{"id":"https://openalex.org/W4221138270","doi":"https://doi.org/10.21437/interspeech.2022-10809","title":"4-bit Conformer with Native Quantization Aware Training for Speech Recognition","display_name":"4-bit Conformer with Native Quantization Aware Training for Speech Recognition","publication_year":2022,"publication_date":"2022-09-16","ids":{"openalex":"https://openalex.org/W4221138270","doi":"https://doi.org/10.21437/interspeech.2022-10809"},"language":"en","primary_location":{"id":"doi:10.21437/interspeech.2022-10809","is_oa":false,"landing_page_url":"https://doi.org/10.21437/interspeech.2022-10809","pdf_url":null,"source":{"id":"https://openalex.org/S4363604309","display_name":"Interspeech 2022","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":"Interspeech 2022","raw_type":"proceedings-article"},"type":"article","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/A5058886181","display_name":"Shaojin Ding","orcid":"https://orcid.org/0000-0002-2108-3111"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Shaojin Ding","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5007839762","display_name":"Phoenix Meadowlark\u200e","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Phoenix Meadowlark\u200e","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101153566","display_name":"Yanzhang He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yanzhang He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104269404","display_name":"\u0141ukasz Lew","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lukasz Lew","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033596864","display_name":"Shivani Agrawal","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shivani Agrawal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5070398306","display_name":"Oleg Rybakov\u200e","orcid":"https://orcid.org/0000-0003-4805-3083"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oleg Rybakov\u200e","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5058886181"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.1838,"has_fulltext":false,"cited_by_count":23,"citation_normalized_percentile":{"value":0.8945416,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1711","last_page":"1715"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9984999895095825,"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":0.9984999895095825,"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":0.9908999800682068,"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"}},{"id":"https://openalex.org/T10860","display_name":"Speech and Audio Processing","score":0.9900000095367432,"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/quantization","display_name":"Quantization (signal processing)","score":0.863140344619751},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7305548191070557},{"id":"https://openalex.org/keywords/lossless-compression","display_name":"Lossless compression","score":0.563318133354187},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5604845285415649},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.5387199521064758},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5218364596366882},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3875812292098999},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.37906116247177124},{"id":"https://openalex.org/keywords/data-compression","display_name":"Data compression","score":0.32708579301834106}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.863140344619751},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7305548191070557},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.563318133354187},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5604845285415649},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.5387199521064758},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5218364596366882},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3875812292098999},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37906116247177124},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.32708579301834106}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.21437/interspeech.2022-10809","is_oa":false,"landing_page_url":"https://doi.org/10.21437/interspeech.2022-10809","pdf_url":null,"source":{"id":"https://openalex.org/S4363604309","display_name":"Interspeech 2022","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":"Interspeech 2022","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.5199999809265137,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2948148442","https://openalex.org/W2461250372","https://openalex.org/W2394342941","https://openalex.org/W2169853506","https://openalex.org/W2547124190","https://openalex.org/W2350586049","https://openalex.org/W2385628723","https://openalex.org/W2032414556","https://openalex.org/W2169871401","https://openalex.org/W2057878850"],"abstract_inverted_index":{"Reducing":[0],"the":[1,42,107,119,135,142,148,184],"latency":[2],"and":[3,36,87,115,122,146,164,174],"model":[4,24,128,170],"size":[5,131,180],"has":[6,26,178],"always":[7],"been":[8],"a":[9,53,124,154,166],"significant":[10],"research":[11],"problem":[12],"for":[13,112,141],"live":[14],"Automatic":[15],"Speech":[16],"Recognition":[17],"(ASR)":[18],"application":[19],"scenarios.":[20],"Along":[21],"this":[22,63],"direction,":[23],"quantization":[25,74,102,117,152],"become":[27],"an":[28],"increasingly":[29],"popular":[30],"approach":[31],"to":[32,67,82,98,134,183],"compress":[33],"neural":[34],"networks":[35],"reduce":[37],"computation":[38],"cost.":[39],"Most":[40],"of":[41,109,150],"existing":[43],"practical":[44,155],"ASR":[45,70,96,156,169],"systems":[46],"apply":[47],"post-training":[48],"8-bit":[49,175],"quantization.":[50],"To":[51],"achieve":[52],"higher":[54],"compression":[55],"rate":[56],"without":[57],"introducing":[58],"additional":[59],"performance":[60],"regression,":[61],"in":[62],"study,":[64],"we":[65,105,140],"propose":[66],"develop":[68],"4-bit":[69,126,151,173],"models":[71,97],"with":[72,129,161,171],"native":[73,79],"aware":[75],"training,":[76],"which":[77],"leverages":[78],"integer":[80],"operations":[81],"effectively":[83],"optimize":[84],"both":[85,113],"training":[86],"inference.":[88],"We":[89],"conducted":[90],"two":[91],"experiments":[92],"on":[93,118,153],"state-of-the-art":[94],"Conformer-based":[95],"evaluate":[99],"our":[100],"proposed":[101],"technique.":[103],"First,":[104],"explored":[106],"impact":[108],"different":[110],"precisions":[111],"weight":[114],"activation":[116],"LibriSpeech":[120],"dataset,":[121],"obtained":[123],"lossless":[125,167],"Conformer":[127,168],"7.7x":[130],"reduction":[132,181],"compared":[133,182],"float32":[136,185],"model.":[137,186],"Following":[138],"this,":[139],"first":[143],"time":[144],"investigated":[145],"revealed":[147],"viability":[149],"system":[157],"that":[158,177],"is":[159],"trained":[160],"large-scale":[162],"datasets,":[163],"produced":[165],"mixed":[172],"weights":[176],"5x":[179]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":7},{"year":2023,"cited_by_count":9},{"year":2022,"cited_by_count":1}],"updated_date":"2026-05-05T08:41:31.759640","created_date":"2025-10-10T00:00:00"}