Matches in SemOpenAlex for { <https://semopenalex.org/work/W3094262219> ?p ?o ?g. }
- W3094262219 abstract "Hybrid Autoregressive Transducer (HAT) is a recently proposed end-to-end acoustic model that extends the standard Recurrent Neural Network Transducer (RNN-T) for the purpose of the external language model (LM) fusion. In HAT, the blank probability and the label probability are estimated using two separate probability distributions, which provides a more accurate solution for internal LM score estimation, and thus works better when combining with an external LM. Previous work mainly focuses on HAT model training with the negative log-likelihood loss, while in this paper, we study the minimum word error rate (MWER) training of HAT -- a criterion that is closer to the evaluation metric for speech recognition, and has been successfully applied to other types of end-to-end models such as sequence-to-sequence (S2S) and RNN-T models. From experiments with around 30,000 hours of training data, we show that MWER training can improve the accuracy of HAT models, while at the same time, improving the robustness of the model against the decoding hyper-parameters such as length normalization and decoding beam during inference." @default.
- W3094262219 created "2020-10-29" @default.
- W3094262219 creator A5016279564 @default.
- W3094262219 creator A5029670581 @default.
- W3094262219 creator A5030625154 @default.
- W3094262219 creator A5051392570 @default.
- W3094262219 creator A5077401426 @default.
- W3094262219 date "2020-10-23" @default.
- W3094262219 modified "2023-09-26" @default.
- W3094262219 title "On Minimum Word Error Rate Training of the Hybrid Autoregressive Transducer" @default.
- W3094262219 cites W1828163288 @default.
- W3094262219 cites W1915251500 @default.
- W3094262219 cites W1922655562 @default.
- W3094262219 cites W1942035323 @default.
- W3094262219 cites W2127141656 @default.
- W3094262219 cites W2131342762 @default.
- W3094262219 cites W2150907703 @default.
- W3094262219 cites W2327501763 @default.
- W3094262219 cites W2403195671 @default.
- W3094262219 cites W2515801922 @default.
- W3094262219 cites W2748679025 @default.
- W3094262219 cites W2889163603 @default.
- W3094262219 cites W2950903920 @default.
- W3094262219 cites W2962760690 @default.
- W3094262219 cites W2962765220 @default.
- W3094262219 cites W2990501692 @default.
- W3094262219 cites W3008037978 @default.
- W3094262219 cites W3016234571 @default.
- W3094262219 cites W3081494755 @default.
- W3094262219 cites W3094979069 @default.
- W3094262219 cites W38527073 @default.
- W3094262219 cites W854541894 @default.
- W3094262219 doi "https://doi.org/10.48550/arxiv.2010.12673" @default.
- W3094262219 hasPublicationYear "2020" @default.
- W3094262219 type Work @default.
- W3094262219 sameAs 3094262219 @default.
- W3094262219 citedByCount "1" @default.
- W3094262219 countsByYear W30942622192021 @default.
- W3094262219 crossrefType "posted-content" @default.
- W3094262219 hasAuthorship W3094262219A5016279564 @default.
- W3094262219 hasAuthorship W3094262219A5029670581 @default.
- W3094262219 hasAuthorship W3094262219A5030625154 @default.
- W3094262219 hasAuthorship W3094262219A5051392570 @default.
- W3094262219 hasAuthorship W3094262219A5077401426 @default.
- W3094262219 hasBestOaLocation W30942622191 @default.
- W3094262219 hasConcept C104317684 @default.
- W3094262219 hasConcept C105795698 @default.
- W3094262219 hasConcept C11413529 @default.
- W3094262219 hasConcept C147168706 @default.
- W3094262219 hasConcept C153180895 @default.
- W3094262219 hasConcept C154945302 @default.
- W3094262219 hasConcept C159877910 @default.
- W3094262219 hasConcept C162324750 @default.
- W3094262219 hasConcept C185592680 @default.
- W3094262219 hasConcept C187736073 @default.
- W3094262219 hasConcept C23224414 @default.
- W3094262219 hasConcept C2524010 @default.
- W3094262219 hasConcept C2776214188 @default.
- W3094262219 hasConcept C2780898871 @default.
- W3094262219 hasConcept C28490314 @default.
- W3094262219 hasConcept C33923547 @default.
- W3094262219 hasConcept C40969351 @default.
- W3094262219 hasConcept C41008148 @default.
- W3094262219 hasConcept C50644808 @default.
- W3094262219 hasConcept C55493867 @default.
- W3094262219 hasConcept C57273362 @default.
- W3094262219 hasConcept C63479239 @default.
- W3094262219 hasConcept C90805587 @default.
- W3094262219 hasConceptScore W3094262219C104317684 @default.
- W3094262219 hasConceptScore W3094262219C105795698 @default.
- W3094262219 hasConceptScore W3094262219C11413529 @default.
- W3094262219 hasConceptScore W3094262219C147168706 @default.
- W3094262219 hasConceptScore W3094262219C153180895 @default.
- W3094262219 hasConceptScore W3094262219C154945302 @default.
- W3094262219 hasConceptScore W3094262219C159877910 @default.
- W3094262219 hasConceptScore W3094262219C162324750 @default.
- W3094262219 hasConceptScore W3094262219C185592680 @default.
- W3094262219 hasConceptScore W3094262219C187736073 @default.
- W3094262219 hasConceptScore W3094262219C23224414 @default.
- W3094262219 hasConceptScore W3094262219C2524010 @default.
- W3094262219 hasConceptScore W3094262219C2776214188 @default.
- W3094262219 hasConceptScore W3094262219C2780898871 @default.
- W3094262219 hasConceptScore W3094262219C28490314 @default.
- W3094262219 hasConceptScore W3094262219C33923547 @default.
- W3094262219 hasConceptScore W3094262219C40969351 @default.
- W3094262219 hasConceptScore W3094262219C41008148 @default.
- W3094262219 hasConceptScore W3094262219C50644808 @default.
- W3094262219 hasConceptScore W3094262219C55493867 @default.
- W3094262219 hasConceptScore W3094262219C57273362 @default.
- W3094262219 hasConceptScore W3094262219C63479239 @default.
- W3094262219 hasConceptScore W3094262219C90805587 @default.
- W3094262219 hasLocation W30942622191 @default.
- W3094262219 hasOpenAccess W3094262219 @default.
- W3094262219 hasPrimaryLocation W30942622191 @default.
- W3094262219 hasRelatedWork W105527591 @default.
- W3094262219 hasRelatedWork W1574295218 @default.
- W3094262219 hasRelatedWork W1717589883 @default.
- W3094262219 hasRelatedWork W1931804893 @default.
- W3094262219 hasRelatedWork W2109705048 @default.
- W3094262219 hasRelatedWork W2134995938 @default.