Matches in SemOpenAlex for { <https://semopenalex.org/work/W3094637009> ?p ?o ?g. }
- W3094637009 abstract "Self-supervised learning (SSL) has shown promise in learning representations of audio that are useful for automatic speech recognition (ASR). But, training SSL models like wav2vec~2.0 requires a two-stage pipeline. In this paper we demonstrate a single-stage training of ASR models that can utilize both unlabeled and labeled data. During training, we alternately minimize two losses: an unsupervised masked Contrastive Predictive Coding (CPC) loss and the supervised audio-to-text alignment loss Connectionist Temporal Classification (CTC). We show that this joint training method directly optimizes performance for the downstream ASR task using unsupervised data while achieving similar word error rates to wav2vec~2.0 on the Librispeech 100-hour dataset. Finally, we postulate that solving the contrastive task is a regularization for the supervised CTC loss." @default.
- W3094637009 created "2020-11-09" @default.
- W3094637009 creator A5041907084 @default.
- W3094637009 creator A5053915453 @default.
- W3094637009 creator A5061907595 @default.
- W3094637009 creator A5074491874 @default.
- W3094637009 date "2020-10-30" @default.
- W3094637009 modified "2023-10-01" @default.
- W3094637009 title "Joint Masked CPC and CTC Training for ASR" @default.
- W3094637009 cites W1494198834 @default.
- W3094637009 cites W1522301498 @default.
- W3094637009 cites W1993660824 @default.
- W3094637009 cites W2127141656 @default.
- W3094637009 cites W2193413348 @default.
- W3094637009 cites W2510867321 @default.
- W3094637009 cites W2526425061 @default.
- W3094637009 cites W2755891984 @default.
- W3094637009 cites W2842511635 @default.
- W3094637009 cites W2936774411 @default.
- W3094637009 cites W2940180244 @default.
- W3094637009 cites W2952976827 @default.
- W3094637009 cites W2953190524 @default.
- W3094637009 cites W2963341956 @default.
- W3094637009 cites W2963403868 @default.
- W3094637009 cites W2991213871 @default.
- W3094637009 cites W2995181338 @default.
- W3094637009 cites W2996159613 @default.
- W3094637009 cites W2998532468 @default.
- W3094637009 cites W3005680577 @default.
- W3094637009 cites W3026041220 @default.
- W3094637009 cites W3027083471 @default.
- W3094637009 cites W3034781633 @default.
- W3094637009 cites W3036601975 @default.
- W3094637009 cites W3039910566 @default.
- W3094637009 cites W3099782249 @default.
- W3094637009 hasPublicationYear "2020" @default.
- W3094637009 type Work @default.
- W3094637009 sameAs 3094637009 @default.
- W3094637009 citedByCount "2" @default.
- W3094637009 countsByYear W30946370092021 @default.
- W3094637009 crossrefType "posted-content" @default.
- W3094637009 hasAuthorship W3094637009A5041907084 @default.
- W3094637009 hasAuthorship W3094637009A5053915453 @default.
- W3094637009 hasAuthorship W3094637009A5061907595 @default.
- W3094637009 hasAuthorship W3094637009A5074491874 @default.
- W3094637009 hasConcept C105795698 @default.
- W3094637009 hasConcept C119857082 @default.
- W3094637009 hasConcept C127413603 @default.
- W3094637009 hasConcept C153180895 @default.
- W3094637009 hasConcept C154945302 @default.
- W3094637009 hasConcept C162324750 @default.
- W3094637009 hasConcept C170154142 @default.
- W3094637009 hasConcept C179518139 @default.
- W3094637009 hasConcept C18555067 @default.
- W3094637009 hasConcept C187736073 @default.
- W3094637009 hasConcept C199360897 @default.
- W3094637009 hasConcept C204321447 @default.
- W3094637009 hasConcept C2776135515 @default.
- W3094637009 hasConcept C2776145971 @default.
- W3094637009 hasConcept C2780451532 @default.
- W3094637009 hasConcept C28490314 @default.
- W3094637009 hasConcept C33923547 @default.
- W3094637009 hasConcept C41008148 @default.
- W3094637009 hasConcept C43521106 @default.
- W3094637009 hasConcept C50644808 @default.
- W3094637009 hasConcept C51632099 @default.
- W3094637009 hasConcept C8521452 @default.
- W3094637009 hasConceptScore W3094637009C105795698 @default.
- W3094637009 hasConceptScore W3094637009C119857082 @default.
- W3094637009 hasConceptScore W3094637009C127413603 @default.
- W3094637009 hasConceptScore W3094637009C153180895 @default.
- W3094637009 hasConceptScore W3094637009C154945302 @default.
- W3094637009 hasConceptScore W3094637009C162324750 @default.
- W3094637009 hasConceptScore W3094637009C170154142 @default.
- W3094637009 hasConceptScore W3094637009C179518139 @default.
- W3094637009 hasConceptScore W3094637009C18555067 @default.
- W3094637009 hasConceptScore W3094637009C187736073 @default.
- W3094637009 hasConceptScore W3094637009C199360897 @default.
- W3094637009 hasConceptScore W3094637009C204321447 @default.
- W3094637009 hasConceptScore W3094637009C2776135515 @default.
- W3094637009 hasConceptScore W3094637009C2776145971 @default.
- W3094637009 hasConceptScore W3094637009C2780451532 @default.
- W3094637009 hasConceptScore W3094637009C28490314 @default.
- W3094637009 hasConceptScore W3094637009C33923547 @default.
- W3094637009 hasConceptScore W3094637009C41008148 @default.
- W3094637009 hasConceptScore W3094637009C43521106 @default.
- W3094637009 hasConceptScore W3094637009C50644808 @default.
- W3094637009 hasConceptScore W3094637009C51632099 @default.
- W3094637009 hasConceptScore W3094637009C8521452 @default.
- W3094637009 hasLocation W30946370091 @default.
- W3094637009 hasOpenAccess W3094637009 @default.
- W3094637009 hasPrimaryLocation W30946370091 @default.
- W3094637009 hasRelatedWork W2137820324 @default.
- W3094637009 hasRelatedWork W2148182949 @default.
- W3094637009 hasRelatedWork W2898132662 @default.
- W3094637009 hasRelatedWork W2939069254 @default.
- W3094637009 hasRelatedWork W2939710050 @default.
- W3094637009 hasRelatedWork W2988736778 @default.
- W3094637009 hasRelatedWork W3003809177 @default.
- W3094637009 hasRelatedWork W3015356564 @default.