{"id":"https://openalex.org/W4401597536","doi":"https://doi.org/10.1109/icasspw62465.2024.10627008","title":"Training Early-Exit Architectures for Automatic Speech Recognition: Fine-Tuning Pre-Trained Models or Training from Scratch","display_name":"Training Early-Exit Architectures for Automatic Speech Recognition: Fine-Tuning Pre-Trained Models or Training from Scratch","publication_year":2024,"publication_date":"2024-04-14","ids":{"openalex":"https://openalex.org/W4401597536","doi":"https://doi.org/10.1109/icasspw62465.2024.10627008"},"language":"en","primary_location":{"id":"doi:10.1109/icasspw62465.2024.10627008","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icasspw62465.2024.10627008","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW)","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/A5109764248","display_name":"George August Wright","orcid":null},"institutions":[{"id":"https://openalex.org/I193223587","display_name":"University of Trento","ror":"https://ror.org/05trd4x28","country_code":"IT","type":"education","lineage":["https://openalex.org/I193223587"]}],"countries":["IT"],"is_corresponding":false,"raw_author_name":"George August Wright","raw_affiliation_strings":["University of Trento"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Trento","institution_ids":["https://openalex.org/I193223587"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043797502","display_name":"Umberto Cappellazzo","orcid":"https://orcid.org/0009-0006-3443-5143"},"institutions":[{"id":"https://openalex.org/I193223587","display_name":"University of Trento","ror":"https://ror.org/05trd4x28","country_code":"IT","type":"education","lineage":["https://openalex.org/I193223587"]}],"countries":["IT"],"is_corresponding":false,"raw_author_name":"Umberto Cappellazzo","raw_affiliation_strings":["University of Trento"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Trento","institution_ids":["https://openalex.org/I193223587"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040954517","display_name":"Salah Zaiem","orcid":null},"institutions":[{"id":"https://openalex.org/I4210165912","display_name":"Laboratoire Traitement et Communication de l\u2019Information","ror":"https://ror.org/057er4c39","country_code":"FR","type":"facility","lineage":["https://openalex.org/I12356871","https://openalex.org/I205703379","https://openalex.org/I4210145102","https://openalex.org/I4210165912"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Salah Zaiem","raw_affiliation_strings":["Institut Polytechnique de Paris,LTCI, Télécom Paris"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institut Polytechnique de Paris,LTCI, Télécom Paris","institution_ids":["https://openalex.org/I4210165912"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004777817","display_name":"Desh Raj","orcid":"https://orcid.org/0000-0002-5038-9400"},"institutions":[{"id":"https://openalex.org/I145311948","display_name":"Johns Hopkins University","ror":"https://ror.org/00za53h95","country_code":"US","type":"education","lineage":["https://openalex.org/I145311948"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Desh Raj","raw_affiliation_strings":["Johns Hopkins University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Johns Hopkins University","institution_ids":["https://openalex.org/I145311948"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112980288","display_name":"Lucas Ondel Yang","orcid":null},"institutions":[{"id":"https://openalex.org/I1294671590","display_name":"Centre National de la Recherche Scientifique","ror":"https://ror.org/02feahw73","country_code":"FR","type":"government","lineage":["https://openalex.org/I1294671590"]},{"id":"https://openalex.org/I277688954","display_name":"Universit\u00e9 Paris-Saclay","ror":"https://ror.org/03xjwb503","country_code":"FR","type":"education","lineage":["https://openalex.org/I277688954"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Lucas Ondel Yang","raw_affiliation_strings":["Universite Paris-Saclay,LISN, CNRS"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Universite Paris-Saclay,LISN, CNRS","institution_ids":["https://openalex.org/I277688954","https://openalex.org/I1294671590"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041102462","display_name":"Daniele Falavigna","orcid":"https://orcid.org/0000-0002-4844-5071"},"institutions":[{"id":"https://openalex.org/I2277624104","display_name":"Fondazione Bruno Kessler","ror":"https://ror.org/01j33xk10","country_code":"IT","type":"facility","lineage":["https://openalex.org/I2277624104"]}],"countries":["IT"],"is_corresponding":false,"raw_author_name":"Daniele Falavigna","raw_affiliation_strings":["Fondazione Bruno Kessler"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fondazione Bruno Kessler","institution_ids":["https://openalex.org/I2277624104"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062830090","display_name":"Mohamed Nabih Ali","orcid":"https://orcid.org/0000-0001-9132-9220"},"institutions":[{"id":"https://openalex.org/I2277624104","display_name":"Fondazione Bruno Kessler","ror":"https://ror.org/01j33xk10","country_code":"IT","type":"facility","lineage":["https://openalex.org/I2277624104"]}],"countries":["IT"],"is_corresponding":false,"raw_author_name":"Mohamed Nabih Ali","raw_affiliation_strings":["Fondazione Bruno Kessler"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fondazione Bruno Kessler","institution_ids":["https://openalex.org/I2277624104"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5066363315","display_name":"Alessio Brutti","orcid":"https://orcid.org/0000-0003-4146-3071"},"institutions":[{"id":"https://openalex.org/I2277624104","display_name":"Fondazione Bruno Kessler","ror":"https://ror.org/01j33xk10","country_code":"IT","type":"facility","lineage":["https://openalex.org/I2277624104"]}],"countries":["IT"],"is_corresponding":false,"raw_author_name":"Alessio Brutti","raw_affiliation_strings":["Fondazione Bruno Kessler"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fondazione Bruno Kessler","institution_ids":["https://openalex.org/I2277624104"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":8,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.9164,"has_fulltext":false,"cited_by_count":3,"citation_normalized_percentile":{"value":0.78885772,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"685","last_page":"689"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9782999753952026,"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":0.9782999753952026,"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/scratch","display_name":"Scratch","score":0.8959658145904541},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7938779592514038},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.7659324407577515},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.690678060054779},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4691157639026642},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.42060378193855286},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.06977325677871704}],"concepts":[{"id":"https://openalex.org/C2781235140","wikidata":"https://www.wikidata.org/wiki/Q275131","display_name":"Scratch","level":2,"score":0.8959658145904541},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7938779592514038},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.7659324407577515},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.690678060054779},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4691157639026642},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.42060378193855286},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.06977325677871704},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","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/icasspw62465.2024.10627008","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icasspw62465.2024.10627008","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2024 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.6800000071525574,"display_name":"Quality Education"}],"awards":[],"funders":[{"id":"https://openalex.org/F4320307764","display_name":"Microsoft","ror":"https://ror.org/00d0nc645"},{"id":"https://openalex.org/F4320308737","display_name":"Facebook","ror":"https://ror.org/01zbnvs85"},{"id":"https://openalex.org/F4320309327","display_name":"Google","ror":"https://ror.org/00njsd438"},{"id":"https://openalex.org/F4320310077","display_name":"National Research Centre","ror":"https://ror.org/02n85j827"},{"id":"https://openalex.org/F4320310145","display_name":"Johns Hopkins University","ror":"https://ror.org/00za53h95"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":37,"referenced_works":["https://openalex.org/W1494198834","https://openalex.org/W1816313093","https://openalex.org/W1989549063","https://openalex.org/W2127141656","https://openalex.org/W2233116163","https://openalex.org/W2526425061","https://openalex.org/W2618530766","https://openalex.org/W2798722023","https://openalex.org/W2799473636","https://openalex.org/W2948981900","https://openalex.org/W2962677625","https://openalex.org/W2981884310","https://openalex.org/W3004034302","https://openalex.org/W3019321633","https://openalex.org/W3034368386","https://openalex.org/W3035030897","https://openalex.org/W3035038672","https://openalex.org/W3036601975","https://openalex.org/W3101648800","https://openalex.org/W3102775700","https://openalex.org/W3119308075","https://openalex.org/W3124540476","https://openalex.org/W3138154797","https://openalex.org/W3167533889","https://openalex.org/W3209059054","https://openalex.org/W3209984917","https://openalex.org/W4224000641","https://openalex.org/W4225285622","https://openalex.org/W4311604301","https://openalex.org/W4372347449","https://openalex.org/W4385484924","https://openalex.org/W4385585286","https://openalex.org/W6623517193","https://openalex.org/W6763608318","https://openalex.org/W6779313456","https://openalex.org/W6780218876","https://openalex.org/W6789394017"],"related_works":["https://openalex.org/W2475116013","https://openalex.org/W2770018148","https://openalex.org/W2358308169","https://openalex.org/W2385135707","https://openalex.org/W2140315382","https://openalex.org/W2059109728","https://openalex.org/W322691623","https://openalex.org/W2494989134","https://openalex.org/W4394050964","https://openalex.org/W2551249631"],"abstract_inverted_index":{"The":[0],"ability":[1],"to":[2,41,70,146,165],"dynamically":[3],"adjust":[4],"the":[5,45,58,115,166,175,185],"computational":[6,24,74],"load":[7],"of":[8,44,60,64,73,178,187],"neural":[9],"models":[10,49,62,85,111,126],"during":[11],"inference":[12],"is":[13,29],"crucial":[14],"for":[15,50,181],"on-device":[16],"processing":[17],"scenarios":[18],"characterised":[19],"by":[20,31],"limited":[21],"and":[22,68,76,109,190],"time-varying":[23],"resources.":[25],"A":[26],"promising":[27],"solution":[28],"presented":[30],"early-exit":[32,55,95,125,179],"architectures,":[33],"in":[34,159],"which":[35],"additional":[36],"exit":[37,155,191],"branches":[38],"are":[39],"appended":[40],"intermediate":[42],"layers":[43,138],"encoder.":[46],"In":[47,97],"self-attention":[48],"automatic":[51],"speech":[52],"recognition":[53],"(ASR),":[54],"architectures":[56,180],"enable":[57],"development":[59],"dynamic":[61],"capable":[63],"adapting":[65],"their":[66],"size":[67],"architecture":[69],"varying":[71],"levels":[72],"resources":[75],"ASR":[77,84,182],"performance":[78,133],"demands.":[79],"Previous":[80],"research":[81],"on":[82,88,120],"early-exiting":[83,116],"has":[86],"relied":[87],"pre-trained":[89,107,149],"self-supervised":[90],"models,":[91,183],"fine-tuned":[92],"with":[93,114],"an":[94,102,154,163],"loss.":[96],"this":[98],"paper,":[99],"we":[100,152],"undertake":[101],"experimental":[103],"comparison":[104],"between":[105],"fine-tuning":[106],"backbones":[108],"training":[110,176,188],"from":[112,128],"scratch":[113,129],"objective.":[117],"Experiments":[118],"conducted":[119],"public":[121],"datasets":[122],"reveal":[123],"that":[124],"trained":[127],"not":[130],"only":[131],"preserve":[132],"when":[134],"using":[135],"fewer":[136],"encoder":[137],"but":[139],"also":[140],"exhibit":[141],"enhanced":[142],"task":[143],"accuracy":[144],"compared":[145],"single-exit":[147],"or":[148],"models.":[150],"Furthermore,":[151],"explore":[153],"selection":[156,192],"strategy":[157],"grounded":[158],"posterior":[160],"probabilities":[161],"as":[162],"alternative":[164],"conventional":[167],"frame-based":[168],"entropy":[169],"approach.":[170],"Results":[171],"provide":[172],"insights":[173],"into":[174],"dynamics":[177],"particularly":[184],"efficacy":[186],"strategies":[189],"methods.":[193]},"counts_by_year":[{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}