{"id":"https://openalex.org/W3203453034","doi":"https://doi.org/10.1109/icassp43922.2022.9746168","title":"Knowledge Distillation for Neural Transducers from Large Self-Supervised Pre-Trained Models","display_name":"Knowledge Distillation for Neural Transducers from Large Self-Supervised Pre-Trained Models","publication_year":2022,"publication_date":"2022-04-27","ids":{"openalex":"https://openalex.org/W3203453034","doi":"https://doi.org/10.1109/icassp43922.2022.9746168","mag":"3203453034"},"language":"en","primary_location":{"id":"doi:10.1109/icassp43922.2022.9746168","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp43922.2022.9746168","pdf_url":null,"source":{"id":"https://openalex.org/S4363607702","display_name":"ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100435428","display_name":"Xiaoyu Yang","orcid":"https://orcid.org/0000-0002-3570-8979"},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Xiaoyu Yang","raw_affiliation_strings":["Cambridge University,Engineering Dept.,Cambridge,U.K.,CB2 1PZ"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Cambridge University,Engineering Dept.,Cambridge,U.K.,CB2 1PZ","institution_ids":["https://openalex.org/I241749"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030757951","display_name":"Qiujia Li","orcid":"https://orcid.org/0000-0003-3074-3692"},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Qiujia Li","raw_affiliation_strings":["Cambridge University,Engineering Dept.,Cambridge,U.K.,CB2 1PZ"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Cambridge University,Engineering Dept.,Cambridge,U.K.,CB2 1PZ","institution_ids":["https://openalex.org/I241749"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5002191410","display_name":"Philip C. Woodland","orcid":"https://orcid.org/0000-0001-9069-0225"},"institutions":[{"id":"https://openalex.org/I241749","display_name":"University of Cambridge","ror":"https://ror.org/013meh722","country_code":"GB","type":"education","lineage":["https://openalex.org/I241749"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Philip C. Woodland","raw_affiliation_strings":["Cambridge University,Engineering Dept.,Cambridge,U.K.,CB2 1PZ"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Cambridge University,Engineering Dept.,Cambridge,U.K.,CB2 1PZ","institution_ids":["https://openalex.org/I241749"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I241749"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":18,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"8527","last_page":"8531"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9976999759674072,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11309","display_name":"Music and Audio Processing","score":0.9915000200271606,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7725493907928467},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.7659021615982056},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.5700768232345581},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5680669546127319},{"id":"https://openalex.org/keywords/streaming-data","display_name":"Streaming data","score":0.5412629842758179},{"id":"https://openalex.org/keywords/test-data","display_name":"Test data","score":0.5056192874908447},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4614187180995941},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.4437096416950226},{"id":"https://openalex.org/keywords/word-error-rate","display_name":"Word error rate","score":0.4291191101074219},{"id":"https://openalex.org/keywords/data-modeling","display_name":"Data modeling","score":0.4220172166824341},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.34960561990737915},{"id":"https://openalex.org/keywords/data-mining","display_name":"Data mining","score":0.20862701535224915},{"id":"https://openalex.org/keywords/database","display_name":"Database","score":0.08265984058380127}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7725493907928467},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.7659021615982056},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5700768232345581},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5680669546127319},{"id":"https://openalex.org/C2777611316","wikidata":"https://www.wikidata.org/wiki/Q39045282","display_name":"Streaming data","level":2,"score":0.5412629842758179},{"id":"https://openalex.org/C16910744","wikidata":"https://www.wikidata.org/wiki/Q7705759","display_name":"Test data","level":2,"score":0.5056192874908447},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4614187180995941},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4437096416950226},{"id":"https://openalex.org/C40969351","wikidata":"https://www.wikidata.org/wiki/Q3516228","display_name":"Word error rate","level":2,"score":0.4291191101074219},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.4220172166824341},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.34960561990737915},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.20862701535224915},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.08265984058380127},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C178790620","wikidata":"https://www.wikidata.org/wiki/Q11351","display_name":"Organic chemistry","level":1,"score":0.0},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icassp43922.2022.9746168","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp43922.2022.9746168","pdf_url":null,"source":{"id":"https://openalex.org/S4363607702","display_name":"ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.8299999833106995,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":54,"referenced_works":["https://openalex.org/W1494198834","https://openalex.org/W1821462560","https://openalex.org/W1828163288","https://openalex.org/W2127141656","https://openalex.org/W2294543795","https://openalex.org/W2312434537","https://openalex.org/W2402040300","https://openalex.org/W2507699225","https://openalex.org/W2889129739","https://openalex.org/W2892008152","https://openalex.org/W2896457183","https://openalex.org/W2913178639","https://openalex.org/W2936774411","https://openalex.org/W2936993002","https://openalex.org/W2952533036","https://openalex.org/W2962760690","https://openalex.org/W2962780374","https://openalex.org/W2962826786","https://openalex.org/W2963240019","https://openalex.org/W2963250244","https://openalex.org/W2963341956","https://openalex.org/W2963414781","https://openalex.org/W2979476256","https://openalex.org/W2981991061","https://openalex.org/W2987019345","https://openalex.org/W2996383576","https://openalex.org/W3016010032","https://openalex.org/W3036601975","https://openalex.org/W3046667470","https://openalex.org/W3093579165","https://openalex.org/W3096518646","https://openalex.org/W3097777922","https://openalex.org/W3098903812","https://openalex.org/W3099782249","https://openalex.org/W3156828761","https://openalex.org/W3160628828","https://openalex.org/W3162649911","https://openalex.org/W3163062345","https://openalex.org/W3198039885","https://openalex.org/W3198715852","https://openalex.org/W4206375145","https://openalex.org/W4285719527","https://openalex.org/W4292779060","https://openalex.org/W6638523607","https://openalex.org/W6638749077","https://openalex.org/W6747398299","https://openalex.org/W6755207826","https://openalex.org/W6769196770","https://openalex.org/W6769238691","https://openalex.org/W6770245836","https://openalex.org/W6778883912","https://openalex.org/W6780218876","https://openalex.org/W6784614252","https://openalex.org/W6792304734"],"related_works":["https://openalex.org/W1566315437","https://openalex.org/W2151348424","https://openalex.org/W2594897229","https://openalex.org/W4290708361","https://openalex.org/W4221142855","https://openalex.org/W2050138804","https://openalex.org/W767271433","https://openalex.org/W2129812225","https://openalex.org/W1507733787","https://openalex.org/W2155620340"],"abstract_inverted_index":{"Self-supervised":[0],"pre-training":[1],"is":[2,28,41,185,198],"an":[3,159],"effective":[4],"approach":[5,96],"to":[6,14,30,43,95,145,171],"leveraging":[7],"a":[8,45,62,91,191],"large":[9,32,101],"amount":[10],"of":[11,20,53,99,130,163],"unlabelled":[12,164],"data":[13,165],"reduce":[15],"word":[16],"error":[17],"rates":[18],"(WERs)":[19],"automatic":[21],"speech":[22],"recognition":[23],"(ASR)":[24],"systems.":[25],"Since":[26],"it":[27,40],"impractical":[29],"use":[31],"pre-trained":[33,55,102],"models":[34,143],"for":[35,69,138,166,175,187],"many":[36],"real-world":[37],"ASR":[38],"applications,":[39],"desirable":[42],"have":[44],"much":[46],"smaller":[47,116],"model":[48,94,182,197],"while":[49],"retaining":[50],"the":[51,54,75,79,97,100,107,118,121,135,146,153,168,195],"performance":[52,98],"model.":[56,104],"In":[57],"this":[58],"paper,":[59],"we":[60],"propose":[61],"simple":[63],"knowledge":[64],"distillation":[65,189],"(KD)":[66],"loss":[67,123],"function":[68],"neural":[70],"transducers":[71,148],"that":[72,111],"focuses":[73],"on":[74,106,134],"one-best":[76],"path":[77],"in":[78,125,194],"output":[80],"probability":[81],"lattice":[82],"under":[83],"both":[84],"streaming":[85,141,178],"and":[86,132,140,173,177],"non-streaming":[87,139,176],"setups,":[88],"which":[89],"allows":[90],"small":[92],"student":[93,142,196],"teacher":[103,119],"Experiments":[105],"LibriSpeech":[108],"dataset":[109],"show":[110],"despite":[112],"being":[113],"10":[114],"times":[115],"than":[117],"model,":[120],"proposed":[122],"results":[124],"relative":[126],"WER":[127],"reductions":[128],"(WERRs)":[129],"11.5%":[131],"6.8%":[133],"test-other":[136],"set":[137],"compared":[144],"baseline":[147],"trained":[149],"without":[150],"KD":[151],"using":[152],"labelled":[154],"100-hour":[155],"clean":[156],"data.":[157],"With":[158],"additional":[160],"860":[161],"hours":[162],"KD,":[167],"WERRs":[169],"increase":[170],"48.2%":[172],"38.5%":[174],"students.":[179],"If":[180],"language":[181],"shallow":[183],"fusion":[184],"used":[186],"producing":[188],"targets,":[190],"further":[192],"improvement":[193],"observed.":[199]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":7},{"year":2024,"cited_by_count":5},{"year":2023,"cited_by_count":4}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}