Unsupervised Data Selection via Discrete Speech Representation for ASR Article Swipe

PDF

Zhiyun Lu , Yongqiang Wang , Yu Zhang , Wei Han , Zhehuai Chen , Parisa Haghani ·

YOU? · · 2022 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2204.01981

Self-supervised learning of speech representations has achieved impressive results in improving automatic speech recognition (ASR). In this paper, we show that data selection is important for self-supervised learning. We propose a simple and effective unsupervised data selection method which selects acoustically similar speech to a target domain. It takes the discrete speech representation available in common self-supervised learning frameworks as input, and applies a contrastive data selection method on the discrete tokens. Through extensive empirical studies we show that our proposed method reduces the amount of required pre-training data and improves the downstream ASR performance. Pre-training on a selected subset of 6% of the general data pool results in 11.8% relative improvements in LibriSpeech test-other compared to pre-training on the full set. On Multilingual LibriSpeech French, German, and Spanish test sets, selecting 6% data for pre-training reduces word error rate by more than 15% relatively compared to the full set, and achieves competitive results compared to current state-of-the-art performances.

Related Topics

Computer Science

Training, Validation, And Test Data Sets

Artificial Intelligence

Concepts

Computer science Selection (genetic algorithm) Word error rate Training set Representation (politics) Speech recognition Set (abstract data type) Labeled data Artificial intelligence Test data Test set Word (group theory) Data set Machine learning Natural language processing Mathematics Politics Programming language Law Political science Geometry

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2204.01981
PDF: https://arxiv.org/pdf/2204.01981
OA Status: green
Cited By: 1
Related Works: 10
OpenAlex ID: https://openalex.org/W4226069388

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4226069388

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2204.01981

Digital Object Identifier
Title: Unsupervised Data Selection via Discrete Speech Representation for ASR

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2022

Year of publication
Publication date: 2022-04-05

Full publication date if available
Authors: Zhiyun Lu, Yongqiang Wang, Yu Zhang, Wei Han, Zhehuai Chen, Parisa Haghani

List of authors in order
Landing page: https://arxiv.org/abs/2204.01981

Publisher landing page
PDF URL: https://arxiv.org/pdf/2204.01981

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: green

Open access status per OpenAlex
OA URL: https://arxiv.org/pdf/2204.01981

Direct OA link when available
Concepts: Computer science, Selection (genetic algorithm), Word error rate, Training set, Representation (politics), Speech recognition, Set (abstract data type), Labeled data, Artificial intelligence, Test data, Test set, Word (group theory), Data set, Machine learning, Natural language processing, Mathematics, Politics, Programming language, Law, Political science, Geometry

Top concepts (fields/topics) attached by OpenAlex
Cited by: 1

Total citation count in OpenAlex
Citations by year (recent): 2023: 1

Per-year citation counts (last 5 years)
Related works (count): 10

Other works algorithmically related by OpenAlex

Full payload

id	https://openalex.org/W4226069388
doi	https://doi.org/10.48550/arxiv.2204.01981
ids.doi	https://doi.org/10.48550/arxiv.2204.01981
ids.openalex	https://openalex.org/W4226069388
fwci
type	preprint
title	Unsupervised Data Selection via Discrete Speech Representation for ASR
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T10201
topics[0].field.id	https://openalex.org/fields/17
topics[0].field.display_name	Computer Science
topics[0].score	0.9994000196456909
topics[0].domain.id	https://openalex.org/domains/3
topics[0].domain.display_name	Physical Sciences
topics[0].subfield.id	https://openalex.org/subfields/1702
topics[0].subfield.display_name	Artificial Intelligence
topics[0].display_name	Speech Recognition and Synthesis
topics[1].id	https://openalex.org/T11309
topics[1].field.id	https://openalex.org/fields/17
topics[1].field.display_name	Computer Science
topics[1].score	0.9855999946594238
topics[1].domain.id	https://openalex.org/domains/3
topics[1].domain.display_name	Physical Sciences
topics[1].subfield.id	https://openalex.org/subfields/1711
topics[1].subfield.display_name	Signal Processing
topics[1].display_name	Music and Audio Processing
topics[2].id	https://openalex.org/T10860
topics[2].field.id	https://openalex.org/fields/17
topics[2].field.display_name	Computer Science
topics[2].score	0.9828000068664551
topics[2].domain.id	https://openalex.org/domains/3
topics[2].domain.display_name	Physical Sciences
topics[2].subfield.id	https://openalex.org/subfields/1711
topics[2].subfield.display_name	Signal Processing
topics[2].display_name	Speech and Audio Processing
is_xpac	False
apc_list
apc_paid
concepts[0].id	https://openalex.org/C41008148
concepts[0].level	0
concepts[0].score	0.7672361135482788
concepts[0].wikidata	https://www.wikidata.org/wiki/Q21198
concepts[0].display_name	Computer science
concepts[1].id	https://openalex.org/C81917197
concepts[1].level	2
concepts[1].score	0.6350440979003906
concepts[1].wikidata	https://www.wikidata.org/wiki/Q628760
concepts[1].display_name	Selection (genetic algorithm)
concepts[2].id	https://openalex.org/C40969351
concepts[2].level	2
concepts[2].score	0.5595206022262573
concepts[2].wikidata	https://www.wikidata.org/wiki/Q3516228
concepts[2].display_name	Word error rate
concepts[3].id	https://openalex.org/C51632099
concepts[3].level	2
concepts[3].score	0.5364644527435303
concepts[3].wikidata	https://www.wikidata.org/wiki/Q3985153
concepts[3].display_name	Training set
concepts[4].id	https://openalex.org/C2776359362
concepts[4].level	3
concepts[4].score	0.5324194431304932
concepts[4].wikidata	https://www.wikidata.org/wiki/Q2145286
concepts[4].display_name	Representation (politics)
concepts[5].id	https://openalex.org/C28490314
concepts[5].level	1
concepts[5].score	0.5297466516494751
concepts[5].wikidata	https://www.wikidata.org/wiki/Q189436
concepts[5].display_name	Speech recognition
concepts[6].id	https://openalex.org/C177264268
concepts[6].level	2
concepts[6].score	0.5140898823738098
concepts[6].wikidata	https://www.wikidata.org/wiki/Q1514741
concepts[6].display_name	Set (abstract data type)
concepts[7].id	https://openalex.org/C2776145971
concepts[7].level	2
concepts[7].score	0.4964771866798401
concepts[7].wikidata	https://www.wikidata.org/wiki/Q30673951
concepts[7].display_name	Labeled data
concepts[8].id	https://openalex.org/C154945302
concepts[8].level	1
concepts[8].score	0.4922894537448883
concepts[8].wikidata	https://www.wikidata.org/wiki/Q11660
concepts[8].display_name	Artificial intelligence
concepts[9].id	https://openalex.org/C16910744
concepts[9].level	2
concepts[9].score	0.45456790924072266
concepts[9].wikidata	https://www.wikidata.org/wiki/Q7705759
concepts[9].display_name	Test data
concepts[10].id	https://openalex.org/C169903167
concepts[10].level	2
concepts[10].score	0.43972715735435486
concepts[10].wikidata	https://www.wikidata.org/wiki/Q3985153
concepts[10].display_name	Test set
concepts[11].id	https://openalex.org/C90805587
concepts[11].level	2
concepts[11].score	0.435777485370636
concepts[11].wikidata	https://www.wikidata.org/wiki/Q10944557
concepts[11].display_name	Word (group theory)
concepts[12].id	https://openalex.org/C58489278
concepts[12].level	2
concepts[12].score	0.4260680675506592
concepts[12].wikidata	https://www.wikidata.org/wiki/Q1172284
concepts[12].display_name	Data set
concepts[13].id	https://openalex.org/C119857082
concepts[13].level	1
concepts[13].score	0.40565037727355957
concepts[13].wikidata	https://www.wikidata.org/wiki/Q2539
concepts[13].display_name	Machine learning
concepts[14].id	https://openalex.org/C204321447
concepts[14].level	1
concepts[14].score	0.32495713233947754
concepts[14].wikidata	https://www.wikidata.org/wiki/Q30642
concepts[14].display_name	Natural language processing
concepts[15].id	https://openalex.org/C33923547
concepts[15].level	0
concepts[15].score	0.12894442677497864
concepts[15].wikidata	https://www.wikidata.org/wiki/Q395
concepts[15].display_name	Mathematics
concepts[16].id	https://openalex.org/C94625758
concepts[16].level	2
concepts[16].score	0.0
concepts[16].wikidata	https://www.wikidata.org/wiki/Q7163
concepts[16].display_name	Politics
concepts[17].id	https://openalex.org/C199360897
concepts[17].level	1
concepts[17].score	0.0
concepts[17].wikidata	https://www.wikidata.org/wiki/Q9143
concepts[17].display_name	Programming language
concepts[18].id	https://openalex.org/C199539241
concepts[18].level	1
concepts[18].score	0.0
concepts[18].wikidata	https://www.wikidata.org/wiki/Q7748
concepts[18].display_name	Law
concepts[19].id	https://openalex.org/C17744445
concepts[19].level	0
concepts[19].score	0.0
concepts[19].wikidata	https://www.wikidata.org/wiki/Q36442
concepts[19].display_name	Political science
concepts[20].id	https://openalex.org/C2524010
concepts[20].level	1
concepts[20].score	0.0
concepts[20].wikidata	https://www.wikidata.org/wiki/Q8087
concepts[20].display_name	Geometry
keywords[0].id	https://openalex.org/keywords/computer-science
keywords[0].score	0.7672361135482788
keywords[0].display_name	Computer science
keywords[1].id	https://openalex.org/keywords/selection
keywords[1].score	0.6350440979003906
keywords[1].display_name	Selection (genetic algorithm)
keywords[2].id	https://openalex.org/keywords/word-error-rate
keywords[2].score	0.5595206022262573
keywords[2].display_name	Word error rate
keywords[3].id	https://openalex.org/keywords/training-set
keywords[3].score	0.5364644527435303
keywords[3].display_name	Training set
keywords[4].id	https://openalex.org/keywords/representation
keywords[4].score	0.5324194431304932
keywords[4].display_name	Representation (politics)
keywords[5].id	https://openalex.org/keywords/speech-recognition
keywords[5].score	0.5297466516494751
keywords[5].display_name	Speech recognition
keywords[6].id	https://openalex.org/keywords/set
keywords[6].score	0.5140898823738098
keywords[6].display_name	Set (abstract data type)
keywords[7].id	https://openalex.org/keywords/labeled-data
keywords[7].score	0.4964771866798401
keywords[7].display_name	Labeled data
keywords[8].id	https://openalex.org/keywords/artificial-intelligence
keywords[8].score	0.4922894537448883
keywords[8].display_name	Artificial intelligence
keywords[9].id	https://openalex.org/keywords/test-data
keywords[9].score	0.45456790924072266
keywords[9].display_name	Test data
keywords[10].id	https://openalex.org/keywords/test-set
keywords[10].score	0.43972715735435486
keywords[10].display_name	Test set
keywords[11].id	https://openalex.org/keywords/word
keywords[11].score	0.435777485370636
keywords[11].display_name	Word (group theory)
keywords[12].id	https://openalex.org/keywords/data-set
keywords[12].score	0.4260680675506592
keywords[12].display_name	Data set
keywords[13].id	https://openalex.org/keywords/machine-learning
keywords[13].score	0.40565037727355957
keywords[13].display_name	Machine learning
keywords[14].id	https://openalex.org/keywords/natural-language-processing
keywords[14].score	0.32495713233947754
keywords[14].display_name	Natural language processing
keywords[15].id	https://openalex.org/keywords/mathematics
keywords[15].score	0.12894442677497864
keywords[15].display_name	Mathematics
language	en
locations[0].id	pmh:oai:arXiv.org:2204.01981
locations[0].is_oa	True
locations[0].source.id	https://openalex.org/S4306400194
locations[0].source.issn
locations[0].source.type	repository
locations[0].source.is_oa	True
locations[0].source.issn_l
locations[0].source.is_core	False
locations[0].source.is_in_doaj	False
locations[0].source.display_name	arXiv (Cornell University)
locations[0].source.host_organization	https://openalex.org/I205783295
locations[0].source.host_organization_name	Cornell University
locations[0].source.host_organization_lineage	https://openalex.org/I205783295
locations[0].license	public-domain
locations[0].pdf_url	https://arxiv.org/pdf/2204.01981
locations[0].version	submittedVersion
locations[0].raw_type	text
locations[0].license_id	https://openalex.org/licenses/public-domain
locations[0].is_accepted	False
locations[0].is_published	False
locations[0].raw_source_name
locations[0].landing_page_url	http://arxiv.org/abs/2204.01981
locations[1].id	doi:10.48550/arxiv.2204.01981
locations[1].is_oa	True
locations[1].source.id	https://openalex.org/S4306400194
locations[1].source.issn
locations[1].source.type	repository
locations[1].source.is_oa	True
locations[1].source.issn_l
locations[1].source.is_core	False
locations[1].source.is_in_doaj	False
locations[1].source.display_name	arXiv (Cornell University)
locations[1].source.host_organization	https://openalex.org/I205783295
locations[1].source.host_organization_name	Cornell University
locations[1].source.host_organization_lineage	https://openalex.org/I205783295
locations[1].license
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article
locations[1].license_id
locations[1].is_accepted	False
locations[1].is_published
locations[1].raw_source_name
locations[1].landing_page_url	https://doi.org/10.48550/arxiv.2204.01981
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5039693533
authorships[0].author.orcid	https://orcid.org/0000-0002-1733-4061
authorships[0].author.display_name	Zhiyun Lu
authorships[0].author_position	first
authorships[0].raw_author_name	Lu, Zhiyun
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5100339989
authorships[1].author.orcid	https://orcid.org/0000-0002-8600-0276
authorships[1].author.display_name	Yongqiang Wang
authorships[1].author_position	middle
authorships[1].raw_author_name	Wang, Yongqiang
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5100433691
authorships[2].author.orcid	https://orcid.org/0000-0003-1100-4835
authorships[2].author.display_name	Yu Zhang
authorships[2].author_position	middle
authorships[2].raw_author_name	Zhang, Yu
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5107931312
authorships[3].author.orcid	https://orcid.org/0000-0003-0809-5465
authorships[3].author.display_name	Wei Han
authorships[3].author_position	middle
authorships[3].raw_author_name	Han, Wei
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5002433660
authorships[4].author.orcid	https://orcid.org/0000-0003-4400-5340
authorships[4].author.display_name	Zhehuai Chen
authorships[4].author_position	middle
authorships[4].raw_author_name	Chen, Zhehuai
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5037705635
authorships[5].author.orcid
authorships[5].author.display_name	Parisa Haghani
authorships[5].author_position	last
authorships[5].raw_author_name	Haghani, Parisa
authorships[5].is_corresponding	False
has_content.pdf	True
has_content.grobid_xml	True
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2204.01981
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Unsupervised Data Selection via Discrete Speech Representation for ASR
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T10201
primary_topic.field.id	https://openalex.org/fields/17
primary_topic.field.display_name	Computer Science
primary_topic.score	0.9994000196456909
primary_topic.domain.id	https://openalex.org/domains/3
primary_topic.domain.display_name	Physical Sciences
primary_topic.subfield.id	https://openalex.org/subfields/1702
primary_topic.subfield.display_name	Artificial Intelligence
primary_topic.display_name	Speech Recognition and Synthesis
related_works	https://openalex.org/W1999699871, https://openalex.org/W2145230572, https://openalex.org/W4225124612, https://openalex.org/W2130553454, https://openalex.org/W2043806667, https://openalex.org/W3022007134, https://openalex.org/W4317548404, https://openalex.org/W2146197305, https://openalex.org/W2021633306, https://openalex.org/W3019226033
cited_by_count	1
counts_by_year[0].year	2023
counts_by_year[0].cited_by_count	1
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2204.01981
best_oa_location.is_oa	True
best_oa_location.source.id	https://openalex.org/S4306400194
best_oa_location.source.issn
best_oa_location.source.type	repository
best_oa_location.source.is_oa	True
best_oa_location.source.issn_l
best_oa_location.source.is_core	False
best_oa_location.source.is_in_doaj	False
best_oa_location.source.display_name	arXiv (Cornell University)
best_oa_location.source.host_organization	https://openalex.org/I205783295
best_oa_location.source.host_organization_name	Cornell University
best_oa_location.source.host_organization_lineage	https://openalex.org/I205783295
best_oa_location.license	public-domain
best_oa_location.pdf_url	https://arxiv.org/pdf/2204.01981
best_oa_location.version	submittedVersion
best_oa_location.raw_type	text
best_oa_location.license_id	https://openalex.org/licenses/public-domain
best_oa_location.is_accepted	False
best_oa_location.is_published	False
best_oa_location.raw_source_name
best_oa_location.landing_page_url	http://arxiv.org/abs/2204.01981
primary_location.id	pmh:oai:arXiv.org:2204.01981
primary_location.is_oa	True
primary_location.source.id	https://openalex.org/S4306400194
primary_location.source.issn
primary_location.source.type	repository
primary_location.source.is_oa	True
primary_location.source.issn_l
primary_location.source.is_core	False
primary_location.source.is_in_doaj	False
primary_location.source.display_name	arXiv (Cornell University)
primary_location.source.host_organization	https://openalex.org/I205783295
primary_location.source.host_organization_name	Cornell University
primary_location.source.host_organization_lineage	https://openalex.org/I205783295
primary_location.license	public-domain
primary_location.pdf_url	https://arxiv.org/pdf/2204.01981
primary_location.version	submittedVersion
primary_location.raw_type	text
primary_location.license_id	https://openalex.org/licenses/public-domain
primary_location.is_accepted	False
primary_location.is_published	False
primary_location.raw_source_name
primary_location.landing_page_url	http://arxiv.org/abs/2204.01981
publication_date	2022-04-05
publication_year	2022
referenced_works_count	0
abstract_inverted_index.a	30, 44, 63, 97
abstract_inverted_index.6%	101, 132
abstract_inverted_index.In	15
abstract_inverted_index.It	47
abstract_inverted_index.On	122
abstract_inverted_index.We	28
abstract_inverted_index.as	59
abstract_inverted_index.by	140
abstract_inverted_index.in	9, 54, 108, 112
abstract_inverted_index.is	23
abstract_inverted_index.of	2, 85, 100, 102
abstract_inverted_index.on	68, 96, 118
abstract_inverted_index.to	43, 116, 146, 155
abstract_inverted_index.we	18, 76
abstract_inverted_index.15%	143
abstract_inverted_index.ASR	93
abstract_inverted_index.and	32, 61, 89, 127, 150
abstract_inverted_index.for	25, 134
abstract_inverted_index.has	5
abstract_inverted_index.our	79
abstract_inverted_index.the	49, 69, 83, 91, 103, 119, 147
abstract_inverted_index.data	21, 35, 65, 88, 105, 133
abstract_inverted_index.full	120, 148
abstract_inverted_index.more	141
abstract_inverted_index.pool	106
abstract_inverted_index.rate	139
abstract_inverted_index.set,	149
abstract_inverted_index.set.	121
abstract_inverted_index.show	19, 77
abstract_inverted_index.test	129
abstract_inverted_index.than	142
abstract_inverted_index.that	20, 78
abstract_inverted_index.this	16
abstract_inverted_index.word	137
abstract_inverted_index.11.8%	109
abstract_inverted_index.error	138
abstract_inverted_index.sets,	130
abstract_inverted_index.takes	48
abstract_inverted_index.which	38
abstract_inverted_index.(ASR).	14
abstract_inverted_index.amount	84
abstract_inverted_index.common	55
abstract_inverted_index.input,	60
abstract_inverted_index.method	37, 67, 81
abstract_inverted_index.paper,	17
abstract_inverted_index.simple	31
abstract_inverted_index.speech	3, 12, 42, 51
abstract_inverted_index.subset	99
abstract_inverted_index.target	45
abstract_inverted_index.French,	125
abstract_inverted_index.German,	126
abstract_inverted_index.Spanish	128
abstract_inverted_index.Through	72
abstract_inverted_index.applies	62
abstract_inverted_index.current	156
abstract_inverted_index.domain.	46
abstract_inverted_index.general	104
abstract_inverted_index.propose	29
abstract_inverted_index.reduces	82, 136
abstract_inverted_index.results	8, 107, 153
abstract_inverted_index.selects	39
abstract_inverted_index.similar	41
abstract_inverted_index.studies	75
abstract_inverted_index.tokens.	71
abstract_inverted_index.achieved	6
abstract_inverted_index.achieves	151
abstract_inverted_index.compared	115, 145, 154
abstract_inverted_index.discrete	50, 70
abstract_inverted_index.improves	90
abstract_inverted_index.learning	1, 57
abstract_inverted_index.proposed	80
abstract_inverted_index.relative	110
abstract_inverted_index.required	86
abstract_inverted_index.selected	98
abstract_inverted_index.automatic	11
abstract_inverted_index.available	53
abstract_inverted_index.effective	33
abstract_inverted_index.empirical	74
abstract_inverted_index.extensive	73
abstract_inverted_index.important	24
abstract_inverted_index.improving	10
abstract_inverted_index.learning.	27
abstract_inverted_index.selecting	131
abstract_inverted_index.selection	22, 36, 66
abstract_inverted_index.downstream	92
abstract_inverted_index.frameworks	58
abstract_inverted_index.impressive	7
abstract_inverted_index.relatively	144
abstract_inverted_index.test-other	114
abstract_inverted_index.LibriSpeech	113, 124
abstract_inverted_index.competitive	152
abstract_inverted_index.contrastive	64
abstract_inverted_index.recognition	13
abstract_inverted_index.Multilingual	123
abstract_inverted_index.Pre-training	95
abstract_inverted_index.acoustically	40
abstract_inverted_index.improvements	111
abstract_inverted_index.performance.	94
abstract_inverted_index.pre-training	87, 117, 135
abstract_inverted_index.unsupervised	34
abstract_inverted_index.performances.	158
abstract_inverted_index.representation	52
abstract_inverted_index.Self-supervised	0
abstract_inverted_index.representations	4
abstract_inverted_index.self-supervised	26, 56
abstract_inverted_index.state-of-the-art	157
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	6
sustainable_development_goals[0].id	https://metadata.un.org/sdg/4
sustainable_development_goals[0].score	0.7900000214576721
sustainable_development_goals[0].display_name	Quality Education
citation_normalized_percentile