{"id":"https://openalex.org/W4319862415","doi":"https://doi.org/10.1109/slt54892.2023.10022821","title":"Sub-8-Bit Quantization for On-Device Speech Recognition: A Regularization-Free Approach","display_name":"Sub-8-Bit Quantization for On-Device Speech Recognition: A Regularization-Free Approach","publication_year":2023,"publication_date":"2023-01-09","ids":{"openalex":"https://openalex.org/W4319862415","doi":"https://doi.org/10.1109/slt54892.2023.10022821"},"language":"en","primary_location":{"id":"doi:10.1109/slt54892.2023.10022821","is_oa":false,"landing_page_url":"https://doi.org/10.1109/slt54892.2023.10022821","pdf_url":null,"source":{"id":"https://openalex.org/S4363605953","display_name":"2022 IEEE Spoken Language Technology Workshop (SLT)","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":"2022 IEEE Spoken Language Technology Workshop (SLT)","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/A5047870570","display_name":"Kai Zhen","orcid":"https://orcid.org/0000-0003-4133-8158"},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kai Zhen","raw_affiliation_strings":["Amazon Alexa AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Alexa AI","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023415990","display_name":"Martin Radfar","orcid":null},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Martin Radfar","raw_affiliation_strings":["Amazon Alexa AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Alexa AI","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101663835","display_name":"Hieu Duy Nguyen","orcid":"https://orcid.org/0000-0002-8774-5122"},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hieu Nguyen","raw_affiliation_strings":["Amazon Alexa AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Alexa AI","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023708721","display_name":"Grant P. Strimel","orcid":null},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Grant P. Strimel","raw_affiliation_strings":["Amazon Alexa AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Alexa AI","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062762746","display_name":"Nathan Susanj","orcid":null},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nathan Susanj","raw_affiliation_strings":["Amazon Alexa AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Alexa AI","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029573067","display_name":"Athanasios Mouchtaris","orcid":"https://orcid.org/0000-0001-7583-0189"},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Athanasios Mouchtaris","raw_affiliation_strings":["Amazon Alexa AI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Alexa AI","institution_ids":["https://openalex.org/I1311688040"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I1311688040"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":8,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"15","last_page":"22"},"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/T10860","display_name":"Speech and Audio Processing","score":0.9997000098228455,"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/T11309","display_name":"Music and Audio Processing","score":0.9987000226974487,"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":0.7289612293243408},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7156258821487427},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.593281626701355},{"id":"https://openalex.org/keywords/centroid","display_name":"Centroid","score":0.5707355737686157},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5521675944328308},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.5213636159896851},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.519393265247345},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3752055764198303},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3439812660217285}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7289612293243408},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7156258821487427},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.593281626701355},{"id":"https://openalex.org/C146599234","wikidata":"https://www.wikidata.org/wiki/Q511093","display_name":"Centroid","level":2,"score":0.5707355737686157},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5521675944328308},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.5213636159896851},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.519393265247345},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3752055764198303},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3439812660217285},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/slt54892.2023.10022821","is_oa":false,"landing_page_url":"https://doi.org/10.1109/slt54892.2023.10022821","pdf_url":null,"source":{"id":"https://openalex.org/S4363605953","display_name":"2022 IEEE Spoken Language Technology Workshop (SLT)","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":"2022 IEEE Spoken Language Technology Workshop (SLT)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.46000000834465027}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":39,"referenced_works":["https://openalex.org/W1522301498","https://openalex.org/W2061715260","https://openalex.org/W2732044853","https://openalex.org/W2936774411","https://openalex.org/W2963414781","https://openalex.org/W2964110616","https://openalex.org/W2972354707","https://openalex.org/W3004392938","https://openalex.org/W3097522836","https://openalex.org/W3097777922","https://openalex.org/W3151287998","https://openalex.org/W3160766462","https://openalex.org/W3163368926","https://openalex.org/W3166859509","https://openalex.org/W3177113534","https://openalex.org/W3177265267","https://openalex.org/W3186609711","https://openalex.org/W3196364802","https://openalex.org/W3197125019","https://openalex.org/W3202442802","https://openalex.org/W3211787299","https://openalex.org/W3215615641","https://openalex.org/W4221138270","https://openalex.org/W4221155876","https://openalex.org/W4224215744","https://openalex.org/W4224918069","https://openalex.org/W4225307083","https://openalex.org/W4284957875","https://openalex.org/W4297841491","https://openalex.org/W4297841771","https://openalex.org/W6631190155","https://openalex.org/W6713134421","https://openalex.org/W6741057705","https://openalex.org/W6772080394","https://openalex.org/W6779405248","https://openalex.org/W6784225549","https://openalex.org/W6789240164","https://openalex.org/W6797854001","https://openalex.org/W6810049602"],"related_works":["https://openalex.org/W2381926679","https://openalex.org/W2007009951","https://openalex.org/W2082644203","https://openalex.org/W3165040664","https://openalex.org/W2350539780","https://openalex.org/W3122652148","https://openalex.org/W1583866266","https://openalex.org/W4242386713","https://openalex.org/W2996505764","https://openalex.org/W2039913993"],"abstract_inverted_index":{"For":[0],"on-device":[1],"automatic":[2],"speech":[3],"recognition":[4],"(ASR),":[5],"quantization":[6,32,69],"aware":[7],"training":[8],"(QAT)":[9],"is":[10,29],"ubiquitous":[11],"to":[12,35,78,114,134],"achieve":[13],"the":[14,31],"trade-off":[15],"between":[16],"model":[17],"predictive":[18],"performance":[19],"and":[20,38,87,93,105,109,118,128],"efficiency.":[21],"Among":[22],"existing":[23],"QAT":[24,136],"methods,":[25],"one":[26],"major":[27],"drawback":[28],"that":[30],"centroids":[33,53],"have":[34],"be":[36],"predetermined":[37],"fixed.":[39],"To":[40],"overcome":[41],"this":[42],"limitation,":[43],"we":[44],"introduce":[45],"a":[46,55,64,123],"regularization-free,":[47],"\u201csoft-to-hard\u201d":[48],"compression":[49],"mechanism":[50],"with":[51],"self-adjustable":[52],"in":[54,63],"<tex":[56],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[57],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$\\mu$</tex>":[58],"-Law":[59],"constrained":[60],"space,":[61],"resulting":[62],"simpler":[65],"yet":[66],"more":[67],"versatile":[68],"scheme,":[70],"called":[71],"General":[72],"Quantizer":[73],"(GQ).":[74],"We":[75,121],"apply":[76],"GQ":[77,100],"ASR":[79],"tasks":[80],"using":[81],"Recurrent":[82],"Neural":[83],"Network":[84],"Transducer":[85],"(RNN-T)":[86],"Conformer":[88,106],"architectures":[89],"on":[90],"both":[91,103],"LibriSpeech":[92],"de-identified":[94],"far-field":[95],"datasets.":[96],"Without":[97],"accuracy":[98],"degradation,":[99],"can":[101],"compress":[102],"RNN-T":[104,112],"into":[107],"sub-8-bit,":[108],"for":[110,116],"some":[111],"layers,":[113],"1-bit":[115],"fast":[117],"accurate":[119],"inference.":[120],"observe":[122],"30.73%":[124],"memory":[125],"footprint":[126],"saving":[127],"31.75%":[129],"user-perceived":[130],"latency":[131],"reduction":[132],"compared":[133],"8-bit":[135],"via":[137],"physical":[138],"device":[139],"benchmarking.":[140]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":2},{"year":2023,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}