{"id":"https://openalex.org/W3163368926","doi":"https://doi.org/10.1109/icassp39728.2021.9414076","title":"Mixed Precision Quantization of Transformer Language Models for Speech Recognition","display_name":"Mixed Precision Quantization of Transformer Language Models for Speech Recognition","publication_year":2021,"publication_date":"2021-05-13","ids":{"openalex":"https://openalex.org/W3163368926","doi":"https://doi.org/10.1109/icassp39728.2021.9414076","mag":"3163368926"},"language":"en","primary_location":{"id":"doi:10.1109/icassp39728.2021.9414076","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp39728.2021.9414076","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2021 - 2021 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/A5011790590","display_name":"Junhao Xu","orcid":"https://orcid.org/0000-0002-9650-0290"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Junhao Xu","raw_affiliation_strings":["The Chinese University of Hong Kong, Hong Kong SAR, China"],"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong, Hong Kong SAR, China","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045355404","display_name":"Shoukang Hu","orcid":"https://orcid.org/0000-0002-3345-6923"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shoukang Hu","raw_affiliation_strings":["The Chinese University of Hong Kong, Hong Kong SAR, China"],"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong, Hong Kong SAR, China","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004643540","display_name":"Jianwei Yu","orcid":"https://orcid.org/0000-0002-2449-1436"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jianwei Yu","raw_affiliation_strings":["The Chinese University of Hong Kong, Hong Kong SAR, China"],"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong, Hong Kong SAR, China","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037109470","display_name":"Xunying Liu","orcid":"https://orcid.org/0000-0001-6725-1160"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xunying Liu","raw_affiliation_strings":["The Chinese University of Hong Kong, Hong Kong SAR, China"],"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong, Hong Kong SAR, China","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5019458385","display_name":"Helen Meng","orcid":"https://orcid.org/0000-0002-4427-3532"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Helen Meng","raw_affiliation_strings":["The Chinese University of Hong Kong, Hong Kong SAR, China"],"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong, Hong Kong SAR, China","institution_ids":["https://openalex.org/I177725633"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5011790590"],"corresponding_institution_ids":["https://openalex.org/I177725633"],"apc_list":null,"apc_paid":null,"fwci":1.0877,"has_fulltext":false,"cited_by_count":14,"citation_normalized_percentile":{"value":0.81437529,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":93,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"7383","last_page":"7387"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9994999766349792,"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":0.9994999766349792,"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":0.9891999959945679,"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":0.9830999970436096,"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":0.7152740955352783},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6722592115402222},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5609980225563049},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.4836243689060211},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.480155885219574},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.46711501479148865},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.4071439504623413},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.34450116753578186},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.10247847437858582}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7152740955352783},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6722592115402222},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5609980225563049},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4836243689060211},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.480155885219574},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.46711501479148865},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4071439504623413},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.34450116753578186},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.10247847437858582},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icassp39728.2021.9414076","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp39728.2021.9414076","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Quality Education","score":0.4399999976158142,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":52,"referenced_works":["https://openalex.org/W1902934009","https://openalex.org/W2057450058","https://openalex.org/W2150769028","https://openalex.org/W2161758346","https://openalex.org/W2164278908","https://openalex.org/W2194775991","https://openalex.org/W2233116163","https://openalex.org/W2267186426","https://openalex.org/W2279098554","https://openalex.org/W2300242332","https://openalex.org/W2405920868","https://openalex.org/W2413794162","https://openalex.org/W2514741789","https://openalex.org/W2524428287","https://openalex.org/W2597655663","https://openalex.org/W2613904329","https://openalex.org/W2739789140","https://openalex.org/W2786951478","https://openalex.org/W2810075754","https://openalex.org/W2899663614","https://openalex.org/W2900694899","https://openalex.org/W2943845043","https://openalex.org/W2951104886","https://openalex.org/W2952265507","https://openalex.org/W2962833442","https://openalex.org/W2963114950","https://openalex.org/W2963386218","https://openalex.org/W2963403868","https://openalex.org/W2963821229","https://openalex.org/W2964228333","https://openalex.org/W2964265128","https://openalex.org/W2982041622","https://openalex.org/W2989530497","https://openalex.org/W2998704965","https://openalex.org/W3008191852","https://openalex.org/W3016230677","https://openalex.org/W3048889251","https://openalex.org/W4285719527","https://openalex.org/W4288026258","https://openalex.org/W4385245566","https://openalex.org/W6680532216","https://openalex.org/W6695314431","https://openalex.org/W6698200048","https://openalex.org/W6714058667","https://openalex.org/W6734062232","https://openalex.org/W6735377749","https://openalex.org/W6737778391","https://openalex.org/W6739901393","https://openalex.org/W6748131973","https://openalex.org/W6752515464","https://openalex.org/W6755977528","https://openalex.org/W6780226713"],"related_works":["https://openalex.org/W4288365749","https://openalex.org/W2936497627","https://openalex.org/W3013624417","https://openalex.org/W4287826556","https://openalex.org/W3098382480","https://openalex.org/W4287598411","https://openalex.org/W3100913109","https://openalex.org/W3198458223","https://openalex.org/W3126642501","https://openalex.org/W2964413124"],"abstract_inverted_index":{"State-of-the-art":[0],"neural":[1,18],"language":[2],"models":[3],"represented":[4],"by":[5],"Transformers":[6],"are":[7,36,68,78,118],"becoming":[8],"increasingly":[9],"complex":[10],"and":[11,41,134],"expensive":[12],"for":[13,45],"practical":[14],"applications.":[15],"Low-bit":[16],"deep":[17],"network":[19],"quantization":[20,34,57,66,90,100,148],"techniques":[21,149],"provides":[22],"a":[23,89,135,175],"powerful":[24],"solution":[25],"to":[26,43,56,120,157,173,191],"dramatically":[27],"reduce":[28],"their":[29],"model":[30,151],"size.":[31],"Current":[32],"low-bit":[33],"methods":[35,67,114],"based":[37,87,105],"on":[38,88,106,130],"uniform":[39],"precision":[40,64,76,108,124,146,163,178],"fail":[42],"account":[44],"the":[46,54,94,143,161],"varying":[47],"performance":[48,168],"sensitivity":[49,91],"at":[50],"different":[51],"parts":[52],"of":[53,96,115,155],"system":[55,141],"errors.":[58],"To":[59],"this":[60,71],"end,":[61],"novel":[62],"mixed":[63,107,123,145],"DNN":[65,126],"proposed":[69,144],"in":[70,93],"paper.":[72],"The":[73,84,102],"optimal":[74],"local":[75],"settings":[77],"automatically":[79],"learned":[80],"using":[81],"two":[82],"techniques.":[83],"first":[85],"is":[86,104],"metric":[92],"form":[95],"Hessian":[97],"trace":[98],"weighted":[99],"perturbation.":[101],"second":[103],"Transformer":[109,147,179],"architecture":[110],"search.":[111],"Alternating":[112],"direction":[113],"multipliers":[116],"(ADMM)":[117],"used":[119,172],"efficiently":[121],"train":[122],"quantized":[125],"systems.":[127],"Experiments":[128],"conducted":[129],"Penn":[131],"Treebank":[132],"(PTB)":[133],"Switchboard":[136],"corpus":[137],"trained":[138],"LF-MMI":[139],"TDNN":[140],"suggest":[142],"achieved":[150],"size":[152],"compression":[153],"ratios":[154],"up":[156,190],"16":[158],"times":[159],"over":[160],"full":[162,177],"baseline":[164],"with":[165,181],"no":[166],"recognition":[167],"degradation.":[169],"When":[170],"being":[171],"compress":[174],"larger":[176],"LM":[180],"more":[182],"layers,":[183],"overall":[184],"word":[185],"error":[186],"rate":[187],"(WER)":[188],"reductions":[189],"1.7%":[192],"absolute":[193],"(18%":[194],"relative)":[195],"were":[196],"obtained.":[197]},"counts_by_year":[{"year":2025,"cited_by_count":6},{"year":2024,"cited_by_count":2},{"year":2023,"cited_by_count":2},{"year":2022,"cited_by_count":2},{"year":2021,"cited_by_count":2}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}