{"id":"https://openalex.org/W3094979069","doi":"https://doi.org/10.21437/interspeech.2020-1557","title":"Efficient Minimum Word Error Rate Training of RNN-Transducer for End-to-End Speech Recognition","display_name":"Efficient Minimum Word Error Rate Training of RNN-Transducer for End-to-End Speech Recognition","publication_year":2020,"publication_date":"2020-10-25","ids":{"openalex":"https://openalex.org/W3094979069","doi":"https://doi.org/10.21437/interspeech.2020-1557","mag":"3094979069"},"language":"en","primary_location":{"id":"doi:10.21437/interspeech.2020-1557","is_oa":false,"landing_page_url":"https://doi.org/10.21437/interspeech.2020-1557","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Interspeech 2020","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/A5103232491","display_name":"Jinxi Guo","orcid":"https://orcid.org/0000-0001-9563-7351"},"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":"Jinxi Guo","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003334789","display_name":"Gautam Tiwari","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":"Gautam Tiwari","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012153296","display_name":"Jasha Droppo","orcid":"https://orcid.org/0000-0001-6097-0090"},"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":"Jasha Droppo","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113701682","display_name":"Maarten Van Segbroeck","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":"Maarten Van Segbroeck","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100891231","display_name":"Che-Wei Huang","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":"Che-Wei Huang","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060979948","display_name":"Andreas Stolcke","orcid":"https://orcid.org/0000-0002-9925-905X"},"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":"Andreas Stolcke","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5090454997","display_name":"Roland Maas","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":"Roland Maas","raw_affiliation_strings":["Amazon.com, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon.com, USA","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":4.5999,"has_fulltext":false,"cited_by_count":52,"citation_normalized_percentile":{"value":0.95942649,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"2807","last_page":"2811"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","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"}},"topics":[{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","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/T10181","display_name":"Natural Language Processing Techniques","score":0.9901999831199646,"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":0.9715999960899353,"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/end-to-end-principle","display_name":"End-to-end principle","score":0.7334920763969421},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7315358519554138},{"id":"https://openalex.org/keywords/word-error-rate","display_name":"Word error rate","score":0.7104937434196472},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.7011170387268066},{"id":"https://openalex.org/keywords/word","display_name":"Word (group theory)","score":0.5745428204536438},{"id":"https://openalex.org/keywords/transducer","display_name":"Transducer","score":0.45747682452201843},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.4491722583770752},{"id":"https://openalex.org/keywords/recurrent-neural-network","display_name":"Recurrent neural network","score":0.42354047298431396},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3708493411540985},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.15011116862297058},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.1352050006389618},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.10531038045883179},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.06902885437011719}],"concepts":[{"id":"https://openalex.org/C74296488","wikidata":"https://www.wikidata.org/wiki/Q2527392","display_name":"End-to-end principle","level":2,"score":0.7334920763969421},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7315358519554138},{"id":"https://openalex.org/C40969351","wikidata":"https://www.wikidata.org/wiki/Q3516228","display_name":"Word error rate","level":2,"score":0.7104937434196472},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.7011170387268066},{"id":"https://openalex.org/C90805587","wikidata":"https://www.wikidata.org/wiki/Q10944557","display_name":"Word (group theory)","level":2,"score":0.5745428204536438},{"id":"https://openalex.org/C56318395","wikidata":"https://www.wikidata.org/wiki/Q215928","display_name":"Transducer","level":2,"score":0.45747682452201843},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.4491722583770752},{"id":"https://openalex.org/C147168706","wikidata":"https://www.wikidata.org/wiki/Q1457734","display_name":"Recurrent neural network","level":3,"score":0.42354047298431396},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3708493411540985},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.15011116862297058},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.1352050006389618},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.10531038045883179},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.06902885437011719},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","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/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.21437/interspeech.2020-1557","is_oa":false,"landing_page_url":"https://doi.org/10.21437/interspeech.2020-1557","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Interspeech 2020","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.5799999833106995,"display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":16,"referenced_works":["https://openalex.org/W1522301498","https://openalex.org/W1828163288","https://openalex.org/W2102113734","https://openalex.org/W2121879602","https://openalex.org/W2131342762","https://openalex.org/W2327501763","https://openalex.org/W2746192915","https://openalex.org/W2935756939","https://openalex.org/W2936774411","https://openalex.org/W2962760690","https://openalex.org/W2962765220","https://openalex.org/W2963414781","https://openalex.org/W2963747784","https://openalex.org/W2964045208","https://openalex.org/W3015927303","https://openalex.org/W3095697114"],"related_works":["https://openalex.org/W3179968364","https://openalex.org/W4225394202","https://openalex.org/W4298287631","https://openalex.org/W2953061907","https://openalex.org/W3032952384","https://openalex.org/W3034302643","https://openalex.org/W1847088711","https://openalex.org/W3036642985","https://openalex.org/W2072454424","https://openalex.org/W2944691285"],"abstract_inverted_index":{"In":[0],"this":[1,22,100],"work,":[2],"we":[3,42,85,152],"propose":[4],"a":[5,116,122,154,173,186],"novel":[6],"and":[7,30,44,61,80,83,90,113,162,185],"efficient":[8],"minimum":[9],"word":[10],"error":[11],"rate":[12],"(MWER)":[13],"training":[14,81,92,128],"method":[15,73,103,109,158],"for":[16,34,52,93,145],"RNN-Transducer":[17],"(RNN-T).Unlike":[18],"previous":[19],"work":[20],"on":[21,177,190],"topic,":[23],"which":[24],"performs":[25],"on-the-fly":[26,108],"limited-size":[27],"beam-search":[28],"decoding":[29,79],"generates":[31],"alignment":[32],"scores":[33,46,60],"expected":[35],"edit-distance":[36],"computation,":[37],"in":[38,55,115],"our":[39],"proposed":[40,72,101,126,155],"method,":[41],"re-calculate":[43],"sum":[45],"of":[47,181],"all":[48],"the":[49,68,71,78,107],"possible":[50],"alignments":[51],"each":[53,94],"hypothesis":[54,58],"N-best":[56],"lists.The":[57],"probability":[59],"back-propagated":[62],"gradients":[63],"are":[64],"calculated":[65],"efficiently":[66],"using":[67],"forward-backward":[69],"algorithm.Moreover,":[70],"allows":[74],"us":[75],"to":[76,159,166],"decouple":[77],"processes,":[82],"thus":[84],"can":[86,104,129,148],"perform":[87,167],"offline":[88],"parallel-decoding":[89],"MWER":[91,127],"subset":[95],"iteratively.Experimental":[96],"results":[97],"show":[98],"that":[99],"semi-on-the-fly":[102],"speed":[105],"up":[106],"by":[110,138],"6":[111],"times":[112],"result":[114],"similar":[117],"WER":[118,188],"improvement":[119,147,176],"(3.6%)":[120],"over":[121],"baseline":[123],"RNN-T":[124,139,156],"model.The":[125],"also":[130],"effectively":[131],"reduce":[132],"high-deletion":[133],"errors":[134],"(9.2%":[135],"WER-reduction)":[136],"introduced":[137],"models":[140],"when":[141],"EOS":[142],"is":[143],"added":[144],"endpointer.Further":[146],"be":[149],"achieved":[150],"if":[151],"use":[153,163],"rescoring":[157],"re-rank":[160],"hypotheses":[161],"external":[164],"RNN-LM":[165],"additional":[168],"rescoring.The":[169],"best":[170],"system":[171],"achieves":[172],"5%":[174],"relative":[175],"an":[178],"English":[179],"test-set":[180],"real":[182],"far-field":[183],"recordings":[184],"11.6%":[187],"reduction":[189],"music-domain":[191],"utterances.":[192]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":8},{"year":2024,"cited_by_count":3},{"year":2023,"cited_by_count":5},{"year":2022,"cited_by_count":13},{"year":2021,"cited_by_count":18},{"year":2020,"cited_by_count":4}],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}