{"id":"https://openalex.org/W4408353464","doi":"https://doi.org/10.1109/icassp49660.2025.10888900","title":"Speech Retrieval-Augmented Generation without Automatic Speech Recognition","display_name":"Speech Retrieval-Augmented Generation without Automatic Speech Recognition","publication_year":2025,"publication_date":"2025-03-12","ids":{"openalex":"https://openalex.org/W4408353464","doi":"https://doi.org/10.1109/icassp49660.2025.10888900"},"language":"en","primary_location":{"id":"doi:10.1109/icassp49660.2025.10888900","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp49660.2025.10888900","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2025 - 2025 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/A5114138481","display_name":"Do June Min","orcid":"https://orcid.org/0000-0001-7830-5638"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Do June Min","raw_affiliation_strings":["University of Michigan"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Michigan","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021276950","display_name":"Karel Mundnich","orcid":"https://orcid.org/0000-0003-4249-8356"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Karel Mundnich","raw_affiliation_strings":["AWS AI Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AWS AI Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5098903353","display_name":"Andy Lapastora","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Andy Lapastora","raw_affiliation_strings":["AWS AI Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AWS AI Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080981075","display_name":"Erfan Soltanmohammadi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Erfan Soltanmohammadi","raw_affiliation_strings":["AWS AI Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AWS AI Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004616142","display_name":"Srikanth Ronanki","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Srikanth Ronanki","raw_affiliation_strings":["AWS AI Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AWS AI Labs","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5109557654","display_name":"Kyu Han","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kyu Han","raw_affiliation_strings":["AWS AI Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"AWS AI Labs","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":7.0351,"has_fulltext":false,"cited_by_count":4,"citation_normalized_percentile":{"value":0.96249071,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":95,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"5"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9886999726295471,"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":0.9886999726295471,"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/T12031","display_name":"Speech and dialogue systems","score":0.914900004863739,"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":0.7975430488586426},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.7594625949859619},{"id":"https://openalex.org/keywords/speech-processing","display_name":"Speech processing","score":0.44643956422805786},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.382192999124527},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.33280467987060547}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7975430488586426},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.7594625949859619},{"id":"https://openalex.org/C61328038","wikidata":"https://www.wikidata.org/wiki/Q3358061","display_name":"Speech processing","level":2,"score":0.44643956422805786},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.382192999124527},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.33280467987060547}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icassp49660.2025.10888900","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp49660.2025.10888900","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2025 - 2025 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":28,"referenced_works":["https://openalex.org/W1494198834","https://openalex.org/W1984076147","https://openalex.org/W2144578941","https://openalex.org/W2148986421","https://openalex.org/W2962854302","https://openalex.org/W2963288440","https://openalex.org/W2963748441","https://openalex.org/W2970641574","https://openalex.org/W3119308075","https://openalex.org/W3198452188","https://openalex.org/W3209059054","https://openalex.org/W4372260310","https://openalex.org/W4372266552","https://openalex.org/W4385569624","https://openalex.org/W4385573236","https://openalex.org/W4386576685","https://openalex.org/W4392902876","https://openalex.org/W4392903115","https://openalex.org/W4392903210","https://openalex.org/W4400231894","https://openalex.org/W4402111429","https://openalex.org/W4402672004","https://openalex.org/W6769457076","https://openalex.org/W6777615688","https://openalex.org/W6783990618","https://openalex.org/W6791353385","https://openalex.org/W6852744543","https://openalex.org/W6859099255"],"related_works":["https://openalex.org/W3204019825","https://openalex.org/W2981428355","https://openalex.org/W1834994814","https://openalex.org/W2041273198","https://openalex.org/W1599055764","https://openalex.org/W2131711534","https://openalex.org/W2149163000","https://openalex.org/W2962858469","https://openalex.org/W2289873871","https://openalex.org/W2559040841"],"abstract_inverted_index":{"One":[0],"common":[1],"approach":[2,69,169],"for":[3,61],"question":[4,122],"answering":[5,63,123],"over":[6,64,133],"speech":[7,13,16,73,77,99,128,149],"data":[8],"is":[9,176],"to":[10,44],"first":[11],"transcribe":[12],"using":[14,142],"automatic":[15],"recognition":[17],"(ASR)":[18],"and":[19,47,96,137],"then":[20],"employ":[21],"text-based":[22,106,135,172],"retrieval-augmented":[23],"generation":[24,48],"(RAG)":[25],"on":[26,120,157],"the":[27,45,91,109,113,134,139,166,180],"transcriptions.":[28],"While":[29],"this":[30,52,168],"cascaded":[31,140,171],"pipeline":[32],"has":[33],"proven":[34],"effective":[35],"in":[36,179],"many":[37],"practical":[38],"settings,":[39],"ASR":[40],"errors":[41],"can":[42],"propagate":[43],"retrieval":[46,87,110,118,129],"steps.":[49],"To":[50],"overcome":[51],"limitation,":[53],"we":[54,146],"introduce":[55],"SpeechRAG,":[56],"a":[57,71,76,81,148,154],"novel":[58],"framework":[59],"designed":[60],"open-question":[62],"spoken":[65,121],"data.":[66],"Our":[67,117],"proposed":[68],"fine-tunes":[70],"pre-trained":[72],"encoder":[74],"into":[75,80],"adapter":[78],"fed":[79],"frozen":[82,114],"large":[83],"language":[84,150],"model":[85,151],"(LLM)\u2013based":[86],"model.":[88],"By":[89],"aligning":[90],"embedding":[92],"spaces":[93],"of":[94,112,165],"text":[95,115],"speech,":[97],"our":[98],"retriever":[100],"directly":[101],"retrieves":[102],"audio":[103,158],"passages":[104,159],"from":[105],"queries,":[107],"leveraging":[108],"capacity":[111],"retriever.":[116],"experiments":[119],"datasets":[124],"show":[125],"that":[126],"direct":[127],"does":[130],"not":[131],"degrade":[132],"baseline,":[136],"outperforms":[138,170],"systems":[141],"ASR.":[143],"For":[144],"generation,":[145],"use":[147],"(SLM)":[152],"as":[153],"generator,":[155],"conditioned":[156],"rather":[160],"than":[161],"transcripts.":[162,181],"Without":[163],"fine-tuning":[164],"SLM,":[167],"models":[173],"when":[174],"there":[175],"high":[177],"WER":[178]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":2}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}