq-14292	297	6	want	want	VERB
fpq-14292	297	7	an	an	DET
fpq-14292	297	8	image	image	NOUN
fpq-14292	297	9	-	-	PUNCT
fpq-14292	297	10	captioning	caption	VERB
fpq-14292	297	11	system	system	NOUN
fpq-14292	297	12	to	to	PART
fpq-14292	297	13	label	label	VERB
fpq-14292	297	14	a	a	DET
fpq-14292	297	15	photo	photo	NOUN
fpq-14292	297	16	of	of	ADP
fpq-14292	297	17	a	a	DET
fpq-14292	297	18	kitesurfing	kitesurfe	VERB
fpq-14292	297	19	woman	woman	NOUN
fpq-14292	297	20	with	with	ADP
fpq-14292	297	21	“	"	PUNCT
fpq-14292	297	22	solar	solar	ADJ
fpq-14292	297	23	system	system	NOUN
fpq-14292	297	24	,	,	PUNCT
fpq-14292	297	25	”	"	PUNCT
fpq-14292	297	26	or	or	CCONJ
fpq-14292	297	27	“	"	PUNCT
fpq-14292	297	28	proton	proton	NOUN
fpq-14292	297	29	,	,	PUNCT
fpq-14292	297	30	”	"	PUNCT
fpq-14292	297	31	since	since	SCONJ
fpq-14292	297	32	these	these	DET
fpq-14292	297	33	captions	caption	NOUN
fpq-14292	297	34	are	be	AUX
fpq-14292	297	35	descriptively	descriptively	ADV
fpq-14292	297	36	inaccurate	inaccurate	ADJ
fpq-14292	297	37	.	.	PUNCT
fpq-14292	298	1	assuming	assume	VERB
fpq-14292	298	2	the	the	DET
fpq-14292	298	3	aim	aim	NOUN
fpq-14292	298	4	is	be	AUX
fpq-14292	298	5	descriptive	descriptive	ADJ
fpq-14292	298	6	accuracy	accuracy	NOUN
fpq-14292	298	7	regarding	regard	VERB
fpq-14292	298	8	the	the	DET
fpq-14292	298	9	image	image	NOUN
fpq-14292	298	10	’s	’s	PART
fpq-14292	298	11	content	content	NOUN
fpq-14292	298	12	,	,	PUNCT
fpq-14292	298	13	the	the	DET
fpq-14292	298	14	caption	caption	NOUN
fpq-14292	298	15	must	must	AUX
fpq-14292	298	16	accurately	accurately	ADV
fpq-14292	298	17	describe	describe	VERB
fpq-14292	298	18	what	what	PRON
fpq-14292	298	19	is	be	AUX
fpq-14292	298	20	in	in	ADP
fpq-14292	298	21	the	the	DET
fpq-14292	298	22	image	image	NOUN
fpq-14292	298	23	.	.	PUNCT
fpq-14292	299	1	because	because	SCONJ
fpq-14292	299	2	images	image	NOUN
fpq-14292	299	3	are	be	AUX
fpq-14292	299	4	typically	typically	ADV
fpq-14292	299	5	more	more	ADV
fpq-14292	299	6	informationally	informationally	ADV
fpq-14292	299	7	rich	rich	ADJ
fpq-14292	299	8	than	than	ADP
fpq-14292	299	9	linguistic	linguistic	ADJ
fpq-14292	299	10	descriptions	description	NOUN
fpq-14292	299	11	of	of	ADP
fpq-14292	299	12	them	they	PRON
fpq-14292	299	13	,	,	PUNCT
fpq-14292	299	14	most	most	ADJ
fpq-14292	299	15	captions	caption	NOUN
fpq-14292	299	16	(	(	PUNCT
fpq-14292	299	17	even	even	ADV
fpq-14292	299	18	if	if	SCONJ
fpq-14292	299	19	they	they	PRON
fpq-14292	299	20	are	be	AUX
fpq-14292	299	21	partly	partly	ADV
fpq-14292	299	22	descriptively	descriptively	ADV
fpq-14292	299	23	accurate	accurate	ADJ
fpq-14292	299	24	)	)	PUNCT
fpq-14292	299	25	will	will	AUX
fpq-14292	299	26	only	only	ADV
fpq-14292	299	27	be	be	AUX
fpq-14292	299	28	minimally	minimally	ADV
fpq-14292	299	29	accurate	accurate	ADJ
fpq-14292	299	30	.	.	PUNCT
fpq-14292	300	1	as	as	ADP
fpq-14292	300	2	a	a	DET
fpq-14292	300	3	result	result	NOUN
fpq-14292	300	4	,	,	PUNCT
fpq-14292	300	5	“	"	PUNCT
fpq-14292	300	6	woman	woman	NOUN
fpq-14292	300	7	sailing	sailing	NOUN
fpq-14292	300	8	”	"	PUNCT
fpq-14292	300	9	might	might	AUX
fpq-14292	300	10	be	be	AUX
fpq-14292	300	11	minimally	minimally	ADV
fpq-14292	300	12	accurate	accurate	ADJ
fpq-14292	300	13	,	,	PUNCT
fpq-14292	300	14	while	while	SCONJ
fpq-14292	300	15	“	"	PUNCT
fpq-14292	300	16	woman	woman	NOUN
fpq-14292	300	17	kitesurfing	kitesurfing	NOUN
fpq-14292	300	18	”	"	PUNCT
fpq-14292	300	19	might	might	AUX
fpq-14292	300	20	be	be	AUX
fpq-14292	300	21	better	well	ADJ
fpq-14292	300	22	.	.	PUNCT
fpq-14292	301	1	but	but	CCONJ
fpq-14292	301	2	“	"	PUNCT
fpq-14292	301	3	woman	woman	NOUN
fpq-14292	301	4	kitesurfing	kitesurfe	VERB
fpq-14292	301	5	on	on	ADP
fpq-14292	301	6	a	a	DET
fpq-14292	301	7	slingshot	slingshot	NOUN
fpq-14292	301	8	kiteboard	kiteboard	NOUN
fpq-14292	301	9	in	in	ADP
fpq-14292	301	10	light	light	ADJ
fpq-14292	301	11	seas	sea	NOUN
fpq-14292	301	12	and	and	CCONJ
fpq-14292	301	13	moderate	moderate	ADJ
fpq-14292	301	14	winds	wind	NOUN
fpq-14292	301	15	”	"	PUNCT
fpq-14292	301	16	might	might	AUX
fpq-14292	301	17	be	be	AUX
fpq-14292	301	18	too	too	ADV
fpq-14292	301	19	accurate	accurate	ADJ
fpq-14292	301	20	for	for	ADP
fpq-14292	301	21	most	most	ADJ
fpq-14292	301	22	contexts	context	NOUN
fpq-14292	301	23	.	.	PUNCT
fpq-14292	302	1	in	in	ADP
fpq-14292	302	2	any	any	DET
fpq-14292	302	3	case	case	NOUN
fpq-14292	302	4	,	,	PUNCT
fpq-14292	302	5	“	"	PUNCT
fpq-14292	302	6	disobedient	disobedient	ADJ
fpq-14292	302	7	wife	wife	NOUN
fpq-14292	302	8	”	"	PUNCT
fpq-14292	302	9	is	be	AUX
fpq-14292	302	10	not	not	PART
fpq-14292	302	11	an	an	DET
fpq-14292	302	12	accurate	accurate	ADJ
fpq-14292	302	13	description	description	NOUN
fpq-14292	302	14	in	in	ADP
fpq-14292	302	15	virtue	virtue	NOUN
fpq-14292	302	16	merely	merely	ADV
fpq-14292	302	17	of	of	ADP
fpq-14292	302	18	the	the	DET
fpq-14292	302	19	informational	informational	ADJ
fpq-14292	302	20	content	content	NOUN
fpq-14292	302	21	of	of	ADP
fpq-14292	302	22	the	the	DET
fpq-14292	302	23	image	image	NOUN
fpq-14292	302	24	.	.	PUNCT
fpq-14292	303	1	this	this	DET
fpq-14292	303	2	caption	caption	NOUN
fpq-14292	303	3	might	might	AUX
fpq-14292	303	4	be	be	AUX
fpq-14292	303	5	descriptively	descriptively	ADV
fpq-14292	303	6	accurate	accurate	ADJ
fpq-14292	303	7	of	of	ADP
fpq-14292	303	8	the	the	DET
fpq-14292	303	9	situation	situation	NOUN
fpq-14292	303	10	being	be	AUX
fpq-14292	303	11	depicted	depict	VERB
fpq-14292	303	12	,	,	PUNCT
fpq-14292	303	13	were	be	AUX
fpq-14292	303	14	we	we	PRON
fpq-14292	303	15	to	to	PART
fpq-14292	303	16	learn	learn	VERB
fpq-14292	303	17	(	(	PUNCT
fpq-14292	303	18	for	for	ADP
fpq-14292	303	19	example	example	NOUN
fpq-14292	303	20	)	)	PUNCT
fpq-14292	303	21	that	that	SCONJ
fpq-14292	303	22	this	this	DET
fpq-14292	303	23	kitesurfing	kitesurfe	VERB
fpq-14292	303	24	woman	woman	NOUN
fpq-14292	303	25	in	in	ADP
fpq-14292	303	26	fact	fact	NOUN
fpq-14292	303	27	disobeyed	disobey	VERB
fpq-14292	303	28	her	her	PRON
fpq-14292	303	29	sexist	sexist	ADJ
fpq-14292	303	30	husband	husband	NOUN
fpq-14292	303	31	’s	’s	PART
fpq-14292	303	32	command	command	NOUN
fpq-14292	303	33	never	never	ADV
fpq-14292	303	34	to	to	ADP
fpq-14292	303	35	kitesurf	kitesurf	VERB
fpq-14292	303	36	,	,	PUNCT
fpq-14292	303	37	since	since	SCONJ
fpq-14292	303	38	the	the	DET
fpq-14292	303	39	husband	husband	NOUN
fpq-14292	303	40	believes	believe	VERB
fpq-14292	303	41	wives	wife	NOUN
fpq-14292	303	42	—	—	PUNCT
fpq-14292	303	43	being	be	AUX
fpq-14292	303	44	women	woman	NOUN
fpq-14292	303	45	—	—	PUNCT
fpq-14292	303	46	should	should	AUX
fpq-14292	303	47	not	not	PART
fpq-14292	303	48	kitesurf	kitesurf	VERB
fpq-14292	303	49	.	.	PUNCT
fpq-14292	304	1	once	once	SCONJ
fpq-14292	304	2	we	we	PRON
fpq-14292	304	3	understand	understand	VERB
fpq-14292	304	4	the	the	DET
fpq-14292	304	5	caption	caption	NOUN
fpq-14292	304	6	“	"	PUNCT
fpq-14292	304	7	disobedient	disobedient	ADJ
fpq-14292	304	8	wife	wife	NOUN
fpq-14292	304	9	”	"	PUNCT
fpq-14292	304	10	in	in	ADP
fpq-14292	304	11	a	a	DET
fpq-14292	304	12	pejorative	pejorative	ADJ
fpq-14292	304	13	sense	sense	NOUN
fpq-14292	304	14	—	—	PUNCT
fpq-14292	304	15	as	as	ADP
fpq-14292	304	16	implying	imply	VERB
fpq-14292	304	17	something	something	PRON
fpq-14292	304	18	like	like	ADP
fpq-14292	304	19	,	,	PUNCT
fpq-14292	304	20	“	"	PUNCT
fpq-14292	304	21	this	this	DET
fpq-14292	304	22	woman	woman	NOUN
fpq-14292	304	23	,	,	PUNCT
fpq-14292	304	24	in	in	ADP
fpq-14292	304	25	virtue	virtue	NOUN
fpq-14292	304	26	of	of	ADP
fpq-14292	304	27	kitesurfing	kitesurfing	NOUN
fpq-14292	304	28	,	,	PUNCT
fpq-14292	304	29	is	be	AUX
fpq-14292	304	30	behaving	behave	VERB
fpq-14292	304	31	as	as	SCONJ
fpq-14292	304	32	no	no	DET
fpq-14292	304	33	woman	woman	NOUN
fpq-14292	304	34	should	should	AUX
fpq-14292	304	35	behave”—we	behave”—we	AUX
fpq-14292	304	36	can	can	AUX
fpq-14292	304	37	see	see	VERB
fpq-14292	304	38	it	it	PRON
fpq-14292	304	39	would	would	AUX
fpq-14292	304	40	be	be	AUX
fpq-14292	304	41	normatively	normatively	ADV
fpq-14292	304	42	incorrect	incorrect	ADJ
fpq-14292	304	43	to	to	PART
fpq-14292	304	44	allow	allow	VERB
fpq-14292	304	45	the	the	DET
fpq-14292	304	46	caption	caption	NOUN
fpq-14292	304	47	.	.	PUNCT
fpq-14292	305	1	most	most	ADJ
fpq-14292	305	2	image	image	NOUN
fpq-14292	305	3	-	-	PUNCT
fpq-14292	305	4	captioning	caption	VERB
fpq-14292	305	5	systems	system	NOUN
fpq-14292	305	6	would	would	AUX
fpq-14292	305	7	be	be	AUX
fpq-14292	305	8	unlikely	unlikely	ADJ
fpq-14292	305	9	to	to	PART
fpq-14292	305	10	provide	provide	VERB
fpq-14292	305	11	this	this	DET
fpq-14292	305	12	caption	caption	NOUN
fpq-14292	305	13	in	in	ADP
fpq-14292	305	14	any	any	DET
fpq-14292	305	15	case	case	NOUN
fpq-14292	305	16	,	,	PUNCT
fpq-14292	305	17	since	since	SCONJ
fpq-14292	305	18	they	they	PRON
fpq-14292	305	19	would	would	AUX
fpq-14292	305	20	not	not	PART
fpq-14292	305	21	have	have	AUX
fpq-14292	305	22	encountered	encounter	VERB
fpq-14292	305	23	it	it	PRON
fpq-14292	305	24	in	in	ADP
fpq-14292	305	25	their	their	PRON
fpq-14292	305	26	training	training	NOUN
fpq-14292	305	27	data	datum	NOUN
fpq-14292	305	28	.	.	PUNCT
fpq-14292	306	1	beyond	beyond	ADP
fpq-14292	306	2	these	these	DET
fpq-14292	306	3	considerations	consideration	NOUN
fpq-14292	306	4	,	,	PUNCT
fpq-14292	306	5	we	we	PRON
fpq-14292	306	6	also	also	ADV
fpq-14292	306	7	want	want	VERB
fpq-14292	306	8	a	a	DET
fpq-14292	306	9	system	system	NOUN
fpq-14292	306	10	to	to	PART
fpq-14292	306	11	be	be	AUX
fpq-14292	306	12	normatively	normatively	ADV
fpq-14292	306	13	correct	correct	ADJ
fpq-14292	306	14	in	in	ADP
fpq-14292	306	15	not	not	PART
fpq-14292	306	16	labeling	label	VERB
fpq-14292	306	17	the	the	DET
fpq-14292	306	18	image	image	NOUN
fpq-14292	306	19	with	with	ADP
fpq-14292	306	20	“	"	PUNCT
fpq-14292	306	21	a	a	DET
fpq-14292	306	22	disobedient	disobedient	ADJ
fpq-14292	306	23	wife	wife	NOUN
fpq-14292	306	24	”	"	PUNCT
fpq-14292	306	25	even	even	ADV
fpq-14292	306	26	if	if	SCONJ
fpq-14292	306	27	the	the	DET
fpq-14292	306	28	system	system	NOUN
fpq-14292	306	29	had	have	AUX
fpq-14292	306	30	been	be	AUX
fpq-14292	306	31	exposed	expose	VERB
fpq-14292	306	32	to	to	ADP
fpq-14292	306	33	this	this	DET
fpq-14292	306	34	label	label	NOUN
fpq-14292	306	35	in	in	ADP
fpq-14292	306	36	its	its	PRON
fpq-14292	306	37	training	training	NOUN
fpq-14292	306	38	,	,	PUNCT
fpq-14292	306	39	since	since	SCONJ
fpq-14292	306	40	the	the	DET
fpq-14292	306	41	label	label	NOUN
fpq-14292	306	42	is	be	AUX
fpq-14292	306	43	sexist	sexist	ADJ
fpq-14292	306	44	—	—	PUNCT
fpq-14292	306	45	that	that	ADV
fpq-14292	306	46	is	is	ADV
fpq-14292	306	47	,	,	PUNCT
fpq-14292	306	48	normatively	normatively	ADV
fpq-14292	306	49	incorrect	incorrect	ADJ
fpq-14292	306	50	.	.	PUNCT
fpq-14292	307	1	finally	finally	ADV
fpq-14292	307	2	,	,	PUNCT
fpq-14292	307	3	consider	consider	VERB
fpq-14292	307	4	dialogue	dialogue	NOUN
fpq-14292	307	5	systems	system	NOUN
fpq-14292	307	6	.	.	PUNCT
fpq-14292	308	1	here	here	ADV
fpq-14292	308	2	,	,	PUNCT
fpq-14292	308	3	there	there	PRON
fpq-14292	308	4	can	can	AUX
fpq-14292	308	5	be	be	AUX
fpq-14292	308	6	appropriate	appropriate	ADJ
fpq-14292	308	7	and	and	CCONJ
fpq-14292	308	8	inappropriate	inappropriate	ADJ
fpq-14292	308	9	ways	way	NOUN
fpq-14292	308	10	to	to	PART
fpq-14292	308	11	respond	respond	VERB
fpq-14292	308	12	to	to	ADP
fpq-14292	308	13	a	a	DET
fpq-14292	308	14	question	question	NOUN
fpq-14292	308	15	,	,	PUNCT
fpq-14292	308	16	even	even	ADV
fpq-14292	308	17	descriptively	descriptively	ADV
fpq-14292	308	18	.	.	PUNCT
fpq-14292	309	1	in	in	ADP
fpq-14292	309	2	answer	answer	NOUN
fpq-14292	309	3	to	to	ADP
fpq-14292	309	4	the	the	DET
fpq-14292	309	5	question	question	NOUN
fpq-14292	309	6	,	,	PUNCT
fpq-14292	309	7	“	"	PUNCT
fpq-14292	309	8	what	what	PRON
fpq-14292	309	9	is	be	AUX
fpq-14292	309	10	the	the	DET
fpq-14292	309	11	weather	weather	NOUN
fpq-14292	309	12	like	like	INTJ
fpq-14292	309	13	in	in	ADP
fpq-14292	309	14	sydney	sydney	NOUN
fpq-14292	309	15	?	?	PUNCT
fpq-14292	309	16	”	"	PUNCT
fpq-14292	310	1	a	a	DET
fpq-14292	310	2	system	system	NOUN
fpq-14292	310	3	can	can	AUX
fpq-14292	310	4	reply	reply	VERB
fpq-14292	310	5	“	"	PUNCT
fpq-14292	310	6	sunny	sunny	ADJ
fpq-14292	310	7	”	"	PUNCT
fpq-14292	310	8	or	or	CCONJ
fpq-14292	310	9	“	"	PUNCT
fpq-14292	310	10	beautiful	beautiful	ADJ
fpq-14292	310	11	,	,	PUNCT
fpq-14292	310	12	”	"	PUNCT
fpq-14292	310	13	but	but	CCONJ
fpq-14292	310	14	presumably	presumably	ADV
fpq-14292	310	15	not	not	PART
fpq-14292	310	16	“	"	PUNCT
fpq-14292	310	17	subatomic	subatomic	PROPN
fpq-14292	310	18	”	"	PUNCT
fpq-14292	310	19	or	or	CCONJ
fpq-14292	310	20	“	"	PUNCT
fpq-14292	310	21	bread	bread	NOUN
fpq-14292	310	22	.	.	PUNCT
fpq-14292	310	23	”	"	PUNCT
fpq-14292	311	1	yet	yet	CCONJ
fpq-14292	311	2	if	if	SCONJ
fpq-14292	311	3	someone	someone	PRON
fpq-14292	311	4	asked	ask	VERB
fpq-14292	311	5	,	,	PUNCT
fpq-14292	311	6	“	"	PUNCT
fpq-14292	311	7	how	how	SCONJ
fpq-14292	311	8	obedient	obedient	ADJ
fpq-14292	311	9	should	should	AUX
fpq-14292	311	10	a	a	DET
fpq-14292	311	11	proper	proper	ADJ
fpq-14292	311	12	wife	wife	NOUN
fpq-14292	311	13	be	be	AUX
fpq-14292	311	14	?	?	PUNCT
fpq-14292	311	15	”	"	PUNCT
fpq-14292	311	16	the	the	DET
fpq-14292	311	17	system	system	NOUN
fpq-14292	311	18	should	should	AUX
fpq-14292	311	19	never	never	ADV
fpq-14292	311	20	answer	answer	VERB
fpq-14292	311	21	,	,	PUNCT
fpq-14292	311	22	“	"	PUNCT
fpq-14292	311	23	a	a	DET
fpq-14292	311	24	proper	proper	ADJ
fpq-14292	311	25	wife	wife	NOUN
fpq-14292	311	26	deery	deery	NOUN
fpq-14292	311	27	and	and	CCONJ
fpq-14292	311	28	bailey	bailey	PROPN
fpq-14292	311	29	–	–	PUNCT
fpq-14292	311	30	the	the	DET
fpq-14292	311	31	bias	bias	NOUN
fpq-14292	311	32	dilemma	dilemma	NOUN
fpq-14292	311	33	:	:	PUNCT
fpq-14292	311	34	the	the	DET
fpq-14292	311	35	ethics	ethic	NOUN
fpq-14292	311	36	of	of	ADP
fpq-14292	311	37	algorithmic	algorithmic	ADJ
fpq-14292	311	38	bias	bias	NOUN
fpq-14292	311	39	in	in	ADP
fpq-14292	311	40	natural	natural	ADJ
fpq-14292	311	41	-	-	PUNCT
fpq-14292	311	42	language	language	NOUN
fpq-14292	311	43	processing	processing	NOUN
fpq-14292	311	44	published	publish	VERB
fpq-14292	311	45	by	by	ADP
fpq-14292	311	46	scholarship@western	scholarship@western	PROPN
fpq-14292	311	47	,	,	PUNCT
fpq-14292	311	48	2022	2022	NUM
fpq-14292	311	49	17	17	NUM
fpq-14292	311	50	should	should	AUX
fpq-14292	311	51	be	be	AUX
fpq-14292	311	52	as	as	ADV
fpq-14292	311	53	obedient	obedient	ADJ
fpq-14292	311	54	as	as	ADP
fpq-14292	311	55	a	a	DET
fpq-14292	311	56	slave	slave	NOUN
fpq-14292	311	57	.	.	PUNCT
fpq-14292	312	1	”29	”29	NOUN
fpq-14292	312	2	what	what	PRON
fpq-14292	312	3	is	be	AUX
fpq-14292	312	4	an	an	DET
fpq-14292	312	5	acceptable	acceptable	ADJ
fpq-14292	312	6	output	output	NOUN
fpq-14292	312	7	in	in	ADP
fpq-14292	312	8	the	the	DET
fpq-14292	312	9	case	case	NOUN
fpq-14292	312	10	of	of	ADP
fpq-14292	312	11	speech	speech	NOUN
fpq-14292	312	12	recognition	recognition	NOUN
fpq-14292	312	13	should	should	AUX
fpq-14292	312	14	never	never	ADV
fpq-14292	312	15	be	be	AUX
fpq-14292	312	16	output	output	VERB
fpq-14292	312	17	by	by	ADP
fpq-14292	312	18	a	a	DET
fpq-14292	312	19	dialogue	dialogue	NOUN
fpq-14292	312	20	system	system	NOUN
fpq-14292	312	21	.	.	PUNCT
fpq-14292	313	1	here	here	ADV
fpq-14292	313	2	,	,	PUNCT
fpq-14292	313	3	we	we	PRON
fpq-14292	313	4	reach	reach	VERB
fpq-14292	313	5	maximal	maximal	ADJ
fpq-14292	313	6	scope	scope	NOUN
fpq-14292	313	7	for	for	ADP
fpq-14292	313	8	imposing	impose	VERB
fpq-14292	313	9	normative	normative	ADJ
fpq-14292	313	10	correctness	correctness	NOUN
fpq-14292	313	11	on	on	ADP
fpq-14292	313	12	a	a	DET
fpq-14292	313	13	system	system	NOUN
fpq-14292	313	14	’s	’s	PART
fpq-14292	313	15	outputs	output	NOUN
fpq-14292	313	16	.	.	PUNCT
fpq-14292	314	1	the	the	DET
fpq-14292	314	2	upshot	upshot	NOUN
fpq-14292	314	3	is	be	AUX
fpq-14292	314	4	,	,	PUNCT
fpq-14292	314	5	first	first	ADV
fpq-14292	314	6	,	,	PUNCT
fpq-14292	314	7	that	that	SCONJ
fpq-14292	314	8	it	it	PRON
fpq-14292	314	9	can	can	AUX
fpq-14292	314	10	be	be	AUX
fpq-14292	314	11	preferable	preferable	ADJ
fpq-14292	314	12	—	—	PUNCT
fpq-14292	314	13	not	not	PART
fpq-14292	314	14	only	only	ADV
fpq-14292	314	15	practically	practically	ADV
fpq-14292	314	16	but	but	CCONJ
fpq-14292	314	17	ethically	ethically	ADV
fpq-14292	314	18	(	(	PUNCT
fpq-14292	314	19	we	we	PRON
fpq-14292	314	20	discuss	discuss	VERB
fpq-14292	314	21	why	why	SCONJ
fpq-14292	314	22	this	this	PRON
fpq-14292	314	23	is	be	AUX
fpq-14292	314	24	an	an	DET
fpq-14292	314	25	ethical	ethical	ADJ
fpq-14292	314	26	issue	issue	NOUN
fpq-14292	314	27	in	in	ADP
fpq-14292	314	28	sections	section	NOUN
fpq-14292	314	29	7	7	NUM
fpq-14292	314	30	and	and	CCONJ
fpq-14292	314	31	8)—to	8)—to	NUM
fpq-14292	314	32	control	control	NOUN
fpq-14292	314	33	for	for	ADP
fpq-14292	314	34	biases	bias	NOUN
fpq-14292	314	35	in	in	ADP
fpq-14292	314	36	ml	ml	NOUN
fpq-14292	314	37	models	model	NOUN
fpq-14292	314	38	and	and	CCONJ
fpq-14292	314	39	systems	system	NOUN
fpq-14292	314	40	at	at	ADP
fpq-14292	314	41	the	the	DET
fpq-14292	314	42	output	output	NOUN
fpq-14292	314	43	stage	stage	NOUN
fpq-14292	314	44	rather	rather	ADV
fpq-14292	314	45	than	than	ADP
fpq-14292	314	46	to	to	PART
fpq-14292	314	47	try	try	VERB
fpq-14292	314	48	to	to	PART
fpq-14292	314	49	debias	debias	VERB
fpq-14292	314	50	a	a	DET
fpq-14292	314	51	model	model	NOUN
fpq-14292	314	52	and	and	CCONJ
fpq-14292	314	53	expect	expect	VERB
fpq-14292	314	54	to	to	PART
fpq-14292	314	55	deploy	deploy	VERB
fpq-14292	314	56	it	it	PRON
fpq-14292	314	57	for	for	ADP
fpq-14292	314	58	downstream	downstream	ADJ
fpq-14292	314	59	tasks	task	NOUN
fpq-14292	314	60	in	in	ADP
fpq-14292	314	61	ways	way	NOUN
fpq-14292	314	62	that	that	PRON
fpq-14292	314	63	will	will	AUX
fpq-14292	314	64	not	not	PART
fpq-14292	314	65	reproduce	reproduce	VERB
fpq-14292	314	66	bias	bias	NOUN
fpq-14292	314	67	.	.	PUNCT
fpq-14292	315	1	second	second	ADJ
fpq-14292	315	2	,	,	PUNCT
fpq-14292	315	3	even	even	ADV
fpq-14292	315	4	when	when	SCONJ
fpq-14292	315	5	we	we	PRON
fpq-14292	315	6	focus	focus	VERB
fpq-14292	315	7	on	on	ADP
fpq-14292	315	8	controlling	control	VERB
fpq-14292	315	9	for	for	ADP
fpq-14292	315	10	biases	bias	NOUN
fpq-14292	315	11	at	at	ADP
fpq-14292	315	12	the	the	DET
fpq-14292	315	13	output	output	NOUN
fpq-14292	315	14	stage	stage	NOUN
fpq-14292	315	15	,	,	PUNCT
fpq-14292	315	16	nlg	nlg	NOUN
fpq-14292	315	17	applications	application	NOUN
fpq-14292	315	18	must	must	AUX
fpq-14292	315	19	be	be	AUX
fpq-14292	315	20	treated	treat	VERB
fpq-14292	315	21	differently	differently	ADV
fpq-14292	315	22	,	,	PUNCT
fpq-14292	315	23	including	include	VERB
fpq-14292	315	24	from	from	ADP
fpq-14292	315	25	an	an	DET
fpq-14292	315	26	ethical	ethical	ADJ
fpq-14292	315	27	perspective	perspective	NOUN
fpq-14292	315	28	.	.	PUNCT
fpq-14292	316	1	in	in	ADP
fpq-14292	316	2	some	some	DET
fpq-14292	316	3	cases	case	NOUN
fpq-14292	316	4	we	we	PRON
fpq-14292	316	5	can	can	AUX
fpq-14292	316	6	reduce	reduce	VERB
fpq-14292	316	7	bias	bias	NOUN
fpq-14292	316	8	;	;	PUNCT
fpq-14292	316	9	in	in	ADP
fpq-14292	316	10	others	other	NOUN
fpq-14292	316	11	we	we	PRON
fpq-14292	316	12	can	can	AUX
fpq-14292	316	13	not	not	PART
fpq-14292	316	14	.	.	PUNCT
fpq-14292	317	1	6.3	6.3	NUM
fpq-14292	317	2	.	.	PUNCT
fpq-14292	318	1	search	search	NOUN
fpq-14292	318	2	search	search	NOUN
fpq-14292	318	3	is	be	AUX
fpq-14292	318	4	a	a	DET
fpq-14292	318	5	different	different	ADJ
fpq-14292	318	6	application	application	NOUN
fpq-14292	318	7	.	.	PUNCT
fpq-14292	319	1	in	in	ADP
fpq-14292	319	2	considering	consider	VERB
fpq-14292	319	3	sources	source	NOUN
fpq-14292	319	4	of	of	ADP
fpq-14292	319	5	bias	bias	NOUN
fpq-14292	319	6	in	in	ADP
fpq-14292	319	7	ml	ml	NOUN
fpq-14292	319	8	systems	system	NOUN
fpq-14292	319	9	,	,	PUNCT
fpq-14292	319	10	harini	harini	PROPN
fpq-14292	319	11	suresh	suresh	PROPN
fpq-14292	319	12	and	and	CCONJ
fpq-14292	319	13	john	john	PROPN
fpq-14292	319	14	guttag	guttag	PROPN
fpq-14292	319	15	(	(	PUNCT
fpq-14292	319	16	2020	2020	NUM
fpq-14292	319	17	)	)	PUNCT
fpq-14292	319	18	discuss	discuss	VERB
fpq-14292	319	19	two	two	NUM
fpq-14292	319	20	that	that	PRON
fpq-14292	319	21	are	be	AUX
fpq-14292	319	22	relevant	relevant	ADJ
fpq-14292	319	23	to	to	ADP
fpq-14292	319	24	both	both	DET
fpq-14292	319	25	search	search	VERB
fpq-14292	319	26	and	and	CCONJ
fpq-14292	319	27	our	our	PRON
fpq-14292	319	28	purposes	purpose	NOUN
fpq-14292	319	29	in	in	ADP
fpq-14292	319	30	this	this	DET
fpq-14292	319	31	paper	paper	NOUN
fpq-14292	319	32	—	—	PUNCT
fpq-14292	319	33	that	that	ADV
fpq-14292	319	34	is	be	AUX
fpq-14292	319	35	,	,	PUNCT
fpq-14292	319	36	historical	historical	ADJ
fpq-14292	319	37	and	and	CCONJ
fpq-14292	319	38	representational	representational	ADJ
fpq-14292	319	39	bias	bias	NOUN
fpq-14292	319	40	(	(	PUNCT
fpq-14292	319	41	see	see	VERB
fpq-14292	319	42	also	also	ADV
fpq-14292	319	43	fazelpour	fazelpour	VERB
fpq-14292	319	44	and	and	CCONJ
fpq-14292	319	45	danks	dank	NOUN
fpq-14292	319	46	2021	2021	NUM
fpq-14292	319	47	)	)	PUNCT
fpq-14292	319	48	.	.	PUNCT
fpq-14292	320	1	the	the	DET
fpq-14292	320	2	latter	latter	ADJ
fpq-14292	320	3	bias	bias	NOUN
fpq-14292	320	4	results	result	VERB
fpq-14292	320	5	from	from	ADP
fpq-14292	320	6	imperfect	imperfect	ADJ
fpq-14292	320	7	measuring	measuring	NOUN
fpq-14292	320	8	or	or	CCONJ
fpq-14292	320	9	sampling	sampling	NOUN
fpq-14292	320	10	.	.	PUNCT
fpq-14292	321	1	by	by	ADP
fpq-14292	321	2	contrast	contrast	NOUN
fpq-14292	321	3	,	,	PUNCT
fpq-14292	321	4	according	accord	VERB
fpq-14292	321	5	to	to	ADP
fpq-14292	321	6	suresh	suresh	NOUN
fpq-14292	321	7	and	and	CCONJ
fpq-14292	321	8	guttag	guttag	NOUN
fpq-14292	321	9	(	(	PUNCT
fpq-14292	321	10	2020	2020	NUM
fpq-14292	321	11	,	,	PUNCT
fpq-14292	321	12	4	4	NUM
fpq-14292	321	13	)	)	PUNCT
fpq-14292	321	14	,	,	PUNCT
fpq-14292	321	15	“	"	PUNCT
fpq-14292	321	16	historical	historical	ADJ
fpq-14292	321	17	bias	bias	NOUN
fpq-14292	321	18	arises	arise	VERB
fpq-14292	321	19	even	even	ADV
fpq-14292	321	20	if	if	SCONJ
fpq-14292	321	21	the	the	DET
fpq-14292	321	22	data	data	NOUN
fpq-14292	321	23	is	be	AUX
fpq-14292	321	24	perfectly	perfectly	ADV
fpq-14292	321	25	measured	measure	VERB
fpq-14292	321	26	and	and	CCONJ
fpq-14292	321	27	sampled	sample	VERB
fpq-14292	321	28	,	,	PUNCT
fpq-14292	321	29	if	if	SCONJ
fpq-14292	321	30	the	the	DET
fpq-14292	321	31	world	world	NOUN
fpq-14292	321	32	as	as	SCONJ
fpq-14292	321	33	it	it	PRON
fpq-14292	321	34	is	be	AUX
fpq-14292	321	35	or	or	CCONJ
fpq-14292	321	36	was	be	AUX
fpq-14292	321	37	leads	lead	VERB
fpq-14292	321	38	a	a	DET
fpq-14292	321	39	model	model	NOUN
fpq-14292	321	40	to	to	PART
fpq-14292	321	41	produce	produce	VERB
fpq-14292	321	42	outcomes	outcome	NOUN
fpq-14292	321	43	that	that	PRON
fpq-14292	321	44	are	be	AUX
fpq-14292	321	45	not	not	PART
fpq-14292	321	46	wanted	want	VERB
fpq-14292	321	47	.	.	PUNCT
fpq-14292	321	48	”	"	PUNCT
fpq-14292	322	1	they	they	PRON
fpq-14292	322	2	provide	provide	VERB
fpq-14292	322	3	an	an	DET
fpq-14292	322	4	example	example	NOUN
fpq-14292	322	5	of	of	ADP
fpq-14292	322	6	this	this	DET
fpq-14292	322	7	bias	bias	NOUN
fpq-14292	322	8	:	:	PUNCT
fpq-14292	322	9	in	in	ADP
fpq-14292	322	10	2018	2018	NUM
fpq-14292	322	11	,	,	PUNCT
fpq-14292	322	12	5	5	NUM
fpq-14292	322	13	%	%	NOUN
fpq-14292	322	14	of	of	ADP
fpq-14292	322	15	fortune	fortune	NOUN
fpq-14292	322	16	500	500	NUM
fpq-14292	322	17	ceos	ceo	NOUN
fpq-14292	322	18	were	be	AUX
fpq-14292	322	19	women	woman	NOUN
fpq-14292	322	20	.	.	PUNCT
fpq-14292	322	21	.	.	PUNCT
fpq-14292	322	22	.	.	PUNCT
fpq-14292	323	1	.	.	PUNCT
fpq-14292	324	1	should	should	AUX
fpq-14292	324	2	image	image	VERB
fpq-14292	324	3	search	search	NOUN
fpq-14292	324	4	results	result	NOUN
fpq-14292	324	5	for	for	ADP
fpq-14292	324	6	“	"	PUNCT
fpq-14292	324	7	ceo	ceo	NOUN
fpq-14292	324	8	”	"	PUNCT
fpq-14292	324	9	reflect	reflect	VERB
fpq-14292	324	10	that	that	DET
fpq-14292	324	11	number	number	NOUN
fpq-14292	324	12	?	?	PUNCT
fpq-14292	325	1	ultimately	ultimately	ADV
fpq-14292	325	2	,	,	PUNCT
fpq-14292	325	3	a	a	DET
fpq-14292	325	4	variety	variety	NOUN
fpq-14292	325	5	of	of	ADP
fpq-14292	325	6	stakeholders	stakeholder	NOUN
fpq-14292	325	7	,	,	PUNCT
fpq-14292	325	8	including	include	VERB
fpq-14292	325	9	affected	affect	VERB
fpq-14292	325	10	members	member	NOUN
fpq-14292	325	11	of	of	ADP
fpq-14292	325	12	society	society	NOUN
fpq-14292	325	13	,	,	PUNCT
fpq-14292	325	14	should	should	AUX
fpq-14292	325	15	evaluate	evaluate	VERB
fpq-14292	325	16	the	the	DET
fpq-14292	325	17	particular	particular	ADJ
fpq-14292	325	18	harms	harm	NOUN
fpq-14292	325	19	that	that	SCONJ
fpq-14292	325	20	this	this	DET
fpq-14292	325	21	result	result	NOUN
fpq-14292	325	22	could	could	AUX
fpq-14292	325	23	cause	cause	VERB
fpq-14292	325	24	and	and	CCONJ
fpq-14292	325	25	make	make	VERB
fpq-14292	325	26	a	a	DET
fpq-14292	325	27	judgment	judgment	NOUN
fpq-14292	325	28	.	.	PUNCT
fpq-14292	326	1	this	this	DET
fpq-14292	326	2	decision	decision	NOUN
fpq-14292	326	3	may	may	AUX
fpq-14292	326	4	be	be	AUX
fpq-14292	326	5	at	at	ADP
fpq-14292	326	6	odds	odd	NOUN
fpq-14292	326	7	with	with	ADP
fpq-14292	326	8	the	the	DET
fpq-14292	326	9	available	available	ADJ
fpq-14292	326	10	data	datum	NOUN
fpq-14292	326	11	even	even	ADV
fpq-14292	326	12	if	if	SCONJ
fpq-14292	326	13	that	that	DET
fpq-14292	326	14	data	data	NOUN
fpq-14292	326	15	is	be	AUX
fpq-14292	326	16	a	a	DET
fpq-14292	326	17	perfect	perfect	ADJ
fpq-14292	326	18	reflection	reflection	NOUN
fpq-14292	326	19	of	of	ADP
fpq-14292	326	20	the	the	DET
fpq-14292	326	21	world	world	NOUN
fpq-14292	326	22	.	.	PUNCT
fpq-14292	327	1	indeed	indeed	ADV
fpq-14292	327	2	,	,	PUNCT
fpq-14292	327	3	google	google	PROPN
fpq-14292	327	4	has	have	AUX
fpq-14292	327	5	recently	recently	ADV
fpq-14292	327	6	changed	change	VERB
fpq-14292	327	7	their	their	PRON
fpq-14292	327	8	image	image	NOUN
fpq-14292	327	9	search	search	NOUN
fpq-14292	327	10	results	result	NOUN
fpq-14292	327	11	for	for	ADP
fpq-14292	327	12	“	"	PUNCT
fpq-14292	327	13	ceo	ceo	NOUN
fpq-14292	327	14	”	"	PUNCT
fpq-14292	327	15	to	to	PART
fpq-14292	327	16	display	display	VERB
fpq-14292	327	17	a	a	DET
fpq-14292	327	18	higher	high	ADJ
fpq-14292	327	19	proportion	proportion	NOUN
fpq-14292	327	20	of	of	ADP
fpq-14292	327	21	women	woman	NOUN
fpq-14292	327	22	.	.	PUNCT
fpq-14292	327	23	”	"	PUNCT
fpq-14292	328	1	(	(	PUNCT
fpq-14292	328	2	suresh	suresh	NOUN
fpq-14292	328	3	and	and	CCONJ
fpq-14292	328	4	guttag	guttag	NOUN
fpq-14292	328	5	2020	2020	NUM
fpq-14292	328	6	,	,	PUNCT
fpq-14292	328	7	5	5	NUM
fpq-14292	328	8	)	)	PUNCT
fpq-14292	328	9	in	in	ADP
fpq-14292	328	10	our	our	PRON
fpq-14292	328	11	terms	term	NOUN
fpq-14292	328	12	,	,	PUNCT
fpq-14292	328	13	suresh	suresh	NOUN
fpq-14292	328	14	and	and	CCONJ
fpq-14292	328	15	guttag	guttag	NOUN
fpq-14292	328	16	claim	claim	NOUN
fpq-14292	328	17	that	that	SCONJ
fpq-14292	328	18	descriptive	descriptive	ADJ
fpq-14292	328	19	accuracy	accuracy	NOUN
fpq-14292	328	20	(	(	PUNCT
fpq-14292	328	21	“	"	PUNCT
fpq-14292	328	22	the	the	DET
fpq-14292	328	23	world	world	NOUN
fpq-14292	328	24	as	as	SCONJ
fpq-14292	328	25	it	it	PRON
fpq-14292	328	26	is	be	AUX
fpq-14292	328	27	or	or	CCONJ
fpq-14292	328	28	was	be	AUX
fpq-14292	328	29	”	"	PUNCT
fpq-14292	328	30	)	)	PUNCT
fpq-14292	328	31	conflicts	conflict	NOUN
fpq-14292	328	32	here	here	ADV
fpq-14292	328	33	with	with	ADP
fpq-14292	328	34	normative	normative	ADJ
fpq-14292	328	35	correctness	correctness	NOUN
fpq-14292	328	36	(	(	PUNCT
fpq-14292	328	37	i.e.	i.e.	X
fpq-14292	328	38	,	,	PUNCT
fpq-14292	328	39	we	we	PRON
fpq-14292	328	40	get	get	VERB
fpq-14292	328	41	“	"	PUNCT
fpq-14292	328	42	outcomes	outcome	NOUN
fpq-14292	328	43	that	that	PRON
fpq-14292	328	44	are	be	AUX
fpq-14292	328	45	not	not	PART
fpq-14292	328	46	wanted	want	VERB
fpq-14292	328	47	”	"	PUNCT
fpq-14292	328	48	)	)	PUNCT
fpq-14292	328	49	.	.	PUNCT
fpq-14292	329	1	that	that	PRON
fpq-14292	329	2	is	be	AUX
fpq-14292	329	3	to	to	PART
fpq-14292	329	4	say	say	VERB
fpq-14292	329	5	,	,	PUNCT
fpq-14292	329	6	imposing	impose	VERB
fpq-14292	329	7	normative	normative	ADJ
fpq-14292	329	8	correctness	correctness	NOUN
fpq-14292	329	9	by	by	ADP
fpq-14292	329	10	displaying	display	VERB
fpq-14292	329	11	a	a	DET
fpq-14292	329	12	higher	high	ADJ
fpq-14292	329	13	proportion	proportion	NOUN
fpq-14292	329	14	of	of	ADP
fpq-14292	329	15	women	woman	NOUN
fpq-14292	329	16	in	in	ADP
fpq-14292	329	17	the	the	DET
fpq-14292	329	18	image	image	NOUN
fpq-14292	329	19	search	search	NOUN
fpq-14292	329	20	results	result	NOUN
fpq-14292	329	21	may	may	AUX
fpq-14292	329	22	be	be	AUX
fpq-14292	329	23	descriptively	descriptively	ADV
fpq-14292	329	24	at	at	ADP
fpq-14292	329	25	odds	odd	NOUN
fpq-14292	329	26	with	with	ADP
fpq-14292	329	27	29	29	NUM
fpq-14292	329	28	if	if	SCONJ
fpq-14292	329	29	we	we	PRON
fpq-14292	329	30	asked	ask	VERB
fpq-14292	329	31	,	,	PUNCT
fpq-14292	329	32	“	"	PUNCT
fpq-14292	329	33	how	how	SCONJ
fpq-14292	329	34	obedient	obedient	ADJ
fpq-14292	329	35	does	do	AUX
fpq-14292	329	36	aristotle	aristotle	PROPN
fpq-14292	329	37	say	say	VERB
fpq-14292	329	38	a	a	DET
fpq-14292	329	39	proper	proper	ADJ
fpq-14292	329	40	wife	wife	NOUN
fpq-14292	329	41	should	should	AUX
fpq-14292	329	42	be	be	AUX
fpq-14292	329	43	?	?	PUNCT
fpq-14292	329	44	”	"	PUNCT
fpq-14292	330	1	the	the	DET
fpq-14292	330	2	appropriate	appropriate	ADJ
fpq-14292	330	3	answer	answer	NOUN
fpq-14292	330	4	might	might	AUX
fpq-14292	330	5	be	be	AUX
fpq-14292	330	6	“	"	PUNCT
fpq-14292	330	7	aristotle	aristotle	PROPN
fpq-14292	330	8	says	say	VERB
fpq-14292	330	9	that	that	SCONJ
fpq-14292	330	10	a	a	DET
fpq-14292	330	11	proper	proper	ADJ
fpq-14292	330	12	wife	wife	NOUN
fpq-14292	330	13	should	should	AUX
fpq-14292	330	14	be	be	AUX
fpq-14292	330	15	as	as	ADV
fpq-14292	330	16	obedient	obedient	ADJ
fpq-14292	330	17	as	as	ADP
fpq-14292	330	18	a	a	DET
fpq-14292	330	19	slave	slave	NOUN
fpq-14292	330	20	.	.	PUNCT
fpq-14292	330	21	”	"	PUNCT
fpq-14292	331	1	yet	yet	CCONJ
fpq-14292	331	2	when	when	SCONJ
fpq-14292	331	3	asking	ask	VERB
fpq-14292	331	4	the	the	DET
fpq-14292	331	5	system	system	NOUN
fpq-14292	331	6	itself	itself	PRON
fpq-14292	331	7	,	,	PUNCT
fpq-14292	331	8	it	it	PRON
fpq-14292	331	9	seems	seem	VERB
fpq-14292	331	10	clear	clear	ADJ
fpq-14292	331	11	that	that	SCONJ
fpq-14292	331	12	imposing	impose	VERB
fpq-14292	331	13	normative	normative	ADJ
fpq-14292	331	14	correctness	correctness	NOUN
fpq-14292	331	15	by	by	ADP
fpq-14292	331	16	forbidding	forbid	VERB
fpq-14292	331	17	such	such	DET
fpq-14292	331	18	a	a	DET
fpq-14292	331	19	response	response	NOUN
fpq-14292	331	20	is	be	AUX
fpq-14292	331	21	justified	justify	VERB
fpq-14292	331	22	.	.	PUNCT
fpq-14292	332	1	feminist	feminist	ADJ
fpq-14292	332	2	philosophy	philosophy	PROPN
fpq-14292	332	3	quarterly	quarterly	ADV
fpq-14292	332	4	,	,	PUNCT
fpq-14292	332	5	2022	2022	NUM
fpq-14292	332	6	,	,	PUNCT
fpq-14292	332	7	vol	vol	NOUN
fpq-14292	332	8	.	.	NOUN
fpq-14292	332	9	8	8	NUM
fpq-14292	332	10	,	,	PUNCT
fpq-14292	332	11	iss	iss	PROPN
fpq-14292	332	12	.	.	PUNCT
fpq-14292	332	13	3/4	3/4	NUM
fpq-14292	332	14	,	,	PUNCT
fpq-14292	332	15	article	article	NOUN
fpq-14292	332	16	1	1	NUM
fpq-14292	332	17	published	publish	VERB
fpq-14292	332	18	by	by	ADP
fpq-14292	332	19	scholarship@western	scholarship@western	PROPN
fpq-14292	332	20	,	,	PUNCT
fpq-14292	332	21	2022	2022	NUM
fpq-14292	332	22	18	18	NUM
fpq-14292	332	23	the	the	DET
fpq-14292	332	24	data	datum	NOUN
fpq-14292	332	25	even	even	ADV
fpq-14292	332	26	if	if	SCONJ
fpq-14292	332	27	the	the	DET
fpq-14292	332	28	data	datum	NOUN
fpq-14292	332	29	are	be	AUX
fpq-14292	332	30	“	"	PUNCT
fpq-14292	332	31	a	a	DET
fpq-14292	332	32	perfect	perfect	ADJ
fpq-14292	332	33	reflection	reflection	NOUN
fpq-14292	332	34	of	of	ADP
fpq-14292	332	35	the	the	DET
fpq-14292	332	36	world	world	NOUN
fpq-14292	332	37	.	.	PUNCT
fpq-14292	333	1	”30	”30	PUNCT
fpq-14292	333	2	by	by	ADP
fpq-14292	333	3	altering	alter	VERB
fpq-14292	333	4	the	the	DET
fpq-14292	333	5	results	result	NOUN
fpq-14292	333	6	,	,	PUNCT
fpq-14292	333	7	google	google	PROPN
fpq-14292	333	8	would	would	AUX
fpq-14292	333	9	seem	seem	VERB
fpq-14292	333	10	to	to	PART
fpq-14292	333	11	sacrifice	sacrifice	VERB
fpq-14292	333	12	descriptive	descriptive	ADJ
fpq-14292	333	13	accuracy	accuracy	NOUN
fpq-14292	333	14	in	in	ADP
fpq-14292	333	15	favor	favor	NOUN
fpq-14292	333	16	of	of	ADP
fpq-14292	333	17	normative	normative	ADJ
fpq-14292	333	18	correctness	correctness	NOUN
fpq-14292	333	19	,	,	PUNCT
fpq-14292	333	20	by	by	ADP
fpq-14292	333	21	glossing	gloss	VERB
fpq-14292	333	22	over	over	ADP
fpq-14292	333	23	the	the	DET
fpq-14292	333	24	actual	actual	ADJ
fpq-14292	333	25	disparity	disparity	NOUN
fpq-14292	333	26	between	between	ADP
fpq-14292	333	27	women	woman	NOUN
fpq-14292	333	28	and	and	CCONJ
fpq-14292	333	29	men	man	NOUN
fpq-14292	333	30	who	who	PRON
fpq-14292	333	31	are	be	AUX
fpq-14292	333	32	ceos	ceo	NOUN
fpq-14292	333	33	.	.	PUNCT
fpq-14292	334	1	this	this	DET
fpq-14292	334	2	inaccuracy	inaccuracy	ADJ
fpq-14292	334	3	highlights	highlight	VERB
fpq-14292	334	4	an	an	DET
fpq-14292	334	5	ethical	ethical	ADJ
fpq-14292	334	6	problem	problem	NOUN
fpq-14292	334	7	.	.	PUNCT
fpq-14292	335	1	to	to	PART
fpq-14292	335	2	possess	possess	VERB
fpq-14292	335	3	such	such	ADJ
fpq-14292	335	4	descriptively	descriptively	ADV
fpq-14292	335	5	accurate	accurate	ADJ
fpq-14292	335	6	information	information	NOUN
fpq-14292	335	7	is	be	AUX
fpq-14292	335	8	often	often	ADV
fpq-14292	335	9	to	to	PART
fpq-14292	335	10	acquire	acquire	VERB
fpq-14292	335	11	ethically	ethically	ADV
fpq-14292	335	12	useful	useful	ADJ
fpq-14292	335	13	information	information	NOUN
fpq-14292	335	14	—	—	PUNCT
fpq-14292	335	15	in	in	ADP
fpq-14292	335	16	this	this	DET
fpq-14292	335	17	case	case	NOUN
fpq-14292	335	18	,	,	PUNCT
fpq-14292	335	19	the	the	DET
fpq-14292	335	20	information	information	NOUN
fpq-14292	335	21	might	might	AUX
fpq-14292	335	22	be	be	AUX
fpq-14292	335	23	used	use	VERB
fpq-14292	335	24	as	as	ADP
fpq-14292	335	25	evidence	evidence	NOUN
fpq-14292	335	26	for	for	ADP
fpq-14292	335	27	the	the	DET
fpq-14292	335	28	claim	claim	NOUN
fpq-14292	335	29	that	that	SCONJ
fpq-14292	335	30	there	there	PRON
fpq-14292	335	31	is	be	VERB
fpq-14292	335	32	indeed	indeed	ADV
fpq-14292	335	33	a	a	DET
fpq-14292	335	34	disparity	disparity	NOUN
fpq-14292	335	35	between	between	ADP
fpq-14292	335	36	women	woman	NOUN
fpq-14292	335	37	and	and	CCONJ
fpq-14292	335	38	men	man	NOUN
fpq-14292	335	39	ceos	ceo	NOUN
fpq-14292	335	40	,	,	PUNCT
fpq-14292	335	41	perhaps	perhaps	ADV
fpq-14292	335	42	in	in	ADP
fpq-14292	335	43	the	the	DET
fpq-14292	335	44	course	course	NOUN
fpq-14292	335	45	of	of	ADP
fpq-14292	335	46	redressing	redress	VERB
fpq-14292	335	47	the	the	DET
fpq-14292	335	48	disparity	disparity	NOUN
fpq-14292	335	49	.	.	PUNCT
fpq-14292	336	1	however	however	ADV
fpq-14292	336	2	,	,	PUNCT
fpq-14292	336	3	once	once	ADV
fpq-14292	336	4	google	google	PROPN
fpq-14292	336	5	alters	alter	VERB
fpq-14292	336	6	the	the	DET
fpq-14292	336	7	results	result	NOUN
fpq-14292	336	8	as	as	SCONJ
fpq-14292	336	9	described	describe	VERB
fpq-14292	336	10	,	,	PUNCT
fpq-14292	336	11	one	one	NUM
fpq-14292	336	12	effect	effect	NOUN
fpq-14292	336	13	is	be	AUX
fpq-14292	336	14	that	that	SCONJ
fpq-14292	336	15	we	we	PRON
fpq-14292	336	16	lack	lack	VERB
fpq-14292	336	17	access	access	NOUN
fpq-14292	336	18	to	to	ADP
fpq-14292	336	19	this	this	DET
fpq-14292	336	20	descriptively	descriptively	ADV
fpq-14292	336	21	accurate	accurate	ADJ
fpq-14292	336	22	and	and	CCONJ
fpq-14292	336	23	ethically	ethically	ADV
fpq-14292	336	24	useful	useful	ADJ
fpq-14292	336	25	information	information	NOUN
fpq-14292	336	26	and	and	CCONJ
fpq-14292	336	27	so	so	ADV
fpq-14292	336	28	injustices	injustice	NOUN
fpq-14292	336	29	of	of	ADP
fpq-14292	336	30	this	this	DET
fpq-14292	336	31	sort	sort	NOUN
fpq-14292	336	32	might	might	AUX
fpq-14292	336	33	go	go	VERB
fpq-14292	336	34	underrecognized	underrecognized	ADJ
fpq-14292	336	35	.	.	PUNCT
fpq-14292	337	1	moreover	moreover	ADV
fpq-14292	337	2	,	,	PUNCT
fpq-14292	337	3	it	it	PRON
fpq-14292	337	4	is	be	AUX
fpq-14292	337	5	implicit	implicit	ADJ
fpq-14292	337	6	in	in	ADP
fpq-14292	337	7	suresh	suresh	PROPN
fpq-14292	337	8	and	and	CCONJ
fpq-14292	337	9	guttag	guttag	NOUN
fpq-14292	337	10	’s	’s	PART
fpq-14292	337	11	discussion	discussion	NOUN
fpq-14292	337	12	of	of	ADP
fpq-14292	337	13	their	their	PRON
fpq-14292	337	14	example	example	NOUN
fpq-14292	337	15	that	that	SCONJ
fpq-14292	337	16	there	there	PRON
fpq-14292	337	17	is	be	VERB
fpq-14292	337	18	a	a	DET
fpq-14292	337	19	value	value	NOUN
fpq-14292	337	20	both	both	PRON
fpq-14292	337	21	to	to	ADP
fpq-14292	337	22	descriptive	descriptive	ADJ
fpq-14292	337	23	accuracy	accuracy	NOUN
fpq-14292	337	24	and	and	CCONJ
fpq-14292	337	25	to	to	ADP
fpq-14292	337	26	normative	normative	ADJ
fpq-14292	337	27	correctness	correctness	NOUN
fpq-14292	337	28	,	,	PUNCT
fpq-14292	337	29	although	although	SCONJ
fpq-14292	337	30	suresh	suresh	PROPN
fpq-14292	337	31	and	and	CCONJ
fpq-14292	337	32	guttag	guttag	NOUN
fpq-14292	337	33	are	be	AUX
fpq-14292	337	34	unclear	unclear	ADJ
fpq-14292	337	35	about	about	ADP
fpq-14292	337	36	whether	whether	SCONJ
fpq-14292	337	37	they	they	PRON
fpq-14292	337	38	think	think	VERB
fpq-14292	337	39	it	it	PRON
fpq-14292	337	40	can	can	AUX
fpq-14292	337	41	be	be	AUX
fpq-14292	337	42	more	more	ADV
fpq-14292	337	43	valuable	valuable	ADJ
fpq-14292	337	44	to	to	PART
fpq-14292	337	45	prioritize	prioritize	VERB
fpq-14292	337	46	one	one	NUM
fpq-14292	337	47	form	form	NOUN
fpq-14292	337	48	of	of	ADP
fpq-14292	337	49	accuracy	accuracy	NOUN
fpq-14292	337	50	or	or	CCONJ
fpq-14292	337	51	correctness	correctness	NOUN
fpq-14292	337	52	over	over	ADP
fpq-14292	337	53	the	the	DET
fpq-14292	337	54	other	other	ADJ
fpq-14292	337	55	.	.	PUNCT
fpq-14292	338	1	after	after	ADV
fpq-14292	338	2	all	all	ADV
fpq-14292	338	3	,	,	PUNCT
fpq-14292	338	4	while	while	SCONJ
fpq-14292	338	5	some	some	DET
fpq-14292	338	6	ethical	ethical	ADJ
fpq-14292	338	7	purposes	purpose	NOUN
fpq-14292	338	8	might	might	AUX
fpq-14292	338	9	best	well	ADV
fpq-14292	338	10	be	be	AUX
fpq-14292	338	11	served	serve	VERB
fpq-14292	338	12	by	by	ADP
fpq-14292	338	13	google	google	PROPN
fpq-14292	338	14	’s	’s	PART
fpq-14292	338	15	decision	decision	NOUN
fpq-14292	338	16	to	to	PART
fpq-14292	338	17	change	change	VERB
fpq-14292	338	18	its	its	PRON
fpq-14292	338	19	results	result	NOUN
fpq-14292	338	20	for	for	ADP
fpq-14292	338	21	“	"	PUNCT
fpq-14292	338	22	ceo	ceo	NOUN
fpq-14292	338	23	”	"	PUNCT
fpq-14292	338	24	to	to	PART
fpq-14292	338	25	display	display	VERB
fpq-14292	338	26	more	more	ADJ
fpq-14292	338	27	images	image	NOUN
fpq-14292	338	28	of	of	ADP
fpq-14292	338	29	women	woman	NOUN
fpq-14292	338	30	(	(	PUNCT
fpq-14292	338	31	thereby	thereby	ADV
fpq-14292	338	32	making	make	VERB
fpq-14292	338	33	the	the	DET
fpq-14292	338	34	results	result	NOUN
fpq-14292	338	35	more	more	ADV
fpq-14292	338	36	normatively	normatively	ADV
fpq-14292	338	37	correct	correct	ADJ
fpq-14292	338	38	)	)	PUNCT
fpq-14292	338	39	,	,	PUNCT
fpq-14292	338	40	other	other	ADJ
fpq-14292	338	41	ethical	ethical	ADJ
fpq-14292	338	42	purposes	purpose	NOUN
fpq-14292	338	43	may	may	AUX
fpq-14292	338	44	be	be	AUX
fpq-14292	338	45	directly	directly	ADV
fpq-14292	338	46	undermined	undermined	ADJ
fpq-14292	338	47	and	and	CCONJ
fpq-14292	338	48	would	would	AUX
fpq-14292	338	49	be	be	AUX
fpq-14292	338	50	better	well	ADV
fpq-14292	338	51	served	serve	VERB
fpq-14292	338	52	by	by	ADP
fpq-14292	338	53	descriptively	descriptively	ADV
fpq-14292	338	54	accurate	accurate	ADJ
fpq-14292	338	55	results	result	NOUN
fpq-14292	338	56	;	;	PUNCT
fpq-14292	338	57	so	so	ADV
fpq-14292	338	58	,	,	PUNCT
fpq-14292	338	59	google	google	PROPN
fpq-14292	338	60	’s	’s	PART
fpq-14292	338	61	strategy	strategy	NOUN
fpq-14292	338	62	may	may	AUX
fpq-14292	338	63	be	be	AUX
fpq-14292	338	64	ethically	ethically	ADV
fpq-14292	338	65	undesirable	undesirable	ADJ
fpq-14292	338	66	in	in	ADP
fpq-14292	338	67	other	other	ADJ
fpq-14292	338	68	respects	respect	NOUN
fpq-14292	338	69	.	.	PUNCT
fpq-14292	339	1	that	that	PRON
fpq-14292	339	2	is	be	AUX
fpq-14292	339	3	not	not	PART
fpq-14292	339	4	to	to	PART
fpq-14292	339	5	say	say	VERB
fpq-14292	339	6	google	google	PROPN
fpq-14292	339	7	made	make	VERB
fpq-14292	339	8	the	the	DET
fpq-14292	339	9	wrong	wrong	ADJ
fpq-14292	339	10	decision	decision	NOUN
fpq-14292	339	11	.	.	PUNCT
fpq-14292	340	1	it	it	PRON
fpq-14292	340	2	is	be	AUX
fpq-14292	340	3	only	only	ADV
fpq-14292	340	4	to	to	PART
fpq-14292	340	5	say	say	VERB
fpq-14292	340	6	that	that	SCONJ
fpq-14292	340	7	there	there	PRON
fpq-14292	340	8	might	might	AUX
fpq-14292	340	9	be	be	AUX
fpq-14292	340	10	a	a	DET
fpq-14292	340	11	cost	cost	NOUN
fpq-14292	340	12	to	to	ADP
fpq-14292	340	13	prioritizing	prioritize	VERB
fpq-14292	340	14	either	either	CCONJ
fpq-14292	340	15	descriptive	descriptive	ADJ
fpq-14292	340	16	accuracy	accuracy	NOUN
fpq-14292	340	17	or	or	CCONJ
fpq-14292	340	18	normative	normative	ADJ
fpq-14292	340	19	correctness	correctness	NOUN
fpq-14292	340	20	in	in	ADP
fpq-14292	340	21	search	search	NOUN
fpq-14292	340	22	,	,	PUNCT
fpq-14292	340	23	as	as	ADP
fpq-14292	340	24	with	with	ADP
fpq-14292	340	25	nlp	nlp	NOUN
fpq-14292	340	26	systems	system	NOUN
fpq-14292	340	27	relying	rely	VERB
fpq-14292	340	28	on	on	ADP
fpq-14292	340	29	wordembeddings	wordembedding	NOUN
fpq-14292	340	30	.	.	PUNCT
fpq-14292	341	1	in	in	ADP
fpq-14292	341	2	fact	fact	NOUN
fpq-14292	341	3	,	,	PUNCT
fpq-14292	341	4	we	we	PRON
fpq-14292	341	5	have	have	VERB
fpq-14292	341	6	reservations	reservation	NOUN
fpq-14292	341	7	about	about	ADP
fpq-14292	341	8	whether	whether	SCONJ
fpq-14292	341	9	our	our	PRON
fpq-14292	341	10	ethical	ethical	ADJ
fpq-14292	341	11	dilemma	dilemma	NOUN
fpq-14292	341	12	applies	apply	VERB
fpq-14292	341	13	in	in	ADP
fpq-14292	341	14	suresh	suresh	PROPN
fpq-14292	341	15	and	and	CCONJ
fpq-14292	341	16	guttag	guttag	NOUN
fpq-14292	341	17	’s	’s	PART
fpq-14292	341	18	ceo	ceo	NOUN
fpq-14292	341	19	case	case	NOUN
fpq-14292	341	20	,	,	PUNCT
fpq-14292	341	21	or	or	CCONJ
fpq-14292	341	22	in	in	ADP
fpq-14292	341	23	search	search	NOUN
fpq-14292	341	24	more	more	ADV
fpq-14292	341	25	widely	widely	ADV
fpq-14292	341	26	.	.	PUNCT
fpq-14292	342	1	that	that	PRON
fpq-14292	342	2	is	be	AUX
fpq-14292	342	3	because	because	SCONJ
fpq-14292	342	4	it	it	PRON
fpq-14292	342	5	is	be	AUX
fpq-14292	342	6	unlikely	unlikely	ADJ
fpq-14292	342	7	that	that	SCONJ
fpq-14292	342	8	google	google	PROPN
fpq-14292	342	9	search	search	NOUN
fpq-14292	342	10	(	(	PUNCT
fpq-14292	342	11	or	or	CCONJ
fpq-14292	342	12	search	search	NOUN
fpq-14292	342	13	engines	engine	NOUN
fpq-14292	342	14	generally	generally	ADV
fpq-14292	342	15	)	)	PUNCT
fpq-14292	342	16	actually	actually	ADV
fpq-14292	342	17	provide	provide	VERB
fpq-14292	342	18	us	we	PRON
fpq-14292	342	19	with	with	ADP
fpq-14292	342	20	results	result	NOUN
fpq-14292	342	21	that	that	PRON
fpq-14292	342	22	are	be	AUX
fpq-14292	342	23	descriptively	descriptively	ADV
fpq-14292	342	24	accurate	accurate	ADJ
fpq-14292	342	25	in	in	ADP
fpq-14292	342	26	the	the	DET
fpq-14292	342	27	first	first	ADJ
fpq-14292	342	28	place	place	NOUN
fpq-14292	342	29	.	.	PUNCT
fpq-14292	343	1	for	for	ADP
fpq-14292	343	2	example	example	NOUN
fpq-14292	343	3	,	,	PUNCT
fpq-14292	343	4	safiya	safiya	PROPN
fpq-14292	343	5	umoja	umoja	NOUN
fpq-14292	343	6	noble	noble	ADJ
fpq-14292	343	7	(	(	PUNCT
fpq-14292	343	8	2018	2018	NUM
fpq-14292	343	9	)	)	PUNCT
fpq-14292	343	10	outlines	outline	VERB
fpq-14292	343	11	how	how	SCONJ
fpq-14292	343	12	google	google	PROPN
fpq-14292	343	13	’s	’s	PART
fpq-14292	343	14	search	search	NOUN
fpq-14292	343	15	results	result	NOUN
fpq-14292	343	16	can	can	AUX
fpq-14292	343	17	be	be	AUX
fpq-14292	343	18	racist	racist	ADJ
fpq-14292	343	19	in	in	ADP
fpq-14292	343	20	ways	way	NOUN
fpq-14292	343	21	that	that	PRON
fpq-14292	343	22	are	be	AUX
fpq-14292	343	23	driven	drive	VERB
fpq-14292	343	24	by	by	ADP
fpq-14292	343	25	internal	internal	ADJ
fpq-14292	343	26	factors	factor	NOUN
fpq-14292	343	27	,	,	PUNCT
fpq-14292	343	28	such	such	ADJ
fpq-14292	343	29	as	as	ADP
fpq-14292	343	30	an	an	DET
fpq-14292	343	31	agenda	agenda	NOUN
fpq-14292	343	32	built	build	VERB
fpq-14292	343	33	into	into	ADP
fpq-14292	343	34	google	google	PROPN
fpq-14292	343	35	’s	’s	PART
fpq-14292	343	36	algorithms	algorithm	NOUN
fpq-14292	343	37	,	,	PUNCT
fpq-14292	343	38	or	or	CCONJ
fpq-14292	343	39	by	by	ADP
fpq-14292	343	40	external	external	ADJ
fpq-14292	343	41	factors	factor	NOUN
fpq-14292	343	42	,	,	PUNCT
fpq-14292	343	43	such	such	ADJ
fpq-14292	343	44	as	as	ADP
fpq-14292	343	45	how	how	SCONJ
fpq-14292	343	46	savvy	savvy	ADJ
fpq-14292	343	47	users	user	NOUN
fpq-14292	343	48	can	can	AUX
fpq-14292	343	49	deliberately	deliberately	ADV
fpq-14292	343	50	influence	influence	VERB
fpq-14292	343	51	results	result	NOUN
fpq-14292	343	52	.	.	PUNCT
fpq-14292	344	1	to	to	ADP
fpq-14292	344	2	the	the	DET
fpq-14292	344	3	extent	extent	NOUN
fpq-14292	344	4	that	that	SCONJ
fpq-14292	344	5	search	search	NOUN
fpq-14292	344	6	results	result	NOUN
fpq-14292	344	7	are	be	AUX
fpq-14292	344	8	not	not	PART
fpq-14292	344	9	descriptively	descriptively	ADV
fpq-14292	344	10	accurate	accurate	ADJ
fpq-14292	344	11	,	,	PUNCT
fpq-14292	344	12	our	our	PRON
fpq-14292	344	13	ethical	ethical	ADJ
fpq-14292	344	14	dilemma	dilemma	NOUN
fpq-14292	344	15	does	do	AUX
fpq-14292	344	16	not	not	PART
fpq-14292	344	17	arise	arise	VERB
fpq-14292	344	18	(	(	PUNCT
fpq-14292	344	19	as	as	SCONJ
fpq-14292	344	20	we	we	PRON
fpq-14292	344	21	explain	explain	VERB
fpq-14292	344	22	in	in	ADP
fpq-14292	344	23	detail	detail	NOUN
fpq-14292	344	24	in	in	ADP
fpq-14292	344	25	section	section	NOUN
fpq-14292	344	26	7	7	NUM
fpq-14292	344	27	)	)	PUNCT
fpq-14292	344	28	.	.	PUNCT
fpq-14292	345	1	we	we	PRON
fpq-14292	345	2	confront	confront	VERB
fpq-14292	345	3	the	the	DET
fpq-14292	345	4	dilemma	dilemma	NOUN
fpq-14292	345	5	only	only	ADV
fpq-14292	345	6	when	when	SCONJ
fpq-14292	345	7	we	we	PRON
fpq-14292	345	8	have	have	VERB
fpq-14292	345	9	descriptively	descriptively	ADV
fpq-14292	345	10	accurate	accurate	ADJ
fpq-14292	345	11	outputs	output	NOUN
fpq-14292	345	12	that	that	PRON
fpq-14292	345	13	are	be	AUX
fpq-14292	345	14	normatively	normatively	ADV
fpq-14292	345	15	incorrect	incorrect	ADJ
fpq-14292	345	16	,	,	PUNCT
fpq-14292	345	17	and	and	CCONJ
fpq-14292	345	18	the	the	DET
fpq-14292	345	19	incorrectness	incorrectness	NOUN
fpq-14292	345	20	is	be	AUX
fpq-14292	345	21	due	due	ADJ
fpq-14292	345	22	to	to	ADP
fpq-14292	345	23	the	the	DET
fpq-14292	345	24	accuracy	accuracy	NOUN
fpq-14292	345	25	.	.	PUNCT
fpq-14292	346	1	in	in	ADP
fpq-14292	346	2	such	such	ADJ
fpq-14292	346	3	cases	case	NOUN
fpq-14292	346	4	,	,	PUNCT
fpq-14292	346	5	the	the	DET
fpq-14292	346	6	dilemma	dilemma	NOUN
fpq-14292	346	7	consists	consist	VERB
fpq-14292	346	8	in	in	ADP
fpq-14292	346	9	our	our	PRON
fpq-14292	346	10	having	have	VERB
fpq-14292	346	11	to	to	PART
fpq-14292	346	12	decide	decide	VERB
fpq-14292	346	13	whether	whether	SCONJ
fpq-14292	346	14	to	to	PART
fpq-14292	346	15	prioritize	prioritize	VERB
fpq-14292	346	16	descriptive	descriptive	ADJ
fpq-14292	346	17	accuracy	accuracy	NOUN
fpq-14292	346	18	(	(	PUNCT
fpq-14292	346	19	including	include	VERB
fpq-14292	346	20	its	its	PRON
fpq-14292	346	21	ethical	ethical	ADJ
fpq-14292	346	22	30	30	NUM
fpq-14292	346	23	in	in	ADP
fpq-14292	346	24	another	another	DET
fpq-14292	346	25	widely	widely	ADV
fpq-14292	346	26	discussed	discuss	VERB
fpq-14292	346	27	case	case	NOUN
fpq-14292	346	28	,	,	PUNCT
fpq-14292	346	29	the	the	DET
fpq-14292	346	30	guardian	guardian	NOUN
fpq-14292	346	31	reported	report	VERB
fpq-14292	346	32	in	in	ADP
fpq-14292	346	33	2016	2016	NUM
fpq-14292	346	34	how	how	SCONJ
fpq-14292	346	35	twitter	twitter	NOUN
fpq-14292	346	36	user	user	NOUN
fpq-14292	346	37	kabir	kabir	PROPN
fpq-14292	346	38	alli	alli	PROPN
fpq-14292	346	39	conducted	conduct	VERB
fpq-14292	346	40	a	a	DET
fpq-14292	346	41	google	google	NOUN
fpq-14292	346	42	image	image	NOUN
fpq-14292	346	43	search	search	NOUN
fpq-14292	346	44	using	use	VERB
fpq-14292	346	45	the	the	DET
fpq-14292	346	46	phrases	phrase	NOUN
fpq-14292	346	47	“	"	PUNCT
fpq-14292	346	48	three	three	NUM
fpq-14292	346	49	black	black	ADJ
fpq-14292	346	50	teenagers	teenager	NOUN
fpq-14292	346	51	”	"	PUNCT
fpq-14292	346	52	and	and	CCONJ
fpq-14292	346	53	“	"	PUNCT
fpq-14292	346	54	three	three	NUM
fpq-14292	346	55	white	white	ADJ
fpq-14292	346	56	teenagers	teenager	NOUN
fpq-14292	346	57	”	"	PUNCT
fpq-14292	346	58	(	(	PUNCT
fpq-14292	346	59	allen	allen	NOUN
fpq-14292	346	60	2016	2016	NUM
fpq-14292	346	61	)	)	PUNCT
fpq-14292	346	62	.	.	PUNCT
fpq-14292	347	1	results	result	NOUN
fpq-14292	347	2	were	be	AUX
fpq-14292	347	3	very	very	ADV
fpq-14292	347	4	different	different	ADJ
fpq-14292	347	5	.	.	PUNCT
fpq-14292	348	1	the	the	DET
fpq-14292	348	2	search	search	NOUN
fpq-14292	348	3	for	for	ADP
fpq-14292	348	4	“	"	PUNCT
fpq-14292	348	5	three	three	NUM
fpq-14292	348	6	black	black	ADJ
fpq-14292	348	7	teenagers	teenager	NOUN
fpq-14292	348	8	”	"	PUNCT
fpq-14292	348	9	returned	return	VERB
fpq-14292	348	10	mostly	mostly	ADV
fpq-14292	348	11	police	police	NOUN
fpq-14292	348	12	mugshots	mugshot	NOUN
fpq-14292	348	13	while	while	SCONJ
fpq-14292	348	14	for	for	ADP
fpq-14292	348	15	“	"	PUNCT
fpq-14292	348	16	three	three	NUM
fpq-14292	348	17	white	white	ADJ
fpq-14292	348	18	teenagers	teenager	NOUN
fpq-14292	348	19	”	"	PUNCT
fpq-14292	348	20	it	it	PRON
fpq-14292	348	21	gave	give	VERB
fpq-14292	348	22	stock	stock	NOUN
fpq-14292	348	23	photographs	photograph	NOUN
fpq-14292	348	24	of	of	ADP
fpq-14292	348	25	smiling	smile	VERB
fpq-14292	348	26	white	white	ADJ
fpq-14292	348	27	youths	youth	NOUN
fpq-14292	348	28	.	.	PUNCT
fpq-14292	349	1	deery	deery	NOUN
fpq-14292	349	2	and	and	CCONJ
fpq-14292	349	3	bailey	bailey	PROPN
fpq-14292	349	4	–	–	PUNCT
fpq-14292	349	5	the	the	DET
fpq-14292	349	6	bias	bias	NOUN
fpq-14292	349	7	dilemma	dilemma	NOUN
fpq-14292	349	8	:	:	PUNCT
fpq-14292	349	9	the	the	DET
fpq-14292	349	10	ethics	ethic	NOUN
fpq-14292	349	11	of	of	ADP
fpq-14292	349	12	algorithmic	algorithmic	ADJ
fpq-14292	349	13	bias	bias	NOUN
fpq-14292	349	14	in	in	ADP
fpq-14292	349	15	natural	natural	ADJ
fpq-14292	349	16	-	-	PUNCT
fpq-14292	349	17	language	language	NOUN
fpq-14292	349	18	processing	processing	NOUN
fpq-14292	349	19	published	publish	VERB
fpq-14292	349	20	by	by	ADP
fpq-14292	349	21	scholarship@western	scholarship@western	PROPN
fpq-14292	349	22	,	,	PUNCT
fpq-14292	349	23	2022	2022	NUM
fpq-14292	349	24	19	19	NUM
fpq-14292	349	25	usefulness	usefulness	NOUN
fpq-14292	349	26	)	)	PUNCT
fpq-14292	349	27	at	at	ADP
fpq-14292	349	28	the	the	DET
fpq-14292	349	29	expense	expense	NOUN
fpq-14292	349	30	of	of	ADP
fpq-14292	349	31	normative	normative	ADJ
fpq-14292	349	32	correctness	correctness	NOUN
fpq-14292	349	33	or	or	CCONJ
fpq-14292	349	34	vice	vice	NOUN
fpq-14292	349	35	versa	versa	ADV
fpq-14292	349	36	.	.	PUNCT
fpq-14292	350	1	even	even	ADV
fpq-14292	350	2	if	if	SCONJ
fpq-14292	350	3	search	search	NOUN
fpq-14292	350	4	results	result	NOUN
fpq-14292	350	5	are	be	AUX
fpq-14292	350	6	normatively	normatively	ADV
fpq-14292	350	7	incorrect	incorrect	ADJ
fpq-14292	350	8	(	(	PUNCT
fpq-14292	350	9	as	as	ADP
fpq-14292	350	10	in	in	ADP
fpq-14292	350	11	suresh	suresh	NOUN
fpq-14292	350	12	and	and	CCONJ
fpq-14292	350	13	guttag	guttag	NOUN
fpq-14292	350	14	’s	’s	PART
fpq-14292	350	15	case	case	NOUN
fpq-14292	350	16	)	)	PUNCT
fpq-14292	350	17	unless	unless	SCONJ
fpq-14292	350	18	they	they	PRON
fpq-14292	350	19	are	be	AUX
fpq-14292	350	20	also	also	ADV
fpq-14292	350	21	descriptively	descriptively	ADV
fpq-14292	350	22	accurate	accurate	ADJ
fpq-14292	350	23	,	,	PUNCT
fpq-14292	350	24	there	there	PRON
fpq-14292	350	25	is	be	VERB
fpq-14292	350	26	no	no	DET
fpq-14292	350	27	dilemma	dilemma	NOUN
fpq-14292	350	28	(	(	PUNCT
fpq-14292	350	29	as	as	SCONJ
fpq-14292	350	30	we	we	PRON
fpq-14292	350	31	explain	explain	VERB
fpq-14292	350	32	in	in	ADP
fpq-14292	350	33	sections	section	NOUN
fpq-14292	350	34	7	7	NUM
fpq-14292	350	35	and	and	CCONJ
fpq-14292	350	36	8	8	NUM
fpq-14292	350	37	below	below	ADV
fpq-14292	350	38	)	)	PUNCT
fpq-14292	350	39	;	;	PUNCT
fpq-14292	350	40	we	we	PRON
fpq-14292	350	41	must	must	AUX
fpq-14292	350	42	simply	simply	ADV
fpq-14292	350	43	address	address	VERB
fpq-14292	350	44	the	the	DET
fpq-14292	350	45	matter	matter	NOUN
fpq-14292	350	46	by	by	ADP
fpq-14292	350	47	trying	try	VERB
fpq-14292	350	48	to	to	PART
fpq-14292	350	49	attain	attain	VERB
fpq-14292	350	50	normative	normative	ADJ
fpq-14292	350	51	correctness.31	correctness.31	NOUN
fpq-14292	350	52	even	even	ADV
fpq-14292	350	53	so	so	ADV
fpq-14292	350	54	,	,	PUNCT
fpq-14292	350	55	in	in	ADP
fpq-14292	350	56	an	an	DET
fpq-14292	350	57	idealized	idealized	ADJ
fpq-14292	350	58	case	case	NOUN
fpq-14292	350	59	where	where	SCONJ
fpq-14292	350	60	search	search	NOUN
fpq-14292	350	61	results	result	NOUN
fpq-14292	350	62	did	do	AUX
fpq-14292	350	63	reflect	reflect	VERB
fpq-14292	350	64	descriptive	descriptive	ADJ
fpq-14292	350	65	accuracy	accuracy	NOUN
fpq-14292	350	66	and	and	CCONJ
fpq-14292	350	67	were	be	AUX
fpq-14292	350	68	normatively	normatively	ADV
fpq-14292	350	69	incorrect	incorrect	ADJ
fpq-14292	350	70	,	,	PUNCT
fpq-14292	350	71	our	our	PRON
fpq-14292	350	72	dilemma	dilemma	NOUN
fpq-14292	350	73	would	would	AUX
fpq-14292	350	74	frequently	frequently	ADV
fpq-14292	350	75	arise	arise	VERB
fpq-14292	350	76	.	.	PUNCT
fpq-14292	351	1	after	after	ADV
fpq-14292	351	2	all	all	ADV
fpq-14292	351	3	,	,	PUNCT
fpq-14292	351	4	descriptively	descriptively	ADV
fpq-14292	351	5	accurate	accurate	ADJ
fpq-14292	351	6	information	information	NOUN
fpq-14292	351	7	can	can	AUX
fpq-14292	351	8	—	—	PUNCT
fpq-14292	351	9	as	as	ADP
fpq-14292	351	10	in	in	ADP
fpq-14292	351	11	the	the	DET
fpq-14292	351	12	case	case	NOUN
fpq-14292	351	13	of	of	ADP
fpq-14292	351	14	word	word	NOUN
fpq-14292	351	15	-	-	PUNCT
fpq-14292	351	16	embeddings	embedding	NOUN
fpq-14292	351	17	models	model	NOUN
fpq-14292	351	18	,	,	PUNCT
fpq-14292	351	19	for	for	ADP
fpq-14292	351	20	instance	instance	NOUN
fpq-14292	351	21	—	—	PUNCT
fpq-14292	351	22	be	be	AUX
fpq-14292	351	23	ethically	ethically	ADV
fpq-14292	351	24	useful	useful	ADJ
fpq-14292	351	25	in	in	ADP
fpq-14292	351	26	studying	study	VERB
fpq-14292	351	27	biases	bias	NOUN
fpq-14292	351	28	(	(	PUNCT
fpq-14292	351	29	as	as	SCONJ
fpq-14292	351	30	expressed	express	VERB
fpq-14292	351	31	in	in	ADP
fpq-14292	351	32	language	language	NOUN
fpq-14292	351	33	)	)	PUNCT
fpq-14292	351	34	,	,	PUNCT
fpq-14292	351	35	how	how	SCONJ
fpq-14292	351	36	they	they	PRON
fpq-14292	351	37	are	be	AUX
fpq-14292	351	38	perpetuated	perpetuate	VERB
fpq-14292	351	39	,	,	PUNCT
fpq-14292	351	40	and	and	CCONJ
fpq-14292	351	41	how	how	SCONJ
fpq-14292	351	42	to	to	PART
fpq-14292	351	43	address	address	VERB
fpq-14292	351	44	them	they	PRON
fpq-14292	351	45	.	.	PUNCT
fpq-14292	352	1	7	7	X
fpq-14292	352	2	.	.	X
fpq-14292	352	3	the	the	DET
fpq-14292	352	4	ethical	ethical	ADJ
fpq-14292	352	5	dilemma	dilemma	NOUN
fpq-14292	352	6	it	it	PRON
fpq-14292	352	7	can	can	AUX
fpq-14292	352	8	sometimes	sometimes	ADV
fpq-14292	352	9	be	be	AUX
fpq-14292	352	10	ethically	ethically	ADV
fpq-14292	352	11	valuable	valuable	ADJ
fpq-14292	352	12	not	not	PART
fpq-14292	352	13	to	to	PART
fpq-14292	352	14	debias	debias	VERB
fpq-14292	352	15	nlp	nlp	NOUN
fpq-14292	352	16	models	model	NOUN
fpq-14292	352	17	or	or	CCONJ
fpq-14292	352	18	their	their	PRON
fpq-14292	352	19	outputs	output	NOUN
fpq-14292	352	20	.	.	PUNCT
fpq-14292	353	1	for	for	ADP
fpq-14292	353	2	one	one	NUM
fpq-14292	353	3	thing	thing	NOUN
fpq-14292	353	4	,	,	PUNCT
fpq-14292	353	5	not	not	PART
fpq-14292	353	6	debiasing	debiase	VERB
fpq-14292	353	7	enables	enable	VERB
fpq-14292	353	8	us	we	PRON
fpq-14292	353	9	to	to	PART
fpq-14292	353	10	uncover	uncover	VERB
fpq-14292	353	11	biases	bias	NOUN
fpq-14292	353	12	and	and	CCONJ
fpq-14292	353	13	can	can	AUX
fpq-14292	353	14	thereby	thereby	ADV
fpq-14292	353	15	prevent	prevent	VERB
fpq-14292	353	16	us	we	PRON
fpq-14292	353	17	from	from	ADP
fpq-14292	353	18	thinking	think	VERB
fpq-14292	353	19	that	that	SCONJ
fpq-14292	353	20	all	all	PRON
fpq-14292	353	21	is	be	AUX
fpq-14292	353	22	right	right	ADJ
fpq-14292	353	23	with	with	ADP
fpq-14292	353	24	the	the	DET
fpq-14292	353	25	world	world	NOUN
fpq-14292	353	26	(	(	PUNCT
fpq-14292	353	27	in	in	ADP
fpq-14292	353	28	some	some	DET
fpq-14292	353	29	regard	regard	NOUN
fpq-14292	353	30	)	)	PUNCT
fpq-14292	353	31	when	when	SCONJ
fpq-14292	353	32	it	it	PRON
fpq-14292	353	33	is	be	AUX
fpq-14292	353	34	not	not	PART
fpq-14292	353	35	.	.	PUNCT
fpq-14292	354	1	however	however	ADV
fpq-14292	354	2	,	,	PUNCT
fpq-14292	354	3	a	a	DET
fpq-14292	354	4	lot	lot	NOUN
fpq-14292	354	5	depends	depend	VERB
fpq-14292	354	6	on	on	ADP
fpq-14292	354	7	the	the	DET
fpq-14292	354	8	application	application	NOUN
fpq-14292	354	9	under	under	ADP
fpq-14292	354	10	consideration	consideration	NOUN
fpq-14292	354	11	.	.	PUNCT
fpq-14292	355	1	with	with	ADP
fpq-14292	355	2	speech	speech	NOUN
fpq-14292	355	3	recognition	recognition	NOUN
fpq-14292	355	4	,	,	PUNCT
fpq-14292	355	5	there	there	PRON
fpq-14292	355	6	is	be	VERB
fpq-14292	355	7	no	no	DET
fpq-14292	355	8	room	room	NOUN
fpq-14292	355	9	for	for	ADP
fpq-14292	355	10	imposing	impose	VERB
fpq-14292	355	11	normative	normative	ADJ
fpq-14292	355	12	correctness	correctness	NOUN
fpq-14292	355	13	,	,	PUNCT
fpq-14292	355	14	and	and	CCONJ
fpq-14292	355	15	doing	do	VERB
fpq-14292	355	16	so	so	ADV
fpq-14292	355	17	might	might	AUX
fpq-14292	355	18	even	even	ADV
fpq-14292	355	19	be	be	AUX
fpq-14292	355	20	ethically	ethically	ADV
fpq-14292	355	21	problematic	problematic	ADJ
fpq-14292	355	22	.	.	PUNCT
fpq-14292	356	1	imagine	imagine	VERB
fpq-14292	356	2	we	we	PRON
fpq-14292	356	3	are	be	AUX
fpq-14292	356	4	listening	listen	VERB
fpq-14292	356	5	to	to	AUX
fpq-14292	356	6	donald	donald	NOUN
fpq-14292	356	7	trump	trump	VERB
fpq-14292	356	8	making	make	VERB
fpq-14292	356	9	a	a	DET
fpq-14292	356	10	speech	speech	NOUN
fpq-14292	356	11	on	on	ADP
fpq-14292	356	12	tv	tv	NOUN
fpq-14292	356	13	and	and	CCONJ
fpq-14292	356	14	our	our	PRON
fpq-14292	356	15	speaker	speaker	NOUN
fpq-14292	356	16	is	be	AUX
fpq-14292	356	17	not	not	PART
fpq-14292	356	18	working	work	VERB
fpq-14292	356	19	.	.	PUNCT
fpq-14292	357	1	instead	instead	ADV
fpq-14292	357	2	,	,	PUNCT
fpq-14292	357	3	we	we	PRON
fpq-14292	357	4	are	be	AUX
fpq-14292	357	5	relying	rely	VERB
fpq-14292	357	6	on	on	ADP
fpq-14292	357	7	a	a	DET
fpq-14292	357	8	livecaptioning	livecaptioning	NOUN
fpq-14292	357	9	feed	feed	NOUN
fpq-14292	357	10	.	.	PUNCT
fpq-14292	358	1	we	we	PRON
fpq-14292	358	2	would	would	AUX
fpq-14292	358	3	be	be	AUX
fpq-14292	358	4	poorly	poorly	ADV
fpq-14292	358	5	served	serve	VERB
fpq-14292	358	6	by	by	ADP
fpq-14292	358	7	this	this	DET
fpq-14292	358	8	system	system	NOUN
fpq-14292	358	9	were	be	AUX
fpq-14292	358	10	it	it	PRON
fpq-14292	358	11	to	to	PART
fpq-14292	358	12	impose	impose	VERB
fpq-14292	358	13	normative	normative	ADJ
fpq-14292	358	14	correctness	correctness	NOUN
fpq-14292	358	15	by	by	ADP
fpq-14292	358	16	producing	produce	VERB
fpq-14292	358	17	an	an	DET
fpq-14292	358	18	eloquent	eloquent	ADJ
fpq-14292	358	19	discourse	discourse	NOUN
fpq-14292	358	20	on	on	ADP
fpq-14292	358	21	feminism	feminism	NOUN
fpq-14292	358	22	.	.	PUNCT
fpq-14292	359	1	we	we	PRON
fpq-14292	359	2	need	need	VERB
fpq-14292	359	3	to	to	PART
fpq-14292	359	4	know	know	VERB
fpq-14292	359	5	what	what	PRON
fpq-14292	359	6	trump	trump	NOUN
fpq-14292	359	7	is	be	AUX
fpq-14292	359	8	actually	actually	ADV
fpq-14292	359	9	saying	say	VERB
fpq-14292	359	10	—	—	PUNCT
fpq-14292	359	11	unpalatable	unpalatable	ADJ
fpq-14292	359	12	as	as	SCONJ
fpq-14292	359	13	it	it	PRON
fpq-14292	359	14	might	might	AUX
fpq-14292	359	15	be	be	AUX
fpq-14292	359	16	—	—	PUNCT
fpq-14292	359	17	in	in	ADP
fpq-14292	359	18	order	order	NOUN
fpq-14292	359	19	to	to	PART
fpq-14292	359	20	hold	hold	VERB
fpq-14292	359	21	him	he	PRON
fpq-14292	359	22	accountable	accountable	ADJ
fpq-14292	359	23	.	.	PUNCT
fpq-14292	360	1	at	at	ADP
fpq-14292	360	2	the	the	DET
fpq-14292	360	3	other	other	ADJ
fpq-14292	360	4	end	end	NOUN
fpq-14292	360	5	of	of	ADP
fpq-14292	360	6	the	the	DET
fpq-14292	360	7	spectrum	spectrum	NOUN
fpq-14292	360	8	is	be	AUX
fpq-14292	360	9	search	search	NOUN
fpq-14292	360	10	.	.	PUNCT
fpq-14292	361	1	if	if	SCONJ
fpq-14292	361	2	search	search	NOUN
fpq-14292	361	3	results	result	NOUN
fpq-14292	361	4	do	do	AUX
fpq-14292	361	5	not	not	PART
fpq-14292	361	6	reflect	reflect	VERB
fpq-14292	361	7	descriptive	descriptive	ADJ
fpq-14292	361	8	accuracy	accuracy	NOUN
fpq-14292	361	9	(	(	PUNCT
fpq-14292	361	10	see	see	VERB
fpq-14292	361	11	,	,	PUNCT
fpq-14292	361	12	e.g.	e.g.	ADV
fpq-14292	361	13	,	,	PUNCT
fpq-14292	361	14	noble	noble	ADJ
fpq-14292	361	15	2018	2018	NUM
fpq-14292	361	16	)	)	PUNCT
fpq-14292	361	17	,	,	PUNCT
fpq-14292	361	18	the	the	DET
fpq-14292	361	19	ethical	ethical	ADJ
fpq-14292	361	20	dilemma	dilemma	NOUN
fpq-14292	361	21	that	that	PRON
fpq-14292	361	22	we	we	PRON
fpq-14292	361	23	outline	outline	VERB
fpq-14292	361	24	does	do	AUX
fpq-14292	361	25	not	not	PART
fpq-14292	361	26	arise	arise	VERB
fpq-14292	361	27	.	.	PUNCT
fpq-14292	362	1	if	if	SCONJ
fpq-14292	362	2	search	search	NOUN
fpq-14292	362	3	results	result	NOUN
fpq-14292	362	4	were	be	AUX
fpq-14292	362	5	somehow	somehow	ADV
fpq-14292	362	6	to	to	PART
fpq-14292	362	7	reflect	reflect	VERB
fpq-14292	362	8	descriptive	descriptive	ADJ
fpq-14292	362	9	accuracy	accuracy	NOUN
fpq-14292	362	10	but	but	CCONJ
fpq-14292	362	11	were	be	AUX
fpq-14292	362	12	normatively	normatively	ADV
fpq-14292	362	13	incorrect	incorrect	ADJ
fpq-14292	362	14	,	,	PUNCT
fpq-14292	362	15	we	we	PRON
fpq-14292	362	16	might	might	AUX
fpq-14292	362	17	confront	confront	VERB
fpq-14292	362	18	the	the	DET
fpq-14292	362	19	dilemma	dilemma	NOUN
fpq-14292	362	20	.	.	PUNCT
fpq-14292	363	1	we	we	PRON
fpq-14292	363	2	would	would	AUX
fpq-14292	363	3	then	then	ADV
fpq-14292	363	4	be	be	AUX
fpq-14292	363	5	faced	face	VERB
fpq-14292	363	6	with	with	ADP
fpq-14292	363	7	two	two	NUM
fpq-14292	363	8	options	option	NOUN
fpq-14292	363	9	:	:	PUNCT
fpq-14292	363	10	either	either	CCONJ
fpq-14292	363	11	prioritize	prioritize	VERB
fpq-14292	363	12	descriptive	descriptive	ADJ
fpq-14292	363	13	accuracy	accuracy	NOUN
fpq-14292	363	14	over	over	ADP
fpq-14292	363	15	normative	normative	ADJ
fpq-14292	363	16	correctness	correctness	NOUN
fpq-14292	363	17	,	,	PUNCT
fpq-14292	363	18	which	which	PRON
fpq-14292	363	19	has	have	VERB
fpq-14292	363	20	the	the	DET
fpq-14292	363	21	potential	potential	ADJ
fpq-14292	363	22	cost	cost	NOUN
fpq-14292	363	23	of	of	ADP
fpq-14292	363	24	perpetuating	perpetuate	VERB
fpq-14292	363	25	or	or	CCONJ
fpq-14292	363	26	amplifying	amplifying	ADJ
fpq-14292	363	27	bias	bias	NOUN
fpq-14292	363	28	,	,	PUNCT
fpq-14292	363	29	or	or	CCONJ
fpq-14292	363	30	instead	instead	ADV
fpq-14292	363	31	prioritize	prioritize	VERB
fpq-14292	363	32	normative	normative	ADJ
fpq-14292	363	33	correctness	correctness	NOUN
fpq-14292	363	34	,	,	PUNCT
fpq-14292	363	35	with	with	ADP
fpq-14292	363	36	the	the	DET
fpq-14292	363	37	potential	potential	ADJ
fpq-14292	363	38	cost	cost	NOUN
fpq-14292	363	39	of	of	ADP
fpq-14292	363	40	withholding	withhold	VERB
fpq-14292	363	41	ethically	ethically	ADV
fpq-14292	363	42	useful	useful	ADJ
fpq-14292	363	43	information	information	NOUN
fpq-14292	363	44	even	even	ADV
fpq-14292	363	45	if	if	SCONJ
fpq-14292	363	46	we	we	PRON
fpq-14292	363	47	thereby	thereby	ADV
fpq-14292	363	48	avoid	avoid	VERB
fpq-14292	363	49	perpetuating	perpetuate	VERB
fpq-14292	363	50	or	or	CCONJ
fpq-14292	363	51	amplifying	amplifying	ADJ
fpq-14292	363	52	bias	bias	NOUN
fpq-14292	363	53	.	.	PUNCT
fpq-14292	364	1	recall	recall	VERB
fpq-14292	364	2	that	that	DET
fpq-14292	364	3	descriptive	descriptive	ADJ
fpq-14292	364	4	accuracy	accuracy	NOUN
fpq-14292	364	5	,	,	PUNCT
fpq-14292	364	6	as	as	SCONJ
fpq-14292	364	7	we	we	PRON
fpq-14292	364	8	use	use	VERB
fpq-14292	364	9	this	this	DET
fpq-14292	364	10	term	term	NOUN
fpq-14292	364	11	,	,	PUNCT
fpq-14292	364	12	means	mean	VERB
fpq-14292	364	13	sufficient	sufficient	ADJ
fpq-14292	364	14	descriptive	descriptive	ADJ
fpq-14292	364	15	accuracy	accuracy	NOUN
fpq-14292	364	16	in	in	ADP
fpq-14292	364	17	a	a	DET
fpq-14292	364	18	relevant	relevant	ADJ
fpq-14292	364	19	regard	regard	NOUN
fpq-14292	364	20	(	(	PUNCT
fpq-14292	364	21	all	all	DET
fpq-14292	364	22	models	model	NOUN
fpq-14292	364	23	are	be	AUX
fpq-14292	364	24	inaccurate	inaccurate	ADJ
fpq-14292	364	25	in	in	ADP
fpq-14292	364	26	some	some	PRON
fpq-14292	364	27	—	—	PUNCT
fpq-14292	364	28	and	and	CCONJ
fpq-14292	364	29	presumably	presumably	ADV
fpq-14292	364	30	many	many	ADJ
fpq-14292	364	31	—	—	PUNCT
fpq-14292	364	32	regards	regard	VERB
fpq-14292	364	33	,	,	PUNCT
fpq-14292	364	34	even	even	ADV
fpq-14292	364	35	with	with	ADP
fpq-14292	364	36	respect	respect	NOUN
fpq-14292	364	37	to	to	ADP
fpq-14292	364	38	a	a	DET
fpq-14292	364	39	target	target	NOUN
fpq-14292	364	40	phenomenon	phenomenon	NOUN
fpq-14292	364	41	)	)	PUNCT
fpq-14292	364	42	.	.	PUNCT
fpq-14292	365	1	without	without	ADP
fpq-14292	365	2	descriptive	descriptive	ADJ
fpq-14292	365	3	accuracy	accuracy	NOUN
fpq-14292	365	4	we	we	PRON
fpq-14292	365	5	can	can	AUX
fpq-14292	365	6	not	not	PART
fpq-14292	365	7	proceed	proceed	VERB
fpq-14292	365	8	usefully	usefully	ADV
fpq-14292	365	9	and	and	CCONJ
fpq-14292	365	10	there	there	PRON
fpq-14292	365	11	is	be	VERB
fpq-14292	365	12	no	no	DET
fpq-14292	365	13	dilemma	dilemma	NOUN
fpq-14292	365	14	.	.	PUNCT
fpq-14292	366	1	call	call	VERB
fpq-14292	366	2	a	a	DET
fpq-14292	366	3	situation	situation	NOUN
fpq-14292	366	4	dilemma	dilemma	NOUN
fpq-14292	366	5	-	-	PUNCT
fpq-14292	366	6	land	land	NOUN
fpq-14292	366	7	when	when	SCONJ
fpq-14292	366	8	we	we	PRON
fpq-14292	366	9	have	have	VERB
fpq-14292	366	10	a	a	DET
fpq-14292	366	11	descriptively	descriptively	ADV
fpq-14292	366	12	accurate	accurate	ADJ
fpq-14292	366	13	model	model	NOUN
fpq-14292	366	14	that	that	SCONJ
fpq-14292	366	15	,	,	PUNCT
fpq-14292	366	16	in	in	ADP
fpq-14292	366	17	virtue	virtue	NOUN
fpq-14292	366	18	of	of	ADP
fpq-14292	366	19	its	its	PRON
fpq-14292	366	20	descriptive	descriptive	ADJ
fpq-14292	366	21	accuracy	accuracy	NOUN
fpq-14292	366	22	,	,	PUNCT
fpq-14292	366	23	gives	give	VERB
fpq-14292	366	24	normatively	normatively	ADV
fpq-14292	366	25	incorrect	incorrect	ADJ
fpq-14292	366	26	(	(	PUNCT
fpq-14292	366	27	i.e.	i.e.	X
fpq-14292	366	28	,	,	PUNCT
fpq-14292	366	29	biased	biased	ADJ
fpq-14292	366	30	)	)	PUNCT
fpq-14292	366	31	outputs	output	NOUN
fpq-14292	366	32	.	.	PUNCT
fpq-14292	367	1	by	by	ADP
fpq-14292	367	2	contrast	contrast	NOUN
fpq-14292	367	3	,	,	PUNCT
fpq-14292	367	4	call	call	VERB
fpq-14292	367	5	a	a	DET
fpq-14292	367	6	situation	situation	NOUN
fpq-14292	367	7	utopia	utopia	NOUN
fpq-14292	367	8	-	-	PUNCT
fpq-14292	367	9	land	land	NOUN
fpq-14292	367	10	when	when	SCONJ
fpq-14292	367	11	we	we	PRON
fpq-14292	367	12	have	have	VERB
fpq-14292	367	13	a	a	DET
fpq-14292	367	14	descriptively	descriptively	ADV
fpq-14292	367	15	accurate	accurate	ADJ
fpq-14292	367	16	model	model	NOUN
fpq-14292	367	17	31	31	NUM
fpq-14292	367	18	alternatively	alternatively	ADV
fpq-14292	367	19	,	,	PUNCT
fpq-14292	367	20	we	we	PRON
fpq-14292	367	21	could	could	AUX
fpq-14292	367	22	strive	strive	VERB
fpq-14292	367	23	for	for	ADP
fpq-14292	367	24	descriptive	descriptive	ADJ
fpq-14292	367	25	accuracy	accuracy	NOUN
fpq-14292	367	26	,	,	PUNCT
fpq-14292	367	27	in	in	ADP
fpq-14292	367	28	which	which	DET
fpq-14292	367	29	case	case	NOUN
fpq-14292	367	30	our	our	PRON
fpq-14292	367	31	dilemma	dilemma	NOUN
fpq-14292	367	32	might	might	AUX
fpq-14292	367	33	apply	apply	VERB
fpq-14292	367	34	after	after	ADV
fpq-14292	367	35	all	all	ADV
fpq-14292	367	36	.	.	PUNCT
fpq-14292	368	1	feminist	feminist	ADJ
fpq-14292	368	2	philosophy	philosophy	PROPN
fpq-14292	368	3	quarterly	quarterly	ADV
fpq-14292	368	4	,	,	PUNCT
fpq-14292	368	5	2022	2022	NUM
fpq-14292	368	6	,	,	PUNCT
fpq-14292	368	7	vol	vol	NOUN
fpq-14292	368	8	.	.	NOUN
fpq-14292	368	9	8	8	NUM
fpq-14292	368	10	,	,	PUNCT
fpq-14292	368	11	iss	iss	PROPN
fpq-14292	368	12	.	.	PUNCT
fpq-14292	368	13	3/4	3/4	NUM
fpq-14292	368	14	,	,	PUNCT
fpq-14292	368	15	article	article	NOUN
fpq-14292	368	16	1	1	NUM
fpq-14292	368	17	published	publish	VERB
fpq-14292	368	18	by	by	ADP
fpq-14292	368	19	scholarship@western	scholarship@western	PROPN
fpq-14292	368	20	,	,	PUNCT
fpq-14292	368	21	2022	2022	NUM
fpq-14292	368	22	20	20	NUM
fpq-14292	368	23	that	that	PRON
fpq-14292	368	24	gives	give	VERB
fpq-14292	368	25	normatively	normatively	ADV
fpq-14292	368	26	correct	correct	ADJ
fpq-14292	368	27	outputs	output	NOUN
fpq-14292	368	28	.	.	PUNCT
fpq-14292	369	1	there	there	PRON
fpq-14292	369	2	is	be	VERB
fpq-14292	369	3	no	no	DET
fpq-14292	369	4	way	way	NOUN
fpq-14292	369	5	of	of	ADP
fpq-14292	369	6	getting	get	VERB
fpq-14292	369	7	to	to	ADP
fpq-14292	369	8	utopia	utopia	NOUN
fpq-14292	369	9	-	-	PUNCT
fpq-14292	369	10	land	land	NOUN
fpq-14292	369	11	until	until	SCONJ
fpq-14292	369	12	people	people	NOUN
fpq-14292	369	13	no	no	ADV
fpq-14292	369	14	longer	long	ADV
fpq-14292	369	15	express	express	VERB
fpq-14292	369	16	biases	bias	NOUN
fpq-14292	369	17	(	(	PUNCT
fpq-14292	369	18	e.g.	e.g.	ADV
fpq-14292	369	19	,	,	PUNCT
fpq-14292	369	20	sexism	sexism	NOUN
fpq-14292	369	21	or	or	CCONJ
fpq-14292	369	22	racism	racism	NOUN
fpq-14292	369	23	)	)	PUNCT
fpq-14292	369	24	in	in	ADP
fpq-14292	369	25	the	the	DET
fpq-14292	369	26	language	language	NOUN
fpq-14292	369	27	corpora	corpora	NOUN
fpq-14292	369	28	on	on	ADP
fpq-14292	369	29	which	which	PRON
fpq-14292	369	30	we	we	PRON
fpq-14292	369	31	train	train	VERB
fpq-14292	369	32	models	model	NOUN
fpq-14292	369	33	.	.	PUNCT
fpq-14292	370	1	what	what	PRON
fpq-14292	370	2	if	if	SCONJ
fpq-14292	370	3	we	we	PRON
fpq-14292	370	4	have	have	VERB
fpq-14292	370	5	descriptively	descriptively	ADV
fpq-14292	370	6	inaccurate	inaccurate	ADJ
fpq-14292	370	7	models	model	NOUN
fpq-14292	370	8	?	?	PUNCT
fpq-14292	371	1	perhaps	perhaps	ADV
fpq-14292	371	2	,	,	PUNCT
fpq-14292	371	3	in	in	ADP
fpq-14292	371	4	that	that	DET
fpq-14292	371	5	case	case	NOUN
fpq-14292	371	6	,	,	PUNCT
fpq-14292	371	7	we	we	PRON
fpq-14292	371	8	only	only	ADV
fpq-14292	371	9	need	need	VERB
fpq-14292	371	10	normative	normative	ADJ
fpq-14292	371	11	correctness	correctness	NOUN
fpq-14292	371	12	.	.	PUNCT
fpq-14292	372	1	however	however	ADV
fpq-14292	372	2	,	,	PUNCT
fpq-14292	372	3	an	an	DET
fpq-14292	372	4	inaccurate	inaccurate	ADJ
fpq-14292	372	5	model	model	NOUN
fpq-14292	372	6	may	may	AUX
fpq-14292	372	7	fail	fail	VERB
fpq-14292	372	8	to	to	PART
fpq-14292	372	9	sufficiently	sufficiently	ADV
fpq-14292	372	10	perform	perform	VERB
fpq-14292	372	11	the	the	DET
fpq-14292	372	12	task	task	NOUN
fpq-14292	372	13	it	it	PRON
fpq-14292	372	14	is	be	AUX
fpq-14292	372	15	meant	mean	VERB
fpq-14292	372	16	to	to	PART
fpq-14292	372	17	perform	perform	VERB
fpq-14292	372	18	,	,	PUNCT
fpq-14292	372	19	since	since	SCONJ
fpq-14292	372	20	inaccuracy	inaccuracy	NOUN
fpq-14292	372	21	often	often	ADV
fpq-14292	372	22	undermines	undermine	VERB
fpq-14292	372	23	usefulness	usefulness	NOUN
fpq-14292	372	24	.	.	PUNCT
fpq-14292	373	1	achieving	achieve	VERB
fpq-14292	373	2	normative	normative	ADJ
fpq-14292	373	3	correctness	correctness	NOUN
fpq-14292	373	4	without	without	ADP
fpq-14292	373	5	descriptive	descriptive	ADJ
fpq-14292	373	6	accuracy	accuracy	NOUN
fpq-14292	373	7	would	would	AUX
fpq-14292	373	8	,	,	PUNCT
fpq-14292	373	9	in	in	ADP
fpq-14292	373	10	any	any	DET
fpq-14292	373	11	case	case	NOUN
fpq-14292	373	12	,	,	PUNCT
fpq-14292	373	13	put	put	VERB
fpq-14292	373	14	us	we	PRON
fpq-14292	373	15	in	in	ADP
fpq-14292	373	16	fantasy	fantasy	NOUN
fpq-14292	373	17	-	-	PUNCT
fpq-14292	373	18	land	land	NOUN
fpq-14292	373	19	;	;	PUNCT
fpq-14292	373	20	if	if	SCONJ
fpq-14292	373	21	a	a	DET
fpq-14292	373	22	model	model	NOUN
fpq-14292	373	23	exhibits	exhibit	VERB
fpq-14292	373	24	normative	normative	ADJ
fpq-14292	373	25	correctness	correctness	NOUN
fpq-14292	373	26	but	but	CCONJ
fpq-14292	373	27	is	be	AUX
fpq-14292	373	28	descriptively	descriptively	ADV
fpq-14292	373	29	inaccurate	inaccurate	ADJ
fpq-14292	373	30	,	,	PUNCT
fpq-14292	373	31	it	it	PRON
fpq-14292	373	32	is	be	AUX
fpq-14292	373	33	out	out	ADP
fpq-14292	373	34	of	of	ADP
fpq-14292	373	35	touch	touch	NOUN
fpq-14292	373	36	with	with	ADP
fpq-14292	373	37	reality	reality	NOUN
fpq-14292	373	38	and	and	CCONJ
fpq-14292	373	39	is	be	AUX
fpq-14292	373	40	thus	thus	ADV
fpq-14292	373	41	a	a	DET
fpq-14292	373	42	fantasy	fantasy	NOUN
fpq-14292	373	43	(	(	PUNCT
fpq-14292	373	44	although	although	SCONJ
fpq-14292	373	45	sometimes	sometimes	ADV
fpq-14292	373	46	,	,	PUNCT
fpq-14292	373	47	as	as	SCONJ
fpq-14292	373	48	we	we	PRON
fpq-14292	373	49	acknowledge	acknowledge	VERB
fpq-14292	373	50	below	below	ADV
fpq-14292	373	51	,	,	PUNCT
fpq-14292	373	52	we	we	PRON
fpq-14292	373	53	may	may	AUX
fpq-14292	373	54	have	have	AUX
fpq-14292	373	55	reason	reason	NOUN
fpq-14292	373	56	to	to	PART
fpq-14292	373	57	prefer	prefer	VERB
fpq-14292	373	58	such	such	ADJ
fpq-14292	373	59	fantasies	fantasy	NOUN
fpq-14292	373	60	to	to	PART
fpq-14292	373	61	descriptive	descriptive	ADJ
fpq-14292	373	62	accuracy	accuracy	NOUN
fpq-14292	373	63	—	—	PUNCT
fpq-14292	373	64	for	for	ADP
fpq-14292	373	65	example	example	NOUN
fpq-14292	373	66	,	,	PUNCT
fpq-14292	373	67	normatively	normatively	ADV
fpq-14292	373	68	correct	correct	ADJ
fpq-14292	373	69	outputs	output	NOUN
fpq-14292	373	70	may	may	AUX
fpq-14292	373	71	serve	serve	VERB
fpq-14292	373	72	to	to	PART
fpq-14292	373	73	work	work	VERB
fpq-14292	373	74	against	against	ADP
fpq-14292	373	75	amplification	amplification	NOUN
fpq-14292	373	76	of	of	ADP
fpq-14292	373	77	bias	bias	NOUN
fpq-14292	373	78	in	in	ADP
fpq-14292	373	79	some	some	DET
fpq-14292	373	80	cases	case	NOUN
fpq-14292	373	81	,	,	PUNCT
fpq-14292	373	82	even	even	ADV
fpq-14292	373	83	if	if	SCONJ
fpq-14292	373	84	the	the	DET
fpq-14292	373	85	model	model	NOUN
fpq-14292	373	86	is	be	AUX
fpq-14292	373	87	descriptively	descriptively	ADV
fpq-14292	373	88	inaccurate	inaccurate	ADJ
fpq-14292	373	89	)	)	PUNCT
fpq-14292	373	90	.	.	PUNCT
fpq-14292	374	1	achieving	achieve	VERB
fpq-14292	374	2	normative	normative	ADJ
fpq-14292	374	3	incorrectness	incorrectness	NOUN
fpq-14292	374	4	and	and	CCONJ
fpq-14292	374	5	descriptive	descriptive	ADJ
fpq-14292	374	6	inaccuracy	inaccuracy	NOUN
fpq-14292	374	7	puts	put	VERB
fpq-14292	374	8	us	we	PRON
fpq-14292	374	9	instead	instead	ADV
fpq-14292	374	10	in	in	ADP
fpq-14292	374	11	disaster	disaster	NOUN
fpq-14292	374	12	-	-	PUNCT
fpq-14292	374	13	land	land	NOUN
fpq-14292	374	14	.	.	PUNCT
fpq-14292	375	1	although	although	SCONJ
fpq-14292	375	2	it	it	PRON
fpq-14292	375	3	may	may	AUX
fpq-14292	375	4	be	be	AUX
fpq-14292	375	5	possible	possible	ADJ
fpq-14292	375	6	to	to	PART
fpq-14292	375	7	get	get	VERB
fpq-14292	375	8	from	from	ADP
fpq-14292	375	9	fantasy	fantasy	NOUN
fpq-14292	375	10	-	-	PUNCT
fpq-14292	375	11	land	land	NOUN
fpq-14292	375	12	or	or	CCONJ
fpq-14292	375	13	disaster	disaster	NOUN
fpq-14292	375	14	-	-	PUNCT
fpq-14292	375	15	land	land	NOUN
fpq-14292	375	16	to	to	ADP
fpq-14292	375	17	utopia	utopia	NOUN
fpq-14292	375	18	-	-	PUNCT
fpq-14292	375	19	land	land	NOUN
fpq-14292	375	20	,	,	PUNCT
fpq-14292	375	21	it	it	PRON
fpq-14292	375	22	is	be	AUX
fpq-14292	375	23	highly	highly	ADV
fpq-14292	375	24	unlikely	unlikely	ADJ
fpq-14292	375	25	.	.	PUNCT
fpq-14292	376	1	(	(	PUNCT
fpq-14292	376	2	we	we	PRON
fpq-14292	376	3	leave	leave	VERB
fpq-14292	376	4	confirmation	confirmation	NOUN
fpq-14292	376	5	to	to	ADP
fpq-14292	376	6	our	our	PRON
fpq-14292	376	7	readers	reader	NOUN
fpq-14292	376	8	.	.	PUNCT
fpq-14292	376	9	)	)	PUNCT
fpq-14292	377	1	more	more	ADV
fpq-14292	377	2	likely	likely	ADV
fpq-14292	377	3	is	be	AUX
fpq-14292	377	4	that	that	SCONJ
fpq-14292	377	5	we	we	PRON
fpq-14292	377	6	simply	simply	ADV
fpq-14292	377	7	cycle	cycle	VERB
fpq-14292	377	8	back	back	ADV
fpq-14292	377	9	and	and	CCONJ
fpq-14292	377	10	forth	forth	ADV
fpq-14292	377	11	between	between	ADP
fpq-14292	377	12	fantasy	fantasy	NOUN
fpq-14292	377	13	-	-	PUNCT
fpq-14292	377	14	land	land	NOUN
fpq-14292	377	15	and	and	CCONJ
fpq-14292	377	16	disaster	disaster	NOUN
fpq-14292	377	17	-	-	PUNCT
fpq-14292	377	18	land	land	NOUN
fpq-14292	377	19	by	by	ADP
fpq-14292	377	20	achieving	achieve	VERB
fpq-14292	377	21	normative	normative	ADJ
fpq-14292	377	22	correctness	correctness	NOUN
fpq-14292	377	23	/	/	SYM
fpq-14292	377	24	incorrectness	incorrectness	NOUN
fpq-14292	377	25	in	in	ADP
fpq-14292	377	26	our	our	PRON
fpq-14292	377	27	model	model	NOUN
fpq-14292	377	28	yet	yet	ADV
fpq-14292	377	29	without	without	ADP
fpq-14292	377	30	descriptive	descriptive	ADJ
fpq-14292	377	31	accuracy	accuracy	NOUN
fpq-14292	377	32	.	.	PUNCT
fpq-14292	378	1	either	either	DET
fpq-14292	378	2	way	way	NOUN
fpq-14292	378	3	,	,	PUNCT
fpq-14292	378	4	until	until	SCONJ
fpq-14292	378	5	we	we	PRON
fpq-14292	378	6	have	have	VERB
fpq-14292	378	7	descriptive	descriptive	ADJ
fpq-14292	378	8	accuracy	accuracy	NOUN
fpq-14292	378	9	,	,	PUNCT
fpq-14292	378	10	we	we	PRON
fpq-14292	378	11	can	can	AUX
fpq-14292	378	12	not	not	PART
fpq-14292	378	13	get	get	VERB
fpq-14292	378	14	to	to	ADP
fpq-14292	378	15	either	either	CCONJ
fpq-14292	378	16	dilemma	dilemma	NOUN
fpq-14292	378	17	-	-	PUNCT
fpq-14292	378	18	land	land	NOUN
fpq-14292	378	19	or	or	CCONJ
fpq-14292	378	20	utopia	utopia	NOUN
fpq-14292	378	21	-	-	PUNCT
fpq-14292	378	22	land	land	NOUN
fpq-14292	378	23	;	;	PUNCT
fpq-14292	378	24	and	and	CCONJ
fpq-14292	378	25	(	(	PUNCT
fpq-14292	378	26	as	as	SCONJ
fpq-14292	378	27	indicated	indicate	VERB
fpq-14292	378	28	above	above	ADV
fpq-14292	378	29	)	)	PUNCT
fpq-14292	378	30	dilemma	dilemma	NOUN
fpq-14292	378	31	-	-	PUNCT
fpq-14292	378	32	land	land	NOUN
fpq-14292	378	33	remains	remain	VERB
fpq-14292	378	34	a	a	DET
fpq-14292	378	35	more	more	ADV
fpq-14292	378	36	likely	likely	ADJ
fpq-14292	378	37	destination	destination	NOUN
fpq-14292	378	38	than	than	ADP
fpq-14292	378	39	utopia	utopia	NOUN
fpq-14292	378	40	-	-	PUNCT
fpq-14292	378	41	land	land	NOUN
fpq-14292	378	42	anyhow	anyhow	ADV
fpq-14292	378	43	.	.	PUNCT
fpq-14292	379	1	only	only	ADV
fpq-14292	379	2	in	in	ADP
fpq-14292	379	3	dilemma	dilemma	NOUN
fpq-14292	379	4	-	-	PUNCT
fpq-14292	379	5	land	land	NOUN
fpq-14292	379	6	does	do	AUX
fpq-14292	379	7	the	the	DET
fpq-14292	379	8	dilemma	dilemma	NOUN
fpq-14292	379	9	arise	arise	VERB
fpq-14292	379	10	,	,	PUNCT
fpq-14292	379	11	since	since	SCONJ
fpq-14292	379	12	for	for	SCONJ
fpq-14292	379	13	it	it	PRON
fpq-14292	379	14	to	to	PART
fpq-14292	379	15	arise	arise	VERB
fpq-14292	379	16	a	a	DET
fpq-14292	379	17	model	model	NOUN
fpq-14292	379	18	must	must	AUX
fpq-14292	379	19	exhibit	exhibit	VERB
fpq-14292	379	20	descriptive	descriptive	ADJ
fpq-14292	379	21	accuracy	accuracy	NOUN
fpq-14292	379	22	and	and	CCONJ
fpq-14292	379	23	,	,	PUNCT
fpq-14292	379	24	in	in	ADP
fpq-14292	379	25	virtue	virtue	NOUN
fpq-14292	379	26	of	of	ADP
fpq-14292	379	27	this	this	DET
fpq-14292	379	28	descriptive	descriptive	ADJ
fpq-14292	379	29	accuracy	accuracy	NOUN
fpq-14292	379	30	,	,	PUNCT
fpq-14292	379	31	produce	produce	VERB
fpq-14292	379	32	normatively	normatively	ADV
fpq-14292	379	33	incorrect	incorrect	ADJ
fpq-14292	379	34	outputs	output	NOUN
fpq-14292	379	35	(	(	PUNCT
fpq-14292	379	36	if	if	SCONJ
fpq-14292	379	37	it	it	PRON
fpq-14292	379	38	gives	give	VERB
fpq-14292	379	39	normatively	normatively	ADV
fpq-14292	379	40	incorrect	incorrect	ADJ
fpq-14292	379	41	outputs	output	NOUN
fpq-14292	379	42	resulting	result	VERB
fpq-14292	379	43	from	from	ADP
fpq-14292	379	44	something	something	PRON
fpq-14292	379	45	else	else	ADV
fpq-14292	379	46	,	,	PUNCT
fpq-14292	379	47	that	that	PRON
fpq-14292	379	48	is	be	AUX
fpq-14292	379	49	a	a	DET
fpq-14292	379	50	different	different	ADJ
fpq-14292	379	51	matter	matter	NOUN
fpq-14292	379	52	that	that	PRON
fpq-14292	379	53	must	must	AUX
fpq-14292	379	54	be	be	AUX
fpq-14292	379	55	addressed	address	VERB
fpq-14292	379	56	on	on	ADP
fpq-14292	379	57	its	its	PRON
fpq-14292	379	58	own	own	ADJ
fpq-14292	379	59	terms	term	NOUN
fpq-14292	379	60	)	)	PUNCT
fpq-14292	379	61	.	.	PUNCT
fpq-14292	380	1	in	in	ADP
fpq-14292	380	2	dilemma	dilemma	NOUN
fpq-14292	380	3	-	-	PUNCT
fpq-14292	380	4	land	land	NOUN
fpq-14292	380	5	,	,	PUNCT
fpq-14292	380	6	we	we	PRON
fpq-14292	380	7	must	must	AUX
fpq-14292	380	8	decide	decide	VERB
fpq-14292	380	9	whether	whether	SCONJ
fpq-14292	380	10	to	to	PART
fpq-14292	380	11	prioritize	prioritize	VERB
fpq-14292	380	12	descriptive	descriptive	ADJ
fpq-14292	380	13	accuracy	accuracy	NOUN
fpq-14292	380	14	or	or	CCONJ
fpq-14292	380	15	instead	instead	ADV
fpq-14292	380	16	aim	aim	VERB
fpq-14292	380	17	for	for	ADP
fpq-14292	380	18	normative	normative	ADJ
fpq-14292	380	19	correctness	correctness	NOUN
fpq-14292	380	20	by	by	ADP
fpq-14292	380	21	debiasing	debiase	VERB
fpq-14292	380	22	.	.	PUNCT
fpq-14292	381	1	debiasing	debiase	VERB
fpq-14292	381	2	may	may	AUX
fpq-14292	381	3	undermine	undermine	VERB
fpq-14292	381	4	the	the	DET
fpq-14292	381	5	usefulness	usefulness	NOUN
fpq-14292	381	6	of	of	ADP
fpq-14292	381	7	the	the	DET
fpq-14292	381	8	model	model	NOUN
fpq-14292	381	9	and	and	CCONJ
fpq-14292	381	10	might	might	AUX
fpq-14292	381	11	undermine	undermine	VERB
fpq-14292	381	12	descriptive	descriptive	ADJ
fpq-14292	381	13	accuracy	accuracy	NOUN
fpq-14292	381	14	and	and	CCONJ
fpq-14292	381	15	thus	thus	ADV
fpq-14292	381	16	withhold	withhold	VERB
fpq-14292	381	17	ethically	ethically	ADV
fpq-14292	381	18	useful	useful	ADJ
fpq-14292	381	19	information	information	NOUN
fpq-14292	381	20	.	.	PUNCT
fpq-14292	382	1	yet	yet	ADV
fpq-14292	382	2	prioritizing	prioritize	VERB
fpq-14292	382	3	descriptive	descriptive	ADJ
fpq-14292	382	4	accuracy	accuracy	NOUN
fpq-14292	382	5	runs	run	VERB
fpq-14292	382	6	the	the	DET
fpq-14292	382	7	risk	risk	NOUN
fpq-14292	382	8	of	of	ADP
fpq-14292	382	9	perpetuating	perpetuate	VERB
fpq-14292	382	10	or	or	CCONJ
fpq-14292	382	11	amplifying	amplify	VERB
fpq-14292	382	12	the	the	DET
fpq-14292	382	13	relevant	relevant	ADJ
fpq-14292	382	14	biases	bias	NOUN
fpq-14292	382	15	.	.	PUNCT
fpq-14292	383	1	if	if	SCONJ
fpq-14292	383	2	a	a	DET
fpq-14292	383	3	model	model	NOUN
fpq-14292	383	4	exhibits	exhibit	VERB
fpq-14292	383	5	descriptive	descriptive	ADJ
fpq-14292	383	6	inaccuracy	inaccuracy	NOUN
fpq-14292	383	7	,	,	PUNCT
fpq-14292	383	8	there	there	PRON
fpq-14292	383	9	is	be	VERB
fpq-14292	383	10	simply	simply	ADV
fpq-14292	383	11	no	no	DET
fpq-14292	383	12	dilemma	dilemma	NOUN
fpq-14292	383	13	to	to	PART
fpq-14292	383	14	confront	confront	VERB
fpq-14292	383	15	given	give	VERB
fpq-14292	383	16	that	that	SCONJ
fpq-14292	383	17	we	we	PRON
fpq-14292	383	18	would	would	AUX
fpq-14292	383	19	never	never	ADV
fpq-14292	383	20	want	want	VERB
fpq-14292	383	21	to	to	PART
fpq-14292	383	22	prioritize	prioritize	VERB
fpq-14292	383	23	descriptive	descriptive	ADJ
fpq-14292	383	24	inaccuracy	inaccuracy	NOUN
fpq-14292	383	25	over	over	ADP
fpq-14292	383	26	either	either	CCONJ
fpq-14292	383	27	normative	normative	ADJ
fpq-14292	383	28	correctness	correctness	NOUN
fpq-14292	383	29	or	or	CCONJ
fpq-14292	383	30	incorrectness	incorrectness	NOUN
fpq-14292	383	31	.	.	PUNCT
fpq-14292	384	1	our	our	PRON
fpq-14292	384	2	aim	aim	NOUN
fpq-14292	384	3	is	be	AUX
fpq-14292	384	4	to	to	PART
fpq-14292	384	5	draw	draw	VERB
fpq-14292	384	6	attention	attention	NOUN
fpq-14292	384	7	to	to	ADP
fpq-14292	384	8	this	this	DET
fpq-14292	384	9	dilemma	dilemma	NOUN
fpq-14292	384	10	.	.	PUNCT
fpq-14292	385	1	language	language	NOUN
fpq-14292	385	2	models	model	NOUN
fpq-14292	385	3	and	and	CCONJ
fpq-14292	385	4	various	various	ADJ
fpq-14292	385	5	applications	application	NOUN
fpq-14292	385	6	(	(	PUNCT
fpq-14292	385	7	even	even	ADV
fpq-14292	385	8	potentially	potentially	ADV
fpq-14292	385	9	including	include	VERB
fpq-14292	385	10	search	search	NOUN
fpq-14292	385	11	)	)	PUNCT
fpq-14292	385	12	that	that	PRON
fpq-14292	385	13	rely	rely	VERB
fpq-14292	385	14	on	on	ADP
fpq-14292	385	15	them	they	PRON
fpq-14292	385	16	can	can	AUX
fpq-14292	385	17	be	be	AUX
fpq-14292	385	18	used	use	VERB
fpq-14292	385	19	for	for	ADP
fpq-14292	385	20	different	different	ADJ
fpq-14292	385	21	purposes	purpose	NOUN
fpq-14292	385	22	.	.	PUNCT
fpq-14292	386	1	sometimes	sometimes	ADV
fpq-14292	386	2	these	these	DET
fpq-14292	386	3	purposes	purpose	NOUN
fpq-14292	386	4	are	be	AUX
fpq-14292	386	5	best	well	ADV
fpq-14292	386	6	served	serve	VERB
fpq-14292	386	7	by	by	ADP
fpq-14292	386	8	having	have	VERB
fpq-14292	386	9	the	the	DET
fpq-14292	386	10	model	model	NOUN
fpq-14292	386	11	reflect	reflect	VERB
fpq-14292	386	12	descriptive	descriptive	ADJ
fpq-14292	386	13	accuracy	accuracy	NOUN
fpq-14292	386	14	even	even	ADV
fpq-14292	386	15	when	when	SCONJ
fpq-14292	386	16	that	that	PRON
fpq-14292	386	17	entails	entail	VERB
fpq-14292	386	18	reflecting	reflect	VERB
fpq-14292	386	19	bias	bias	NOUN
fpq-14292	386	20	.	.	PUNCT
fpq-14292	387	1	those	those	DET
fpq-14292	387	2	purposes	purpose	NOUN
fpq-14292	387	3	will	will	AUX
fpq-14292	387	4	,	,	PUNCT
fpq-14292	387	5	moreover	moreover	ADV
fpq-14292	387	6	,	,	PUNCT
fpq-14292	387	7	often	often	ADV
fpq-14292	387	8	be	be	AUX
fpq-14292	387	9	served	serve	VERB
fpq-14292	387	10	precisely	precisely	ADV
fpq-14292	387	11	because	because	SCONJ
fpq-14292	387	12	the	the	DET
fpq-14292	387	13	biased	biased	ADJ
fpq-14292	387	14	results	result	NOUN
fpq-14292	387	15	provide	provide	VERB
fpq-14292	387	16	us	we	PRON
fpq-14292	387	17	with	with	ADP
fpq-14292	387	18	ethically	ethically	ADV
fpq-14292	387	19	useful	useful	ADJ
fpq-14292	387	20	information	information	NOUN
fpq-14292	387	21	.	.	PUNCT
fpq-14292	388	1	however	however	ADV
fpq-14292	388	2	,	,	PUNCT
fpq-14292	388	3	we	we	PRON
fpq-14292	388	4	do	do	AUX
fpq-14292	388	5	run	run	VERB
fpq-14292	388	6	the	the	DET
fpq-14292	388	7	risk	risk	NOUN
fpq-14292	388	8	of	of	ADP
fpq-14292	388	9	amplifying	amplifying	ADJ
fpq-14292	388	10	bias	bias	NOUN
fpq-14292	388	11	.	.	PUNCT
fpq-14292	389	1	in	in	ADP
fpq-14292	389	2	other	other	ADJ
fpq-14292	389	3	cases	case	NOUN
fpq-14292	389	4	,	,	PUNCT
fpq-14292	389	5	an	an	DET
fpq-14292	389	6	application	application	NOUN
fpq-14292	389	7	might	might	AUX
fpq-14292	389	8	be	be	AUX
fpq-14292	389	9	used	use	VERB
fpq-14292	389	10	for	for	ADP
fpq-14292	389	11	a	a	DET
fpq-14292	389	12	different	different	ADJ
fpq-14292	389	13	purpose	purpose	NOUN
fpq-14292	389	14	,	,	PUNCT
fpq-14292	389	15	such	such	ADJ
fpq-14292	389	16	that	that	SCONJ
fpq-14292	389	17	it	it	PRON
fpq-14292	389	18	is	be	AUX
fpq-14292	389	19	more	more	ADV
fpq-14292	389	20	deery	deery	NOUN
fpq-14292	389	21	and	and	CCONJ
fpq-14292	389	22	bailey	bailey	PROPN
fpq-14292	389	23	–	–	PUNCT
fpq-14292	389	24	the	the	DET
fpq-14292	389	25	bias	bias	NOUN
fpq-14292	389	26	dilemma	dilemma	NOUN
fpq-14292	389	27	:	:	PUNCT
fpq-14292	389	28	the	the	DET
fpq-14292	389	29	ethics	ethic	NOUN
fpq-14292	389	30	of	of	ADP
fpq-14292	389	31	algorithmic	algorithmic	ADJ
fpq-14292	389	32	bias	bias	NOUN
fpq-14292	389	33	in	in	ADP
fpq-14292	389	34	natural	natural	ADJ
fpq-14292	389	35	-	-	PUNCT
fpq-14292	389	36	language	language	NOUN
fpq-14292	389	37	processing	processing	NOUN
fpq-14292	389	38	published	publish	VERB
fpq-14292	389	39	by	by	ADP
fpq-14292	389	40	scholarship@western	scholarship@western	PROPN
fpq-14292	389	41	,	,	PUNCT
fpq-14292	389	42	2022	2022	NUM
fpq-14292	389	43	21	21	NUM
fpq-14292	389	44	important	important	ADJ
fpq-14292	389	45	to	to	PART
fpq-14292	389	46	avoid	avoid	VERB
fpq-14292	389	47	amplifying	amplifying	ADJ
fpq-14292	389	48	bias	bias	NOUN
fpq-14292	389	49	than	than	SCONJ
fpq-14292	389	50	to	to	PART
fpq-14292	389	51	retain	retain	VERB
fpq-14292	389	52	ethically	ethically	ADV
fpq-14292	389	53	useful	useful	ADJ
fpq-14292	389	54	information	information	NOUN
fpq-14292	389	55	.	.	PUNCT
fpq-14292	390	1	then	then	ADV
fpq-14292	390	2	,	,	PUNCT
fpq-14292	390	3	we	we	PRON
fpq-14292	390	4	may	may	AUX
fpq-14292	390	5	prefer	prefer	VERB
fpq-14292	390	6	normatively	normatively	ADV
fpq-14292	390	7	,	,	PUNCT
fpq-14292	390	8	not	not	PART
fpq-14292	390	9	descriptively	descriptively	ADV
fpq-14292	390	10	,	,	PUNCT
fpq-14292	390	11	correct	correct	ADJ
fpq-14292	390	12	results	result	NOUN
fpq-14292	390	13	.	.	PUNCT
fpq-14292	391	1	either	either	DET
fpq-14292	391	2	way	way	NOUN
fpq-14292	391	3	,	,	PUNCT
fpq-14292	391	4	there	there	PRON
fpq-14292	391	5	is	be	VERB
fpq-14292	391	6	a	a	DET
fpq-14292	391	7	cost	cost	NOUN
fpq-14292	391	8	.	.	PUNCT
fpq-14292	392	1	we	we	PRON
fpq-14292	392	2	also	also	ADV
fpq-14292	392	3	maintain	maintain	VERB
fpq-14292	392	4	that	that	SCONJ
fpq-14292	392	5	there	there	PRON
fpq-14292	392	6	is	be	VERB
fpq-14292	392	7	sometimes	sometimes	ADV
fpq-14292	392	8	an	an	DET
fpq-14292	392	9	ethical	ethical	ADJ
fpq-14292	392	10	value	value	NOUN
fpq-14292	392	11	to	to	ADP
fpq-14292	392	12	the	the	DET
fpq-14292	392	13	less	less	ADV
fpq-14292	392	14	intuitive	intuitive	ADJ
fpq-14292	392	15	horn	horn	NOUN
fpq-14292	392	16	of	of	ADP
fpq-14292	392	17	the	the	DET
fpq-14292	392	18	dilemma	dilemma	NOUN
fpq-14292	392	19	,	,	PUNCT
fpq-14292	392	20	on	on	ADP
fpq-14292	392	21	which	which	PRON
fpq-14292	392	22	we	we	PRON
fpq-14292	392	23	retain	retain	VERB
fpq-14292	392	24	ethically	ethically	ADV
fpq-14292	392	25	useful	useful	ADJ
fpq-14292	392	26	information	information	NOUN
fpq-14292	392	27	by	by	ADP
fpq-14292	392	28	not	not	PART
fpq-14292	392	29	imposing	impose	VERB
fpq-14292	392	30	normative	normative	ADJ
fpq-14292	392	31	correctness	correctness	NOUN
fpq-14292	392	32	.	.	PUNCT
fpq-14292	393	1	even	even	ADV
fpq-14292	393	2	so	so	ADV
fpq-14292	393	3	,	,	PUNCT
fpq-14292	393	4	we	we	PRON
fpq-14292	393	5	recognize	recognize	VERB
fpq-14292	393	6	that	that	SCONJ
fpq-14292	393	7	in	in	ADP
fpq-14292	393	8	doing	do	VERB
fpq-14292	393	9	so	so	ADV
fpq-14292	393	10	we	we	PRON
fpq-14292	393	11	can	can	AUX
fpq-14292	393	12	amplify	amplify	VERB
fpq-14292	393	13	bias	bias	NOUN
fpq-14292	393	14	.	.	PUNCT
fpq-14292	394	1	8	8	X
fpq-14292	394	2	.	.	X
fpq-14292	394	3	rethinking	rethink	VERB
fpq-14292	394	4	the	the	DET
fpq-14292	394	5	debate	debate	NOUN
fpq-14292	394	6	about	about	ADP
fpq-14292	394	7	bias	bias	NOUN
fpq-14292	394	8	in	in	ADP
fpq-14292	394	9	nlp	nlp	NOUN
fpq-14292	394	10	our	our	PRON
fpq-14292	394	11	way	way	NOUN
fpq-14292	394	12	of	of	ADP
fpq-14292	394	13	distinguishing	distinguish	VERB
fpq-14292	394	14	normative	normative	ADJ
fpq-14292	394	15	correctness	correctness	NOUN
fpq-14292	394	16	and	and	CCONJ
fpq-14292	394	17	descriptive	descriptive	ADJ
fpq-14292	394	18	accuracy	accuracy	NOUN
fpq-14292	394	19	adds	add	VERB
fpq-14292	394	20	nuance	nuance	NOUN
fpq-14292	394	21	to	to	ADP
fpq-14292	394	22	recent	recent	ADJ
fpq-14292	394	23	debates	debate	NOUN
fpq-14292	394	24	about	about	ADP
fpq-14292	394	25	algorithmic	algorithmic	ADJ
fpq-14292	394	26	bias	bias	NOUN
fpq-14292	394	27	in	in	ADP
fpq-14292	394	28	nlp	nlp	NOUN
fpq-14292	394	29	and	and	CCONJ
fpq-14292	394	30	can	can	AUX
fpq-14292	394	31	help	help	VERB
fpq-14292	394	32	us	we	PRON
fpq-14292	394	33	to	to	PART
fpq-14292	394	34	understand	understand	VERB
fpq-14292	394	35	these	these	DET
fpq-14292	394	36	debates	debate	NOUN
fpq-14292	394	37	better	well	ADV
fpq-14292	394	38	.	.	PUNCT
fpq-14292	395	1	for	for	ADP
fpq-14292	395	2	instance	instance	NOUN
fpq-14292	395	3	,	,	PUNCT
fpq-14292	395	4	as	as	SCONJ
fpq-14292	395	5	outlined	outline	VERB
fpq-14292	395	6	above	above	ADV
fpq-14292	395	7	,	,	PUNCT
fpq-14292	395	8	suresh	suresh	NOUN
fpq-14292	395	9	and	and	CCONJ
fpq-14292	395	10	guttag	guttag	NOUN
fpq-14292	395	11	’s	’s	PART
fpq-14292	395	12	(	(	PUNCT
fpq-14292	395	13	2020	2020	NUM
fpq-14292	395	14	)	)	PUNCT
fpq-14292	395	15	historical	historical	ADJ
fpq-14292	395	16	bias	bias	NOUN
fpq-14292	395	17	can	can	AUX
fpq-14292	395	18	be	be	AUX
fpq-14292	395	19	understood	understand	VERB
fpq-14292	395	20	as	as	ADP
fpq-14292	395	21	arising	arise	VERB
fpq-14292	395	22	when	when	SCONJ
fpq-14292	395	23	we	we	PRON
fpq-14292	395	24	have	have	VERB
fpq-14292	395	25	descriptively	descriptively	ADV
fpq-14292	395	26	accurate	accurate	ADJ
fpq-14292	395	27	outputs	output	NOUN
fpq-14292	395	28	(	(	PUNCT
fpq-14292	395	29	regarding	regard	VERB
fpq-14292	395	30	people	people	NOUN
fpq-14292	395	31	’s	’s	PART
fpq-14292	395	32	use	use	NOUN
fpq-14292	395	33	of	of	ADP
fpq-14292	395	34	language	language	NOUN
fpq-14292	395	35	)	)	PUNCT
fpq-14292	395	36	about	about	ADP
fpq-14292	395	37	how	how	SCONJ
fpq-14292	395	38	the	the	DET
fpq-14292	395	39	world	world	NOUN
fpq-14292	395	40	is	be	AUX
fpq-14292	395	41	yet	yet	ADV
fpq-14292	395	42	these	these	DET
fpq-14292	395	43	outputs	output	NOUN
fpq-14292	395	44	are	be	AUX
fpq-14292	395	45	normatively	normatively	ADV
fpq-14292	395	46	incorrect	incorrect	ADJ
fpq-14292	395	47	.	.	PUNCT
fpq-14292	396	1	thus	thus	ADV
fpq-14292	396	2	,	,	PUNCT
fpq-14292	396	3	our	our	PRON
fpq-14292	396	4	distinction	distinction	NOUN
fpq-14292	396	5	can	can	AUX
fpq-14292	396	6	help	help	VERB
fpq-14292	396	7	to	to	PART
fpq-14292	396	8	clarify	clarify	VERB
fpq-14292	396	9	the	the	DET
fpq-14292	396	10	points	point	NOUN
fpq-14292	396	11	of	of	ADP
fpq-14292	396	12	contention	contention	NOUN
fpq-14292	396	13	among	among	ADP
fpq-14292	396	14	these	these	DET
fpq-14292	396	15	debates	debate	NOUN
fpq-14292	396	16	’	'	PUNCT
fpq-14292	396	17	participants	participant	NOUN
fpq-14292	396	18	and	and	CCONJ
fpq-14292	396	19	the	the	DET
fpq-14292	396	20	implications	implication	NOUN
fpq-14292	396	21	of	of	ADP
fpq-14292	396	22	prioritizing	prioritize	VERB
fpq-14292	396	23	one	one	NUM
fpq-14292	396	24	form	form	NOUN
fpq-14292	396	25	of	of	ADP
fpq-14292	396	26	accuracy	accuracy	NOUN
fpq-14292	396	27	over	over	ADP
fpq-14292	396	28	another	another	PRON
fpq-14292	396	29	.	.	PUNCT
fpq-14292	397	1	once	once	SCONJ
fpq-14292	397	2	we	we	PRON
fpq-14292	397	3	acknowledge	acknowledge	VERB
fpq-14292	397	4	the	the	DET
fpq-14292	397	5	ethical	ethical	ADJ
fpq-14292	397	6	dilemma	dilemma	NOUN
fpq-14292	397	7	that	that	PRON
fpq-14292	397	8	arises	arise	VERB
fpq-14292	397	9	when	when	SCONJ
fpq-14292	397	10	descriptive	descriptive	ADJ
fpq-14292	397	11	and	and	CCONJ
fpq-14292	397	12	normative	normative	ADJ
fpq-14292	397	13	accuracy	accuracy	NOUN
fpq-14292	397	14	conflict	conflict	NOUN
fpq-14292	397	15	in	in	ADP
fpq-14292	397	16	the	the	DET
fpq-14292	397	17	way	way	NOUN
fpq-14292	397	18	we	we	PRON
fpq-14292	397	19	have	have	AUX
fpq-14292	397	20	outlined	outline	VERB
fpq-14292	397	21	,	,	PUNCT
fpq-14292	397	22	we	we	PRON
fpq-14292	397	23	can	can	AUX
fpq-14292	397	24	see	see	VERB
fpq-14292	397	25	clearly	clearly	ADV
fpq-14292	397	26	the	the	DET
fpq-14292	397	27	ethical	ethical	ADJ
fpq-14292	397	28	costs	cost	NOUN
fpq-14292	397	29	of	of	ADP
fpq-14292	397	30	prioritizing	prioritize	VERB
fpq-14292	397	31	one	one	NUM
fpq-14292	397	32	form	form	NOUN
fpq-14292	397	33	of	of	ADP
fpq-14292	397	34	accuracy	accuracy	NOUN
fpq-14292	397	35	over	over	ADP
fpq-14292	397	36	the	the	DET
fpq-14292	397	37	other	other	ADJ
fpq-14292	397	38	.	.	PUNCT
fpq-14292	398	1	to	to	PART
fpq-14292	398	2	take	take	VERB
fpq-14292	398	3	another	another	DET
fpq-14292	398	4	example	example	NOUN
fpq-14292	398	5	,	,	PUNCT
fpq-14292	398	6	emily	emily	PROPN
fpq-14292	398	7	bender	bender	PROPN
fpq-14292	398	8	,	,	PUNCT
fpq-14292	398	9	timnit	timnit	NOUN
fpq-14292	398	10	gebru	gebru	NOUN
fpq-14292	398	11	,	,	PUNCT
fpq-14292	398	12	and	and	CCONJ
fpq-14292	398	13	colleagues	colleague	NOUN
fpq-14292	398	14	argue	argue	VERB
fpq-14292	398	15	that	that	SCONJ
fpq-14292	398	16	large	large	ADJ
fpq-14292	398	17	language	language	NOUN
fpq-14292	398	18	models	model	NOUN
fpq-14292	398	19	like	like	ADP
fpq-14292	398	20	bert	bert	NOUN
fpq-14292	398	21	or	or	CCONJ
fpq-14292	398	22	gpt-2/3	gpt-2/3	PROPN
fpq-14292	398	23	not	not	PART
fpq-14292	398	24	only	only	ADV
fpq-14292	398	25	reproduce	reproduce	VERB
fpq-14292	398	26	bias	bias	NOUN
fpq-14292	398	27	but	but	CCONJ
fpq-14292	398	28	do	do	AUX
fpq-14292	398	29	not	not	PART
fpq-14292	398	30	reflect	reflect	VERB
fpq-14292	398	31	“	"	PUNCT
fpq-14292	398	32	the	the	DET
fpq-14292	398	33	world	world	NOUN
fpq-14292	398	34	as	as	SCONJ
fpq-14292	398	35	it	it	PRON
fpq-14292	398	36	is	be	AUX
fpq-14292	398	37	”	"	PUNCT
fpq-14292	398	38	(	(	PUNCT
fpq-14292	398	39	suresh	suresh	NOUN
fpq-14292	398	40	and	and	CCONJ
fpq-14292	398	41	guttag	guttag	NOUN
fpq-14292	398	42	2020	2020	NUM
fpq-14292	398	43	,	,	PUNCT
fpq-14292	398	44	4	4	NUM
fpq-14292	398	45	)	)	PUNCT
fpq-14292	398	46	,	,	PUNCT
fpq-14292	398	47	since	since	SCONJ
fpq-14292	398	48	the	the	DET
fpq-14292	398	49	corpora	corpora	NOUN
fpq-14292	398	50	on	on	ADP
fpq-14292	398	51	which	which	PRON
fpq-14292	398	52	they	they	PRON
fpq-14292	398	53	are	be	AUX
fpq-14292	398	54	trained	train	VERB
fpq-14292	398	55	do	do	AUX
fpq-14292	398	56	not	not	PART
fpq-14292	398	57	reflect	reflect	VERB
fpq-14292	398	58	marginalized	marginalized	ADJ
fpq-14292	398	59	voices	voice	NOUN
fpq-14292	398	60	(	(	PUNCT
fpq-14292	398	61	bender	bender	NOUN
fpq-14292	398	62	,	,	PUNCT
fpq-14292	398	63	gebru	gebru	VERB
fpq-14292	398	64	et	et	PROPN
fpq-14292	398	65	al	al	PROPN
fpq-14292	398	66	.	.	PUNCT
fpq-14292	399	1	2021).32	2021).32	NUM
fpq-14292	399	2	thus	thus	ADV
fpq-14292	399	3	,	,	PUNCT
fpq-14292	399	4	one	one	PRON
fpq-14292	399	5	can	can	AUX
fpq-14292	399	6	not	not	PART
fpq-14292	399	7	argue	argue	VERB
fpq-14292	399	8	on	on	ADP
fpq-14292	399	9	the	the	DET
fpq-14292	399	10	basis	basis	NOUN
fpq-14292	399	11	of	of	ADP
fpq-14292	399	12	a	a	DET
fpq-14292	399	13	model	model	NOUN
fpq-14292	399	14	’s	’s	PART
fpq-14292	399	15	descriptive	descriptive	ADJ
fpq-14292	399	16	accuracy	accuracy	NOUN
fpq-14292	399	17	for	for	ADP
fpq-14292	399	18	not	not	PART
fpq-14292	399	19	debiasing	debiase	VERB
fpq-14292	399	20	it	it	PRON
fpq-14292	399	21	if	if	SCONJ
fpq-14292	399	22	it	it	PRON
fpq-14292	399	23	is	be	AUX
fpq-14292	399	24	not	not	PART
fpq-14292	399	25	descriptively	descriptively	ADV
fpq-14292	399	26	accurate	accurate	ADJ
fpq-14292	399	27	.	.	PUNCT
fpq-14292	400	1	part	part	NOUN
fpq-14292	400	2	of	of	ADP
fpq-14292	400	3	the	the	DET
fpq-14292	400	4	problem	problem	NOUN
fpq-14292	400	5	is	be	AUX
fpq-14292	400	6	that	that	SCONJ
fpq-14292	400	7	while	while	SCONJ
fpq-14292	400	8	“	"	PUNCT
fpq-14292	400	9	user	user	NOUN
fpq-14292	400	10	-	-	PUNCT
fpq-14292	400	11	generated	generate	VERB
fpq-14292	400	12	content	content	NOUN
fpq-14292	400	13	sites	site	NOUN
fpq-14292	400	14	like	like	ADP
fpq-14292	400	15	reddit	reddit	NOUN
fpq-14292	400	16	,	,	PUNCT
fpq-14292	400	17	twitter	twitter	NOUN
fpq-14292	400	18	,	,	PUNCT
fpq-14292	400	19	and	and	CCONJ
fpq-14292	400	20	wikipedia	wikipedia	PROPN
fpq-14292	400	21	present	present	VERB
fpq-14292	400	22	themselves	themselves	PRON
fpq-14292	400	23	as	as	ADV
fpq-14292	400	24	open	open	ADJ
fpq-14292	400	25	and	and	CCONJ
fpq-14292	400	26	accessible	accessible	ADJ
fpq-14292	400	27	to	to	ADP
fpq-14292	400	28	anyone	anyone	PRON
fpq-14292	400	29	,	,	PUNCT
fpq-14292	400	30	there	there	PRON
fpq-14292	400	31	are	be	VERB
fpq-14292	400	32	structural	structural	ADJ
fpq-14292	400	33	factors	factor	NOUN
fpq-14292	400	34	including	include	VERB
fpq-14292	400	35	moderation	moderation	NOUN
fpq-14292	400	36	practices	practice	NOUN
fpq-14292	400	37	which	which	PRON
fpq-14292	400	38	make	make	VERB
fpq-14292	400	39	them	they	PRON
fpq-14292	400	40	less	less	ADV
fpq-14292	400	41	welcoming	welcoming	ADJ
fpq-14292	400	42	to	to	ADP
fpq-14292	400	43	marginalized	marginalize	VERB
fpq-14292	400	44	populations	population	NOUN
fpq-14292	400	45	”	"	PUNCT
fpq-14292	400	46	(	(	PUNCT
fpq-14292	400	47	bender	bender	NOUN
fpq-14292	400	48	,	,	PUNCT
fpq-14292	400	49	gebru	gebru	VERB
fpq-14292	401	1	et	et	PROPN
fpq-14292	401	2	al	al	PROPN
fpq-14292	402	1	.	.	PROPN
fpq-14292	402	2	2021	2021	NUM
fpq-14292	402	3	,	,	PUNCT
fpq-14292	402	4	613	613	NUM
fpq-14292	402	5	)	)	PUNCT
fpq-14292	402	6	,	,	PUNCT
fpq-14292	402	7	resulting	result	VERB
fpq-14292	402	8	in	in	ADP
fpq-14292	402	9	marginalized	marginalize	VERB
fpq-14292	402	10	voices	voice	NOUN
fpq-14292	402	11	’	'	PUNCT
fpq-14292	402	12	being	be	AUX
fpq-14292	402	13	underrepresented	underrepresente	VERB
fpq-14292	402	14	in	in	ADP
fpq-14292	402	15	the	the	DET
fpq-14292	402	16	models	model	NOUN
fpq-14292	402	17	.	.	PUNCT
fpq-14292	403	1	simultaneously	simultaneously	ADV
fpq-14292	403	2	,	,	PUNCT
fpq-14292	403	3	“	"	PUNCT
fpq-14292	403	4	white	white	ADJ
fpq-14292	403	5	supremacist	supremacist	ADJ
fpq-14292	403	6	and	and	CCONJ
fpq-14292	403	7	misogynistic	misogynistic	ADJ
fpq-14292	403	8	,	,	PUNCT
fpq-14292	403	9	ageist	ageist	ADJ
fpq-14292	403	10	”	"	PUNCT
fpq-14292	403	11	voices	voice	NOUN
fpq-14292	403	12	tend	tend	VERB
fpq-14292	403	13	to	to	PART
fpq-14292	403	14	be	be	AUX
fpq-14292	403	15	“	"	PUNCT
fpq-14292	403	16	overrepresented	overrepresente	VERB
fpq-14292	403	17	in	in	ADP
fpq-14292	403	18	the	the	DET
fpq-14292	403	19	training	training	NOUN
fpq-14292	403	20	data	datum	NOUN
fpq-14292	403	21	”	"	PUNCT
fpq-14292	403	22	(	(	PUNCT
fpq-14292	403	23	613	613	NUM
fpq-14292	403	24	)	)	PUNCT
fpq-14292	403	25	due	due	ADP
fpq-14292	403	26	to	to	ADP
fpq-14292	403	27	how	how	SCONJ
fpq-14292	403	28	the	the	DET
fpq-14292	403	29	data	datum	NOUN
fpq-14292	403	30	are	be	AUX
fpq-14292	403	31	filtered	filter	VERB
fpq-14292	403	32	.	.	PUNCT
fpq-14292	404	1	bender	bender	NOUN
fpq-14292	404	2	/	/	SYM
fpq-14292	404	3	gebru	gebru	PROPN
fpq-14292	404	4	and	and	CCONJ
fpq-14292	404	5	colleagues	colleague	NOUN
fpq-14292	404	6	’	'	PUNCT
fpq-14292	404	7	description	description	NOUN
fpq-14292	404	8	of	of	ADP
fpq-14292	404	9	this	this	DET
fpq-14292	404	10	problem	problem	NOUN
fpq-14292	404	11	can	can	AUX
fpq-14292	404	12	be	be	AUX
fpq-14292	404	13	usefully	usefully	ADV
fpq-14292	404	14	framed	frame	VERB
fpq-14292	404	15	in	in	ADP
fpq-14292	404	16	terms	term	NOUN
fpq-14292	404	17	of	of	ADP
fpq-14292	404	18	normative	normative	ADJ
fpq-14292	404	19	and	and	CCONJ
fpq-14292	404	20	descriptive	descriptive	ADJ
fpq-14292	404	21	correctness	correctness	NOUN
fpq-14292	404	22	.	.	PUNCT
fpq-14292	405	1	large	large	ADJ
fpq-14292	405	2	language	language	NOUN
fpq-14292	405	3	models	model	NOUN
fpq-14292	405	4	,	,	PUNCT
fpq-14292	405	5	bender	bender	NOUN
fpq-14292	405	6	/	/	SYM
fpq-14292	405	7	gebru	gebru	PROPN
fpq-14292	405	8	and	and	CCONJ
fpq-14292	405	9	colleagues	colleague	NOUN
fpq-14292	405	10	appear	appear	VERB
fpq-14292	405	11	to	to	PART
fpq-14292	405	12	say	say	VERB
fpq-14292	405	13	,	,	PUNCT
fpq-14292	405	14	are	be	AUX
fpq-14292	405	15	not	not	PART
fpq-14292	405	16	only	only	ADV
fpq-14292	405	17	normatively	normatively	ADV
fpq-14292	405	18	incorrect	incorrect	ADJ
fpq-14292	405	19	but	but	CCONJ
fpq-14292	405	20	also	also	ADV
fpq-14292	405	21	descriptively	descriptively	ADV
fpq-14292	405	22	inaccurate	inaccurate	ADJ
fpq-14292	405	23	in	in	ADP
fpq-14292	405	24	important	important	ADJ
fpq-14292	405	25	ways	way	NOUN
fpq-14292	405	26	,	,	PUNCT
fpq-14292	405	27	and	and	CCONJ
fpq-14292	405	28	they	they	PRON
fpq-14292	405	29	are	be	AUX
fpq-14292	405	30	normatively	normatively	ADV
fpq-14292	405	31	incorrect	incorrect	ADJ
fpq-14292	405	32	partly	partly	ADV
fpq-14292	405	33	because	because	SCONJ
fpq-14292	405	34	they	they	PRON
fpq-14292	405	35	are	be	AUX
fpq-14292	405	36	descriptively	descriptively	ADV
fpq-14292	405	37	inaccurate	inaccurate	ADJ
fpq-14292	405	38	.	.	PUNCT
fpq-14292	406	1	32	32	NUM
fpq-14292	406	2	bender	bender	NOUN
fpq-14292	406	3	and	and	CCONJ
fpq-14292	406	4	gebru	gebru	NOUN
fpq-14292	406	5	are	be	AUX
fpq-14292	406	6	listed	list	VERB
fpq-14292	406	7	as	as	ADP
fpq-14292	406	8	joint	joint	ADJ
fpq-14292	406	9	lead	lead	NOUN
fpq-14292	406	10	authors	author	NOUN
fpq-14292	406	11	on	on	ADP
fpq-14292	406	12	their	their	PRON
fpq-14292	406	13	2021	2021	NUM
fpq-14292	406	14	paper	paper	NOUN
fpq-14292	406	15	.	.	PUNCT
fpq-14292	407	1	we	we	PRON
fpq-14292	407	2	have	have	VERB
fpq-14292	407	3	slightly	slightly	ADV
fpq-14292	407	4	modified	modify	VERB
fpq-14292	407	5	citation	citation	NOUN
fpq-14292	407	6	format	format	NOUN
fpq-14292	407	7	for	for	SCONJ
fpq-14292	407	8	that	that	DET
fpq-14292	407	9	paper	paper	NOUN
fpq-14292	407	10	to	to	PART
fpq-14292	407	11	reflect	reflect	VERB
fpq-14292	407	12	this	this	PRON
fpq-14292	407	13	.	.	PUNCT
fpq-14292	408	1	feminist	feminist	PROPN
fpq-14292	408	2	philosophy	philosophy	PROPN
fpq-14292	408	3	quarterly	quarterly	ADV
fpq-14292	408	4	,	,	PUNCT
fpq-14292	408	5	2022	2022	NUM
fpq-14292	408	6	,	,	PUNCT
fpq-14292	408	7	vol	vol	NOUN
fpq-14292	408	8	.	.	NOUN
fpq-14292	408	9	8	8	NUM
fpq-14292	408	10	,	,	PUNCT
fpq-14292	408	11	iss	iss	PROPN
fpq-14292	408	12	.	.	PUNCT
fpq-14292	408	13	3/4	3/4	NUM
fpq-14292	408	14	,	,	PUNCT
fpq-14292	408	15	article	article	NOUN
fpq-14292	408	16	1	1	NUM
fpq-14292	408	17	published	publish	VERB
fpq-14292	408	18	by	by	ADP
fpq-14292	408	19	scholarship@western	scholarship@western	PROPN
fpq-14292	408	20	,	,	PUNCT
fpq-14292	408	21	2022	2022	NUM
fpq-14292	408	22	22	22	NUM
fpq-14292	408	23	even	even	ADV
fpq-14292	408	24	criticisms	criticism	NOUN
fpq-14292	408	25	of	of	ADP
fpq-14292	408	26	bender	bender	NOUN
fpq-14292	408	27	/	/	SYM
fpq-14292	408	28	gebru	gebru	PROPN
fpq-14292	408	29	and	and	CCONJ
fpq-14292	408	30	colleagues	colleague	NOUN
fpq-14292	408	31	’	'	PUNCT
fpq-14292	408	32	claims	claim	NOUN
fpq-14292	408	33	can	can	AUX
fpq-14292	408	34	be	be	AUX
fpq-14292	408	35	usefully	usefully	ADV
fpq-14292	408	36	framed	frame	VERB
fpq-14292	408	37	in	in	ADP
fpq-14292	408	38	terms	term	NOUN
fpq-14292	408	39	of	of	ADP
fpq-14292	408	40	normative	normative	ADJ
fpq-14292	408	41	correctness	correctness	NOUN
fpq-14292	408	42	and	and	CCONJ
fpq-14292	408	43	descriptive	descriptive	ADJ
fpq-14292	408	44	accuracy	accuracy	NOUN
fpq-14292	408	45	.	.	PUNCT
fpq-14292	409	1	for	for	ADP
fpq-14292	409	2	instance	instance	NOUN
fpq-14292	409	3	,	,	PUNCT
fpq-14292	409	4	yoav	yoav	PROPN
fpq-14292	409	5	goldberg	goldberg	PROPN
fpq-14292	409	6	(	(	PUNCT
fpq-14292	409	7	2021	2021	NUM
fpq-14292	409	8	)	)	PUNCT
fpq-14292	409	9	claims	claim	VERB
fpq-14292	409	10	that	that	SCONJ
fpq-14292	409	11	bender	bender	NOUN
fpq-14292	409	12	/	/	SYM
fpq-14292	409	13	gebru	gebru	PROPN
fpq-14292	409	14	and	and	CCONJ
fpq-14292	409	15	colleagues	colleague	NOUN
fpq-14292	409	16	“	"	PUNCT
fpq-14292	409	17	suggest	suggest	VERB
fpq-14292	409	18	that	that	SCONJ
fpq-14292	409	19	good	good	ADJ
fpq-14292	409	20	(=	(=	NOUN
fpq-14292	409	21	not	not	PART
fpq-14292	409	22	dangerous	dangerous	ADJ
fpq-14292	409	23	)	)	PUNCT
fpq-14292	409	24	language	language	NOUN
fpq-14292	409	25	models	model	NOUN
fpq-14292	409	26	are	be	AUX
fpq-14292	409	27	.	.	PUNCT
fpq-14292	409	28	.	.	PUNCT
fpq-14292	409	29	.	.	PUNCT
fpq-14292	410	1	models	model	NOUN
fpq-14292	410	2	which	which	PRON
fpq-14292	410	3	reflect	reflect	VERB
fpq-14292	410	4	the	the	DET
fpq-14292	410	5	world	world	NOUN
fpq-14292	410	6	as	as	SCONJ
fpq-14292	410	7	they	they	PRON
fpq-14292	410	8	think	think	VERB
fpq-14292	410	9	the	the	DET
fpq-14292	410	10	world	world	NOUN
fpq-14292	410	11	should	should	AUX
fpq-14292	410	12	be	be	AUX
fpq-14292	410	13	,	,	PUNCT
fpq-14292	410	14	”	"	PUNCT
fpq-14292	410	15	yet	yet	ADV
fpq-14292	410	16	“	"	PUNCT
fpq-14292	410	17	an	an	DET
fpq-14292	410	18	alternative	alternative	ADJ
fpq-14292	410	19	view	view	NOUN
fpq-14292	410	20	by	by	ADP
fpq-14292	410	21	which	which	PRON
fpq-14292	410	22	language	language	NOUN
fpq-14292	410	23	models	model	NOUN
fpq-14292	410	24	should	should	AUX
fpq-14292	410	25	reflect	reflect	VERB
fpq-14292	410	26	language	language	NOUN
fpq-14292	410	27	as	as	SCONJ
fpq-14292	410	28	it	it	PRON
fpq-14292	410	29	is	be	AUX
fpq-14292	410	30	being	be	AUX
fpq-14292	410	31	used	use	VERB
fpq-14292	410	32	in	in	ADP
fpq-14292	410	33	a	a	DET
fpq-14292	410	34	training	training	NOUN
fpq-14292	410	35	corpus	corpus	NOUN
fpq-14292	410	36	is	be	AUX
fpq-14292	410	37	at	at	ADV
fpq-14292	410	38	least	least	ADJ
fpq-14292	410	39	as	as	ADP
fpq-14292	410	40	valid	valid	ADJ
fpq-14292	410	41	,	,	PUNCT
fpq-14292	410	42	and	and	CCONJ
fpq-14292	410	43	should	should	AUX
fpq-14292	410	44	be	be	AUX
fpq-14292	410	45	acknowledged	acknowledge	VERB
fpq-14292	410	46	.	.	PUNCT
fpq-14292	410	47	”	"	PUNCT
fpq-14292	411	1	goldberg	goldberg	PROPN
fpq-14292	411	2	picks	pick	VERB
fpq-14292	411	3	up	up	ADP
fpq-14292	411	4	here	here	ADV
fpq-14292	411	5	on	on	ADP
fpq-14292	411	6	an	an	DET
fpq-14292	411	7	ambiguity	ambiguity	NOUN
fpq-14292	411	8	in	in	ADP
fpq-14292	411	9	how	how	SCONJ
fpq-14292	411	10	“	"	PUNCT
fpq-14292	411	11	good	good	ADJ
fpq-14292	411	12	”	"	PUNCT
fpq-14292	411	13	(	(	PUNCT
fpq-14292	411	14	or	or	CCONJ
fpq-14292	411	15	its	its	PRON
fpq-14292	411	16	cognates	cognate	NOUN
fpq-14292	411	17	)	)	PUNCT
fpq-14292	411	18	can	can	AUX
fpq-14292	411	19	be	be	AUX
fpq-14292	411	20	used	use	VERB
fpq-14292	411	21	in	in	ADP
fpq-14292	411	22	this	this	DET
fpq-14292	411	23	context	context	NOUN
fpq-14292	411	24	—	—	PUNCT
fpq-14292	411	25	that	that	ADV
fpq-14292	411	26	is	is	ADV
fpq-14292	411	27	,	,	PUNCT
fpq-14292	411	28	as	as	ADP
fpq-14292	411	29	picking	pick	VERB
fpq-14292	411	30	out	out	ADP
fpq-14292	411	31	normative	normative	ADJ
fpq-14292	411	32	correctness	correctness	NOUN
fpq-14292	411	33	(	(	PUNCT
fpq-14292	411	34	“	"	PUNCT
fpq-14292	411	35	not	not	PART
fpq-14292	411	36	dangerous	dangerous	ADJ
fpq-14292	411	37	”	"	PUNCT
fpq-14292	411	38	)	)	PUNCT
fpq-14292	411	39	or	or	CCONJ
fpq-14292	411	40	descriptive	descriptive	ADJ
fpq-14292	411	41	accuracy	accuracy	NOUN
fpq-14292	411	42	(	(	PUNCT
fpq-14292	411	43	“	"	PUNCT
fpq-14292	411	44	language	language	NOUN
fpq-14292	411	45	as	as	SCONJ
fpq-14292	411	46	it	it	PRON
fpq-14292	411	47	is	be	AUX
fpq-14292	411	48	being	be	AUX
fpq-14292	411	49	used	use	VERB
fpq-14292	411	50	”	"	PUNCT
fpq-14292	411	51	)	)	PUNCT
fpq-14292	411	52	.	.	PUNCT
fpq-14292	412	1	in	in	ADP
fpq-14292	412	2	our	our	PRON
fpq-14292	412	3	terms	term	NOUN
fpq-14292	412	4	,	,	PUNCT
fpq-14292	412	5	goldberg	goldberg	PROPN
fpq-14292	412	6	claims	claim	VERB
fpq-14292	412	7	that	that	SCONJ
fpq-14292	412	8	bender	bender	NOUN
fpq-14292	412	9	/	/	SYM
fpq-14292	412	10	gebru	gebru	PROPN
fpq-14292	412	11	and	and	CCONJ
fpq-14292	412	12	colleagues	colleague	NOUN
fpq-14292	412	13	fail	fail	VERB
fpq-14292	412	14	adequately	adequately	ADV
fpq-14292	412	15	to	to	PART
fpq-14292	412	16	acknowledge	acknowledge	VERB
fpq-14292	412	17	their	their	PRON
fpq-14292	412	18	preference	preference	NOUN
fpq-14292	412	19	for	for	ADP
fpq-14292	412	20	prioritizing	prioritize	VERB
fpq-14292	412	21	the	the	DET
fpq-14292	412	22	normative	normative	NOUN
fpq-14292	412	23	over	over	ADP
fpq-14292	412	24	the	the	DET
fpq-14292	412	25	descriptive	descriptive	ADJ
fpq-14292	412	26	,	,	PUNCT
fpq-14292	412	27	and	and	CCONJ
fpq-14292	412	28	in	in	ADP
fpq-14292	412	29	a	a	DET
fpq-14292	412	30	particular	particular	ADJ
fpq-14292	412	31	way	way	NOUN
fpq-14292	412	32	.	.	PUNCT
fpq-14292	413	1	indeed	indeed	ADV
fpq-14292	413	2	,	,	PUNCT
fpq-14292	413	3	goldberg	goldberg	PROPN
fpq-14292	413	4	thinks	think	VERB
fpq-14292	413	5	that	that	SCONJ
fpq-14292	413	6	bender	bender	NOUN
fpq-14292	413	7	/	/	SYM
fpq-14292	413	8	gebru	gebru	PROPN
fpq-14292	413	9	and	and	CCONJ
fpq-14292	413	10	colleagues	colleague	NOUN
fpq-14292	413	11	prioritize	prioritize	VERB
fpq-14292	413	12	normative	normative	ADJ
fpq-14292	413	13	correctness	correctness	NOUN
fpq-14292	413	14	in	in	ADP
fpq-14292	413	15	a	a	DET
fpq-14292	413	16	way	way	NOUN
fpq-14292	413	17	that	that	PRON
fpq-14292	413	18	can	can	AUX
fpq-14292	413	19	result	result	VERB
fpq-14292	413	20	in	in	ADP
fpq-14292	413	21	our	our	PRON
fpq-14292	413	22	adopting	adopt	VERB
fpq-14292	413	23	certain	certain	ADJ
fpq-14292	413	24	views	view	NOUN
fpq-14292	413	25	on	on	ADP
fpq-14292	413	26	ethical	ethical	ADJ
fpq-14292	413	27	issues	issue	NOUN
fpq-14292	413	28	even	even	ADV
fpq-14292	413	29	when	when	SCONJ
fpq-14292	413	30	there	there	PRON
fpq-14292	413	31	is	be	VERB
fpq-14292	413	32	legitimate	legitimate	ADJ
fpq-14292	413	33	debate	debate	NOUN
fpq-14292	413	34	about	about	ADP
fpq-14292	413	35	whether	whether	SCONJ
fpq-14292	413	36	they	they	PRON
fpq-14292	413	37	are	be	AUX
fpq-14292	413	38	the	the	DET
fpq-14292	413	39	right	right	ADJ
fpq-14292	413	40	views	view	NOUN
fpq-14292	413	41	to	to	PART
fpq-14292	413	42	adopt	adopt	VERB
fpq-14292	413	43	.	.	PUNCT
fpq-14292	414	1	as	as	SCONJ
fpq-14292	414	2	goldberg	goldberg	PROPN
fpq-14292	414	3	(	(	PUNCT
fpq-14292	414	4	2021	2021	NUM
fpq-14292	414	5	)	)	PUNCT
fpq-14292	414	6	puts	put	VERB
fpq-14292	414	7	it	it	PRON
fpq-14292	414	8	,	,	PUNCT
fpq-14292	414	9	the	the	DET
fpq-14292	414	10	fact	fact	NOUN
fpq-14292	414	11	that	that	SCONJ
fpq-14292	414	12	there	there	PRON
fpq-14292	414	13	is	be	VERB
fpq-14292	414	14	ongoing	ongoing	ADJ
fpq-14292	414	15	debate	debate	NOUN
fpq-14292	414	16	on	on	ADP
fpq-14292	414	17	such	such	ADJ
fpq-14292	414	18	issues	issue	NOUN
fpq-14292	414	19	“	"	PUNCT
fpq-14292	414	20	should	should	AUX
fpq-14292	414	21	be	be	AUX
fpq-14292	414	22	made	make	VERB
fpq-14292	414	23	explicit	explicit	ADJ
fpq-14292	414	24	”	"	PUNCT
fpq-14292	414	25	by	by	ADP
fpq-14292	414	26	bender	bender	NOUN
fpq-14292	414	27	/	/	SYM
fpq-14292	414	28	gebru	gebru	PROPN
fpq-14292	414	29	and	and	CCONJ
fpq-14292	414	30	colleagues	colleague	NOUN
fpq-14292	414	31	.	.	PUNCT
fpq-14292	415	1	goldberg	goldberg	PROPN
fpq-14292	415	2	also	also	ADV
fpq-14292	415	3	thinks	think	VERB
fpq-14292	415	4	there	there	PRON
fpq-14292	415	5	can	can	AUX
fpq-14292	415	6	be	be	AUX
fpq-14292	415	7	practical	practical	ADJ
fpq-14292	415	8	and	and	CCONJ
fpq-14292	415	9	ethical	ethical	ADJ
fpq-14292	415	10	reasons	reason	NOUN
fpq-14292	415	11	for	for	ADP
fpq-14292	415	12	prioritizing	prioritize	VERB
fpq-14292	415	13	descriptive	descriptive	ADJ
fpq-14292	415	14	accuracy	accuracy	NOUN
fpq-14292	415	15	over	over	ADP
fpq-14292	415	16	normative	normative	ADJ
fpq-14292	415	17	correctness	correctness	NOUN
fpq-14292	415	18	,	,	PUNCT
fpq-14292	415	19	contra	contra	PROPN
fpq-14292	415	20	what	what	PRON
fpq-14292	415	21	he	he	PRON
fpq-14292	415	22	takes	take	VERB
fpq-14292	415	23	bender	bender	NOUN
fpq-14292	415	24	/	/	SYM
fpq-14292	415	25	gebru	gebru	NOUN
fpq-14292	415	26	and	and	CCONJ
fpq-14292	415	27	colleagues	colleague	NOUN
fpq-14292	415	28	to	to	PART
fpq-14292	415	29	implicitly	implicitly	ADV
fpq-14292	415	30	maintain	maintain	VERB
fpq-14292	415	31	.	.	PUNCT
fpq-14292	416	1	a	a	DET
fpq-14292	416	2	central	central	ADJ
fpq-14292	416	3	target	target	NOUN
fpq-14292	416	4	of	of	ADP
fpq-14292	416	5	bender	bender	NOUN
fpq-14292	416	6	/	/	SYM
fpq-14292	416	7	gebru	gebru	PROPN
fpq-14292	416	8	and	and	CCONJ
fpq-14292	416	9	colleagues	colleague	NOUN
fpq-14292	416	10	’	'	PUNCT
fpq-14292	416	11	concern	concern	NOUN
fpq-14292	416	12	seems	seem	VERB
fpq-14292	416	13	not	not	PART
fpq-14292	416	14	to	to	PART
fpq-14292	416	15	be	be	AUX
fpq-14292	416	16	what	what	PRON
fpq-14292	416	17	suresh	suresh	NOUN
fpq-14292	416	18	and	and	CCONJ
fpq-14292	416	19	guttag	guttag	NOUN
fpq-14292	416	20	call	call	NOUN
fpq-14292	416	21	historical	historical	ADJ
fpq-14292	416	22	bias	bias	NOUN
fpq-14292	416	23	but	but	CCONJ
fpq-14292	416	24	rather	rather	ADV
fpq-14292	416	25	what	what	PRON
fpq-14292	416	26	they	they	PRON
fpq-14292	416	27	call	call	VERB
fpq-14292	416	28	representation	representation	NOUN
fpq-14292	416	29	bias	bias	NOUN
fpq-14292	416	30	,	,	PUNCT
fpq-14292	416	31	which	which	PRON
fpq-14292	416	32	occurs	occur	VERB
fpq-14292	416	33	“	"	PUNCT
fpq-14292	416	34	when	when	SCONJ
fpq-14292	416	35	certain	certain	ADJ
fpq-14292	416	36	parts	part	NOUN
fpq-14292	416	37	of	of	ADP
fpq-14292	416	38	the	the	DET
fpq-14292	416	39	input	input	NOUN
fpq-14292	416	40	space	space	NOUN
fpq-14292	416	41	are	be	AUX
fpq-14292	416	42	underrepresented	underrepresented	ADJ
fpq-14292	416	43	”	"	PUNCT
fpq-14292	416	44	(	(	PUNCT
fpq-14292	416	45	suresh	suresh	NOUN
fpq-14292	416	46	and	and	CCONJ
fpq-14292	416	47	guttag	guttag	NOUN
fpq-14292	416	48	2020	2020	NUM
fpq-14292	416	49	,	,	PUNCT
fpq-14292	416	50	5	5	NUM
fpq-14292	416	51	)	)	PUNCT
fpq-14292	416	52	.	.	PUNCT
fpq-14292	417	1	for	for	ADP
fpq-14292	417	2	example	example	NOUN
fpq-14292	417	3	,	,	PUNCT
fpq-14292	417	4	with	with	ADP
fpq-14292	417	5	language	language	NOUN
fpq-14292	417	6	models	model	NOUN
fpq-14292	417	7	,	,	PUNCT
fpq-14292	417	8	“	"	PUNCT
fpq-14292	417	9	datasets	dataset	NOUN
fpq-14292	417	10	collected	collect	VERB
fpq-14292	417	11	through	through	ADP
fpq-14292	417	12	smartphone	smartphone	NOUN
fpq-14292	417	13	apps	app	NOUN
fpq-14292	417	14	can	can	AUX
fpq-14292	417	15	under	under	AUX
fpq-14292	417	16	-	-	PUNCT
fpq-14292	417	17	represent	represent	VERB
fpq-14292	417	18	lower	low	ADJ
fpq-14292	417	19	-	-	PUNCT
fpq-14292	417	20	income	income	NOUN
fpq-14292	417	21	or	or	CCONJ
fpq-14292	417	22	older	old	ADJ
fpq-14292	417	23	groups	group	NOUN
fpq-14292	417	24	,	,	PUNCT
fpq-14292	417	25	who	who	PRON
fpq-14292	417	26	are	be	AUX
fpq-14292	417	27	less	less	ADV
fpq-14292	417	28	likely	likely	ADJ
fpq-14292	417	29	to	to	PART
fpq-14292	417	30	own	own	VERB
fpq-14292	417	31	smartphones	smartphone	NOUN
fpq-14292	417	32	”	"	PUNCT
fpq-14292	417	33	(	(	PUNCT
fpq-14292	417	34	suresh	suresh	NOUN
fpq-14292	417	35	and	and	CCONJ
fpq-14292	417	36	guttag	guttag	NOUN
fpq-14292	417	37	2020	2020	NUM
fpq-14292	417	38	,	,	PUNCT
fpq-14292	417	39	5	5	NUM
fpq-14292	417	40	)	)	PUNCT
fpq-14292	417	41	.	.	PUNCT
fpq-14292	418	1	this	this	DET
fpq-14292	418	2	form	form	NOUN
fpq-14292	418	3	of	of	ADP
fpq-14292	418	4	bias	bias	NOUN
fpq-14292	418	5	is	be	AUX
fpq-14292	418	6	one	one	NUM
fpq-14292	418	7	of	of	ADP
fpq-14292	418	8	the	the	DET
fpq-14292	418	9	central	central	ADJ
fpq-14292	418	10	focuses	focus	NOUN
fpq-14292	418	11	of	of	ADP
fpq-14292	418	12	bender	bender	NOUN
fpq-14292	418	13	/	/	SYM
fpq-14292	418	14	gebru	gebru	PROPN
fpq-14292	418	15	and	and	CCONJ
fpq-14292	418	16	colleagues	colleague	NOUN
fpq-14292	418	17	’	'	PUNCT
fpq-14292	418	18	attention	attention	NOUN
fpq-14292	418	19	,	,	PUNCT
fpq-14292	418	20	rather	rather	ADV
fpq-14292	418	21	than	than	ADP
fpq-14292	418	22	historical	historical	ADJ
fpq-14292	418	23	bias	bias	NOUN
fpq-14292	418	24	,	,	PUNCT
fpq-14292	418	25	given	give	VERB
fpq-14292	418	26	that	that	SCONJ
fpq-14292	418	27	a	a	DET
fpq-14292	418	28	central	central	ADJ
fpq-14292	418	29	claim	claim	NOUN
fpq-14292	418	30	of	of	ADP
fpq-14292	418	31	theirs	theirs	PRON
fpq-14292	418	32	is	be	AUX
fpq-14292	418	33	that	that	SCONJ
fpq-14292	418	34	often	often	ADV
fpq-14292	418	35	the	the	DET
fpq-14292	418	36	relevant	relevant	ADJ
fpq-14292	418	37	data	datum	NOUN
fpq-14292	418	38	have	have	AUX
fpq-14292	418	39	not	not	PART
fpq-14292	418	40	been	be	AUX
fpq-14292	418	41	perfectly	perfectly	ADV
fpq-14292	418	42	measured	measure	VERB
fpq-14292	418	43	and	and	CCONJ
fpq-14292	418	44	sampled	sample	VERB
fpq-14292	418	45	.	.	PUNCT
fpq-14292	419	1	by	by	ADP
fpq-14292	419	2	contrast	contrast	NOUN
fpq-14292	419	3	,	,	PUNCT
fpq-14292	419	4	historical	historical	ADJ
fpq-14292	419	5	bias	bias	NOUN
fpq-14292	419	6	arises	arise	VERB
fpq-14292	419	7	even	even	ADV
fpq-14292	419	8	when	when	SCONJ
fpq-14292	419	9	the	the	DET
fpq-14292	419	10	data	datum	NOUN
fpq-14292	419	11	have	have	AUX
fpq-14292	419	12	been	be	AUX
fpq-14292	419	13	“	"	PUNCT
fpq-14292	419	14	perfectly	perfectly	ADV
fpq-14292	419	15	measured	measure	VERB
fpq-14292	419	16	and	and	CCONJ
fpq-14292	419	17	sampled	sample	VERB
fpq-14292	419	18	”	"	PUNCT
fpq-14292	419	19	and	and	CCONJ
fpq-14292	419	20	“	"	PUNCT
fpq-14292	419	21	the	the	DET
fpq-14292	419	22	world	world	NOUN
fpq-14292	419	23	as	as	SCONJ
fpq-14292	419	24	it	it	PRON
fpq-14292	419	25	is	be	AUX
fpq-14292	419	26	or	or	CCONJ
fpq-14292	419	27	was	be	AUX
fpq-14292	419	28	leads	lead	VERB
fpq-14292	419	29	a	a	DET
fpq-14292	419	30	model	model	NOUN
fpq-14292	419	31	to	to	PART
fpq-14292	419	32	produce	produce	VERB
fpq-14292	419	33	outcomes	outcome	NOUN
fpq-14292	419	34	that	that	PRON
fpq-14292	419	35	are	be	AUX
fpq-14292	419	36	not	not	PART
fpq-14292	419	37	wanted	want	VERB
fpq-14292	419	38	”	"	PUNCT
fpq-14292	419	39	(	(	PUNCT
fpq-14292	419	40	suresh	suresh	NOUN
fpq-14292	419	41	and	and	CCONJ
fpq-14292	419	42	guttag	guttag	NOUN
fpq-14292	419	43	2020	2020	NUM
fpq-14292	419	44	,	,	PUNCT
fpq-14292	419	45	4	4	NUM
fpq-14292	419	46	)	)	PUNCT
fpq-14292	419	47	.	.	PUNCT
fpq-14292	420	1	our	our	PRON
fpq-14292	420	2	normative	normative	ADJ
fpq-14292	420	3	/	/	SYM
fpq-14292	420	4	descriptive	descriptive	ADJ
fpq-14292	420	5	distinction	distinction	NOUN
fpq-14292	420	6	again	again	ADV
fpq-14292	420	7	helps	help	VERB
fpq-14292	420	8	to	to	PART
fpq-14292	420	9	clarify	clarify	VERB
fpq-14292	420	10	the	the	DET
fpq-14292	420	11	points	point	NOUN
fpq-14292	420	12	of	of	ADP
fpq-14292	420	13	disagreement	disagreement	NOUN
fpq-14292	420	14	and	and	CCONJ
fpq-14292	420	15	agreement	agreement	NOUN
fpq-14292	420	16	between	between	ADP
fpq-14292	420	17	bender	bender	NOUN
fpq-14292	420	18	/	/	SYM
fpq-14292	420	19	gebru	gebru	PROPN
fpq-14292	420	20	and	and	CCONJ
fpq-14292	420	21	colleagues	colleague	NOUN
fpq-14292	420	22	and	and	CCONJ
fpq-14292	420	23	their	their	PRON
fpq-14292	420	24	critics	critic	NOUN
fpq-14292	420	25	such	such	ADJ
fpq-14292	420	26	as	as	ADP
fpq-14292	420	27	goldberg	goldberg	PROPN
fpq-14292	420	28	,	,	PUNCT
fpq-14292	420	29	while	while	SCONJ
fpq-14292	420	30	additionally	additionally	ADV
fpq-14292	420	31	clarifying	clarify	VERB
fpq-14292	420	32	how	how	SCONJ
fpq-14292	420	33	language	language	NOUN
fpq-14292	420	34	models	model	NOUN
fpq-14292	420	35	can	can	AUX
fpq-14292	420	36	involve	involve	VERB
fpq-14292	420	37	representation	representation	NOUN
fpq-14292	420	38	bias	bias	NOUN
fpq-14292	420	39	in	in	ADP
fpq-14292	420	40	suresh	suresh	PROPN
fpq-14292	420	41	and	and	CCONJ
fpq-14292	420	42	guttag	guttag	NOUN
fpq-14292	420	43	’s	’s	PART
fpq-14292	420	44	sense	sense	NOUN
fpq-14292	420	45	.	.	PUNCT
fpq-14292	421	1	bender	bender	NOUN
fpq-14292	421	2	/	/	SYM
fpq-14292	421	3	gebru	gebru	PROPN
fpq-14292	421	4	and	and	CCONJ
fpq-14292	421	5	colleagues	colleague	NOUN
fpq-14292	421	6	’	'	PUNCT
fpq-14292	421	7	claim	claim	NOUN
fpq-14292	421	8	is	be	AUX
fpq-14292	421	9	that	that	SCONJ
fpq-14292	421	10	large	large	ADJ
fpq-14292	421	11	language	language	NOUN
fpq-14292	421	12	models	model	NOUN
fpq-14292	421	13	are	be	AUX
fpq-14292	421	14	descriptively	descriptively	ADV
fpq-14292	421	15	inaccurate	inaccurate	ADJ
fpq-14292	421	16	relative	relative	ADJ
fpq-14292	421	17	to	to	ADP
fpq-14292	421	18	the	the	DET
fpq-14292	421	19	wider	wide	ADJ
fpq-14292	421	20	population	population	NOUN
fpq-14292	421	21	of	of	ADP
fpq-14292	421	22	language	language	NOUN
fpq-14292	421	23	users	user	NOUN
fpq-14292	421	24	,	,	PUNCT
fpq-14292	421	25	as	as	ADP
fpq-14292	421	26	a	a	DET
fpq-14292	421	27	result	result	NOUN
fpq-14292	421	28	of	of	ADP
fpq-14292	421	29	their	their	PRON
fpq-14292	421	30	being	be	AUX
fpq-14292	421	31	built	build	VERB
fpq-14292	421	32	around	around	ADP
fpq-14292	421	33	an	an	DET
fpq-14292	421	34	imperfect	imperfect	ADJ
fpq-14292	421	35	measuring	measuring	NOUN
fpq-14292	421	36	and	and	CCONJ
fpq-14292	421	37	sampling	sampling	NOUN
fpq-14292	421	38	of	of	ADP
fpq-14292	421	39	data	datum	NOUN
fpq-14292	421	40	.	.	PUNCT
fpq-14292	422	1	such	such	ADJ
fpq-14292	422	2	models	model	NOUN
fpq-14292	422	3	can	can	AUX
fpq-14292	422	4	be	be	AUX
fpq-14292	422	5	normatively	normatively	ADV
fpq-14292	422	6	incorrect	incorrect	ADJ
fpq-14292	422	7	in	in	ADP
fpq-14292	422	8	ways	way	NOUN
fpq-14292	422	9	deery	deery	NOUN
fpq-14292	422	10	and	and	CCONJ
fpq-14292	422	11	bailey	bailey	PROPN
fpq-14292	422	12	–	–	PUNCT
fpq-14292	422	13	the	the	DET
fpq-14292	422	14	bias	bias	NOUN
fpq-14292	422	15	dilemma	dilemma	NOUN
fpq-14292	422	16	:	:	PUNCT
fpq-14292	422	17	the	the	DET
fpq-14292	422	18	ethics	ethic	NOUN
fpq-14292	422	19	of	of	ADP
fpq-14292	422	20	algorithmic	algorithmic	ADJ
fpq-14292	422	21	bias	bias	NOUN
fpq-14292	422	22	in	in	ADP
fpq-14292	422	23	natural	natural	ADJ
fpq-14292	422	24	-	-	PUNCT
fpq-14292	422	25	language	language	NOUN
fpq-14292	422	26	processing	processing	NOUN
fpq-14292	422	27	published	publish	VERB
fpq-14292	422	28	by	by	ADP
fpq-14292	422	29	scholarship@western	scholarship@western	PROPN
fpq-14292	422	30	,	,	PUNCT
fpq-14292	422	31	2022	2022	NUM
fpq-14292	422	32	23	23	NUM
fpq-14292	422	33	that	that	PRON
fpq-14292	422	34	result	result	VERB
fpq-14292	422	35	from	from	ADP
fpq-14292	422	36	such	such	ADJ
fpq-14292	422	37	inaccuracy.33	inaccuracy.33	PROPN
fpq-14292	422	38	thus	thus	ADV
fpq-14292	422	39	,	,	PUNCT
fpq-14292	422	40	suresh	suresh	PROPN
fpq-14292	422	41	and	and	CCONJ
fpq-14292	422	42	guttag	guttag	NOUN
fpq-14292	422	43	’s	’s	PART
fpq-14292	422	44	representation	representation	NOUN
fpq-14292	422	45	bias	bias	NOUN
fpq-14292	422	46	is	be	AUX
fpq-14292	422	47	one	one	NUM
fpq-14292	422	48	way	way	NOUN
fpq-14292	422	49	in	in	ADP
fpq-14292	422	50	which	which	PRON
fpq-14292	422	51	a	a	DET
fpq-14292	422	52	model	model	NOUN
fpq-14292	422	53	can	can	AUX
fpq-14292	422	54	be	be	AUX
fpq-14292	422	55	descriptively	descriptively	ADV
fpq-14292	422	56	inaccurate	inaccurate	ADJ
fpq-14292	422	57	.	.	PUNCT
fpq-14292	423	1	when	when	SCONJ
fpq-14292	423	2	such	such	ADJ
fpq-14292	423	3	inaccuracy	inaccuracy	NOUN
fpq-14292	423	4	is	be	AUX
fpq-14292	423	5	distinguished	distinguish	VERB
fpq-14292	423	6	from	from	ADP
fpq-14292	423	7	normative	normative	ADJ
fpq-14292	423	8	incorrectness	incorrectness	NOUN
fpq-14292	423	9	,	,	PUNCT
fpq-14292	423	10	the	the	DET
fpq-14292	423	11	points	point	NOUN
fpq-14292	423	12	of	of	ADP
fpq-14292	423	13	disagreement	disagreement	NOUN
fpq-14292	423	14	among	among	ADP
fpq-14292	423	15	goldberg	goldberg	PROPN
fpq-14292	423	16	and	and	CCONJ
fpq-14292	423	17	bender	bender	PROPN
fpq-14292	423	18	/	/	SYM
fpq-14292	423	19	gebru	gebru	PROPN
fpq-14292	423	20	and	and	CCONJ
fpq-14292	423	21	colleagues	colleague	NOUN
fpq-14292	423	22	become	become	VERB
fpq-14292	423	23	clearer	clear	ADJ
fpq-14292	423	24	.	.	PUNCT
fpq-14292	424	1	it	it	PRON
fpq-14292	424	2	might	might	AUX
fpq-14292	424	3	be	be	AUX
fpq-14292	424	4	suggested	suggest	VERB
fpq-14292	424	5	that	that	SCONJ
fpq-14292	424	6	bender	bender	NOUN
fpq-14292	424	7	/	/	SYM
fpq-14292	424	8	gebru	gebru	PROPN
fpq-14292	424	9	and	and	CCONJ
fpq-14292	424	10	colleagues	colleague	NOUN
fpq-14292	424	11	underestimate	underestimate	VERB
fpq-14292	424	12	the	the	DET
fpq-14292	424	13	extent	extent	NOUN
fpq-14292	424	14	to	to	PART
fpq-14292	424	15	which	which	PRON
fpq-14292	424	16	language	language	NOUN
fpq-14292	424	17	models	model	NOUN
fpq-14292	424	18	like	like	ADP
fpq-14292	424	19	gtp-2/3	gtp-2/3	PROPN
fpq-14292	424	20	are	be	AUX
fpq-14292	424	21	descriptively	descriptively	ADV
fpq-14292	424	22	accurate	accurate	ADJ
fpq-14292	424	23	.	.	PUNCT
fpq-14292	425	1	after	after	ADV
fpq-14292	425	2	all	all	ADV
fpq-14292	425	3	,	,	PUNCT
fpq-14292	425	4	even	even	ADV
fpq-14292	425	5	if	if	SCONJ
fpq-14292	425	6	they	they	PRON
fpq-14292	425	7	are	be	AUX
fpq-14292	425	8	trained	train	VERB
fpq-14292	425	9	on	on	ADP
fpq-14292	425	10	datasets	dataset	NOUN
fpq-14292	425	11	that	that	PRON
fpq-14292	425	12	come	come	VERB
fpq-14292	425	13	in	in	ADP
fpq-14292	425	14	significant	significant	ADJ
fpq-14292	425	15	part	part	NOUN
fpq-14292	425	16	from	from	ADP
fpq-14292	425	17	platforms	platform	NOUN
fpq-14292	425	18	from	from	ADP
fpq-14292	425	19	which	which	PRON
fpq-14292	425	20	marginalized	marginalize	VERB
fpq-14292	425	21	voices	voice	NOUN
fpq-14292	425	22	have	have	AUX
fpq-14292	425	23	been	be	AUX
fpq-14292	425	24	excluded	exclude	VERB
fpq-14292	425	25	,	,	PUNCT
fpq-14292	425	26	and	and	CCONJ
fpq-14292	425	27	so	so	ADV
fpq-14292	425	28	the	the	DET
fpq-14292	425	29	models	model	NOUN
fpq-14292	425	30	are	be	AUX
fpq-14292	425	31	descriptively	descriptively	ADV
fpq-14292	425	32	inaccurate	inaccurate	ADJ
fpq-14292	425	33	at	at	ADP
fpq-14292	425	34	least	least	ADJ
fpq-14292	425	35	to	to	ADP
fpq-14292	425	36	that	that	DET
fpq-14292	425	37	extent	extent	NOUN
fpq-14292	425	38	,	,	PUNCT
fpq-14292	425	39	the	the	DET
fpq-14292	425	40	models	model	NOUN
fpq-14292	425	41	might	might	AUX
fpq-14292	425	42	still	still	ADV
fpq-14292	425	43	be	be	AUX
fpq-14292	425	44	descriptively	descriptively	ADV
fpq-14292	425	45	accurate	accurate	ADJ
fpq-14292	425	46	about	about	ADP
fpq-14292	425	47	how	how	SCONJ
fpq-14292	425	48	language	language	NOUN
fpq-14292	425	49	is	be	AUX
fpq-14292	425	50	used	use	VERB
fpq-14292	425	51	on	on	ADP
fpq-14292	425	52	those	those	DET
fpq-14292	425	53	platforms	platform	NOUN
fpq-14292	425	54	.	.	PUNCT
fpq-14292	426	1	bender	bender	NOUN
fpq-14292	426	2	/	/	SYM
fpq-14292	426	3	gebru	gebru	PROPN
fpq-14292	426	4	and	and	CCONJ
fpq-14292	426	5	colleagues	colleague	NOUN
fpq-14292	426	6	’	'	PUNCT
fpq-14292	426	7	response	response	NOUN
fpq-14292	426	8	might	might	AUX
fpq-14292	426	9	be	be	AUX
fpq-14292	426	10	that	that	SCONJ
fpq-14292	426	11	the	the	DET
fpq-14292	426	12	aim	aim	NOUN
fpq-14292	426	13	of	of	ADP
fpq-14292	426	14	descriptive	descriptive	ADJ
fpq-14292	426	15	accuracy	accuracy	NOUN
fpq-14292	426	16	is	be	AUX
fpq-14292	426	17	to	to	PART
fpq-14292	426	18	capture	capture	VERB
fpq-14292	426	19	patterns	pattern	NOUN
fpq-14292	426	20	that	that	PRON
fpq-14292	426	21	exist	exist	VERB
fpq-14292	426	22	in	in	ADP
fpq-14292	426	23	the	the	DET
fpq-14292	426	24	entire	entire	ADJ
fpq-14292	426	25	population	population	NOUN
fpq-14292	426	26	(	(	PUNCT
fpq-14292	426	27	or	or	CCONJ
fpq-14292	426	28	most	most	ADJ
fpq-14292	426	29	of	of	ADP
fpq-14292	426	30	it	it	PRON
fpq-14292	426	31	)	)	PUNCT
fpq-14292	426	32	,	,	PUNCT
fpq-14292	426	33	not	not	PART
fpq-14292	426	34	simply	simply	ADV
fpq-14292	426	35	a	a	DET
fpq-14292	426	36	subpopulation	subpopulation	NOUN
fpq-14292	426	37	.	.	PUNCT
fpq-14292	427	1	for	for	ADP
fpq-14292	427	2	large	large	ADJ
fpq-14292	427	3	language	language	NOUN
fpq-14292	427	4	models	model	NOUN
fpq-14292	427	5	,	,	PUNCT
fpq-14292	427	6	the	the	DET
fpq-14292	427	7	aim	aim	NOUN
fpq-14292	427	8	(	(	PUNCT
fpq-14292	427	9	in	in	ADP
fpq-14292	427	10	that	that	DET
fpq-14292	427	11	case	case	NOUN
fpq-14292	427	12	)	)	PUNCT
fpq-14292	427	13	would	would	AUX
fpq-14292	427	14	be	be	AUX
fpq-14292	427	15	that	that	SCONJ
fpq-14292	427	16	models	model	NOUN
fpq-14292	427	17	accurately	accurately	ADV
fpq-14292	427	18	reflect	reflect	VERB
fpq-14292	427	19	how	how	SCONJ
fpq-14292	427	20	humans	human	NOUN
fpq-14292	427	21	in	in	ADP
fpq-14292	427	22	general	general	ADJ
fpq-14292	427	23	,	,	PUNCT
fpq-14292	427	24	including	include	VERB
fpq-14292	427	25	those	those	PRON
fpq-14292	427	26	in	in	ADP
fpq-14292	427	27	marginalized	marginalize	VERB
fpq-14292	427	28	groups	group	NOUN
fpq-14292	427	29	,	,	PUNCT
fpq-14292	427	30	use	use	VERB
fpq-14292	427	31	language	language	NOUN
fpq-14292	427	32	.	.	PUNCT
fpq-14292	428	1	if	if	SCONJ
fpq-14292	428	2	we	we	PRON
fpq-14292	428	3	take	take	VERB
fpq-14292	428	4	language	language	NOUN
fpq-14292	428	5	from	from	ADP
fpq-14292	428	6	only	only	ADV
fpq-14292	428	7	a	a	DET
fpq-14292	428	8	specific	specific	ADJ
fpq-14292	428	9	subpopulation	subpopulation	NOUN
fpq-14292	428	10	,	,	PUNCT
fpq-14292	428	11	we	we	PRON
fpq-14292	428	12	can	can	AUX
fpq-14292	428	13	not	not	PART
fpq-14292	428	14	make	make	VERB
fpq-14292	428	15	claims	claim	NOUN
fpq-14292	428	16	about	about	ADP
fpq-14292	428	17	descriptive	descriptive	ADJ
fpq-14292	428	18	accuracy	accuracy	NOUN
fpq-14292	428	19	(	(	PUNCT
fpq-14292	428	20	except	except	SCONJ
fpq-14292	428	21	,	,	PUNCT
fpq-14292	428	22	perhaps	perhaps	ADV
fpq-14292	428	23	,	,	PUNCT
fpq-14292	428	24	regarding	regard	VERB
fpq-14292	428	25	that	that	DET
fpq-14292	428	26	specific	specific	ADJ
fpq-14292	428	27	subpopulation	subpopulation	NOUN
fpq-14292	428	28	’s	’s	PART
fpq-14292	428	29	use	use	NOUN
fpq-14292	428	30	of	of	ADP
fpq-14292	428	31	language	language	NOUN
fpq-14292	428	32	)	)	PUNCT
fpq-14292	428	33	.	.	PUNCT
fpq-14292	429	1	the	the	DET
fpq-14292	429	2	most	most	ADV
fpq-14292	429	3	obvious	obvious	ADJ
fpq-14292	429	4	bias	bias	NOUN
fpq-14292	429	5	here	here	ADV
fpq-14292	429	6	would	would	AUX
fpq-14292	429	7	seem	seem	VERB
fpq-14292	429	8	to	to	PART
fpq-14292	429	9	be	be	AUX
fpq-14292	429	10	representational	representational	ADJ
fpq-14292	429	11	bias	bias	NOUN
fpq-14292	429	12	,	,	PUNCT
fpq-14292	429	13	given	give	VERB
fpq-14292	429	14	that	that	SCONJ
fpq-14292	429	15	the	the	DET
fpq-14292	429	16	model	model	NOUN
fpq-14292	429	17	is	be	AUX
fpq-14292	429	18	only	only	ADV
fpq-14292	429	19	representative	representative	NOUN
fpq-14292	429	20	of	of	ADP
fpq-14292	429	21	a	a	DET
fpq-14292	429	22	specific	specific	ADJ
fpq-14292	429	23	subgroup	subgroup	NOUN
fpq-14292	429	24	while	while	SCONJ
fpq-14292	429	25	ignoring	ignore	VERB
fpq-14292	429	26	others	other	NOUN
fpq-14292	429	27	.	.	PUNCT
fpq-14292	430	1	even	even	ADV
fpq-14292	430	2	practically	practically	ADV
fpq-14292	430	3	,	,	PUNCT
fpq-14292	430	4	the	the	DET
fpq-14292	430	5	model	model	NOUN
fpq-14292	430	6	’s	’s	PART
fpq-14292	430	7	usefulness	usefulness	NOUN
fpq-14292	430	8	for	for	ADP
fpq-14292	430	9	downstream	downstream	ADJ
fpq-14292	430	10	tasks	task	NOUN
fpq-14292	430	11	on	on	ADP
fpq-14292	430	12	texts	text	NOUN
fpq-14292	430	13	written	write	VERB
fpq-14292	430	14	by	by	ADP
fpq-14292	430	15	other	other	ADJ
fpq-14292	430	16	language	language	NOUN
fpq-14292	430	17	users	user	NOUN
fpq-14292	430	18	will	will	AUX
fpq-14292	430	19	thus	thus	ADV
fpq-14292	430	20	be	be	AUX
fpq-14292	430	21	limited	limit	VERB
fpq-14292	430	22	.	.	PUNCT
fpq-14292	431	1	yet	yet	ADV
fpq-14292	431	2	suppose	suppose	VERB
fpq-14292	431	3	we	we	PRON
fpq-14292	431	4	trained	train	VERB
fpq-14292	431	5	a	a	DET
fpq-14292	431	6	hypothetical	hypothetical	ADJ
fpq-14292	431	7	model	model	NOUN
fpq-14292	431	8	on	on	ADP
fpq-14292	431	9	all	all	PRON
fpq-14292	431	10	of	of	ADP
fpq-14292	431	11	the	the	DET
fpq-14292	431	12	language	language	NOUN
fpq-14292	431	13	ever	ever	ADV
fpq-14292	431	14	produced	produce	VERB
fpq-14292	431	15	by	by	ADP
fpq-14292	431	16	people	people	NOUN
fpq-14292	431	17	across	across	ADP
fpq-14292	431	18	all	all	DET
fpq-14292	431	19	groups.34	groups.34	NOUN
fpq-14292	431	20	we	we	PRON
fpq-14292	431	21	might	might	AUX
fpq-14292	431	22	then	then	ADV
fpq-14292	431	23	have	have	VERB
fpq-14292	431	24	descriptive	descriptive	ADJ
fpq-14292	431	25	accuracy	accuracy	NOUN
fpq-14292	431	26	.	.	PUNCT
fpq-14292	432	1	even	even	ADV
fpq-14292	432	2	so	so	ADV
fpq-14292	432	3	,	,	PUNCT
fpq-14292	432	4	bias	bias	NOUN
fpq-14292	432	5	might	might	AUX
fpq-14292	432	6	remain	remain	VERB
fpq-14292	432	7	—	—	PUNCT
fpq-14292	432	8	not	not	PART
fpq-14292	432	9	representational	representational	ADJ
fpq-14292	432	10	bias	bias	NOUN
fpq-14292	432	11	but	but	CCONJ
fpq-14292	432	12	historical	historical	ADJ
fpq-14292	432	13	bias	bias	NOUN
fpq-14292	432	14	.	.	PUNCT
fpq-14292	433	1	in	in	ADP
fpq-14292	433	2	our	our	PRON
fpq-14292	433	3	terms	term	NOUN
fpq-14292	433	4	,	,	PUNCT
fpq-14292	433	5	we	we	PRON
fpq-14292	433	6	would	would	AUX
fpq-14292	433	7	have	have	VERB
fpq-14292	433	8	descriptive	descriptive	ADJ
fpq-14292	433	9	accuracy	accuracy	NOUN
fpq-14292	433	10	yet	yet	CCONJ
fpq-14292	433	11	normative	normative	ADJ
fpq-14292	433	12	incorrectness	incorrectness	NOUN
fpq-14292	433	13	(	(	PUNCT
fpq-14292	433	14	and	and	CCONJ
fpq-14292	433	15	the	the	DET
fpq-14292	433	16	normative	normative	ADJ
fpq-14292	433	17	incorrectness	incorrectness	NOUN
fpq-14292	433	18	would	would	AUX
fpq-14292	433	19	be	be	AUX
fpq-14292	433	20	due	due	ADJ
fpq-14292	433	21	to	to	ADP
fpq-14292	433	22	the	the	DET
fpq-14292	433	23	descriptive	descriptive	ADJ
fpq-14292	433	24	accuracy	accuracy	NOUN
fpq-14292	433	25	)	)	PUNCT
fpq-14292	433	26	.	.	PUNCT
fpq-14292	434	1	for	for	ADP
fpq-14292	434	2	example	example	NOUN
fpq-14292	434	3	,	,	PUNCT
fpq-14292	434	4	we	we	PRON
fpq-14292	434	5	would	would	AUX
fpq-14292	434	6	still	still	ADV
fpq-14292	434	7	have	have	VERB
fpq-14292	434	8	the	the	DET
fpq-14292	434	9	problem	problem	NOUN
fpq-14292	434	10	of	of	ADP
fpq-14292	434	11	gendered	gendere	VERB
fpq-14292	434	12	pronouns	pronoun	NOUN
fpq-14292	434	13	in	in	ADP
fpq-14292	434	14	translation	translation	NOUN
fpq-14292	434	15	.	.	PUNCT
fpq-14292	435	1	if	if	SCONJ
fpq-14292	435	2	people	people	NOUN
fpq-14292	435	3	speak	speak	VERB
fpq-14292	435	4	as	as	SCONJ
fpq-14292	435	5	though	though	SCONJ
fpq-14292	435	6	there	there	PRON
fpq-14292	435	7	are	be	VERB
fpq-14292	435	8	more	more	ADV
fpq-14292	435	9	female	female	ADJ
fpq-14292	435	10	than	than	ADP
fpq-14292	435	11	male	male	ADJ
fpq-14292	435	12	nurses	nurse	NOUN
fpq-14292	435	13	,	,	PUNCT
fpq-14292	435	14	the	the	DET
fpq-14292	435	15	pronoun	pronoun	NOUN
fpq-14292	435	16	“	"	PUNCT
fpq-14292	435	17	she	she	PRON
fpq-14292	435	18	”	"	PUNCT
fpq-14292	435	19	will	will	AUX
fpq-14292	435	20	still	still	ADV
fpq-14292	435	21	show	show	VERB
fpq-14292	435	22	up	up	ADP
fpq-14292	435	23	more	more	ADV
fpq-14292	435	24	frequently	frequently	ADV
fpq-14292	435	25	near	near	ADP
fpq-14292	435	26	“	"	PUNCT
fpq-14292	435	27	nurse	nurse	NOUN
fpq-14292	435	28	”	"	PUNCT
fpq-14292	435	29	than	than	ADP
fpq-14292	435	30	“	"	PUNCT
fpq-14292	435	31	he	he	PRON
fpq-14292	435	32	”	"	PUNCT
fpq-14292	435	33	in	in	ADP
fpq-14292	435	34	the	the	DET
fpq-14292	435	35	relevant	relevant	ADJ
fpq-14292	435	36	vector	vector	NOUN
fpq-14292	435	37	space	space	NOUN
fpq-14292	435	38	,	,	PUNCT
fpq-14292	435	39	even	even	ADV
fpq-14292	435	40	across	across	ADP
fpq-14292	435	41	the	the	DET
fpq-14292	435	42	enormous	enormous	ADJ
fpq-14292	435	43	dataset	dataset	NOUN
fpq-14292	435	44	of	of	ADP
fpq-14292	435	45	all	all	DET
fpq-14292	435	46	the	the	DET
fpq-14292	435	47	language	language	NOUN
fpq-14292	435	48	ever	ever	ADV
fpq-14292	435	49	produced	produce	VERB
fpq-14292	435	50	by	by	ADP
fpq-14292	435	51	all	all	DET
fpq-14292	435	52	groups	group	NOUN
fpq-14292	435	53	.	.	PUNCT
fpq-14292	436	1	likewise	likewise	ADV
fpq-14292	436	2	,	,	PUNCT
fpq-14292	436	3	if	if	SCONJ
fpq-14292	436	4	people	people	NOUN
fpq-14292	436	5	speak	speak	VERB
fpq-14292	436	6	as	as	SCONJ
fpq-14292	436	7	though	though	SCONJ
fpq-14292	436	8	there	there	PRON
fpq-14292	436	9	are	be	VERB
fpq-14292	436	10	more	more	ADV
fpq-14292	436	11	male	male	ADJ
fpq-14292	436	12	than	than	ADP
fpq-14292	436	13	female	female	ADJ
fpq-14292	436	14	programmers	programmer	NOUN
fpq-14292	436	15	,	,	PUNCT
fpq-14292	436	16	“	"	PUNCT
fpq-14292	436	17	he	he	PRON
fpq-14292	436	18	”	"	PUNCT
fpq-14292	436	19	will	will	AUX
fpq-14292	436	20	show	show	VERB
fpq-14292	436	21	up	up	ADP
fpq-14292	436	22	more	more	ADV
fpq-14292	436	23	frequently	frequently	ADV
fpq-14292	436	24	near	near	ADP
fpq-14292	436	25	“	"	PUNCT
fpq-14292	436	26	computer	computer	NOUN
fpq-14292	436	27	programmer	programmer	NOUN
fpq-14292	436	28	”	"	PUNCT
fpq-14292	436	29	than	than	ADP
fpq-14292	436	30	“	"	PUNCT
fpq-14292	436	31	she	she	PRON
fpq-14292	436	32	.	.	PUNCT
fpq-14292	436	33	”	"	PUNCT
fpq-14292	437	1	this	this	DET
fpq-14292	437	2	bias	bias	NOUN
fpq-14292	437	3	comes	come	VERB
fpq-14292	437	4	from	from	ADP
fpq-14292	437	5	descriptive	descriptive	ADJ
fpq-14292	437	6	,	,	PUNCT
fpq-14292	437	7	historical	historical	ADJ
fpq-14292	437	8	facts	fact	NOUN
fpq-14292	437	9	about	about	ADP
fpq-14292	437	10	language	language	NOUN
fpq-14292	437	11	use	use	NOUN
fpq-14292	437	12	,	,	PUNCT
fpq-14292	437	13	not	not	PART
fpq-14292	437	14	from	from	ADP
fpq-14292	437	15	how	how	SCONJ
fpq-14292	437	16	the	the	DET
fpq-14292	437	17	dataset	dataset	NOUN
fpq-14292	437	18	was	be	AUX
fpq-14292	437	19	built	build	VERB
fpq-14292	437	20	.	.	PUNCT
fpq-14292	438	1	33	33	NUM
fpq-14292	438	2	bender	bender	NOUN
fpq-14292	438	3	,	,	PUNCT
fpq-14292	438	4	gebru	gebru	VERB
fpq-14292	438	5	et	et	PROPN
fpq-14292	438	6	al	al	PROPN
fpq-14292	438	7	.	.	PROPN
fpq-14292	438	8	(	(	PUNCT
fpq-14292	438	9	2021	2021	NUM
fpq-14292	438	10	)	)	PUNCT
fpq-14292	438	11	thereby	thereby	ADV
fpq-14292	438	12	seem	seem	VERB
fpq-14292	438	13	to	to	PART
fpq-14292	438	14	be	be	AUX
fpq-14292	438	15	saying	say	VERB
fpq-14292	438	16	that	that	SCONJ
fpq-14292	438	17	normative	normative	ADJ
fpq-14292	438	18	incorrectness	incorrectness	NOUN
fpq-14292	438	19	can	can	AUX
fpq-14292	438	20	result	result	VERB
fpq-14292	438	21	from	from	ADP
fpq-14292	438	22	descriptive	descriptive	ADJ
fpq-14292	438	23	inaccuracy	inaccuracy	NOUN
fpq-14292	438	24	.	.	PUNCT
fpq-14292	439	1	we	we	PRON
fpq-14292	439	2	do	do	AUX
fpq-14292	439	3	not	not	PART
fpq-14292	439	4	deny	deny	VERB
fpq-14292	439	5	this	this	PRON
fpq-14292	439	6	,	,	PUNCT
fpq-14292	439	7	although	although	SCONJ
fpq-14292	439	8	our	our	PRON
fpq-14292	439	9	focus	focus	NOUN
fpq-14292	439	10	is	be	AUX
fpq-14292	439	11	instead	instead	ADV
fpq-14292	439	12	on	on	ADP
fpq-14292	439	13	how	how	SCONJ
fpq-14292	439	14	descriptive	descriptive	ADJ
fpq-14292	439	15	accuracy	accuracy	NOUN
fpq-14292	439	16	can	can	AUX
fpq-14292	439	17	conflict	conflict	VERB
fpq-14292	439	18	with	with	ADP
fpq-14292	439	19	normative	normative	ADJ
fpq-14292	439	20	accuracy	accuracy	NOUN
fpq-14292	439	21	—	—	PUNCT
fpq-14292	439	22	i.e.	i.e.	X
fpq-14292	439	23	,	,	PUNCT
fpq-14292	439	24	how	how	SCONJ
fpq-14292	439	25	normative	normative	ADJ
fpq-14292	439	26	incorrectness	incorrectness	NOUN
fpq-14292	439	27	can	can	AUX
fpq-14292	439	28	result	result	VERB
fpq-14292	439	29	from	from	ADP
fpq-14292	439	30	descriptive	descriptive	ADJ
fpq-14292	439	31	accuracy	accuracy	NOUN
fpq-14292	439	32	.	.	PUNCT
fpq-14292	440	1	34	34	NUM
fpq-14292	440	2	falbo	falbo	PROPN
fpq-14292	440	3	and	and	CCONJ
fpq-14292	440	4	lacroix	lacroix	PROPN
fpq-14292	440	5	(	(	PUNCT
fpq-14292	440	6	2022	2022	NUM
fpq-14292	440	7	)	)	PUNCT
fpq-14292	440	8	note	note	VERB
fpq-14292	440	9	that	that	SCONJ
fpq-14292	440	10	minority	minority	NOUN
fpq-14292	440	11	groups	group	NOUN
fpq-14292	440	12	often	often	ADV
fpq-14292	440	13	“	"	PUNCT
fpq-14292	440	14	code	code	NOUN
fpq-14292	440	15	-	-	PUNCT
fpq-14292	440	16	switch	switch	NOUN
fpq-14292	440	17	”	"	PUNCT
fpq-14292	440	18	by	by	ADP
fpq-14292	440	19	conforming	conform	VERB
fpq-14292	440	20	to	to	ADP
fpq-14292	440	21	majority	majority	NOUN
fpq-14292	440	22	norms	norm	NOUN
fpq-14292	440	23	of	of	ADP
fpq-14292	440	24	expression	expression	NOUN
fpq-14292	440	25	;	;	PUNCT
fpq-14292	440	26	thus	thus	ADV
fpq-14292	440	27	,	,	PUNCT
fpq-14292	440	28	the	the	DET
fpq-14292	440	29	content	content	NOUN
fpq-14292	440	30	they	they	PRON
fpq-14292	440	31	do	do	AUX
fpq-14292	440	32	generate	generate	VERB
fpq-14292	440	33	might	might	AUX
fpq-14292	440	34	still	still	ADV
fpq-14292	440	35	not	not	PART
fpq-14292	440	36	reflect	reflect	VERB
fpq-14292	440	37	marginalized	marginalized	ADJ
fpq-14292	440	38	voices	voice	NOUN
fpq-14292	440	39	.	.	PUNCT
fpq-14292	441	1	feminist	feminist	ADJ
fpq-14292	441	2	philosophy	philosophy	PROPN
fpq-14292	441	3	quarterly	quarterly	ADV
fpq-14292	441	4	,	,	PUNCT
fpq-14292	441	5	2022	2022	NUM
fpq-14292	441	6	,	,	PUNCT
fpq-14292	441	7	vol	vol	NOUN
fpq-14292	441	8	.	.	NOUN
fpq-14292	441	9	8	8	NUM
fpq-14292	441	10	,	,	PUNCT
fpq-14292	441	11	iss	iss	PROPN
fpq-14292	441	12	.	.	PUNCT
fpq-14292	441	13	3/4	3/4	NUM
fpq-14292	441	14	,	,	PUNCT
fpq-14292	441	15	article	article	NOUN
fpq-14292	441	16	1	1	NUM
fpq-14292	441	17	published	publish	VERB
fpq-14292	441	18	by	by	ADP
fpq-14292	441	19	scholarship@western	scholarship@western	PROPN
fpq-14292	441	20	,	,	PUNCT
fpq-14292	441	21	2022	2022	NUM
fpq-14292	441	22	24	24	NUM
fpq-14292	441	23	the	the	DET
fpq-14292	441	24	distinction	distinction	NOUN
fpq-14292	441	25	between	between	ADP
fpq-14292	441	26	representation	representation	NOUN
fpq-14292	441	27	bias	bias	NOUN
fpq-14292	441	28	and	and	CCONJ
fpq-14292	441	29	historical	historical	ADJ
fpq-14292	441	30	bias	bias	NOUN
fpq-14292	441	31	is	be	AUX
fpq-14292	441	32	also	also	ADV
fpq-14292	441	33	important	important	ADJ
fpq-14292	441	34	when	when	SCONJ
fpq-14292	441	35	we	we	PRON
fpq-14292	441	36	consider	consider	VERB
fpq-14292	441	37	whether	whether	SCONJ
fpq-14292	441	38	there	there	PRON
fpq-14292	441	39	can	can	AUX
fpq-14292	441	40	be	be	AUX
fpq-14292	441	41	a	a	DET
fpq-14292	441	42	value	value	NOUN
fpq-14292	441	43	to	to	ADP
fpq-14292	441	44	not	not	PART
fpq-14292	441	45	debiasing	debiase	VERB
fpq-14292	441	46	models	model	NOUN
fpq-14292	441	47	.	.	PUNCT
fpq-14292	442	1	historical	historical	ADJ
fpq-14292	442	2	bias	bias	NOUN
fpq-14292	442	3	often	often	ADV
fpq-14292	442	4	indicates	indicate	VERB
fpq-14292	442	5	something	something	PRON
fpq-14292	442	6	interesting	interesting	ADJ
fpq-14292	442	7	about	about	ADP
fpq-14292	442	8	the	the	DET
fpq-14292	442	9	world	world	NOUN
fpq-14292	442	10	,	,	PUNCT
fpq-14292	442	11	whereas	whereas	SCONJ
fpq-14292	442	12	representational	representational	ADJ
fpq-14292	442	13	bias	bias	NOUN
fpq-14292	442	14	might	might	AUX
fpq-14292	442	15	—	—	PUNCT
fpq-14292	442	16	in	in	ADP
fpq-14292	442	17	the	the	DET
fpq-14292	442	18	worst	bad	ADJ
fpq-14292	442	19	cases	case	NOUN
fpq-14292	442	20	—	—	PUNCT
fpq-14292	442	21	only	only	ADV
fpq-14292	442	22	indicate	indicate	VERB
fpq-14292	442	23	lazy	lazy	ADJ
fpq-14292	442	24	data	data	NOUN
fpq-14292	442	25	-	-	PUNCT
fpq-14292	442	26	gathering	gathering	NOUN
fpq-14292	442	27	.	.	PUNCT
fpq-14292	443	1	the	the	DET
fpq-14292	443	2	more	more	ADV
fpq-14292	443	3	descriptively	descriptively	ADV
fpq-14292	443	4	accurate	accurate	ADJ
fpq-14292	443	5	a	a	DET
fpq-14292	443	6	model	model	NOUN
fpq-14292	443	7	is	be	AUX
fpq-14292	443	8	,	,	PUNCT
fpq-14292	443	9	the	the	PRON
fpq-14292	443	10	more	more	ADV
fpq-14292	443	11	useful	useful	ADJ
fpq-14292	443	12	it	it	PRON
fpq-14292	443	13	will	will	AUX
fpq-14292	443	14	be	be	AUX
fpq-14292	443	15	for	for	ADP
fpq-14292	443	16	downstream	downstream	ADJ
fpq-14292	443	17	tasks	task	NOUN
fpq-14292	443	18	and	and	CCONJ
fpq-14292	443	19	the	the	DET
fpq-14292	443	20	more	more	ADV
fpq-14292	443	21	accurately	accurately	ADV
fpq-14292	443	22	it	it	PRON
fpq-14292	443	23	will	will	AUX
fpq-14292	443	24	reflect	reflect	VERB
fpq-14292	443	25	historical	historical	ADJ
fpq-14292	443	26	(	(	PUNCT
fpq-14292	443	27	and	and	CCONJ
fpq-14292	443	28	current	current	ADJ
fpq-14292	443	29	)	)	PUNCT
fpq-14292	443	30	biases	bias	NOUN
fpq-14292	443	31	in	in	ADP
fpq-14292	443	32	language	language	NOUN
fpq-14292	443	33	use	use	NOUN
fpq-14292	443	34	.	.	PUNCT
fpq-14292	444	1	the	the	DET
fpq-14292	444	2	choice	choice	NOUN
fpq-14292	444	3	to	to	ADP
fpq-14292	444	4	debias	debia	NOUN
fpq-14292	444	5	must	must	AUX
fpq-14292	444	6	start	start	VERB
fpq-14292	444	7	with	with	ADP
fpq-14292	444	8	a	a	DET
fpq-14292	444	9	descriptively	descriptively	ADV
fpq-14292	444	10	accurate	accurate	ADJ
fpq-14292	444	11	model	model	NOUN
fpq-14292	444	12	.	.	PUNCT
fpq-14292	445	1	in	in	ADP
fpq-14292	445	2	that	that	DET
fpq-14292	445	3	case	case	NOUN
fpq-14292	445	4	,	,	PUNCT
fpq-14292	445	5	we	we	PRON
fpq-14292	445	6	encounter	encounter	VERB
fpq-14292	445	7	the	the	DET
fpq-14292	445	8	dilemma	dilemma	NOUN
fpq-14292	445	9	:	:	PUNCT
fpq-14292	445	10	either	either	CCONJ
fpq-14292	445	11	we	we	PRON
fpq-14292	445	12	debias	debia	VERB
fpq-14292	445	13	the	the	DET
fpq-14292	445	14	accurate	accurate	ADJ
fpq-14292	445	15	model	model	NOUN
fpq-14292	445	16	and	and	CCONJ
fpq-14292	445	17	avoid	avoid	VERB
fpq-14292	445	18	amplifying	amplifying	ADJ
fpq-14292	445	19	bias	bias	NOUN
fpq-14292	445	20	yet	yet	ADV
fpq-14292	445	21	lose	lose	VERB
fpq-14292	445	22	accuracy	accuracy	NOUN
fpq-14292	445	23	and	and	CCONJ
fpq-14292	445	24	ethically	ethically	ADV
fpq-14292	445	25	useful	useful	ADJ
fpq-14292	445	26	information	information	NOUN
fpq-14292	445	27	,	,	PUNCT
fpq-14292	445	28	or	or	CCONJ
fpq-14292	445	29	we	we	PRON
fpq-14292	445	30	do	do	AUX
fpq-14292	445	31	not	not	PART
fpq-14292	445	32	debias	debia	VERB
fpq-14292	445	33	and	and	CCONJ
fpq-14292	445	34	we	we	PRON
fpq-14292	445	35	use	use	VERB
fpq-14292	445	36	the	the	DET
fpq-14292	445	37	model	model	NOUN
fpq-14292	445	38	to	to	PART
fpq-14292	445	39	learn	learn	VERB
fpq-14292	445	40	about	about	ADP
fpq-14292	445	41	biases	bias	NOUN
fpq-14292	445	42	,	,	PUNCT
fpq-14292	445	43	for	for	ADP
fpq-14292	445	44	example	example	NOUN
fpq-14292	445	45	,	,	PUNCT
fpq-14292	445	46	but	but	CCONJ
fpq-14292	445	47	risk	risk	VERB
fpq-14292	445	48	amplifying	amplify	VERB
fpq-14292	445	49	bias	bias	NOUN
fpq-14292	445	50	.	.	PUNCT
fpq-14292	446	1	9	9	X
fpq-14292	446	2	.	.	X
fpq-14292	446	3	conclusion	conclusion	NOUN
fpq-14292	446	4	what	what	PRON
fpq-14292	446	5	should	should	AUX
fpq-14292	446	6	we	we	PRON
fpq-14292	446	7	do	do	VERB
fpq-14292	446	8	about	about	ADP
fpq-14292	446	9	bias	bias	NOUN
fpq-14292	446	10	in	in	ADP
fpq-14292	446	11	nlp	nlp	NOUN
fpq-14292	446	12	systems	system	NOUN
fpq-14292	446	13	?	?	PUNCT
fpq-14292	447	1	it	it	PRON
fpq-14292	447	2	depends	depend	VERB
fpq-14292	447	3	,	,	PUNCT
fpq-14292	447	4	of	of	ADP
fpq-14292	447	5	course	course	NOUN
fpq-14292	447	6	,	,	PUNCT
fpq-14292	447	7	on	on	ADP
fpq-14292	447	8	the	the	DET
fpq-14292	447	9	source	source	NOUN
fpq-14292	447	10	of	of	ADP
fpq-14292	447	11	such	such	ADJ
fpq-14292	447	12	bias	bias	NOUN
fpq-14292	447	13	.	.	PUNCT
fpq-14292	448	1	but	but	CCONJ
fpq-14292	448	2	even	even	ADV
fpq-14292	448	3	for	for	ADP
fpq-14292	448	4	bias	bias	NOUN
fpq-14292	448	5	from	from	ADP
fpq-14292	448	6	properly	properly	ADV
fpq-14292	448	7	sampled	sample	VERB
fpq-14292	448	8	data	datum	NOUN
fpq-14292	448	9	,	,	PUNCT
fpq-14292	448	10	it	it	PRON
fpq-14292	448	11	also	also	ADV
fpq-14292	448	12	depends	depend	VERB
fpq-14292	448	13	on	on	ADP
fpq-14292	448	14	what	what	PRON
fpq-14292	448	15	we	we	PRON
fpq-14292	448	16	want	want	VERB
fpq-14292	448	17	from	from	ADP
fpq-14292	448	18	such	such	ADJ
fpq-14292	448	19	systems	system	NOUN
fpq-14292	448	20	.	.	PUNCT
fpq-14292	449	1	do	do	AUX
fpq-14292	449	2	we	we	PRON
fpq-14292	449	3	want	want	VERB
fpq-14292	449	4	them	they	PRON
fpq-14292	449	5	to	to	PART
fpq-14292	449	6	be	be	AUX
fpq-14292	449	7	more	more	ADV
fpq-14292	449	8	like	like	ADP
fpq-14292	449	9	cameras	camera	NOUN
fpq-14292	449	10	or	or	CCONJ
fpq-14292	449	11	more	more	ADJ
fpq-14292	449	12	like	like	ADP
fpq-14292	449	13	people	people	NOUN
fpq-14292	449	14	?	?	PUNCT
fpq-14292	450	1	we	we	PRON
fpq-14292	450	2	expect	expect	VERB
fpq-14292	450	3	cameras	camera	NOUN
fpq-14292	450	4	to	to	PART
fpq-14292	450	5	get	get	VERB
fpq-14292	450	6	things	thing	NOUN
fpq-14292	450	7	descriptively	descriptively	ADV
fpq-14292	450	8	right	right	ADV
fpq-14292	450	9	in	in	ADP
fpq-14292	450	10	certain	certain	ADJ
fpq-14292	450	11	relevant	relevant	ADJ
fpq-14292	450	12	regards	regard	NOUN
fpq-14292	450	13	,	,	PUNCT
fpq-14292	450	14	but	but	CCONJ
fpq-14292	450	15	not	not	PART
fpq-14292	450	16	normatively	normatively	ADV
fpq-14292	450	17	right	right	ADJ
fpq-14292	450	18	.	.	PUNCT
fpq-14292	451	1	yet	yet	ADV
fpq-14292	451	2	we	we	PRON
fpq-14292	451	3	expect	expect	VERB
fpq-14292	451	4	people	people	NOUN
fpq-14292	451	5	to	to	PART
fpq-14292	451	6	get	get	VERB
fpq-14292	451	7	things	thing	NOUN
fpq-14292	451	8	normatively	normatively	ADV
fpq-14292	451	9	right	right	ADJ
fpq-14292	451	10	,	,	PUNCT
fpq-14292	451	11	even	even	ADV
fpq-14292	451	12	if	if	SCONJ
fpq-14292	451	13	they	they	PRON
fpq-14292	451	14	get	get	VERB
fpq-14292	451	15	things	thing	NOUN
fpq-14292	451	16	descriptively	descriptively	ADV
fpq-14292	451	17	wrong	wrong	ADJ
fpq-14292	451	18	.	.	PUNCT
fpq-14292	452	1	for	for	ADP
fpq-14292	452	2	example	example	NOUN
fpq-14292	452	3	,	,	PUNCT
fpq-14292	452	4	we	we	PRON
fpq-14292	452	5	do	do	AUX
fpq-14292	452	6	not	not	PART
fpq-14292	452	7	excuse	excuse	VERB
fpq-14292	452	8	someone	someone	PRON
fpq-14292	452	9	for	for	ADP
fpq-14292	452	10	making	make	VERB
fpq-14292	452	11	a	a	DET
fpq-14292	452	12	sexist	sexist	ADJ
fpq-14292	452	13	comment	comment	NOUN
fpq-14292	452	14	even	even	ADV
fpq-14292	452	15	if	if	SCONJ
fpq-14292	452	16	their	their	PRON
fpq-14292	452	17	making	make	VERB
fpq-14292	452	18	that	that	DET
fpq-14292	452	19	comment	comment	NOUN
fpq-14292	452	20	is	be	AUX
fpq-14292	452	21	a	a	DET
fpq-14292	452	22	descriptively	descriptively	ADV
fpq-14292	452	23	accurate	accurate	ADJ
fpq-14292	452	24	reflection	reflection	NOUN
fpq-14292	452	25	of	of	ADP
fpq-14292	452	26	how	how	SCONJ
fpq-14292	452	27	language	language	NOUN
fpq-14292	452	28	is	be	AUX
fpq-14292	452	29	actually	actually	ADV
fpq-14292	452	30	used	use	VERB
fpq-14292	452	31	in	in	ADP
fpq-14292	452	32	their	their	PRON
fpq-14292	452	33	language	language	NOUN
fpq-14292	452	34	community	community	NOUN
fpq-14292	452	35	—	—	PUNCT
fpq-14292	452	36	we	we	PRON
fpq-14292	452	37	want	want	VERB
fpq-14292	452	38	the	the	DET
fpq-14292	452	39	person	person	NOUN
fpq-14292	452	40	to	to	PART
fpq-14292	452	41	get	get	VERB
fpq-14292	452	42	things	thing	NOUN
fpq-14292	452	43	normatively	normatively	ADV
fpq-14292	452	44	right	right	ADJ
fpq-14292	452	45	,	,	PUNCT
fpq-14292	452	46	not	not	PART
fpq-14292	452	47	descriptively	descriptively	ADV
fpq-14292	452	48	right	right	ADJ
fpq-14292	452	49	.	.	PUNCT
fpq-14292	453	1	should	should	AUX
fpq-14292	453	2	we	we	PRON
fpq-14292	453	3	expect	expect	VERB
fpq-14292	453	4	nlp	nlp	NOUN
fpq-14292	453	5	systems	system	NOUN
fpq-14292	453	6	to	to	PART
fpq-14292	453	7	get	get	VERB
fpq-14292	453	8	things	thing	NOUN
fpq-14292	453	9	normatively	normatively	ADV
fpq-14292	453	10	right	right	ADJ
fpq-14292	453	11	,	,	PUNCT
fpq-14292	453	12	even	even	ADV
fpq-14292	453	13	if	if	SCONJ
fpq-14292	453	14	they	they	PRON
fpq-14292	453	15	get	get	VERB
fpq-14292	453	16	things	thing	NOUN
fpq-14292	453	17	descriptively	descriptively	ADV
fpq-14292	453	18	wrong	wrong	ADJ
fpq-14292	453	19	?	?	PUNCT
fpq-14292	454	1	such	such	ADJ
fpq-14292	454	2	systems	system	NOUN
fpq-14292	454	3	could	could	AUX
fpq-14292	454	4	work	work	VERB
fpq-14292	454	5	like	like	ADP
fpq-14292	454	6	the	the	DET
fpq-14292	454	7	iblur	iblur	ADJ
fpq-14292	454	8	camera	camera	NOUN
fpq-14292	454	9	that	that	PRON
fpq-14292	454	10	we	we	PRON
fpq-14292	454	11	described	describe	VERB
fpq-14292	454	12	in	in	ADP
fpq-14292	454	13	section	section	NOUN
fpq-14292	454	14	3	3	NUM
fpq-14292	454	15	,	,	PUNCT
fpq-14292	454	16	or	or	CCONJ
fpq-14292	454	17	like	like	ADP
fpq-14292	454	18	the	the	DET
fpq-14292	454	19	speech	speech	NOUN
fpq-14292	454	20	-	-	PUNCT
fpq-14292	454	21	filter	filter	NOUN
fpq-14292	454	22	in	in	ADP
fpq-14292	454	23	the	the	DET
fpq-14292	454	24	tv	tv	NOUN
fpq-14292	454	25	show	show	VERB
fpq-14292	454	26	the	the	DET
fpq-14292	454	27	good	good	ADJ
fpq-14292	454	28	place	place	NOUN
fpq-14292	454	29	,	,	PUNCT
fpq-14292	454	30	where	where	SCONJ
fpq-14292	454	31	eleanor	eleanor	NOUN
fpq-14292	454	32	says	say	VERB
fpq-14292	454	33	things	thing	NOUN
fpq-14292	454	34	like	like	ADP
fpq-14292	454	35	“	"	PUNCT
fpq-14292	454	36	holy	holy	ADJ
fpq-14292	454	37	motherforking	motherforking	ADJ
fpq-14292	454	38	shirtballs	shirtball	NOUN
fpq-14292	454	39	!	!	PUNCT
fpq-14292	454	40	”	"	PUNCT
fpq-14292	455	1	despite	despite	SCONJ
fpq-14292	455	2	presumably	presumably	ADV
fpq-14292	455	3	trying	try	VERB
fpq-14292	455	4	to	to	PART
fpq-14292	455	5	say	say	VERB
fpq-14292	455	6	something	something	PRON
fpq-14292	455	7	else	else	ADV
fpq-14292	455	8	.	.	PUNCT
fpq-14292	456	1	this	this	DET
fpq-14292	456	2	filter	filter	NOUN
fpq-14292	456	3	“	"	PUNCT
fpq-14292	456	4	translates	translate	VERB
fpq-14292	456	5	out	out	ADP
fpq-14292	456	6	”	"	PUNCT
fpq-14292	456	7	things	thing	NOUN
fpq-14292	456	8	that	that	PRON
fpq-14292	456	9	we	we	PRON
fpq-14292	456	10	do	do	AUX
fpq-14292	456	11	not	not	PART
fpq-14292	456	12	want	want	VERB
fpq-14292	456	13	to	to	PART
fpq-14292	456	14	hear	hear	VERB
fpq-14292	456	15	or	or	CCONJ
fpq-14292	456	16	are	be	AUX
fpq-14292	456	17	not	not	PART
fpq-14292	456	18	permitted	permit	VERB
fpq-14292	456	19	to	to	PART
fpq-14292	456	20	hear	hear	VERB
fpq-14292	456	21	.	.	PUNCT
fpq-14292	457	1	alternatively	alternatively	ADV
fpq-14292	457	2	,	,	PUNCT
fpq-14292	457	3	is	be	AUX
fpq-14292	457	4	it	it	PRON
fpq-14292	457	5	enough	enough	ADJ
fpq-14292	457	6	for	for	SCONJ
fpq-14292	457	7	nlp	nlp	ADJ
fpq-14292	457	8	systems	system	NOUN
fpq-14292	457	9	to	to	PART
fpq-14292	457	10	be	be	AUX
fpq-14292	457	11	like	like	ADP
fpq-14292	457	12	cameras	camera	NOUN
fpq-14292	457	13	in	in	ADP
fpq-14292	457	14	getting	get	VERB
fpq-14292	457	15	things	thing	NOUN
fpq-14292	457	16	descriptively	descriptively	ADV
fpq-14292	457	17	right	right	ADJ
fpq-14292	457	18	,	,	PUNCT
fpq-14292	457	19	so	so	SCONJ
fpq-14292	457	20	that	that	SCONJ
fpq-14292	457	21	they	they	PRON
fpq-14292	457	22	reflect	reflect	VERB
fpq-14292	457	23	the	the	DET
fpq-14292	457	24	actual	actual	ADJ
fpq-14292	457	25	world	world	NOUN
fpq-14292	457	26	and	and	CCONJ
fpq-14292	457	27	let	let	VERB
fpq-14292	457	28	us	we	PRON
fpq-14292	457	29	decide	decide	VERB
fpq-14292	457	30	what	what	PRON
fpq-14292	457	31	to	to	PART
fpq-14292	457	32	do	do	VERB
fpq-14292	457	33	about	about	ADP
fpq-14292	457	34	it	it	PRON
fpq-14292	457	35	in	in	ADP
fpq-14292	457	36	particular	particular	ADJ
fpq-14292	457	37	contexts	contexts	NOUN
fpq-14292	457	38	?	?	PUNCT
fpq-14292	458	1	we	we	PRON
fpq-14292	458	2	maintain	maintain	VERB
fpq-14292	458	3	that	that	SCONJ
fpq-14292	458	4	there	there	PRON
fpq-14292	458	5	can	can	AUX
fpq-14292	458	6	be	be	AUX
fpq-14292	458	7	a	a	DET
fpq-14292	458	8	value	value	NOUN
fpq-14292	458	9	to	to	ADP
fpq-14292	458	10	having	have	VERB
fpq-14292	458	11	nlp	nlp	NOUN
fpq-14292	458	12	systems	system	NOUN
fpq-14292	458	13	function	function	VERB
fpq-14292	458	14	more	more	ADV
fpq-14292	458	15	like	like	ADP
fpq-14292	458	16	cameras	camera	NOUN
fpq-14292	458	17	.	.	PUNCT
fpq-14292	459	1	for	for	ADP
fpq-14292	459	2	one	one	NUM
fpq-14292	459	3	thing	thing	NOUN
fpq-14292	459	4	,	,	PUNCT
fpq-14292	459	5	it	it	PRON
fpq-14292	459	6	can	can	AUX
fpq-14292	459	7	provide	provide	VERB
fpq-14292	459	8	ethically	ethically	ADV
fpq-14292	459	9	useful	useful	ADJ
fpq-14292	459	10	information	information	NOUN
fpq-14292	459	11	,	,	PUNCT
fpq-14292	459	12	which	which	PRON
fpq-14292	459	13	we	we	PRON
fpq-14292	459	14	may	may	AUX
fpq-14292	459	15	even	even	ADV
fpq-14292	459	16	have	have	VERB
fpq-14292	459	17	an	an	DET
fpq-14292	459	18	ethical	ethical	ADJ
fpq-14292	459	19	obligation	obligation	NOUN
fpq-14292	459	20	to	to	PART
fpq-14292	459	21	provide	provide	VERB
fpq-14292	459	22	.	.	PUNCT
fpq-14292	460	1	then	then	ADV
fpq-14292	460	2	we	we	PRON
fpq-14292	460	3	face	face	VERB
fpq-14292	460	4	a	a	DET
fpq-14292	460	5	dilemma	dilemma	NOUN
fpq-14292	460	6	,	,	PUNCT
fpq-14292	460	7	since	since	SCONJ
fpq-14292	460	8	allowing	allow	VERB
fpq-14292	460	9	a	a	DET
fpq-14292	460	10	system	system	NOUN
fpq-14292	460	11	to	to	PART
fpq-14292	460	12	get	get	VERB
fpq-14292	460	13	things	thing	NOUN
fpq-14292	460	14	descriptively	descriptively	ADV
fpq-14292	460	15	right	right	ADJ
fpq-14292	460	16	by	by	ADP
fpq-14292	460	17	not	not	PART
fpq-14292	460	18	debiasing	debiase	VERB
fpq-14292	460	19	it	it	PRON
fpq-14292	460	20	can	can	AUX
fpq-14292	460	21	perpetuate	perpetuate	VERB
fpq-14292	460	22	or	or	CCONJ
fpq-14292	460	23	amplify	amplify	VERB
fpq-14292	460	24	bias	bias	NOUN
fpq-14292	460	25	.	.	PUNCT
fpq-14292	461	1	we	we	PRON
fpq-14292	461	2	think	think	VERB
fpq-14292	461	3	that	that	SCONJ
fpq-14292	461	4	deciding	decide	VERB
fpq-14292	461	5	which	which	DET
fpq-14292	461	6	way	way	NOUN
fpq-14292	461	7	to	to	PART
fpq-14292	461	8	go	go	VERB
fpq-14292	461	9	can	can	AUX
fpq-14292	461	10	only	only	ADV
fpq-14292	461	11	be	be	AUX
fpq-14292	461	12	done	do	VERB
fpq-14292	461	13	on	on	ADP
fpq-14292	461	14	a	a	DET
fpq-14292	461	15	case	case	NOUN
fpq-14292	461	16	-	-	PUNCT
fpq-14292	461	17	by	by	ADP
fpq-14292	461	18	-	-	PUNCT
fpq-14292	461	19	case	case	NOUN
fpq-14292	461	20	basis	basis	NOUN
fpq-14292	461	21	,	,	PUNCT
fpq-14292	461	22	by	by	ADP
fpq-14292	461	23	weighing	weigh	VERB
fpq-14292	461	24	the	the	DET
fpq-14292	461	25	ethical	ethical	ADJ
fpq-14292	461	26	costs	cost	NOUN
fpq-14292	461	27	and	and	CCONJ
fpq-14292	461	28	benefits	benefit	NOUN
fpq-14292	461	29	.	.	PUNCT
fpq-14292	462	1	our	our	PRON
fpq-14292	462	2	primary	primary	ADJ
fpq-14292	462	3	aim	aim	NOUN
fpq-14292	462	4	has	have	AUX
fpq-14292	462	5	been	be	AUX
fpq-14292	462	6	to	to	PART
fpq-14292	462	7	draw	draw	VERB
fpq-14292	462	8	attention	attention	NOUN
fpq-14292	462	9	to	to	ADP
fpq-14292	462	10	the	the	DET
fpq-14292	462	11	dilemma	dilemma	NOUN
fpq-14292	462	12	,	,	PUNCT
fpq-14292	462	13	which	which	PRON
fpq-14292	462	14	arises	arise	VERB
fpq-14292	462	15	once	once	SCONJ
fpq-14292	462	16	we	we	PRON
fpq-14292	462	17	acknowledge	acknowledge	VERB
fpq-14292	462	18	that	that	SCONJ
fpq-14292	462	19	there	there	PRON
fpq-14292	462	20	can	can	AUX
fpq-14292	462	21	often	often	ADV
fpq-14292	462	22	be	be	AUX
fpq-14292	462	23	a	a	DET
fpq-14292	462	24	value	value	NOUN
fpq-14292	462	25	,	,	PUNCT
fpq-14292	462	26	however	however	ADV
fpq-14292	462	27	counterintuitive	counterintuitive	ADJ
fpq-14292	462	28	,	,	PUNCT
fpq-14292	462	29	to	to	ADP
fpq-14292	462	30	not	not	PART
fpq-14292	462	31	debiasing	debiase	VERB
fpq-14292	462	32	.	.	PUNCT
fpq-14292	463	1	we	we	PRON
fpq-14292	463	2	have	have	AUX
fpq-14292	463	3	also	also	ADV
fpq-14292	463	4	argued	argue	VERB
fpq-14292	463	5	that	that	SCONJ
fpq-14292	463	6	it	it	PRON
fpq-14292	463	7	is	be	AUX
fpq-14292	463	8	frequently	frequently	ADV
fpq-14292	463	9	best	good	ADJ
fpq-14292	463	10	to	to	PART
fpq-14292	463	11	control	control	VERB
fpq-14292	463	12	for	for	ADP
fpq-14292	463	13	the	the	DET
fpq-14292	463	14	negative	negative	ADJ
fpq-14292	463	15	effects	effect	NOUN
fpq-14292	463	16	of	of	ADP
fpq-14292	463	17	bias	bias	NOUN
fpq-14292	463	18	at	at	ADP
fpq-14292	463	19	the	the	DET
fpq-14292	463	20	output	output	NOUN
fpq-14292	463	21	stage	stage	NOUN
fpq-14292	463	22	rather	rather	ADV
fpq-14292	463	23	than	than	ADP
fpq-14292	463	24	by	by	ADP
fpq-14292	463	25	debiasing	debiase	VERB
fpq-14292	463	26	models	model	NOUN
fpq-14292	463	27	themselves	themselves	PRON
fpq-14292	463	28	.	.	PUNCT
fpq-14292	464	1	yet	yet	ADV
fpq-14292	464	2	deery	deery	VERB
fpq-14292	464	3	and	and	CCONJ
fpq-14292	464	4	bailey	bailey	PROPN
fpq-14292	464	5	–	–	PUNCT
fpq-14292	464	6	the	the	DET
fpq-14292	464	7	bias	bias	NOUN
fpq-14292	464	8	dilemma	dilemma	NOUN
fpq-14292	464	9	:	:	PUNCT
fpq-14292	464	10	the	the	DET
fpq-14292	464	11	ethics	ethic	NOUN
fpq-14292	464	12	of	of	ADP
fpq-14292	464	13	algorithmic	algorithmic	ADJ
fpq-14292	464	14	bias	bias	NOUN
fpq-14292	464	15	in	in	ADP
fpq-14292	464	16	natural	natural	ADJ
fpq-14292	464	17	-	-	PUNCT
fpq-14292	464	18	language	language	NOUN
fpq-14292	464	19	processing	processing	NOUN
fpq-14292	464	20	published	publish	VERB
fpq-14292	464	21	by	by	ADP
fpq-14292	464	22	scholarship@western	scholarship@western	PROPN
fpq-14292	464	23	,	,	PUNCT
fpq-14292	464	24	2022	2022	NUM
fpq-14292	464	25	25	25	NUM
fpq-14292	464	26	even	even	ADV
fpq-14292	464	27	then	then	ADV
fpq-14292	464	28	,	,	PUNCT
fpq-14292	464	29	we	we	PRON
fpq-14292	464	30	can	can	AUX
fpq-14292	464	31	not	not	PART
fpq-14292	464	32	easily	easily	ADV
fpq-14292	464	33	avoid	avoid	VERB
fpq-14292	464	34	the	the	DET
fpq-14292	464	35	dilemma	dilemma	NOUN
fpq-14292	464	36	,	,	PUNCT
fpq-14292	464	37	since	since	SCONJ
fpq-14292	464	38	we	we	PRON
fpq-14292	464	39	must	must	AUX
fpq-14292	464	40	still	still	ADV
fpq-14292	464	41	decide	decide	VERB
fpq-14292	464	42	whether	whether	SCONJ
fpq-14292	464	43	normative	normative	ADJ
fpq-14292	464	44	correctness	correctness	NOUN
fpq-14292	464	45	should	should	AUX
fpq-14292	464	46	overrule	overrule	VERB
fpq-14292	464	47	descriptive	descriptive	ADJ
fpq-14292	464	48	accuracy	accuracy	NOUN
fpq-14292	464	49	in	in	ADP
fpq-14292	464	50	cases	case	NOUN
fpq-14292	464	51	where	where	SCONJ
fpq-14292	464	52	descriptive	descriptive	ADJ
fpq-14292	464	53	accuracy	accuracy	NOUN
fpq-14292	464	54	provides	provide	VERB
fpq-14292	464	55	us	we	PRON
fpq-14292	464	56	with	with	ADP
fpq-14292	464	57	ethically	ethically	ADV
fpq-14292	464	58	useful	useful	ADJ
fpq-14292	464	59	information	information	NOUN
fpq-14292	464	60	.	.	PUNCT
fpq-14292	465	1	finally	finally	ADV
fpq-14292	465	2	,	,	PUNCT
fpq-14292	465	3	our	our	PRON
fpq-14292	465	4	distinction	distinction	NOUN
fpq-14292	465	5	between	between	ADP
fpq-14292	465	6	normative	normative	ADJ
fpq-14292	465	7	correctness	correctness	NOUN
fpq-14292	465	8	and	and	CCONJ
fpq-14292	465	9	descriptive	descriptive	ADJ
fpq-14292	465	10	accuracy	accuracy	NOUN
fpq-14292	465	11	provides	provide	VERB
fpq-14292	465	12	help	help	NOUN
fpq-14292	465	13	in	in	ADP
fpq-14292	465	14	better	well	ADJ
fpq-14292	465	15	understanding	understand	VERB
fpq-14292	465	16	recent	recent	ADJ
fpq-14292	465	17	debates	debate	NOUN
fpq-14292	465	18	about	about	ADP
fpq-14292	465	19	algorithmic	algorithmic	ADJ
fpq-14292	465	20	bias	bias	NOUN
fpq-14292	465	21	in	in	ADP
fpq-14292	465	22	natural	natural	ADJ
fpq-14292	465	23	-	-	PUNCT
fpq-14292	465	24	language	language	NOUN
fpq-14292	465	25	processing	processing	NOUN
fpq-14292	465	26	.	.	PUNCT
fpq-14292	466	1	in	in	ADP
fpq-14292	466	2	particular	particular	ADJ
fpq-14292	466	3	,	,	PUNCT
fpq-14292	466	4	the	the	DET
fpq-14292	466	5	question	question	NOUN
fpq-14292	466	6	of	of	ADP
fpq-14292	466	7	whether	whether	SCONJ
fpq-14292	466	8	to	to	PART
fpq-14292	466	9	debias	debias	PROPN
fpq-14292	466	10	helps	help	VERB
fpq-14292	466	11	us	we	PRON
fpq-14292	466	12	to	to	PART
fpq-14292	466	13	better	well	ADV
fpq-14292	466	14	understand	understand	VERB
fpq-14292	466	15	these	these	DET
fpq-14292	466	16	debates	debate	NOUN
fpq-14292	466	17	by	by	ADP
fpq-14292	466	18	bringing	bring	VERB
fpq-14292	466	19	the	the	DET
fpq-14292	466	20	consequences	consequence	NOUN
fpq-14292	466	21	of	of	ADP
fpq-14292	466	22	adopting	adopt	VERB
fpq-14292	466	23	one	one	NUM
fpq-14292	466	24	view	view	NOUN
fpq-14292	466	25	over	over	ADP
fpq-14292	466	26	another	another	PRON
fpq-14292	466	27	into	into	ADP
fpq-14292	466	28	clearer	clear	ADJ
fpq-14292	466	29	focus	focus	NOUN
fpq-14292	466	30	.	.	PUNCT
fpq-14292	467	1	references	reference	NOUN
fpq-14292	467	2	allen	allen	PROPN
fpq-14292	467	3	,	,	PUNCT
fpq-14292	467	4	antoine	antoine	PROPN
fpq-14292	467	5	.	.	PUNCT
fpq-14292	468	1	2016	2016	NUM
fpq-14292	468	2	.	.	PUNCT
fpq-14292	469	1	“	"	PUNCT
fpq-14292	469	2	the	the	DET
fpq-14292	469	3	‘	'	PUNCT
fpq-14292	469	4	three	three	NUM
fpq-14292	469	5	black	black	ADJ
fpq-14292	469	6	teenagers	teenager	NOUN
fpq-14292	469	7	’	'	PUNCT
fpq-14292	469	8	search	search	NOUN
fpq-14292	469	9	shows	show	VERB
fpq-14292	469	10	it	it	PRON
fpq-14292	469	11	is	be	AUX
fpq-14292	469	12	society	society	NOUN
fpq-14292	469	13	,	,	PUNCT
fpq-14292	469	14	not	not	PART
fpq-14292	469	15	google	google	NOUN
fpq-14292	469	16	,	,	PUNCT
fpq-14292	469	17	that	that	PRON
fpq-14292	469	18	is	be	AUX
fpq-14292	469	19	racist	racist	ADJ
fpq-14292	469	20	.	.	PUNCT
fpq-14292	469	21	”	"	PUNCT
fpq-14292	470	1	guardian	guardian	PROPN
fpq-14292	470	2	,	,	PUNCT
fpq-14292	470	3	june	june	PROPN
fpq-14292	470	4	10	10	NUM
fpq-14292	470	5	,	,	PUNCT
fpq-14292	470	6	2016	2016	NUM
fpq-14292	470	7	,	,	PUNCT
fpq-14292	470	8	https://www.theguardian	https://www.theguardian	PROPN
fpq-14292	470	9	.com	.com	PROPN
fpq-14292	470	10	/	/	SYM
fpq-14292	470	11	commentisfree/2016	commentisfree/2016	PROPN
fpq-14292	470	12	/	/	SYM
fpq-14292	470	13	jun/10	jun/10	ADJ
fpq-14292	470	14	/	/	SYM
fpq-14292	470	15	three	three	NUM
fpq-14292	470	16	-	-	PUNCT
fpq-14292	470	17	black	black	ADJ
fpq-14292	470	18	-	-	PUNCT
fpq-14292	470	19	teenagers	teenager	NOUN
fpq-14292	470	20	-	-	PUNCT
fpq-14292	470	21	google	google	NOUN
fpq-14292	470	22	-	-	PUNCT
fpq-14292	470	23	racisttweet	racisttweet	NOUN
fpq-14292	470	24	.	.	PUNCT
fpq-14292	471	1	barber	barber	PROPN
fpq-14292	471	2	,	,	PUNCT
fpq-14292	471	3	anjuli	anjuli	PROPN
fpq-14292	471	4	l.	l.	PROPN
fpq-14292	471	5	a.	a.	PROPN
fpq-14292	471	6	,	,	PUNCT
fpq-14292	471	7	daniel	daniel	PROPN
fpq-14292	471	8	mills	mills	PROPN
fpq-14292	471	9	s.	s.	PROPN
fpq-14292	471	10	,	,	PUNCT
fpq-14292	471	11	fernando	fernando	PROPN
fpq-14292	471	12	montealegre	montealegre	PROPN
fpq-14292	471	13	-	-	PUNCT
fpq-14292	471	14	z	z	PROPN
fpq-14292	471	15	,	,	PUNCT
fpq-14292	471	16	victoria	victoria	PROPN
fpq-14292	471	17	f.	f.	PROPN
fpq-14292	471	18	ratcliffe	ratcliffe	PROPN
fpq-14292	471	19	,	,	PUNCT
fpq-14292	471	20	kun	kun	PROPN
fpq-14292	471	21	guo	guo	PROPN
fpq-14292	471	22	,	,	PUNCT
fpq-14292	471	23	and	and	CCONJ
fpq-14292	471	24	anna	anna	PROPN
fpq-14292	471	25	wilkinson	wilkinson	PROPN
fpq-14292	471	26	.	.	PUNCT
fpq-14292	471	27	2020	2020	NUM
fpq-14292	471	28	.	.	PUNCT
fpq-14292	472	1	“	"	PUNCT
fpq-14292	472	2	functional	functional	ADJ
fpq-14292	472	3	performance	performance	NOUN
fpq-14292	472	4	of	of	ADP
fpq-14292	472	5	the	the	DET
fpq-14292	472	6	visual	visual	ADJ
fpq-14292	472	7	system	system	NOUN
fpq-14292	472	8	in	in	ADP
fpq-14292	472	9	dogs	dog	NOUN
fpq-14292	472	10	and	and	CCONJ
fpq-14292	472	11	humans	human	NOUN
fpq-14292	472	12	:	:	PUNCT
fpq-14292	472	13	a	a	DET
fpq-14292	472	14	comparative	comparative	ADJ
fpq-14292	472	15	perspective	perspective	NOUN
fpq-14292	472	16	.	.	PUNCT
fpq-14292	472	17	”	"	PUNCT
fpq-14292	473	1	comparative	comparative	ADJ
fpq-14292	473	2	cognition	cognition	NOUN
fpq-14292	473	3	&	&	CCONJ
fpq-14292	473	4	behavior	behavior	NOUN
fpq-14292	473	5	reviews	review	VERB
fpq-14292	473	6	15:1–44	15:1–44	NUM
fpq-14292	473	7	.	.	PUNCT
fpq-14292	474	1	https://doi.org/10.3819/ccbr.2020.150002	https://doi.org/10.3819/ccbr.2020.150002	PROPN
fpq-14292	474	2	.	.	PUNCT
fpq-14292	475	1	basu	basu	PROPN
fpq-14292	475	2	,	,	PUNCT
fpq-14292	475	3	rima	rima	PROPN
fpq-14292	475	4	.	.	PUNCT
fpq-14292	476	1	2020	2020	NUM
fpq-14292	476	2	.	.	PUNCT
fpq-14292	477	1	“	"	PUNCT
fpq-14292	477	2	the	the	DET
fpq-14292	477	3	specter	specter	NOUN
fpq-14292	477	4	of	of	ADP
fpq-14292	477	5	normative	normative	ADJ
fpq-14292	477	6	conflict	conflict	NOUN
fpq-14292	477	7	?	?	PUNCT
fpq-14292	478	1	does	do	AUX
fpq-14292	478	2	fairness	fairness	NOUN
fpq-14292	478	3	require	require	VERB
fpq-14292	478	4	inaccuracy	inaccuracy	NOUN
fpq-14292	478	5	.	.	PUNCT
fpq-14292	478	6	”	"	PUNCT
fpq-14292	479	1	in	in	ADP
fpq-14292	479	2	an	an	DET
fpq-14292	479	3	introduction	introduction	NOUN
fpq-14292	479	4	to	to	ADP
fpq-14292	479	5	implicit	implicit	ADJ
fpq-14292	479	6	bias	bias	NOUN
fpq-14292	479	7	:	:	PUNCT
fpq-14292	479	8	knowledge	knowledge	NOUN
fpq-14292	479	9	,	,	PUNCT
fpq-14292	479	10	justice	justice	NOUN
fpq-14292	479	11	,	,	PUNCT
fpq-14292	479	12	and	and	CCONJ
fpq-14292	479	13	the	the	DET
fpq-14292	479	14	social	social	ADJ
fpq-14292	479	15	mind	mind	NOUN
fpq-14292	479	16	,	,	PUNCT
fpq-14292	479	17	edited	edit	VERB
fpq-14292	479	18	by	by	ADP
fpq-14292	479	19	erin	erin	PROPN
fpq-14292	479	20	beeghly	beeghly	PROPN
fpq-14292	479	21	and	and	CCONJ
fpq-14292	479	22	alex	alex	PROPN
fpq-14292	479	23	madva	madva	PROPN
fpq-14292	479	24	,	,	PUNCT
fpq-14292	479	25	191–210	191–210	NUM
fpq-14292	479	26	.	.	PUNCT
fpq-14292	480	1	new	new	PROPN
fpq-14292	480	2	york	york	PROPN
fpq-14292	480	3	:	:	PUNCT
fpq-14292	481	1	routledge	routledge	PROPN
fpq-14292	481	2	.	.	PUNCT
fpq-14292	482	1	bender	bender	PROPN
fpq-14292	482	2	,	,	PUNCT
fpq-14292	482	3	emily	emily	PROPN
fpq-14292	482	4	m.	m.	PROPN
fpq-14292	482	5	,	,	PUNCT
fpq-14292	482	6	timnit	timnit	NOUN
fpq-14292	482	7	gebru	gebru	PROPN
fpq-14292	482	8	,	,	PUNCT
fpq-14292	482	9	angelina	angelina	PROPN
fpq-14292	482	10	mcmillan	mcmillan	PROPN
fpq-14292	482	11	-	-	PUNCT
fpq-14292	482	12	major	major	PROPN
fpq-14292	482	13	,	,	PUNCT
fpq-14292	482	14	and	and	CCONJ
fpq-14292	482	15	shmargaret	shmargaret	PROPN
fpq-14292	482	16	shmitchell	shmitchell	PROPN
fpq-14292	482	17	.	.	PUNCT
fpq-14292	482	18	2021	2021	NUM
fpq-14292	482	19	.	.	PUNCT
fpq-14292	483	1	“	"	PUNCT
fpq-14292	483	2	on	on	ADP
fpq-14292	483	3	the	the	DET
fpq-14292	483	4	dangers	danger	NOUN
fpq-14292	483	5	of	of	ADP
fpq-14292	483	6	stochastic	stochastic	ADJ
fpq-14292	483	7	parrots	parrot	NOUN
fpq-14292	483	8	:	:	PUNCT
fpq-14292	483	9	can	can	AUX
fpq-14292	483	10	language	language	NOUN
fpq-14292	483	11	models	model	NOUN
fpq-14292	483	12	be	be	AUX
fpq-14292	483	13	too	too	ADV
fpq-14292	483	14	big	big	ADJ
fpq-14292	483	15	?	?	PUNCT
fpq-14292	483	16	”	"	PUNCT
fpq-14292	483	17	in	in	ADP
fpq-14292	483	18	facct	facct	ADJ
fpq-14292	483	19	’	'	PUNCT
fpq-14292	483	20	21	21	NUM
fpq-14292	483	21	:	:	PUNCT
fpq-14292	483	22	proceedings	proceeding	NOUN
fpq-14292	483	23	of	of	ADP
fpq-14292	483	24	the	the	DET
fpq-14292	483	25	2021	2021	NUM
fpq-14292	483	26	acm	acm	NOUN
fpq-14292	483	27	conference	conference	NOUN
fpq-14292	483	28	on	on	ADP
fpq-14292	483	29	fairness	fairness	NOUN
fpq-14292	483	30	,	,	PUNCT
fpq-14292	483	31	accountability	accountability	NOUN
fpq-14292	483	32	,	,	PUNCT
fpq-14292	483	33	and	and	CCONJ
fpq-14292	483	34	transparency	transparency	NOUN
fpq-14292	483	35	,	,	PUNCT
fpq-14292	483	36	610–23	610–23	PROPN
fpq-14292	483	37	.	.	PUNCT
fpq-14292	484	1	new	new	PROPN
fpq-14292	484	2	york	york	PROPN
fpq-14292	484	3	:	:	PUNCT
fpq-14292	484	4	association	association	NOUN
fpq-14292	484	5	for	for	ADP
fpq-14292	484	6	computing	compute	VERB
fpq-14292	484	7	machinery	machinery	NOUN
fpq-14292	484	8	.	.	PUNCT
fpq-14292	485	1	https://doi.org/10.1145/3442188.3445922	https://doi.org/10.1145/3442188.3445922	PROPN
fpq-14292	485	2	.	.	PUNCT
fpq-14292	486	1	benjamin	benjamin	PROPN
fpq-14292	486	2	,	,	PUNCT
fpq-14292	486	3	ruha	ruha	NOUN
fpq-14292	486	4	.	.	PUNCT
fpq-14292	486	5	2019	2019	NUM
fpq-14292	486	6	.	.	PUNCT
fpq-14292	487	1	race	race	NOUN
fpq-14292	487	2	after	after	ADP
fpq-14292	487	3	technology	technology	NOUN
fpq-14292	487	4	:	:	PUNCT
fpq-14292	487	5	abolitionist	abolitionist	NOUN
fpq-14292	487	6	tools	tool	NOUN
fpq-14292	487	7	for	for	ADP
fpq-14292	487	8	the	the	DET
fpq-14292	487	9	new	new	ADJ
fpq-14292	487	10	jim	jim	PROPN
fpq-14292	487	11	code	code	PROPN
fpq-14292	487	12	.	.	PUNCT
fpq-14292	488	1	medford	medford	PROPN
fpq-14292	488	2	,	,	PUNCT
fpq-14292	488	3	ma	ma	PROPN
fpq-14292	488	4	:	:	PUNCT
fpq-14292	488	5	polity	polity	NOUN
fpq-14292	488	6	press	press	NOUN
fpq-14292	488	7	.	.	PUNCT
fpq-14292	489	1	blodgett	blodgett	PROPN
fpq-14292	489	2	,	,	PUNCT
fpq-14292	489	3	su	su	PROPN
fpq-14292	489	4	lin	lin	PROPN
fpq-14292	489	5	,	,	PUNCT
fpq-14292	489	6	solon	solon	PROPN
fpq-14292	489	7	barocas	barocas	PROPN
fpq-14292	489	8	,	,	PUNCT
fpq-14292	489	9	hal	hal	PROPN
fpq-14292	489	10	daumé	daumé	PROPN
fpq-14292	489	11	iii	iii	PROPN
fpq-14292	489	12	,	,	PUNCT
fpq-14292	489	13	and	and	CCONJ
fpq-14292	489	14	hanna	hanna	PROPN
fpq-14292	489	15	wallach	wallach	PROPN
fpq-14292	489	16	.	.	PUNCT
fpq-14292	490	1	2020	2020	NUM
fpq-14292	490	2	.	.	PUNCT
fpq-14292	491	1	“	"	PUNCT
fpq-14292	491	2	language	language	NOUN
fpq-14292	491	3	(	(	PUNCT
fpq-14292	491	4	technology	technology	NOUN
fpq-14292	491	5	)	)	PUNCT
fpq-14292	491	6	is	be	AUX
fpq-14292	491	7	power	power	NOUN
fpq-14292	491	8	:	:	PUNCT
fpq-14292	491	9	a	a	DET
fpq-14292	491	10	critical	critical	ADJ
fpq-14292	491	11	survey	survey	NOUN
fpq-14292	491	12	of	of	ADP
fpq-14292	491	13	‘	'	PUNCT
fpq-14292	491	14	bias	bias	NOUN
fpq-14292	491	15	’	'	PUNCT
fpq-14292	491	16	in	in	ADP
fpq-14292	491	17	nlp	nlp	NOUN
fpq-14292	491	18	.	.	PUNCT
fpq-14292	491	19	”	"	PUNCT
fpq-14292	492	1	in	in	ADP
fpq-14292	492	2	proceedings	proceeding	NOUN
fpq-14292	492	3	of	of	ADP
fpq-14292	492	4	the	the	DET
fpq-14292	492	5	58th	58th	ADJ
fpq-14292	492	6	annual	annual	ADJ
fpq-14292	492	7	meeting	meeting	NOUN
fpq-14292	492	8	of	of	ADP
fpq-14292	492	9	the	the	DET
fpq-14292	492	10	association	association	NOUN
fpq-14292	492	11	for	for	ADP
fpq-14292	492	12	computational	computational	ADJ
fpq-14292	492	13	linguistics	linguistic	NOUN
fpq-14292	492	14	,	,	PUNCT
fpq-14292	492	15	edited	edit	VERB
fpq-14292	492	16	by	by	ADP
fpq-14292	492	17	dan	dan	PROPN
fpq-14292	492	18	jurafsky	jurafsky	PROPN
fpq-14292	492	19	,	,	PUNCT
fpq-14292	492	20	joyce	joyce	PROPN
fpq-14292	492	21	chai	chai	PROPN
fpq-14292	492	22	,	,	PUNCT
fpq-14292	492	23	natalie	natalie	PROPN
fpq-14292	492	24	schluter	schluter	PROPN
fpq-14292	492	25	,	,	PUNCT
fpq-14292	492	26	and	and	CCONJ
fpq-14292	492	27	joel	joel	PROPN
fpq-14292	492	28	tetreault	tetreault	PROPN
fpq-14292	492	29	,	,	PUNCT
fpq-14292	492	30	5454–76	5454–76	PROPN
fpq-14292	492	31	.	.	PUNCT
fpq-14292	492	32	stroudsburg	stroudsburg	PROPN
fpq-14292	492	33	,	,	PUNCT
fpq-14292	492	34	pa	pa	PROPN
fpq-14292	492	35	:	:	PUNCT
fpq-14292	492	36	association	association	NOUN
fpq-14292	492	37	for	for	ADP
fpq-14292	492	38	computational	computational	ADJ
fpq-14292	492	39	linguistics	linguistic	NOUN
fpq-14292	492	40	.	.	PUNCT
fpq-14292	493	1	https://doi.org/10	https://doi.org/10	PROPN
fpq-14292	493	2	.18653	.18653	PROPN
fpq-14292	493	3	/	/	SYM
fpq-14292	493	4	v1/2020.acl	v1/2020.acl	PROPN
fpq-14292	493	5	-	-	PUNCT
fpq-14292	493	6	main.485	main.485	NOUN
fpq-14292	493	7	.	.	PUNCT
fpq-14292	494	1	bolukbasi	bolukbasi	NOUN
fpq-14292	494	2	,	,	PUNCT
fpq-14292	494	3	tolga	tolga	PROPN
fpq-14292	494	4	,	,	PUNCT
fpq-14292	494	5	kai	kai	PROPN
fpq-14292	494	6	-	-	PUNCT
fpq-14292	494	7	wei	wei	PROPN
fpq-14292	494	8	chang	chang	PROPN
fpq-14292	494	9	,	,	PUNCT
fpq-14292	494	10	james	james	PROPN
fpq-14292	494	11	zou	zou	PROPN
fpq-14292	494	12	,	,	PUNCT
fpq-14292	494	13	venkatesh	venkatesh	NOUN
fpq-14292	494	14	saligrama	saligrama	NOUN
fpq-14292	494	15	,	,	PUNCT
fpq-14292	494	16	and	and	CCONJ
fpq-14292	494	17	adam	adam	PROPN
fpq-14292	494	18	kalai	kalai	PROPN
fpq-14292	494	19	.	.	PROPN
fpq-14292	494	20	2016	2016	NUM
fpq-14292	494	21	.	.	PUNCT
fpq-14292	495	1	“	"	PUNCT
fpq-14292	495	2	man	man	NOUN
fpq-14292	495	3	is	be	AUX
fpq-14292	495	4	to	to	PART
fpq-14292	495	5	computer	computer	NOUN
fpq-14292	495	6	programmer	programmer	NOUN
fpq-14292	495	7	as	as	SCONJ
fpq-14292	495	8	woman	woman	NOUN
fpq-14292	495	9	is	be	AUX
fpq-14292	495	10	to	to	ADP
fpq-14292	495	11	homemaker	homemaker	NOUN
fpq-14292	495	12	?	?	PUNCT
fpq-14292	496	1	debiasing	debiase	VERB
fpq-14292	496	2	word	word	NOUN
fpq-14292	496	3	embeddings	embedding	NOUN
fpq-14292	496	4	.	.	PUNCT
fpq-14292	496	5	”	"	PUNCT
fpq-14292	497	1	arxiv	arxiv	PROPN
fpq-14292	497	2	preprint	preprint	NOUN
fpq-14292	497	3	,	,	PUNCT
fpq-14292	497	4	arxiv:1607.06520	arxiv:1607.06520	X
fpq-14292	498	1	[	[	X
fpq-14292	498	2	cs.cl	cs.cl	PROPN
fpq-14292	498	3	]	]	X
fpq-14292	498	4	.	.	PUNCT
fpq-14292	499	1	https://doi.org/10.48550/arxiv.1607.06520	https://doi.org/10.48550/arxiv.1607.06520	PROPN
fpq-14292	499	2	.	.	PROPN
fpq-14292	499	3	feminist	feminist	PROPN
fpq-14292	499	4	philosophy	philosophy	PROPN
fpq-14292	499	5	quarterly	quarterly	ADV
fpq-14292	499	6	,	,	PUNCT
fpq-14292	499	7	2022	2022	NUM
fpq-14292	499	8	,	,	PUNCT
fpq-14292	499	9	vol	vol	NOUN
fpq-14292	499	10	.	.	NOUN
fpq-14292	499	11	8	8	NUM
fpq-14292	499	12	,	,	PUNCT
fpq-14292	499	13	iss	iss	PROPN
fpq-14292	499	14	.	.	PUNCT
fpq-14292	499	15	3/4	3/4	NUM
fpq-14292	499	16	,	,	PUNCT
fpq-14292	499	17	article	article	NOUN
fpq-14292	499	18	1	1	NUM
fpq-14292	499	19	published	publish	VERB
fpq-14292	499	20	by	by	ADP
fpq-14292	499	21	scholarship@western	scholarship@western	PROPN
fpq-14292	499	22	,	,	PUNCT
fpq-14292	499	23	2022	2022	NUM
fpq-14292	499	24	26	26	NUM
fpq-14292	499	25	box	box	NOUN
fpq-14292	499	26	,	,	PUNCT
fpq-14292	499	27	george	george	PROPN
fpq-14292	499	28	e.	e.	PROPN
fpq-14292	499	29	p.	p.	PROPN
fpq-14292	499	30	1976	1976	NUM
fpq-14292	499	31	.	.	PUNCT
fpq-14292	500	1	“	"	PUNCT
fpq-14292	500	2	science	science	NOUN
fpq-14292	500	3	and	and	CCONJ
fpq-14292	500	4	statistics	statistic	NOUN
fpq-14292	500	5	.	.	PUNCT
fpq-14292	500	6	”	"	PUNCT
fpq-14292	500	7	journal	journal	NOUN
fpq-14292	500	8	of	of	ADP
fpq-14292	500	9	the	the	DET
fpq-14292	500	10	american	american	PROPN
fpq-14292	500	11	statistical	statistical	ADJ
fpq-14292	500	12	association	association	NOUN
fpq-14292	500	13	71	71	NUM
fpq-14292	500	14	(	(	PUNCT
fpq-14292	500	15	356	356	NUM
fpq-14292	500	16	):	):	PUNCT
fpq-14292	500	17	791–99	791–99	NOUN
fpq-14292	500	18	.	.	PUNCT
fpq-14292	501	1	brownstein	brownstein	PROPN
fpq-14292	501	2	,	,	PUNCT
fpq-14292	501	3	michael	michael	PROPN
fpq-14292	501	4	.	.	PROPN
fpq-14292	501	5	2018	2018	NUM
fpq-14292	501	6	.	.	PUNCT
fpq-14292	502	1	the	the	DET
fpq-14292	502	2	implicit	implicit	ADJ
fpq-14292	502	3	mind	mind	NOUN
fpq-14292	502	4	:	:	PUNCT
fpq-14292	502	5	cognitive	cognitive	ADJ
fpq-14292	502	6	architecture	architecture	NOUN
fpq-14292	502	7	,	,	PUNCT
fpq-14292	502	8	the	the	DET
fpq-14292	502	9	self	self	NOUN
fpq-14292	502	10	,	,	PUNCT
fpq-14292	502	11	and	and	CCONJ
fpq-14292	502	12	ethics	ethic	NOUN
fpq-14292	502	13	.	.	PUNCT
fpq-14292	503	1	new	new	PROPN
fpq-14292	503	2	york	york	PROPN
fpq-14292	503	3	:	:	PUNCT
fpq-14292	503	4	oxford	oxford	PROPN
fpq-14292	503	5	university	university	PROPN
fpq-14292	503	6	press	press	NOUN
fpq-14292	503	7	.	.	PUNCT
fpq-14292	504	1	brownstein	brownstein	PROPN
fpq-14292	504	2	,	,	PUNCT
fpq-14292	504	3	michael	michael	PROPN
fpq-14292	504	4	,	,	PUNCT
fpq-14292	504	5	and	and	CCONJ
fpq-14292	504	6	jennifer	jennifer	PROPN
fpq-14292	504	7	saul	saul	PROPN
fpq-14292	504	8	,	,	PUNCT
fpq-14292	504	9	eds	eds	PROPN
fpq-14292	504	10	.	.	PROPN
fpq-14292	504	11	2016a	2016a	NOUN
fpq-14292	504	12	.	.	PUNCT
fpq-14292	505	1	implicit	implicit	ADJ
fpq-14292	505	2	bias	bias	NOUN
fpq-14292	505	3	and	and	CCONJ
fpq-14292	505	4	philosophy	philosophy	NOUN
fpq-14292	505	5	:	:	PUNCT
fpq-14292	505	6	volume	volume	NOUN
fpq-14292	505	7	i	i	PRON
fpq-14292	505	8	,	,	PUNCT
fpq-14292	505	9	metaphysics	metaphysics	PROPN
fpq-14292	505	10	and	and	CCONJ
fpq-14292	505	11	epistemology	epistemology	NOUN
fpq-14292	505	12	.	.	PUNCT
fpq-14292	506	1	oxford	oxford	NOUN
fpq-14292	506	2	:	:	PUNCT
fpq-14292	506	3	oxford	oxford	PROPN
fpq-14292	506	4	university	university	PROPN
fpq-14292	506	5	press	press	NOUN
fpq-14292	506	6	.	.	PUNCT
fpq-14292	507	1	—	—	PUNCT
fpq-14292	507	2	—	—	PUNCT
fpq-14292	507	3	—	—	PUNCT
fpq-14292	507	4	,	,	PUNCT
fpq-14292	507	5	eds	ed	NOUN
fpq-14292	507	6	.	.	NOUN
fpq-14292	507	7	2016b	2016b	NUM
fpq-14292	507	8	.	.	PUNCT
fpq-14292	508	1	implicit	implicit	ADJ
fpq-14292	508	2	bias	bias	NOUN
fpq-14292	508	3	and	and	CCONJ
fpq-14292	508	4	philosophy	philosophy	NOUN
fpq-14292	508	5	:	:	PUNCT
fpq-14292	508	6	volume	volume	NOUN
fpq-14292	508	7	2	2	NUM
fpq-14292	508	8	,	,	PUNCT
fpq-14292	508	9	moral	moral	ADJ
fpq-14292	508	10	responsibility	responsibility	NOUN
fpq-14292	508	11	,	,	PUNCT
fpq-14292	508	12	structural	structural	ADJ
fpq-14292	508	13	injustice	injustice	NOUN
fpq-14292	508	14	,	,	PUNCT
fpq-14292	508	15	and	and	CCONJ
fpq-14292	508	16	ethics	ethic	NOUN
fpq-14292	508	17	.	.	PUNCT
fpq-14292	509	1	oxford	oxford	NOUN
fpq-14292	509	2	:	:	PUNCT
fpq-14292	509	3	oxford	oxford	PROPN
fpq-14292	509	4	university	university	PROPN
fpq-14292	509	5	press	press	NOUN
fpq-14292	509	6	.	.	PUNCT
fpq-14292	510	1	caliskan	caliskan	PROPN
fpq-14292	510	2	,	,	PUNCT
fpq-14292	510	3	aylin	aylin	PROPN
fpq-14292	510	4	,	,	PUNCT
fpq-14292	510	5	joanna	joanna	PROPN
fpq-14292	510	6	j.	j.	PROPN
fpq-14292	510	7	bryson	bryson	PROPN
fpq-14292	510	8	,	,	PUNCT
fpq-14292	510	9	and	and	CCONJ
fpq-14292	510	10	arvind	arvind	PROPN
fpq-14292	510	11	narayanan	narayanan	PROPN
fpq-14292	510	12	.	.	PROPN
fpq-14292	510	13	2017	2017	NUM
fpq-14292	510	14	.	.	PUNCT
fpq-14292	511	1	“	"	PUNCT
fpq-14292	511	2	semantics	semantic	NOUN
fpq-14292	511	3	derived	derive	VERB
fpq-14292	511	4	automatically	automatically	ADV
fpq-14292	511	5	from	from	ADP
fpq-14292	511	6	language	language	NOUN
fpq-14292	511	7	corpora	corpora	NOUN
fpq-14292	511	8	contain	contain	VERB
fpq-14292	511	9	human	human	ADJ
fpq-14292	511	10	-	-	PUNCT
fpq-14292	511	11	like	like	ADJ
fpq-14292	511	12	biases	bias	NOUN
fpq-14292	511	13	.	.	PUNCT
fpq-14292	511	14	”	"	PUNCT
fpq-14292	512	1	science	science	NOUN
fpq-14292	512	2	356	356	NUM
fpq-14292	512	3	,	,	PUNCT
fpq-14292	512	4	no	no	INTJ
fpq-14292	512	5	.	.	NOUN
fpq-14292	512	6	6334	6334	NUM
fpq-14292	512	7	(	(	PUNCT
fpq-14292	512	8	april	april	PROPN
fpq-14292	512	9	14	14	NUM
fpq-14292	512	10	):	):	PUNCT
fpq-14292	512	11	183–86	183–86	NUM
fpq-14292	512	12	.	.	PUNCT
fpq-14292	513	1	https://doi.org/10.1126/science.aal4230	https://doi.org/10.1126/science.aal4230	PROPN
fpq-14292	513	2	.	.	PUNCT
fpq-14292	514	1	deery	deery	NOUN
fpq-14292	514	2	,	,	PUNCT
fpq-14292	514	3	oisín	oisín	NOUN
fpq-14292	514	4	.	.	PROPN
fpq-14292	514	5	2021	2021	NUM
fpq-14292	514	6	.	.	PUNCT
fpq-14292	515	1	naturally	naturally	ADV
fpq-14292	515	2	free	free	ADJ
fpq-14292	515	3	action	action	NOUN
fpq-14292	515	4	.	.	PUNCT
fpq-14292	516	1	oxford	oxford	NOUN
fpq-14292	516	2	:	:	PUNCT
fpq-14292	516	3	oxford	oxford	PROPN
fpq-14292	516	4	university	university	PROPN
fpq-14292	516	5	press	press	NOUN
fpq-14292	516	6	.	.	PUNCT
fpq-14292	517	1	devlin	devlin	PROPN
fpq-14292	517	2	,	,	PUNCT
fpq-14292	517	3	jacob	jacob	PROPN
fpq-14292	517	4	,	,	PUNCT
fpq-14292	517	5	ming	ming	PROPN
fpq-14292	517	6	-	-	PUNCT
fpq-14292	517	7	wei	wei	PROPN
fpq-14292	517	8	chang	chang	PROPN
fpq-14292	517	9	,	,	PUNCT
fpq-14292	517	10	kenton	kenton	PROPN
fpq-14292	517	11	lee	lee	PROPN
fpq-14292	517	12	,	,	PUNCT
fpq-14292	517	13	and	and	CCONJ
fpq-14292	517	14	kristina	kristina	PROPN
fpq-14292	517	15	toutanova	toutanova	PROPN
fpq-14292	517	16	.	.	PROPN
fpq-14292	517	17	2019	2019	NUM
fpq-14292	517	18	.	.	PUNCT
fpq-14292	518	1	“	"	PUNCT
fpq-14292	518	2	bert	bert	PROPN
fpq-14292	518	3	:	:	PUNCT
fpq-14292	518	4	pretraining	pretraine	VERB
fpq-14292	518	5	of	of	ADP
fpq-14292	518	6	deep	deep	ADJ
fpq-14292	518	7	bidirectional	bidirectional	ADJ
fpq-14292	518	8	transformers	transformer	NOUN
fpq-14292	518	9	for	for	ADP
fpq-14292	518	10	language	language	NOUN
fpq-14292	518	11	understanding	understanding	NOUN
fpq-14292	518	12	.	.	PUNCT
fpq-14292	518	13	”	"	PUNCT
fpq-14292	518	14	in	in	ADP
fpq-14292	518	15	proceedings	proceeding	NOUN
fpq-14292	518	16	of	of	ADP
fpq-14292	518	17	the	the	DET
fpq-14292	518	18	2019	2019	NUM
fpq-14292	518	19	conference	conference	NOUN
fpq-14292	518	20	of	of	ADP
fpq-14292	518	21	the	the	DET
fpq-14292	518	22	north	north	ADJ
fpq-14292	518	23	american	american	ADJ
fpq-14292	518	24	chapter	chapter	NOUN
fpq-14292	518	25	of	of	ADP
fpq-14292	518	26	the	the	DET
fpq-14292	518	27	association	association	NOUN
fpq-14292	518	28	for	for	ADP
fpq-14292	518	29	computational	computational	ADJ
fpq-14292	518	30	linguistics	linguistic	NOUN
fpq-14292	518	31	:	:	PUNCT
fpq-14292	518	32	human	human	ADJ
fpq-14292	518	33	language	language	NOUN
fpq-14292	518	34	technologies	technology	NOUN
fpq-14292	518	35	,	,	PUNCT
fpq-14292	518	36	volume	volume	NOUN
fpq-14292	518	37	1	1	NUM
fpq-14292	518	38	,	,	PUNCT
fpq-14292	518	39	edited	edit	VERB
fpq-14292	518	40	by	by	ADP
fpq-14292	518	41	jill	jill	PROPN
fpq-14292	518	42	burstein	burstein	PROPN
fpq-14292	518	43	,	,	PUNCT
fpq-14292	518	44	christy	christy	PROPN
fpq-14292	518	45	doran	doran	PROPN
fpq-14292	518	46	,	,	PUNCT
fpq-14292	518	47	and	and	CCONJ
fpq-14292	518	48	thamar	thamar	PROPN
fpq-14292	518	49	solorio	solorio	PROPN
fpq-14292	518	50	,	,	PUNCT
fpq-14292	518	51	4171–86	4171–86	PROPN
fpq-14292	518	52	.	.	PUNCT
fpq-14292	518	53	stroudsburg	stroudsburg	PROPN
fpq-14292	518	54	,	,	PUNCT
fpq-14292	518	55	pa	pa	PROPN
fpq-14292	518	56	:	:	PUNCT
fpq-14292	518	57	association	association	NOUN
fpq-14292	518	58	for	for	ADP
fpq-14292	518	59	computational	computational	ADJ
fpq-14292	518	60	linguistics	linguistic	NOUN
fpq-14292	518	61	.	.	PUNCT
fpq-14292	519	1	https://doi.org/10	https://doi.org/10	PROPN
fpq-14292	519	2	.18653	.18653	PROPN
fpq-14292	519	3	/	/	SYM
fpq-14292	519	4	v1	v1	NOUN
fpq-14292	519	5	/	/	SYM
fpq-14292	519	6	n19	n19	NOUN
fpq-14292	519	7	-	-	PUNCT
fpq-14292	519	8	1423	1423	NUM
fpq-14292	519	9	.	.	PUNCT
fpq-14292	520	1	falbo	falbo	PROPN
fpq-14292	520	2	,	,	PUNCT
fpq-14292	520	3	arianna	arianna	PROPN
fpq-14292	520	4	,	,	PUNCT
fpq-14292	520	5	and	and	CCONJ
fpq-14292	520	6	travis	travis	PROPN
fpq-14292	520	7	lacroix	lacroix	PROPN
fpq-14292	520	8	.	.	PUNCT
fpq-14292	520	9	2022	2022	NUM
fpq-14292	520	10	.	.	PUNCT
fpq-14292	521	1	“	"	PUNCT
fpq-14292	521	2	est	est	X
fpq-14292	521	3	-	-	PUNCT
fpq-14292	521	4	ce	ce	PROPN
fpq-14292	521	5	que	que	PROPN
fpq-14292	521	6	vous	vous	PROPN
fpq-14292	521	7	compute	compute	PROPN
fpq-14292	521	8	?	?	PUNCT
fpq-14292	521	9	code	code	NOUN
fpq-14292	521	10	-	-	PUNCT
fpq-14292	521	11	switching	switching	NOUN
fpq-14292	521	12	,	,	PUNCT
fpq-14292	521	13	cultural	cultural	ADJ
fpq-14292	521	14	identity	identity	NOUN
fpq-14292	521	15	,	,	PUNCT
fpq-14292	521	16	and	and	CCONJ
fpq-14292	521	17	ai	ai	VERB
fpq-14292	521	18	.	.	PUNCT
fpq-14292	521	19	”	"	PUNCT
fpq-14292	522	1	feminist	feminist	ADJ
fpq-14292	522	2	philosophy	philosophy	NOUN
fpq-14292	522	3	quarterly	quarterly	ADJ
fpq-14292	522	4	8	8	NUM
fpq-14292	522	5	(	(	PUNCT
fpq-14292	522	6	3/4	3/4	NUM
fpq-14292	522	7	):	):	PUNCT
fpq-14292	522	8	article	article	NOUN
fpq-14292	522	9	9	9	NUM
fpq-14292	522	10	.	.	PUNCT
fpq-14292	522	11	fazelpour	fazelpour	ADJ
fpq-14292	522	12	,	,	PUNCT
fpq-14292	522	13	sina	sina	PROPN
fpq-14292	522	14	,	,	PUNCT
fpq-14292	522	15	and	and	CCONJ
fpq-14292	522	16	david	david	PROPN
fpq-14292	522	17	danks	danks	PROPN
fpq-14292	522	18	.	.	PUNCT
fpq-14292	522	19	2021	2021	NUM
fpq-14292	522	20	.	.	PUNCT
fpq-14292	523	1	“	"	PUNCT
fpq-14292	523	2	algorithmic	algorithmic	ADJ
fpq-14292	523	3	bias	bias	NOUN
fpq-14292	523	4	:	:	PUNCT
fpq-14292	523	5	senses	sense	NOUN
fpq-14292	523	6	,	,	PUNCT
fpq-14292	523	7	sources	source	NOUN
fpq-14292	523	8	,	,	PUNCT
fpq-14292	523	9	solutions	solution	NOUN
fpq-14292	523	10	.	.	PUNCT
fpq-14292	523	11	”	"	PUNCT
fpq-14292	524	1	philosophy	philosophy	NOUN
fpq-14292	524	2	compass	compass	NOUN
fpq-14292	524	3	16	16	NUM
fpq-14292	524	4	,	,	PUNCT
fpq-14292	524	5	no	no	INTJ
fpq-14292	524	6	.	.	NOUN
fpq-14292	524	7	8	8	NUM
fpq-14292	524	8	(	(	PUNCT
fpq-14292	524	9	august	august	PROPN
fpq-14292	524	10	):	):	PUNCT
fpq-14292	524	11	e12760	e12760	PROPN
fpq-14292	524	12	.	.	PUNCT
fpq-14292	525	1	https://doi.org/10	https://doi.org/10	PROPN
fpq-14292	525	2	.1111	.1111	PROPN
fpq-14292	525	3	/	/	SYM
fpq-14292	525	4	phc3.12760	phc3.12760	PROPN
fpq-14292	525	5	.	.	PUNCT
fpq-14292	526	1	gendler	gendler	PROPN
fpq-14292	526	2	,	,	PUNCT
fpq-14292	526	3	tamar	tamar	PROPN
fpq-14292	526	4	szabó	szabó	PROPN
fpq-14292	526	5	.	.	PUNCT
fpq-14292	527	1	2011	2011	NUM
fpq-14292	527	2	.	.	PUNCT
fpq-14292	528	1	“	"	PUNCT
fpq-14292	528	2	on	on	ADP
fpq-14292	528	3	the	the	DET
fpq-14292	528	4	epistemic	epistemic	ADJ
fpq-14292	528	5	costs	cost	NOUN
fpq-14292	528	6	of	of	ADP
fpq-14292	528	7	implicit	implicit	ADJ
fpq-14292	528	8	bias	bias	NOUN
fpq-14292	528	9	.	.	PUNCT
fpq-14292	528	10	”	"	PUNCT
fpq-14292	529	1	philosophical	philosophical	ADJ
fpq-14292	529	2	studies	study	NOUN
fpq-14292	529	3	156	156	NUM
fpq-14292	529	4	,	,	PUNCT
fpq-14292	529	5	no	no	INTJ
fpq-14292	529	6	.	.	NOUN
fpq-14292	529	7	1	1	NUM
fpq-14292	529	8	(	(	PUNCT
fpq-14292	529	9	october	october	PROPN
fpq-14292	529	10	):	):	PUNCT
fpq-14292	529	11	33–63	33–63	NUM
fpq-14292	529	12	.	.	PUNCT
fpq-14292	530	1	goldberg	goldberg	PROPN
fpq-14292	530	2	,	,	PUNCT
fpq-14292	530	3	yoav	yoav	PROPN
fpq-14292	530	4	.	.	PROPN
fpq-14292	530	5	2021	2021	NUM
fpq-14292	530	6	.	.	PUNCT
fpq-14292	531	1	“	"	PUNCT
fpq-14292	531	2	a	a	DET
fpq-14292	531	3	criticism	criticism	NOUN
fpq-14292	531	4	of	of	ADP
fpq-14292	531	5	‘	'	PUNCT
fpq-14292	531	6	on	on	ADP
fpq-14292	531	7	the	the	DET
fpq-14292	531	8	dangers	danger	NOUN
fpq-14292	531	9	of	of	ADP
fpq-14292	531	10	stochastic	stochastic	ADJ
fpq-14292	531	11	parrots	parrot	NOUN
fpq-14292	531	12	:	:	PUNCT
fpq-14292	531	13	can	can	AUX
fpq-14292	531	14	language	language	NOUN
fpq-14292	531	15	models	model	NOUN
fpq-14292	531	16	be	be	AUX
fpq-14292	531	17	too	too	ADV
fpq-14292	531	18	big	big	ADJ
fpq-14292	531	19	?	?	PUNCT
fpq-14292	531	20	’	'	PUNCT
fpq-14292	531	21	”	"	PUNCT
fpq-14292	531	22	github	github	PROPN
fpq-14292	531	23	post	post	PROPN
fpq-14292	531	24	,	,	PUNCT
fpq-14292	531	25	january	january	PROPN
fpq-14292	531	26	23	23	NUM
fpq-14292	531	27	,	,	PUNCT
fpq-14292	531	28	2021	2021	NUM
fpq-14292	531	29	.	.	PUNCT
fpq-14292	532	1	https://gist.git	https://gist.git	PROPN
fpq-14292	532	2	hub.com/yoavg/9fc9be2f98b47c189a513573d902fb27	hub.com/yoavg/9fc9be2f98b47c189a513573d902fb27	PROPN
fpq-14292	532	3	.	.	PUNCT
fpq-14292	533	1	goldin	goldin	PROPN
fpq-14292	533	2	,	,	PUNCT
fpq-14292	533	3	claudia	claudia	NOUN
fpq-14292	533	4	,	,	PUNCT
fpq-14292	533	5	and	and	CCONJ
fpq-14292	533	6	cecelia	cecelia	PROPN
fpq-14292	533	7	rouse	rouse	NOUN
fpq-14292	533	8	.	.	PUNCT
fpq-14292	534	1	2000	2000	NUM
fpq-14292	534	2	.	.	PUNCT
fpq-14292	535	1	“	"	PUNCT
fpq-14292	535	2	orchestrating	orchestrate	VERB
fpq-14292	535	3	impartiality	impartiality	NOUN
fpq-14292	535	4	:	:	PUNCT
fpq-14292	535	5	the	the	DET
fpq-14292	535	6	impact	impact	NOUN
fpq-14292	535	7	of	of	ADP
fpq-14292	535	8	‘	'	PUNCT
fpq-14292	535	9	blind	blind	ADJ
fpq-14292	535	10	’	'	PUNCT
fpq-14292	535	11	auditions	audition	NOUN
fpq-14292	535	12	on	on	ADP
fpq-14292	535	13	female	female	ADJ
fpq-14292	535	14	musicians	musician	NOUN
fpq-14292	535	15	.	.	PUNCT
fpq-14292	535	16	”	"	PUNCT
fpq-14292	536	1	american	american	PROPN
fpq-14292	536	2	economic	economic	PROPN
fpq-14292	536	3	review	review	NOUN
fpq-14292	536	4	90	90	NUM
fpq-14292	536	5	,	,	PUNCT
fpq-14292	536	6	no	no	INTJ
fpq-14292	536	7	.	.	NOUN
fpq-14292	536	8	4	4	NUM
fpq-14292	536	9	(	(	PUNCT
fpq-14292	536	10	september	september	PROPN
fpq-14292	536	11	):	):	PUNCT
fpq-14292	536	12	715–41	715–41	NUM
fpq-14292	536	13	.	.	PUNCT
fpq-14292	537	1	https://doi.org/10.1257/aer.90.4.715	https://doi.org/10.1257/aer.90.4.715	PROPN
fpq-14292	537	2	.	.	PUNCT
fpq-14292	538	1	gonen	gonen	PROPN
fpq-14292	538	2	,	,	PUNCT
fpq-14292	538	3	hila	hila	PROPN
fpq-14292	538	4	,	,	PUNCT
fpq-14292	538	5	and	and	CCONJ
fpq-14292	538	6	yoav	yoav	PROPN
fpq-14292	538	7	goldberg	goldberg	PROPN
fpq-14292	538	8	.	.	PROPN
fpq-14292	538	9	2019	2019	NUM
fpq-14292	538	10	.	.	PUNCT
fpq-14292	539	1	“	"	PUNCT
fpq-14292	539	2	lipstick	lipstick	VERB
fpq-14292	539	3	on	on	ADP
fpq-14292	539	4	a	a	DET
fpq-14292	539	5	pig	pig	NOUN
fpq-14292	539	6	:	:	PUNCT
fpq-14292	539	7	debiasing	debiase	VERB
fpq-14292	539	8	methods	method	NOUN
fpq-14292	539	9	cover	cover	VERB
fpq-14292	539	10	up	up	ADP
fpq-14292	539	11	systematic	systematic	ADJ
fpq-14292	539	12	gender	gender	NOUN
fpq-14292	539	13	biases	bias	NOUN
fpq-14292	539	14	in	in	ADP
fpq-14292	539	15	word	word	NOUN
fpq-14292	539	16	embeddings	embedding	NOUN
fpq-14292	539	17	but	but	CCONJ
fpq-14292	539	18	do	do	AUX
fpq-14292	539	19	not	not	PART
fpq-14292	539	20	remove	remove	VERB
fpq-14292	539	21	them	they	PRON
fpq-14292	539	22	.	.	PUNCT
fpq-14292	539	23	”	"	PUNCT
fpq-14292	540	1	in	in	ADP
fpq-14292	540	2	proceedings	proceeding	NOUN
fpq-14292	540	3	of	of	ADP
fpq-14292	540	4	the	the	DET
fpq-14292	540	5	2019	2019	NUM
fpq-14292	540	6	conference	conference	NOUN
fpq-14292	540	7	of	of	ADP
fpq-14292	540	8	the	the	DET
fpq-14292	540	9	north	north	ADJ
fpq-14292	540	10	american	american	ADJ
fpq-14292	540	11	chapter	chapter	NOUN
fpq-14292	540	12	of	of	ADP
fpq-14292	540	13	the	the	DET
fpq-14292	540	14	association	association	NOUN
fpq-14292	540	15	for	for	ADP
fpq-14292	540	16	computational	computational	ADJ
fpq-14292	540	17	linguistics	linguistic	NOUN
fpq-14292	540	18	:	:	PUNCT
fpq-14292	540	19	human	human	ADJ
fpq-14292	540	20	language	language	NOUN
fpq-14292	540	21	technologies	technology	NOUN
fpq-14292	540	22	,	,	PUNCT
fpq-14292	540	23	volume	volume	NOUN
fpq-14292	540	24	1	1	NUM
fpq-14292	540	25	,	,	PUNCT
fpq-14292	540	26	edited	edit	VERB
fpq-14292	540	27	by	by	ADP
fpq-14292	540	28	jill	jill	PROPN
fpq-14292	540	29	burstein	burstein	PROPN
fpq-14292	540	30	,	,	PUNCT
fpq-14292	540	31	christy	christy	PROPN
fpq-14292	540	32	doran	doran	PROPN
fpq-14292	540	33	,	,	PUNCT
fpq-14292	540	34	and	and	CCONJ
fpq-14292	540	35	thamar	thamar	PROPN
fpq-14292	540	36	solorio	solorio	PROPN
fpq-14292	540	37	,	,	PUNCT
fpq-14292	540	38	609–14	609–14	PROPN
fpq-14292	540	39	.	.	PUNCT
fpq-14292	540	40	stroudsburg	stroudsburg	PROPN
fpq-14292	540	41	,	,	PUNCT
fpq-14292	540	42	pa	pa	PROPN
fpq-14292	540	43	:	:	PUNCT
fpq-14292	540	44	association	association	NOUN
fpq-14292	540	45	for	for	ADP
fpq-14292	540	46	computational	computational	ADJ
fpq-14292	540	47	linguistics	linguistic	NOUN
fpq-14292	540	48	.	.	PUNCT
fpq-14292	541	1	https://doi.org/10	https://doi.org/10	PROPN
fpq-14292	541	2	.18653	.18653	PROPN
fpq-14292	541	3	/	/	SYM
fpq-14292	541	4	v1	v1	NOUN
fpq-14292	541	5	/	/	SYM
fpq-14292	541	6	n19	n19	NOUN
fpq-14292	541	7	-	-	PUNCT
fpq-14292	541	8	1061	1061	NUM
fpq-14292	541	9	.	.	PUNCT
fpq-14292	542	1	deery	deery	NOUN
fpq-14292	542	2	and	and	CCONJ
fpq-14292	542	3	bailey	bailey	PROPN
fpq-14292	542	4	–	–	PUNCT
fpq-14292	542	5	the	the	DET
fpq-14292	542	6	bias	bias	NOUN
fpq-14292	542	7	dilemma	dilemma	NOUN
fpq-14292	542	8	:	:	PUNCT
fpq-14292	542	9	the	the	DET
fpq-14292	542	10	ethics	ethic	NOUN
fpq-14292	542	11	of	of	ADP
fpq-14292	542	12	algorithmic	algorithmic	ADJ
fpq-14292	542	13	bias	bias	NOUN
fpq-14292	542	14	in	in	ADP
fpq-14292	542	15	natural	natural	ADJ
fpq-14292	542	16	-	-	PUNCT
fpq-14292	542	17	language	language	NOUN
fpq-14292	542	18	processing	processing	NOUN
fpq-14292	542	19	published	publish	VERB
fpq-14292	542	20	by	by	ADP
fpq-14292	542	21	scholarship@western	scholarship@western	PROPN
fpq-14292	542	22	,	,	PUNCT
fpq-14292	542	23	2022	2022	NUM
fpq-14292	542	24	27	27	NUM
fpq-14292	542	25	hitchcock	hitchcock	PROPN
fpq-14292	542	26	,	,	PUNCT
fpq-14292	542	27	christopher	christopher	PROPN
fpq-14292	542	28	.	.	PROPN
fpq-14292	542	29	2004	2004	NUM
fpq-14292	542	30	.	.	PUNCT
fpq-14292	543	1	“	"	PUNCT
fpq-14292	543	2	introduction	introduction	NOUN
fpq-14292	543	3	:	:	PUNCT
fpq-14292	543	4	what	what	PRON
fpq-14292	543	5	is	be	AUX
fpq-14292	543	6	the	the	DET
fpq-14292	543	7	philosophy	philosophy	NOUN
fpq-14292	543	8	of	of	ADP
fpq-14292	543	9	science	science	NOUN
fpq-14292	543	10	?	?	PUNCT
fpq-14292	543	11	”	"	PUNCT
fpq-14292	544	1	in	in	ADP
fpq-14292	544	2	contemporary	contemporary	ADJ
fpq-14292	544	3	debates	debate	NOUN
fpq-14292	544	4	in	in	ADP
fpq-14292	544	5	the	the	DET
fpq-14292	544	6	philosophy	philosophy	NOUN
fpq-14292	544	7	of	of	ADP
fpq-14292	544	8	science	science	NOUN
fpq-14292	544	9	,	,	PUNCT
fpq-14292	544	10	edited	edit	VERB
fpq-14292	544	11	by	by	ADP
fpq-14292	544	12	christopher	christopher	PROPN
fpq-14292	544	13	hitchcock	hitchcock	PROPN
fpq-14292	544	14	,	,	PUNCT
fpq-14292	544	15	1–19	1–19	PROPN
fpq-14292	544	16	.	.	PUNCT
fpq-14292	544	17	malden	malden	PROPN
fpq-14292	544	18	,	,	PUNCT
fpq-14292	544	19	ma	ma	PROPN
fpq-14292	544	20	:	:	PUNCT
fpq-14292	544	21	blackwell	blackwell	PROPN
fpq-14292	544	22	.	.	PUNCT
fpq-14292	545	1	hofstadter	hofstadter	PROPN
fpq-14292	545	2	,	,	PUNCT
fpq-14292	545	3	douglas	douglas	PROPN
fpq-14292	545	4	.	.	PROPN
fpq-14292	545	5	2018	2018	NUM
fpq-14292	545	6	.	.	PUNCT
fpq-14292	546	1	“	"	PUNCT
fpq-14292	546	2	the	the	DET
fpq-14292	546	3	shallowness	shallowness	NOUN
fpq-14292	546	4	of	of	ADP
fpq-14292	546	5	google	google	PROPN
fpq-14292	546	6	translate	translate	NOUN
fpq-14292	546	7	.	.	PUNCT
fpq-14292	546	8	”	"	PUNCT
fpq-14292	547	1	atlantic	atlantic	PROPN
fpq-14292	547	2	,	,	PUNCT
fpq-14292	547	3	january	january	PROPN
fpq-14292	547	4	30	30	NUM
fpq-14292	547	5	,	,	PUNCT
fpq-14292	547	6	2018	2018	NUM
fpq-14292	547	7	.	.	PUNCT
fpq-14292	548	1	https://www.theatlantic.com/technology/archive/2018/01/theshallowness-of-google-translate/551570/.	https://www.theatlantic.com/technology/archive/2018/01/theshallowness-of-google-translate/551570/.	PROPN
fpq-14292	548	2	hu	hu	PROPN
fpq-14292	548	3	,	,	PUNCT
fpq-14292	548	4	lily	lily	ADV
fpq-14292	548	5	,	,	PUNCT
fpq-14292	548	6	and	and	CCONJ
fpq-14292	548	7	yiling	yiling	PROPN
fpq-14292	548	8	chen	chen	PROPN
fpq-14292	548	9	.	.	PROPN
fpq-14292	548	10	2017	2017	NUM
fpq-14292	548	11	.	.	PUNCT
fpq-14292	549	1	“	"	PUNCT
fpq-14292	549	2	fairness	fairness	NOUN
fpq-14292	549	3	at	at	ADP
fpq-14292	549	4	equilibrium	equilibrium	NOUN
fpq-14292	549	5	in	in	ADP
fpq-14292	549	6	the	the	DET
fpq-14292	549	7	labor	labor	NOUN
fpq-14292	549	8	market	market	NOUN
fpq-14292	549	9	.	.	PUNCT
fpq-14292	549	10	”	"	PUNCT
fpq-14292	550	1	arxiv	arxiv	PROPN
fpq-14292	550	2	preprint	preprint	NOUN
fpq-14292	550	3	,	,	PUNCT
fpq-14292	550	4	arxiv:1707.01590	arxiv:1707.01590	PROPN
fpq-14292	551	1	[	[	X
fpq-14292	551	2	cs.gt	cs.gt	NOUN
fpq-14292	551	3	]	]	PUNCT
fpq-14292	551	4	.	.	PUNCT
fpq-14292	552	1	https://doi.org/10.48550/arxiv.1707	https://doi.org/10.48550/arxiv.1707	NOUN
fpq-14292	552	2	.01590	.01590	PROPN
fpq-14292	552	3	.	.	PUNCT
fpq-14292	553	1	kehl	kehl	PROPN
fpq-14292	553	2	,	,	PUNCT
fpq-14292	553	3	danielle	danielle	PROPN
fpq-14292	553	4	,	,	PUNCT
fpq-14292	553	5	priscilla	priscilla	PROPN
fpq-14292	553	6	guo	guo	PROPN
fpq-14292	553	7	,	,	PUNCT
fpq-14292	553	8	and	and	CCONJ
fpq-14292	553	9	samuel	samuel	PROPN
fpq-14292	553	10	kessler	kessler	PROPN
fpq-14292	553	11	.	.	PUNCT
fpq-14292	553	12	2017	2017	NUM
fpq-14292	553	13	.	.	PUNCT
fpq-14292	554	1	“	"	PUNCT
fpq-14292	554	2	algorithms	algorithm	NOUN
fpq-14292	554	3	in	in	ADP
fpq-14292	554	4	the	the	DET
fpq-14292	554	5	criminal	criminal	ADJ
fpq-14292	554	6	justice	justice	NOUN
fpq-14292	554	7	system	system	NOUN
fpq-14292	554	8	:	:	PUNCT
fpq-14292	554	9	assessing	assess	VERB
fpq-14292	554	10	the	the	DET
fpq-14292	554	11	use	use	NOUN
fpq-14292	554	12	of	of	ADP
fpq-14292	554	13	risk	risk	NOUN
fpq-14292	554	14	assessments	assessment	NOUN
fpq-14292	554	15	in	in	ADP
fpq-14292	554	16	sentencing	sentencing	NOUN
fpq-14292	554	17	.	.	PUNCT
fpq-14292	554	18	”	"	PUNCT
fpq-14292	555	1	responsive	responsive	ADJ
fpq-14292	555	2	communities	community	NOUN
fpq-14292	555	3	initiative/	initiative/	NUM
fpq-14292	555	4	berkman	berkman	PROPN
fpq-14292	555	5	klein	klein	PROPN
fpq-14292	555	6	center	center	PROPN
fpq-14292	555	7	for	for	ADP
fpq-14292	555	8	internet	internet	NOUN
fpq-14292	555	9	&	&	CCONJ
fpq-14292	555	10	society	society	NOUN
fpq-14292	555	11	,	,	PUNCT
fpq-14292	555	12	harvard	harvard	PROPN
fpq-14292	555	13	law	law	PROPN
fpq-14292	555	14	school	school	NOUN
fpq-14292	555	15	.	.	PUNCT
fpq-14292	556	1	http://nrs.harvard.edu/urn-3:hul.instrepos	http://nrs.harvard.edu/urn-3:hul.instrepo	NOUN
fpq-14292	556	2	:	:	PUNCT
fpq-14292	556	3	33746041	33746041	NUM
fpq-14292	556	4	.	.	PUNCT
fpq-14292	557	1	lacassagne	lacassagne	PROPN
fpq-14292	557	2	,	,	PUNCT
fpq-14292	557	3	doris	doris	PROPN
fpq-14292	557	4	,	,	PUNCT
fpq-14292	557	5	jérémy	jérémy	PROPN
fpq-14292	557	6	béna	béna	PROPN
fpq-14292	557	7	,	,	PUNCT
fpq-14292	557	8	and	and	CCONJ
fpq-14292	557	9	olivier	olivier	PROPN
fpq-14292	557	10	corneille	corneille	PROPN
fpq-14292	557	11	.	.	PUNCT
fpq-14292	557	12	2022	2022	NUM
fpq-14292	557	13	.	.	PUNCT
fpq-14292	558	1	“	"	PUNCT
fpq-14292	558	2	is	be	AUX
fpq-14292	558	3	earth	earth	NOUN
fpq-14292	558	4	a	a	DET
fpq-14292	558	5	perfect	perfect	ADJ
fpq-14292	558	6	square	square	NOUN
fpq-14292	558	7	?	?	PUNCT
fpq-14292	559	1	repetition	repetition	NOUN
fpq-14292	559	2	increases	increase	VERB
fpq-14292	559	3	the	the	DET
fpq-14292	559	4	perceived	perceive	VERB
fpq-14292	559	5	truth	truth	NOUN
fpq-14292	559	6	of	of	ADP
fpq-14292	559	7	highly	highly	ADV
fpq-14292	559	8	implausible	implausible	ADJ
fpq-14292	559	9	statements	statement	NOUN
fpq-14292	559	10	.	.	PUNCT
fpq-14292	559	11	”	"	PUNCT
fpq-14292	560	1	cognition	cognition	NOUN
fpq-14292	560	2	223	223	NUM
fpq-14292	560	3	(	(	PUNCT
fpq-14292	560	4	june	june	PROPN
fpq-14292	560	5	):	):	PUNCT
fpq-14292	560	6	105052	105052	NUM
fpq-14292	560	7	.	.	PUNCT
fpq-14292	561	1	https://doi.org/10.1016/j.cogni	https://doi.org/10.1016/j.cogni	PROPN
fpq-14292	561	2	tion.2022.105052	tion.2022.105052	PROPN
fpq-14292	561	3	.	.	PUNCT
fpq-14292	561	4	levy	levy	PROPN
fpq-14292	561	5	,	,	PUNCT
fpq-14292	561	6	neil	neil	PROPN
fpq-14292	561	7	.	.	PROPN
fpq-14292	561	8	2017	2017	NUM
fpq-14292	561	9	.	.	PUNCT
fpq-14292	562	1	“	"	PUNCT
fpq-14292	562	2	implicit	implicit	ADJ
fpq-14292	562	3	bias	bias	NOUN
fpq-14292	562	4	and	and	CCONJ
fpq-14292	562	5	moral	moral	ADJ
fpq-14292	562	6	responsibility	responsibility	NOUN
fpq-14292	562	7	:	:	PUNCT
fpq-14292	562	8	probing	probe	VERB
fpq-14292	562	9	the	the	DET
fpq-14292	562	10	data	datum	NOUN
fpq-14292	562	11	.	.	PUNCT
fpq-14292	562	12	”	"	PUNCT
fpq-14292	562	13	philosophy	philosophy	NOUN
fpq-14292	562	14	and	and	CCONJ
fpq-14292	562	15	phenomenological	phenomenological	ADJ
fpq-14292	562	16	research	research	NOUN
fpq-14292	562	17	94	94	NUM
fpq-14292	562	18	,	,	PUNCT
fpq-14292	562	19	no	no	INTJ
fpq-14292	562	20	.	.	NOUN
fpq-14292	562	21	1	1	NUM
fpq-14292	562	22	(	(	PUNCT
fpq-14292	562	23	january	january	PROPN
fpq-14292	562	24	):	):	PUNCT
fpq-14292	562	25	3–26	3–26	NUM
fpq-14292	562	26	.	.	PUNCT
fpq-14292	563	1	https://doi.org/10.1111/phpr.12352	https://doi.org/10.1111/phpr.12352	PROPN
fpq-14292	563	2	.	.	PUNCT
fpq-14292	564	1	liao	liao	PROPN
fpq-14292	564	2	,	,	PUNCT
fpq-14292	564	3	shen	shen	PROPN
fpq-14292	564	4	-	-	PUNCT
fpq-14292	564	5	yi	yi	PROPN
fpq-14292	564	6	,	,	PUNCT
fpq-14292	564	7	and	and	CCONJ
fpq-14292	564	8	bryce	bryce	PROPN
fpq-14292	564	9	huebner	huebner	PROPN
fpq-14292	564	10	.	.	PUNCT
fpq-14292	565	1	“	"	PUNCT
fpq-14292	565	2	oppressive	oppressive	ADJ
fpq-14292	565	3	things	thing	NOUN
fpq-14292	565	4	.	.	PUNCT
fpq-14292	565	5	”	"	PUNCT
fpq-14292	566	1	philosophy	philosophy	NOUN
fpq-14292	566	2	and	and	CCONJ
fpq-14292	566	3	phenomenological	phenomenological	ADJ
fpq-14292	566	4	research	research	NOUN
fpq-14292	566	5	103	103	NUM
fpq-14292	566	6	,	,	PUNCT
fpq-14292	566	7	no	no	INTJ
fpq-14292	566	8	.	.	NOUN
fpq-14292	566	9	1	1	NUM
fpq-14292	566	10	(	(	PUNCT
fpq-14292	566	11	july	july	PROPN
fpq-14292	566	12	):	):	PUNCT
fpq-14292	566	13	92–113	92–113	PROPN
fpq-14292	566	14	.	.	PUNCT
fpq-14292	567	1	https://doi.org/10.1111	https://doi.org/10.1111	PROPN
fpq-14292	567	2	/phpr.12701	/phpr.12701	PROPN
fpq-14292	567	3	.	.	PROPN
fpq-14292	568	1	mikolov	mikolov	PROPN
fpq-14292	568	2	,	,	PUNCT
fpq-14292	568	3	tomas	tomas	PROPN
fpq-14292	568	4	,	,	PUNCT
fpq-14292	568	5	kai	kai	PROPN
fpq-14292	568	6	chen	chen	PROPN
fpq-14292	568	7	,	,	PUNCT
fpq-14292	568	8	greg	greg	PROPN
fpq-14292	568	9	corrado	corrado	PROPN
fpq-14292	568	10	,	,	PUNCT
fpq-14292	568	11	and	and	CCONJ
fpq-14292	568	12	jeffrey	jeffrey	PROPN
fpq-14292	568	13	dean	dean	PROPN
fpq-14292	568	14	.	.	PUNCT
fpq-14292	569	1	2013	2013	NUM
fpq-14292	569	2	.	.	PUNCT
fpq-14292	570	1	“	"	PUNCT
fpq-14292	570	2	efficient	efficient	ADJ
fpq-14292	570	3	estimation	estimation	NOUN
fpq-14292	570	4	of	of	ADP
fpq-14292	570	5	word	word	NOUN
fpq-14292	570	6	representations	representation	NOUN
fpq-14292	570	7	in	in	ADP
fpq-14292	570	8	vector	vector	NOUN
fpq-14292	570	9	space	space	NOUN
fpq-14292	570	10	.	.	PUNCT
fpq-14292	570	11	”	"	PUNCT
fpq-14292	571	1	arxiv	arxiv	PROPN
fpq-14292	571	2	preprint	preprint	PROPN
fpq-14292	571	3	,	,	PUNCT
fpq-14292	571	4	arxiv	arxiv	PROPN
fpq-14292	571	5	:	:	PUNCT
fpq-14292	571	6	1301.3781	1301.3781	NUM
fpq-14292	572	1	[	[	X
fpq-14292	572	2	cs.cl	cs.cl	NOUN
fpq-14292	572	3	]	]	X
fpq-14292	572	4	.	.	PUNCT
fpq-14292	573	1	https://doi.org/10.48550/arxiv.1301.3781	https://doi.org/10.48550/arxiv.1301.3781	PROPN
fpq-14292	573	2	.	.	PROPN
fpq-14292	573	3	nissim	nissim	PROPN
fpq-14292	573	4	,	,	PUNCT
fpq-14292	573	5	malvina	malvina	PROPN
fpq-14292	573	6	,	,	PUNCT
fpq-14292	573	7	rik	rik	PROPN
fpq-14292	573	8	van	van	PROPN
fpq-14292	573	9	noord	noord	PROPN
fpq-14292	573	10	,	,	PUNCT
fpq-14292	573	11	and	and	CCONJ
fpq-14292	573	12	rob	rob	PROPN
fpq-14292	573	13	van	van	PROPN
fpq-14292	573	14	der	der	PROPN
fpq-14292	573	15	goot	goot	PROPN
fpq-14292	573	16	.	.	PUNCT
fpq-14292	574	1	2019	2019	NUM
fpq-14292	574	2	.	.	PUNCT
fpq-14292	575	1	“	"	PUNCT
fpq-14292	575	2	fair	fair	ADJ
fpq-14292	575	3	is	be	AUX
fpq-14292	575	4	better	well	ADJ
fpq-14292	575	5	than	than	ADP
fpq-14292	575	6	sensational	sensational	ADJ
fpq-14292	575	7	:	:	PUNCT
fpq-14292	575	8	man	man	NOUN
fpq-14292	575	9	is	be	AUX
fpq-14292	575	10	to	to	PART
fpq-14292	575	11	doctor	doctor	VERB
fpq-14292	575	12	as	as	SCONJ
fpq-14292	575	13	woman	woman	NOUN
fpq-14292	575	14	is	be	AUX
fpq-14292	575	15	to	to	PART
fpq-14292	575	16	doctor	doctor	VERB
fpq-14292	575	17	.	.	PUNCT
fpq-14292	575	18	”	"	PUNCT
fpq-14292	576	1	arxiv	arxiv	PROPN
fpq-14292	576	2	preprint	preprint	NOUN
fpq-14292	576	3	,	,	PUNCT
fpq-14292	576	4	arxiv:1905.09866v2	arxiv:1905.09866v2	VERB
fpq-14292	576	5	[	[	X
fpq-14292	576	6	cs.cl	cs.cl	PROPN
fpq-14292	576	7	]	]	X
fpq-14292	576	8	.	.	PUNCT
fpq-14292	577	1	https://doi.org/10.48550/arxiv.1905.09866	https://doi.org/10.48550/arxiv.1905.09866	PROPN
fpq-14292	577	2	.	.	PUNCT
fpq-14292	578	1	noble	noble	PROPN
fpq-14292	578	2	,	,	PUNCT
fpq-14292	578	3	safiya	safiya	NOUN
fpq-14292	578	4	umoja	umoja	NOUN
fpq-14292	578	5	.	.	PUNCT
fpq-14292	579	1	2018	2018	NUM
fpq-14292	579	2	.	.	PUNCT
fpq-14292	580	1	algorithms	algorithm	NOUN
fpq-14292	580	2	of	of	ADP
fpq-14292	580	3	oppression	oppression	NOUN
fpq-14292	580	4	:	:	PUNCT
fpq-14292	580	5	how	how	SCONJ
fpq-14292	580	6	search	search	NOUN
fpq-14292	580	7	engines	engine	NOUN
fpq-14292	580	8	reinforce	reinforce	VERB
fpq-14292	580	9	racism	racism	NOUN
fpq-14292	580	10	.	.	PUNCT
fpq-14292	581	1	new	new	PROPN
fpq-14292	581	2	york	york	PROPN
fpq-14292	581	3	:	:	PUNCT
fpq-14292	581	4	new	new	PROPN
fpq-14292	581	5	york	york	PROPN
fpq-14292	581	6	university	university	PROPN
fpq-14292	581	7	press	press	NOUN
fpq-14292	581	8	.	.	PUNCT
fpq-14292	582	1	puddifoot	puddifoot	NOUN
fpq-14292	582	2	,	,	PUNCT
fpq-14292	582	3	katherine	katherine	PROPN
fpq-14292	582	4	.	.	PROPN
fpq-14292	582	5	2017	2017	NUM
fpq-14292	582	6	.	.	PUNCT
fpq-14292	583	1	“	"	PUNCT
fpq-14292	583	2	dissolving	dissolve	VERB
fpq-14292	583	3	the	the	DET
fpq-14292	583	4	epistemic	epistemic	ADJ
fpq-14292	583	5	/	/	SYM
fpq-14292	583	6	ethical	ethical	ADJ
fpq-14292	583	7	dilemma	dilemma	NOUN
fpq-14292	583	8	over	over	ADP
fpq-14292	583	9	implicit	implicit	ADJ
fpq-14292	583	10	bias	bias	NOUN
fpq-14292	583	11	.	.	PUNCT
fpq-14292	583	12	”	"	PUNCT
fpq-14292	584	1	philosophical	philosophical	ADJ
fpq-14292	584	2	explorations	exploration	NOUN
fpq-14292	584	3	20	20	NUM
fpq-14292	584	4	(	(	PUNCT
fpq-14292	584	5	supplement	supplement	NOUN
fpq-14292	584	6	1	1	NUM
fpq-14292	584	7	):	):	PUNCT
fpq-14292	584	8	s73	s73	PROPN
fpq-14292	584	9	–	–	PUNCT
fpq-14292	584	10	s93	s93	PROPN
fpq-14292	584	11	.	.	PUNCT
fpq-14292	585	1	https://doi.org	https://doi.org	PROPN
fpq-14292	585	2	/10.1080/13869795.2017.1287295	/10.1080/13869795.2017.1287295	PUNCT
fpq-14292	585	3	.	.	PUNCT
fpq-14292	586	1	roth	roth	PROPN
fpq-14292	586	2	,	,	PUNCT
fpq-14292	586	3	lorna	lorna	PROPN
fpq-14292	586	4	.	.	PUNCT
fpq-14292	586	5	2009	2009	NUM
fpq-14292	586	6	.	.	PUNCT
fpq-14292	587	1	“	"	PUNCT
fpq-14292	587	2	looking	look	VERB
fpq-14292	587	3	at	at	ADP
fpq-14292	587	4	shirley	shirley	NOUN
fpq-14292	587	5	,	,	PUNCT
fpq-14292	587	6	the	the	DET
fpq-14292	587	7	ultimate	ultimate	ADJ
fpq-14292	587	8	norm	norm	NOUN
fpq-14292	587	9	:	:	PUNCT
fpq-14292	587	10	colour	colour	NOUN
fpq-14292	587	11	balance	balance	NOUN
fpq-14292	587	12	,	,	PUNCT
fpq-14292	587	13	image	image	NOUN
fpq-14292	587	14	technologies	technology	NOUN
fpq-14292	587	15	,	,	PUNCT
fpq-14292	587	16	and	and	CCONJ
fpq-14292	587	17	cognitive	cognitive	ADJ
fpq-14292	587	18	equity	equity	NOUN
fpq-14292	587	19	.	.	PUNCT
fpq-14292	587	20	”	"	PUNCT
fpq-14292	588	1	canadian	canadian	ADJ
fpq-14292	588	2	journal	journal	NOUN
fpq-14292	588	3	of	of	ADP
fpq-14292	588	4	communication	communication	NOUN
fpq-14292	588	5	34	34	NUM
fpq-14292	588	6	,	,	PUNCT
fpq-14292	588	7	no	no	INTJ
fpq-14292	588	8	.	.	NOUN
fpq-14292	588	9	1	1	NUM
fpq-14292	588	10	(	(	PUNCT
fpq-14292	588	11	march	march	PROPN
fpq-14292	588	12	):	):	PUNCT
fpq-14292	588	13	111–36	111–36	NUM
fpq-14292	588	14	.	.	PUNCT
fpq-14292	589	1	https://doi.org/10.22230/cjc.2009v34n1a2196	https://doi.org/10.22230/cjc.2009v34n1a2196	ADJ
fpq-14292	589	2	.	.	PUNCT
fpq-14292	590	1	schluter	schluter	PROPN
fpq-14292	590	2	,	,	PUNCT
fpq-14292	590	3	natalie	natalie	PROPN
fpq-14292	590	4	.	.	PROPN
fpq-14292	590	5	2018	2018	NUM
fpq-14292	590	6	.	.	PUNCT
fpq-14292	591	1	“	"	PUNCT
fpq-14292	591	2	the	the	DET
fpq-14292	591	3	word	word	NOUN
fpq-14292	591	4	analogy	analogy	NOUN
fpq-14292	591	5	testing	testing	NOUN
fpq-14292	591	6	caveat	caveat	NOUN
fpq-14292	591	7	.	.	PUNCT
fpq-14292	591	8	”	"	PUNCT
fpq-14292	592	1	proceedings	proceeding	NOUN
fpq-14292	592	2	of	of	ADP
fpq-14292	592	3	the	the	DET
fpq-14292	592	4	2018	2018	NUM
fpq-14292	592	5	conference	conference	NOUN
fpq-14292	592	6	of	of	ADP
fpq-14292	592	7	the	the	DET
fpq-14292	592	8	north	north	ADJ
fpq-14292	592	9	american	american	ADJ
fpq-14292	592	10	chapter	chapter	NOUN
fpq-14292	592	11	of	of	ADP
fpq-14292	592	12	the	the	DET
fpq-14292	592	13	association	association	NOUN
fpq-14292	592	14	for	for	ADP
fpq-14292	592	15	compufeminist	compufeminist	ADJ
fpq-14292	592	16	philosophy	philosophy	NOUN
fpq-14292	592	17	quarterly	quarterly	ADV
fpq-14292	592	18	,	,	PUNCT
fpq-14292	592	19	2022	2022	NUM
fpq-14292	592	20	,	,	PUNCT
fpq-14292	592	21	vol	vol	NOUN
fpq-14292	592	22	.	.	NOUN
fpq-14292	592	23	8	8	NUM
fpq-14292	592	24	,	,	PUNCT
fpq-14292	592	25	iss	iss	PROPN
fpq-14292	592	26	.	.	PUNCT
fpq-14292	592	27	3/4	3/4	NUM
fpq-14292	592	28	,	,	PUNCT
fpq-14292	592	29	article	article	NOUN
fpq-14292	592	30	1	1	NUM
fpq-14292	592	31	published	publish	VERB
fpq-14292	592	32	by	by	ADP
fpq-14292	592	33	scholarship@western	scholarship@western	PROPN
fpq-14292	592	34	,	,	PUNCT
fpq-14292	592	35	2022	2022	NUM
fpq-14292	592	36	28	28	NUM
fpq-14292	592	37	tational	tational	ADJ
fpq-14292	592	38	linguistics	linguistic	NOUN
fpq-14292	592	39	:	:	PUNCT
fpq-14292	592	40	human	human	ADJ
fpq-14292	592	41	language	language	NOUN
fpq-14292	592	42	technologies	technology	NOUN
fpq-14292	592	43	,	,	PUNCT
fpq-14292	592	44	volume	volume	NOUN
fpq-14292	592	45	2	2	NUM
fpq-14292	592	46	,	,	PUNCT
fpq-14292	592	47	edited	edit	VERB
fpq-14292	592	48	by	by	ADP
fpq-14292	592	49	marilyn	marilyn	PROPN
fpq-14292	592	50	walker	walker	PROPN
fpq-14292	592	51	,	,	PUNCT
fpq-14292	592	52	heng	heng	PROPN
fpq-14292	592	53	ji	ji	PROPN
fpq-14292	592	54	,	,	PUNCT
fpq-14292	592	55	and	and	CCONJ
fpq-14292	592	56	amanda	amanda	PROPN
fpq-14292	592	57	stent	stent	NOUN
fpq-14292	592	58	,	,	PUNCT
fpq-14292	592	59	242–46	242–46	PROPN
fpq-14292	592	60	.	.	PUNCT
fpq-14292	592	61	stroudsburg	stroudsburg	PROPN
fpq-14292	592	62	,	,	PUNCT
fpq-14292	592	63	pa	pa	PROPN
fpq-14292	592	64	:	:	PUNCT
fpq-14292	592	65	association	association	NOUN
fpq-14292	592	66	for	for	ADP
fpq-14292	592	67	computational	computational	ADJ
fpq-14292	592	68	linguistics	linguistic	NOUN
fpq-14292	592	69	.	.	PUNCT
fpq-14292	593	1	https://doi.org/10.18653/v1/n182039	https://doi.org/10.18653/v1/n182039	NOUN
fpq-14292	593	2	.	.	PUNCT
fpq-14292	594	1	smith	smith	PROPN
fpq-14292	594	2	,	,	PUNCT
fpq-14292	594	3	angela	angela	PROPN
fpq-14292	594	4	m.	m.	PROPN
fpq-14292	594	5	2008	2008	NUM
fpq-14292	594	6	.	.	PUNCT
fpq-14292	595	1	“	"	PUNCT
fpq-14292	595	2	control	control	NOUN
fpq-14292	595	3	,	,	PUNCT
fpq-14292	595	4	responsibility	responsibility	NOUN
fpq-14292	595	5	,	,	PUNCT
fpq-14292	595	6	and	and	CCONJ
fpq-14292	595	7	moral	moral	ADJ
fpq-14292	595	8	assessment	assessment	NOUN
fpq-14292	595	9	.	.	PUNCT
fpq-14292	595	10	”	"	PUNCT
fpq-14292	596	1	philosophical	philosophical	ADJ
fpq-14292	596	2	studies	study	NOUN
fpq-14292	596	3	138	138	NUM
fpq-14292	596	4	,	,	PUNCT
fpq-14292	596	5	no	no	INTJ
fpq-14292	596	6	.	.	NOUN
fpq-14292	596	7	3	3	NUM
fpq-14292	596	8	(	(	PUNCT
fpq-14292	596	9	april	april	PROPN
fpq-14292	596	10	):	):	PUNCT
fpq-14292	596	11	367–92	367–92	PROPN
fpq-14292	596	12	.	.	PUNCT
fpq-14292	597	1	https://doi.org/10.1007/s11	https://doi.org/10.1007/s11	PROPN
fpq-14292	597	2	098	098	NUM
fpq-14292	597	3	-	-	PUNCT
fpq-14292	597	4	006	006	NUM
fpq-14292	597	5	-	-	PUNCT
fpq-14292	597	6	9048	9048	NUM
fpq-14292	597	7	-	-	PUNCT
fpq-14292	597	8	x.	x.	NOUN
fpq-14292	597	9	smith	smith	PROPN
fpq-14292	597	10	,	,	PUNCT
fpq-14292	597	11	david	david	PROPN
fpq-14292	597	12	.	.	PROPN
fpq-14292	597	13	2013	2013	NUM
fpq-14292	597	14	.	.	PUNCT
fpq-14292	598	1	“	"	PUNCT
fpq-14292	598	2	‘	'	PUNCT
fpq-14292	598	3	racism	racism	NOUN
fpq-14292	598	4	’	'	PUNCT
fpq-14292	598	5	of	of	ADP
fpq-14292	598	6	early	early	ADJ
fpq-14292	598	7	colour	colour	NOUN
fpq-14292	598	8	photography	photography	NOUN
fpq-14292	598	9	explored	explore	VERB
fpq-14292	598	10	in	in	ADP
fpq-14292	598	11	art	art	NOUN
fpq-14292	598	12	exhibition	exhibition	NOUN
fpq-14292	598	13	.	.	PUNCT
fpq-14292	598	14	”	"	PUNCT
fpq-14292	599	1	guardian	guardian	PROPN
fpq-14292	599	2	,	,	PUNCT
fpq-14292	599	3	january	january	PROPN
fpq-14292	599	4	25	25	NUM
fpq-14292	599	5	,	,	PUNCT
fpq-14292	599	6	2013	2013	NUM
fpq-14292	599	7	.	.	PUNCT
fpq-14292	600	1	http://www.theguardian.com/artanddesign/20	http://www.theguardian.com/artanddesign/20	PROPN
fpq-14292	600	2	13	13	NUM
fpq-14292	600	3	/	/	SYM
fpq-14292	600	4	jan/25	jan/25	NOUN
fpq-14292	600	5	/	/	SYM
fpq-14292	600	6	racism	racism	NOUN
fpq-14292	600	7	-	-	PUNCT
fpq-14292	600	8	colour	colour	NOUN
fpq-14292	600	9	-	-	PUNCT
fpq-14292	600	10	photography	photography	NOUN
fpq-14292	600	11	-	-	NOUN
fpq-14292	600	12	exhibition	exhibition	NOUN
fpq-14292	600	13	.	.	PUNCT
fpq-14292	601	1	suresh	suresh	PROPN
fpq-14292	601	2	,	,	PUNCT
fpq-14292	601	3	harini	harini	PROPN
fpq-14292	601	4	,	,	PUNCT
fpq-14292	601	5	and	and	CCONJ
fpq-14292	601	6	john	john	PROPN
fpq-14292	601	7	v.	v.	PROPN
fpq-14292	601	8	guttag	guttag	NOUN
fpq-14292	601	9	.	.	PUNCT
fpq-14292	602	1	2020	2020	NUM
fpq-14292	602	2	.	.	PUNCT
fpq-14292	603	1	“	"	PUNCT
fpq-14292	603	2	a	a	DET
fpq-14292	603	3	framework	framework	NOUN
fpq-14292	603	4	for	for	ADP
fpq-14292	603	5	understanding	understand	VERB
fpq-14292	603	6	unintended	unintended	ADJ
fpq-14292	603	7	consequences	consequence	NOUN
fpq-14292	603	8	of	of	ADP
fpq-14292	603	9	machine	machine	NOUN
fpq-14292	603	10	learning	learning	NOUN
fpq-14292	603	11	.	.	PUNCT
fpq-14292	603	12	”	"	PUNCT
fpq-14292	604	1	arxiv	arxiv	PROPN
fpq-14292	604	2	preprint	preprint	NOUN
fpq-14292	604	3	,	,	PUNCT
fpq-14292	604	4	february	february	PROPN
fpq-14292	604	5	17	17	NUM
fpq-14292	604	6	,	,	PUNCT
fpq-14292	604	7	2020	2020	NUM
fpq-14292	604	8	,	,	PUNCT
fpq-14292	604	9	revision	revision	NOUN
fpq-14292	604	10	,	,	PUNCT
fpq-14292	604	11	arxiv:1901.10002v3	arxiv:1901.10002v3	PROPN
fpq-14292	605	1	[	[	X
fpq-14292	605	2	cs.lg	cs.lg	PROPN
fpq-14292	605	3	]	]	PUNCT
fpq-14292	605	4	.	.	PUNCT
fpq-14292	606	1	https://doi.org/10.48550/arxiv	https://doi.org/10.48550/arxiv	PROPN
fpq-14292	606	2	.1901.10002	.1901.10002	PROPN
fpq-14292	606	3	.	.	PUNCT
fpq-14292	607	1	wu	wu	PROPN
fpq-14292	607	2	,	,	PUNCT
fpq-14292	607	3	yonghui	yonghui	PROPN
fpq-14292	607	4	,	,	PUNCT
fpq-14292	607	5	mike	mike	PROPN
fpq-14292	607	6	schuster	schuster	PROPN
fpq-14292	607	7	,	,	PUNCT
fpq-14292	607	8	zhifeng	zhifeng	PROPN
fpq-14292	607	9	chen	chen	PROPN
fpq-14292	607	10	,	,	PUNCT
fpq-14292	607	11	quoc	quoc	PROPN
fpq-14292	607	12	v.	v.	ADP
fpq-14292	607	13	le	le	X
fpq-14292	607	14	,	,	PUNCT
fpq-14292	607	15	mohammad	mohammad	PROPN
fpq-14292	607	16	norouzi	norouzi	PROPN
fpq-14292	607	17	,	,	PUNCT
fpq-14292	607	18	wolfgang	wolfgang	PROPN
fpq-14292	607	19	macherey	macherey	PROPN
fpq-14292	607	20	,	,	PUNCT
fpq-14292	607	21	maxim	maxim	NOUN
fpq-14292	607	22	krikun	krikun	NOUN
fpq-14292	607	23	,	,	PUNCT
fpq-14292	607	24	et	et	PROPN
fpq-14292	607	25	al	al	PROPN
fpq-14292	607	26	.	.	PROPN
fpq-14292	607	27	2016	2016	NUM
fpq-14292	607	28	.	.	PUNCT
fpq-14292	608	1	“	"	PUNCT
fpq-14292	608	2	google	google	PROPN
fpq-14292	608	3	’s	’s	PART
fpq-14292	608	4	neural	neural	ADJ
fpq-14292	608	5	machine	machine	NOUN
fpq-14292	608	6	translation	translation	NOUN
fpq-14292	608	7	system	system	NOUN
fpq-14292	608	8	:	:	PUNCT
fpq-14292	608	9	bridging	bridge	VERB
fpq-14292	608	10	the	the	DET
fpq-14292	608	11	gap	gap	NOUN
fpq-14292	608	12	between	between	ADP
fpq-14292	608	13	human	human	ADJ
fpq-14292	608	14	and	and	CCONJ
fpq-14292	608	15	machine	machine	NOUN
fpq-14292	608	16	translation	translation	NOUN
fpq-14292	608	17	.	.	PUNCT
fpq-14292	608	18	”	"	PUNCT
fpq-14292	609	1	arxiv	arxiv	PROPN
fpq-14292	609	2	preprint	preprint	NOUN
fpq-14292	609	3	,	,	PUNCT
fpq-14292	609	4	arxiv:1609.08144	arxiv:1609.08144	NOUN
fpq-14292	610	1	[	[	X
fpq-14292	610	2	cs.cl	cs.cl	NOUN
fpq-14292	610	3	]	]	X
fpq-14292	610	4	.	.	PUNCT
fpq-14292	611	1	https://doi.org/10.485	https://doi.org/10.485	VERB
fpq-14292	611	2	50	50	NUM
fpq-14292	611	3	/	/	SYM
fpq-14292	611	4	arxiv.1609.08144	arxiv.1609.08144	PROPN
fpq-14292	611	5	.	.	PUNCT
fpq-14292	612	1	oisín	oisín	PROPN
fpq-14292	612	2	deery	deery	PROPN
fpq-14292	612	3	is	be	AUX
fpq-14292	612	4	an	an	DET
fpq-14292	612	5	assistant	assistant	ADJ
fpq-14292	612	6	professor	professor	NOUN
fpq-14292	612	7	of	of	ADP
fpq-14292	612	8	philosophy	philosophy	NOUN
fpq-14292	612	9	at	at	ADP
fpq-14292	612	10	york	york	PROPN
fpq-14292	612	11	university	university	PROPN
fpq-14292	612	12	,	,	PUNCT
fpq-14292	612	13	toronto	toronto	PROPN
fpq-14292	612	14	,	,	PUNCT
fpq-14292	612	15	canada	canada	PROPN
fpq-14292	612	16	,	,	PUNCT
fpq-14292	612	17	and	and	CCONJ
fpq-14292	612	18	a	a	DET
fpq-14292	612	19	lecturer	lecturer	NOUN
fpq-14292	612	20	and	and	CCONJ
fpq-14292	612	21	arc	arc	NOUN
fpq-14292	612	22	decra	decra	VERB
fpq-14292	612	23	fellow	fellow	NOUN
fpq-14292	612	24	at	at	ADP
fpq-14292	612	25	macquarie	macquarie	PROPN
fpq-14292	612	26	university	university	PROPN
fpq-14292	612	27	in	in	ADP
fpq-14292	612	28	sydney	sydney	PROPN
fpq-14292	612	29	,	,	PUNCT
fpq-14292	612	30	australia	australia	PROPN
fpq-14292	612	31	.	.	PUNCT
fpq-14292	613	1	he	he	PRON
fpq-14292	613	2	has	have	AUX
fpq-14292	613	3	previously	previously	ADV
fpq-14292	613	4	held	hold	VERB
fpq-14292	613	5	positions	position	NOUN
fpq-14292	613	6	at	at	ADP
fpq-14292	613	7	monash	monash	PROPN
fpq-14292	613	8	university	university	PROPN
fpq-14292	613	9	,	,	PUNCT
fpq-14292	613	10	florida	florida	PROPN
fpq-14292	613	11	state	state	PROPN
fpq-14292	613	12	university	university	PROPN
fpq-14292	613	13	,	,	PUNCT
fpq-14292	613	14	the	the	DET
fpq-14292	613	15	university	university	PROPN
fpq-14292	613	16	of	of	ADP
fpq-14292	613	17	arizona	arizona	PROPN
fpq-14292	613	18	,	,	PUNCT
fpq-14292	613	19	and	and	CCONJ
fpq-14292	613	20	the	the	DET
fpq-14292	613	21	university	university	NOUN
fpq-14292	613	22	of	of	ADP
fpq-14292	613	23	montreal	montreal	PROPN
fpq-14292	613	24	.	.	PUNCT
fpq-14292	614	1	oisín	oisín	PROPN
fpq-14292	614	2	’s	’s	PART
fpq-14292	614	3	work	work	NOUN
fpq-14292	614	4	straddles	straddle	VERB
fpq-14292	614	5	the	the	DET
fpq-14292	614	6	philosophy	philosophy	NOUN
fpq-14292	614	7	of	of	ADP
fpq-14292	614	8	mind	mind	NOUN
fpq-14292	614	9	and	and	CCONJ
fpq-14292	614	10	action	action	NOUN
fpq-14292	614	11	,	,	PUNCT
fpq-14292	614	12	moral	moral	ADJ
fpq-14292	614	13	psychology	psychology	NOUN
fpq-14292	614	14	,	,	PUNCT
fpq-14292	614	15	and	and	CCONJ
fpq-14292	614	16	metaphysics	metaphysic	NOUN
fpq-14292	614	17	.	.	PUNCT
fpq-14292	615	1	his	his	PRON
fpq-14292	615	2	work	work	NOUN
fpq-14292	615	3	has	have	AUX
fpq-14292	615	4	focused	focus	VERB
fpq-14292	615	5	primarily	primarily	ADV
fpq-14292	615	6	on	on	ADP
fpq-14292	615	7	developing	develop	VERB
fpq-14292	615	8	a	a	DET
fpq-14292	615	9	naturalistic	naturalistic	ADJ
fpq-14292	615	10	understanding	understanding	NOUN
fpq-14292	615	11	of	of	ADP
fpq-14292	615	12	human	human	ADJ
fpq-14292	615	13	agency	agency	NOUN
fpq-14292	615	14	and	and	CCONJ
fpq-14292	615	15	responsibility	responsibility	NOUN
fpq-14292	615	16	,	,	PUNCT
fpq-14292	615	17	but	but	CCONJ
fpq-14292	615	18	recently	recently	ADV
fpq-14292	615	19	he	he	PRON
fpq-14292	615	20	has	have	AUX
fpq-14292	615	21	also	also	ADV
fpq-14292	615	22	been	be	AUX
fpq-14292	615	23	working	work	VERB
fpq-14292	615	24	on	on	ADP
fpq-14292	615	25	issues	issue	NOUN
fpq-14292	615	26	related	relate	VERB
fpq-14292	615	27	to	to	ADP
fpq-14292	615	28	artificial	artificial	ADJ
fpq-14292	615	29	intelligence	intelligence	NOUN
fpq-14292	615	30	and	and	CCONJ
fpq-14292	615	31	agency	agency	NOUN
fpq-14292	615	32	.	.	PUNCT
fpq-14292	616	1	oisín	oisín	PROPN
fpq-14292	616	2	has	have	AUX
fpq-14292	616	3	published	publish	VERB
fpq-14292	616	4	widely	widely	ADV
fpq-14292	616	5	in	in	ADP
fpq-14292	616	6	journals	journal	NOUN
fpq-14292	616	7	such	such	ADJ
fpq-14292	616	8	as	as	ADP
fpq-14292	616	9	philosophical	philosophical	ADJ
fpq-14292	616	10	studies	study	NOUN
fpq-14292	616	11	,	,	PUNCT
fpq-14292	616	12	the	the	DET
fpq-14292	616	13	australasian	australasian	ADJ
fpq-14292	616	14	journal	journal	NOUN
fpq-14292	616	15	of	of	ADP
fpq-14292	616	16	philosophy	philosophy	NOUN
fpq-14292	616	17	,	,	PUNCT
fpq-14292	616	18	and	and	CCONJ
fpq-14292	616	19	philosophical	philosophical	ADJ
fpq-14292	616	20	psychology	psychology	NOUN
fpq-14292	616	21	.	.	PUNCT
fpq-14292	617	1	he	he	PRON
fpq-14292	617	2	is	be	AUX
fpq-14292	617	3	the	the	DET
fpq-14292	617	4	author	author	NOUN
fpq-14292	617	5	of	of	ADP
fpq-14292	617	6	a	a	DET
fpq-14292	617	7	monograph	monograph	NOUN
fpq-14292	617	8	entitled	entitle	VERB
fpq-14292	617	9	naturally	naturally	ADV
fpq-14292	617	10	free	free	ADJ
fpq-14292	617	11	action	action	NOUN
fpq-14292	617	12	,	,	PUNCT
fpq-14292	617	13	published	publish	VERB
fpq-14292	617	14	by	by	ADP
fpq-14292	617	15	oxford	oxford	PROPN
fpq-14292	617	16	university	university	PROPN
fpq-14292	617	17	press	press	NOUN
fpq-14292	617	18	in	in	ADP
fpq-14292	617	19	2021	2021	NUM
fpq-14292	617	20	.	.	PUNCT
fpq-14292	618	1	katherine	katherine	PROPN
fpq-14292	618	2	bailey	bailey	PROPN
fpq-14292	618	3	has	have	AUX
fpq-14292	618	4	worked	work	VERB
fpq-14292	618	5	in	in	ADP
fpq-14292	618	6	a	a	DET
fpq-14292	618	7	number	number	NOUN
fpq-14292	618	8	of	of	ADP
fpq-14292	618	9	roles	role	NOUN
fpq-14292	618	10	as	as	ADP
fpq-14292	618	11	a	a	DET
fpq-14292	618	12	data	data	NOUN
fpq-14292	618	13	scientist	scientist	NOUN
fpq-14292	618	14	and	and	CCONJ
fpq-14292	618	15	software	software	NOUN
fpq-14292	618	16	engineer	engineer	NOUN
fpq-14292	618	17	.	.	PUNCT
fpq-14292	619	1	she	she	PRON
fpq-14292	619	2	is	be	AUX
fpq-14292	619	3	currently	currently	ADV
fpq-14292	619	4	employed	employ	VERB
fpq-14292	619	5	by	by	ADP
fpq-14292	619	6	shopify	shopify	PROPN
fpq-14292	619	7	inc	inc	PROPN
fpq-14292	619	8	.	.	PROPN
fpq-14292	619	9	,	,	PUNCT
fpq-14292	619	10	in	in	ADP
fpq-14292	619	11	toronto	toronto	PROPN
fpq-14292	619	12	,	,	PUNCT
fpq-14292	619	13	canada	canada	PROPN
fpq-14292	619	14	.	.	PUNCT
fpq-14292	620	1	katherine	katherine	PROPN
fpq-14292	620	2	has	have	VERB
fpq-14292	620	3	a	a	DET
fpq-14292	620	4	longstanding	longstanding	ADJ
fpq-14292	620	5	interest	interest	NOUN
fpq-14292	620	6	in	in	ADP
fpq-14292	620	7	issues	issue	NOUN
fpq-14292	620	8	related	relate	VERB
fpq-14292	620	9	to	to	ADP
fpq-14292	620	10	artificial	artificial	ADJ
fpq-14292	620	11	intelligence	intelligence	NOUN
fpq-14292	620	12	and	and	CCONJ
fpq-14292	620	13	natural	natural	ADJ
fpq-14292	620	14	language	language	NOUN
fpq-14292	620	15	processing	processing	NOUN
fpq-14292	620	16	.	.	PUNCT
fpq-14292	621	1	she	she	PRON
fpq-14292	621	2	has	have	AUX
fpq-14292	621	3	written	write	VERB
fpq-14292	621	4	on	on	ADP
fpq-14292	621	5	these	these	DET
fpq-14292	621	6	issues	issue	NOUN
fpq-14292	621	7	in	in	ADP
fpq-14292	621	8	blog	blog	NOUN
fpq-14292	621	9	posts	post	NOUN
fpq-14292	621	10	and	and	CCONJ
fpq-14292	621	11	in	in	ADP
fpq-14292	621	12	publications	publication	NOUN
fpq-14292	621	13	such	such	ADJ
fpq-14292	621	14	as	as	ADP
fpq-14292	621	15	techcrunch	techcrunch	NOUN
fpq-14292	621	16	and	and	CCONJ
fpq-14292	621	17	has	have	AUX
fpq-14292	621	18	been	be	AUX
fpq-14292	621	19	invited	invite	VERB
fpq-14292	621	20	to	to	PART
fpq-14292	621	21	speak	speak	VERB
fpq-14292	621	22	at	at	ADP
fpq-14292	621	23	various	various	ADJ
fpq-14292	621	24	venues	venue	NOUN
fpq-14292	621	25	,	,	PUNCT
fpq-14292	621	26	including	include	VERB
fpq-14292	621	27	google	google	PROPN
fpq-14292	621	28	and	and	CCONJ
fpq-14292	621	29	the	the	DET
fpq-14292	621	30	north	north	ADJ
fpq-14292	621	31	american	american	PROPN
fpq-14292	621	32	association	association	PROPN
fpq-14292	621	33	for	for	ADP
fpq-14292	621	34	computational	computational	ADJ
fpq-14292	621	35	linguistics	linguistic	NOUN
fpq-14292	621	36	.	.	PUNCT
fpq-14292	622	1	14292	14292	NUM
fpq-14292	622	2	deery	deery	NOUN
fpq-14292	622	3	and	and	CCONJ
fpq-14292	622	4	bailey	bailey	PROPN
fpq-14292	622	5	title	title	NOUN
fpq-14292	622	6	page	page	NOUN
fpq-14292	622	7	14292	14292	NUM
fpq-14292	622	8	deery	deery	PROPN
fpq-14292	622	9	&	&	CCONJ
fpq-14292	622	10	bailey	bailey	PROPN
fpq-14292	622	11	final	final	ADJ
fpq-14292	622	12	format	format	NOUN