{"id":"https://openalex.org/W3196364802","doi":"https://doi.org/10.21437/interspeech.2021-1962","title":"4-Bit Quantization of LSTM-Based Speech Recognition Models","display_name":"4-Bit Quantization of LSTM-Based Speech Recognition Models","publication_year":2021,"publication_date":"2021-08-27","ids":{"openalex":"https://openalex.org/W3196364802","doi":"https://doi.org/10.21437/interspeech.2021-1962","mag":"3196364802"},"language":"en","primary_location":{"id":"doi:10.21437/interspeech.2021-1962","is_oa":false,"landing_page_url":"https://doi.org/10.21437/interspeech.2021-1962","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Interspeech 2021","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/A5026800272","display_name":"Andrea Fasoli","orcid":"https://orcid.org/0000-0001-6892-5139"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Andrea Fasoli","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101406569","display_name":"Chia\u2010Yu Chen","orcid":"https://orcid.org/0000-0001-5542-7149"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chia-Yu Chen","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042998510","display_name":"Maur\u00edcio Serrano","orcid":"https://orcid.org/0000-0003-0250-5881"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mauricio Serrano","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101886784","display_name":"Xiao Sun","orcid":"https://orcid.org/0000-0002-8385-7020"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao Sun","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082043392","display_name":"Naigang Wang","orcid":"https://orcid.org/0000-0001-7664-0061"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Naigang Wang","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010094713","display_name":"Swagath Venkataramani","orcid":"https://orcid.org/0000-0002-0470-6364"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Swagath Venkataramani","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079994647","display_name":"George Saon","orcid":"https://orcid.org/0009-0004-6837-5009"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"George Saon","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102014291","display_name":"Xiaodong Cui","orcid":"https://orcid.org/0000-0003-4865-1307"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaodong Cui","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003725957","display_name":"Brian Kingsbury","orcid":"https://orcid.org/0000-0002-1343-6837"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Brian Kingsbury","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5057842914","display_name":"Wei Zhang","orcid":"https://orcid.org/0000-0002-8415-1062"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei Zhang","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047353872","display_name":"Zolt\u00e1n T\u00fcske","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zolt\u00e1n T\u00fcske","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5109912930","display_name":"Kailash Gopalakrishnan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kailash Gopalakrishnan","raw_affiliation_strings":["{cchen"],"affiliations":[{"raw_affiliation_string":"{cchen","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":12,"corresponding_author_ids":["https://openalex.org/A5026800272"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.2357,"has_fulltext":false,"cited_by_count":19,"citation_normalized_percentile":{"value":0.90008534,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":93,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"2586","last_page":"2590"},"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":0.9988999962806702,"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":0.9900000095367432,"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/computer-science","display_name":"Computer science","score":0.7826876640319824},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.7708836197853088},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.701021134853363},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.49241459369659424},{"id":"https://openalex.org/keywords/bit","display_name":"Bit (key)","score":0.43431344628334045},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.1334156095981598},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.05516546964645386}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7826876640319824},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7708836197853088},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.701021134853363},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49241459369659424},{"id":"https://openalex.org/C117011727","wikidata":"https://www.wikidata.org/wiki/Q1278488","display_name":"Bit (key)","level":2,"score":0.43431344628334045},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.1334156095981598},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.05516546964645386}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.21437/interspeech.2021-1962","is_oa":false,"landing_page_url":"https://doi.org/10.21437/interspeech.2021-1962","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Interspeech 2021","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","score":0.6399999856948853,"display_name":"Peace, Justice and strong institutions"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":31,"referenced_works":["https://openalex.org/W2058641082","https://openalex.org/W2166637769","https://openalex.org/W2242818861","https://openalex.org/W2394932179","https://openalex.org/W2407080277","https://openalex.org/W2507699225","https://openalex.org/W2508418541","https://openalex.org/W2524428287","https://openalex.org/W2557257847","https://openalex.org/W2622263826","https://openalex.org/W2786951478","https://openalex.org/W2809624076","https://openalex.org/W2936774411","https://openalex.org/W2937780860","https://openalex.org/W2962760690","https://openalex.org/W2963266252","https://openalex.org/W2963381607","https://openalex.org/W2970233825","https://openalex.org/W2973188344","https://openalex.org/W2975044525","https://openalex.org/W2981680926","https://openalex.org/W2988070836","https://openalex.org/W3005538384","https://openalex.org/W3007227084","https://openalex.org/W3016230677","https://openalex.org/W3017746288","https://openalex.org/W3097522836","https://openalex.org/W3163300396","https://openalex.org/W3172967059","https://openalex.org/W4289478342","https://openalex.org/W4289752678"],"related_works":["https://openalex.org/W3107474891","https://openalex.org/W2893763841","https://openalex.org/W2368779261","https://openalex.org/W2794438528","https://openalex.org/W4249640049","https://openalex.org/W4246522322","https://openalex.org/W2778699561","https://openalex.org/W2319139559","https://openalex.org/W2312116756","https://openalex.org/W4238216420"],"abstract_inverted_index":{"We":[0],"investigate":[1],"the":[2,48,62,88,92,105,114,135],"impact":[3],"of":[4,8,15,51,77,91,113,124],"aggressive":[5],"low-precision":[6],"representations":[7],"weights":[9],"and":[10,32,79,108,129],"activations":[11],"in":[12,55,145],"two":[13],"families":[14],"large":[16],"LSTM-based":[17],"architectures":[18],"for":[19],"Automatic":[20],"Speech":[21],"Recognition":[22],"(ASR):":[23],"hybrid":[24],"Deep":[25],"Bidirectional":[26],"LSTM":[27,49],"-Hidden":[28],"Markov":[29],"Models":[30],"(DBLSTM-HMMs)":[31],"Recurrent":[33],"Neural":[34],"Network":[35],"-Transducers":[36],"(RNN-Ts).Using":[37],"a":[38,42],"4-bit":[39,146],"integer":[40],"representation,":[41],"na\u00efve":[43],"quantization":[44,84,141],"approach":[45],"applied":[46],"to":[47,148],"portion":[50],"these":[52],"models":[53],"results":[54],"significant":[56],"Word":[57],"Error":[58],"Rate":[59],"(WER)":[60],"degradation.On":[61],"other":[63],"hand,":[64],"we":[65,82],"show":[66],"that":[67],"minimal":[68],"accuracy":[69],"loss":[70],"is":[71],"achievable":[72],"with":[73,119],"an":[74],"appropriate":[75],"choice":[76],"quantizers":[78],"initializations.In":[80],"particular,":[81],"customize":[83],"schemes":[85],"depending":[86],"on":[87,104],"local":[89],"properties":[90],"network,":[93],"improving":[94],"recognition":[95],"performance":[96],"while":[97],"limiting":[98],"computational":[99],"time.We":[100],"demonstrate":[101],"our":[102,140],"solution":[103],"Switchboard":[106],"(SWB)":[107],"CallHome":[109],"(CH)":[110],"test":[111],"sets":[112],"NIST":[115],"Hub5-2000":[116],"evaluation.DBLSTM-HMMs":[117],"trained":[118],"300":[120],"or":[121],"2000":[122],"hours":[123],"SWB":[125],"data":[126],"achieves":[127],"<0.5%":[128],"<1%":[130],"average":[131],"WER":[132],"degradation,":[133],"respectively.On":[134],"more":[136],"challenging":[137],"RNN-T":[138],"models,":[139],"strategy":[142],"limits":[143],"degradation":[144],"inference":[147],"1.3%.":[149]},"counts_by_year":[{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":3},{"year":2023,"cited_by_count":6},{"year":2022,"cited_by_count":5},{"year":2021,"cited_by_count":2}],"updated_date":"2026-03-09T08:58:05.943551","created_date":"2025-10-10T00:00:00"}