{"id":"https://openalex.org/W4392903749","doi":"https://doi.org/10.1109/icassp48485.2024.10446240","title":"Effective Internal Language Model Training and Fusion for Factorized Transducer Model","display_name":"Effective Internal Language Model Training and Fusion for Factorized Transducer Model","publication_year":2024,"publication_date":"2024-03-18","ids":{"openalex":"https://openalex.org/W4392903749","doi":"https://doi.org/10.1109/icassp48485.2024.10446240"},"language":"en","primary_location":{"id":"doi:10.1109/icassp48485.2024.10446240","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp48485.2024.10446240","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","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/A5103232491","display_name":"Jinxi Guo","orcid":"https://orcid.org/0000-0001-9563-7351"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Jinxi Guo","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013781881","display_name":"Niko Moritz","orcid":"https://orcid.org/0000-0003-4352-8115"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Niko Moritz","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103207430","display_name":"Yingyi Ma","orcid":"https://orcid.org/0000-0002-0205-0197"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yingyi Ma","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072932051","display_name":"Frank Seide","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Frank Seide","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103012144","display_name":"Chunyang Wu","orcid":"https://orcid.org/0000-0002-5796-8288"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chunyang Wu","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074237839","display_name":"Jay Mahadeokar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jay Mahadeokar","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066166549","display_name":"Ozlem Kalinli","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ozlem Kalinli","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047073253","display_name":"Christian Fuegen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Christian Fuegen","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5113773386","display_name":"Mike Seltzer","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mike Seltzer","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5103232491"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.6891,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.72657832,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"12687","last_page":"12691"},"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":0.9995999932289124,"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/T10028","display_name":"Topic Modeling","score":0.9980000257492065,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8004040718078613},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.7455596327781677},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.7122650146484375},{"id":"https://openalex.org/keywords/transducer","display_name":"Transducer","score":0.554653525352478},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5311191082000732},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5043991804122925},{"id":"https://openalex.org/keywords/word-error-rate","display_name":"Word error rate","score":0.4905058741569519},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.4291737675666809},{"id":"https://openalex.org/keywords/acoustic-model","display_name":"Acoustic model","score":0.4182223081588745},{"id":"https://openalex.org/keywords/rendering","display_name":"Rendering (computer graphics)","score":0.41078630089759827},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.3994576930999756},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.35997673869132996},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.30234062671661377},{"id":"https://openalex.org/keywords/speech-processing","display_name":"Speech processing","score":0.15423175692558289},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.09366849064826965}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8004040718078613},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.7455596327781677},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.7122650146484375},{"id":"https://openalex.org/C56318395","wikidata":"https://www.wikidata.org/wiki/Q215928","display_name":"Transducer","level":2,"score":0.554653525352478},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5311191082000732},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5043991804122925},{"id":"https://openalex.org/C40969351","wikidata":"https://www.wikidata.org/wiki/Q3516228","display_name":"Word error rate","level":2,"score":0.4905058741569519},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4291737675666809},{"id":"https://openalex.org/C155635449","wikidata":"https://www.wikidata.org/wiki/Q4674699","display_name":"Acoustic model","level":3,"score":0.4182223081588745},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.41078630089759827},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3994576930999756},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35997673869132996},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.30234062671661377},{"id":"https://openalex.org/C61328038","wikidata":"https://www.wikidata.org/wiki/Q3358061","display_name":"Speech processing","level":2,"score":0.15423175692558289},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.09366849064826965},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icassp48485.2024.10446240","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp48485.2024.10446240","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":22,"referenced_works":["https://openalex.org/W1494198834","https://openalex.org/W2127141656","https://openalex.org/W2131342762","https://openalex.org/W2327501763","https://openalex.org/W2936774411","https://openalex.org/W2962760690","https://openalex.org/W2962765220","https://openalex.org/W3008037978","https://openalex.org/W3016234571","https://openalex.org/W3094979069","https://openalex.org/W3149509723","https://openalex.org/W3152221657","https://openalex.org/W3162665866","https://openalex.org/W3184976814","https://openalex.org/W3197507772","https://openalex.org/W3202419788","https://openalex.org/W4223954171","https://openalex.org/W4319862418","https://openalex.org/W4372266720","https://openalex.org/W4375869444","https://openalex.org/W6638749077","https://openalex.org/W6675365184"],"related_works":["https://openalex.org/W2594897229","https://openalex.org/W2151348424","https://openalex.org/W4221142855","https://openalex.org/W2050138804","https://openalex.org/W4290708361","https://openalex.org/W2129812225","https://openalex.org/W26527944","https://openalex.org/W2523799048","https://openalex.org/W2155620340","https://openalex.org/W2126322296"],"abstract_inverted_index":{"The":[0,160],"internal":[1,54],"language":[2,38,55,171],"model":[3,56,144,162],"(ILM)":[4],"of":[5,42,66],"the":[6,23,64,98,112,122,142,183],"neural":[7],"transducer":[8,44,68,93],"has":[9,72],"been":[10,47,73],"widely":[11],"studied.":[12],"In":[13,79],"most":[14],"prior":[15],"work,":[16],"it":[17,174],"is":[18,27],"mainly":[19],"used":[20],"for":[21,57,91,157,177],"estimating":[22],"ILM":[24,86,102,120,201],"score":[25],"and":[26,88,101,121,152,189],"subsequently":[28],"subtracted":[29],"during":[30],"inference":[31],"to":[32,76,132],"facilitate":[33],"improved":[34],"integration":[35,202],"with":[36,63,138],"external":[37,139,170],"models.":[39],"Recently,":[40],"various":[41],"factorized":[43,67,92],"models":[45],"have":[46],"proposed,":[48],"which":[49,95,199],"explicitly":[50],"embrace":[51],"a":[52,84,107,118,133,146,187],"standalone":[53],"non-blank":[58],"token":[59],"prediction.":[60],"However,":[61],"even":[62],"adoption":[65],"models,":[69,94,172],"limited":[70],"improvement":[71,110,149],"observed":[74],"compared":[75,131],"shallow":[77],"fusion.":[78],"this":[80],"paper,":[81],"we":[82,185],"propose":[83,186],"novel":[85,188],"training":[87,197],"decoding":[89,114,124],"strategy":[90,125],"effectively":[96],"combines":[97],"blank,":[99],"acoustic":[100],"scores.":[103],"Our":[104],"experiments":[105],"show":[106],"17%":[108],"relative":[109,148],"over":[111],"standard":[113],"method":[115,198],"when":[116,130],"utilizing":[117],"well-trained":[119],"proposed":[123,143,161],"on":[126,150,169],"LibriSpeech":[127],"datasets.":[128],"Furthermore,":[129],"strong":[134],"RNN-T":[135],"baseline":[136],"enhanced":[137],"LM":[140],"fusion,":[141],"yields":[145],"5.5%":[147],"general-sets":[151],"an":[153],"8.9%":[154],"WER":[155],"reduction":[156],"rare":[158],"words.":[159],"can":[163],"achieve":[164],"superior":[165],"performance":[166],"without":[167],"relying":[168],"rendering":[173],"highly":[175],"efficient":[176],"production":[178],"use-cases.":[179],"To":[180],"further":[181],"improve":[182],"performance,":[184],"memory-efficient":[190],"ILM-fusion-aware":[191],"minimum":[192],"word":[193],"error":[194],"rate":[195],"(MWER)":[196],"improves":[200],"significantly.":[203]},"counts_by_year":[{"year":2025,"cited_by_count":2}],"updated_date":"2026-04-03T22:45:19.894376","created_date":"2025-10-10T00:00:00"}