Matches in SemOpenAlex for { <https://semopenalex.org/work/W4311731008> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4311731008 abstract "Direct speech-to-speech translation (S2ST), in which all components can be optimized jointly, is advantageous over cascaded approaches to achieve fast inference with a simplified pipeline. We present a novel two-pass direct S2ST architecture, UnitY, which first generates textual representations and predicts discrete acoustic units subsequently. We enhance the model performance by subword prediction in the first-pass decoder, advanced two-pass decoder architecture design and search strategy, and better training regularization. To leverage large amounts of unlabeled text data, we pre-train the first-pass text decoder based on the self-supervised denoising auto-encoding task. Experimental evaluations on benchmark datasets at various data scales demonstrate that UnitY outperforms a single-pass speech-to-unit translation model by 2.5-4.2 ASR-BLEU with 2.83x decoding speed-up. We show that the proposed methods boost the performance even when predicting spectrogram in the second pass. However, predicting discrete units achieves 2.51x decoding speed-up compared to that case." @default.
- W4311731008 created "2022-12-28" @default.
- W4311731008 creator A5008107321 @default.
- W4311731008 creator A5040282669 @default.
- W4311731008 creator A5052049405 @default.
- W4311731008 creator A5054479105 @default.
- W4311731008 creator A5058915697 @default.
- W4311731008 creator A5064261515 @default.
- W4311731008 creator A5072019513 @default.
- W4311731008 creator A5083886640 @default.
- W4311731008 creator A5084509956 @default.
- W4311731008 creator A5087491225 @default.
- W4311731008 date "2022-12-15" @default.
- W4311731008 modified "2023-09-27" @default.
- W4311731008 title "UnitY: Two-pass Direct Speech-to-speech Translation with Discrete Units" @default.
- W4311731008 doi "https://doi.org/10.48550/arxiv.2212.08055" @default.
- W4311731008 hasPublicationYear "2022" @default.
- W4311731008 type Work @default.
- W4311731008 citedByCount "0" @default.
- W4311731008 crossrefType "posted-content" @default.
- W4311731008 hasAuthorship W4311731008A5008107321 @default.
- W4311731008 hasAuthorship W4311731008A5040282669 @default.
- W4311731008 hasAuthorship W4311731008A5052049405 @default.
- W4311731008 hasAuthorship W4311731008A5054479105 @default.
- W4311731008 hasAuthorship W4311731008A5058915697 @default.
- W4311731008 hasAuthorship W4311731008A5064261515 @default.
- W4311731008 hasAuthorship W4311731008A5072019513 @default.
- W4311731008 hasAuthorship W4311731008A5083886640 @default.
- W4311731008 hasAuthorship W4311731008A5084509956 @default.
- W4311731008 hasAuthorship W4311731008A5087491225 @default.
- W4311731008 hasBestOaLocation W43117310081 @default.
- W4311731008 hasConcept C111472728 @default.
- W4311731008 hasConcept C11413529 @default.
- W4311731008 hasConcept C13280743 @default.
- W4311731008 hasConcept C138885662 @default.
- W4311731008 hasConcept C153083717 @default.
- W4311731008 hasConcept C154945302 @default.
- W4311731008 hasConcept C163294075 @default.
- W4311731008 hasConcept C185798385 @default.
- W4311731008 hasConcept C199360897 @default.
- W4311731008 hasConcept C203005215 @default.
- W4311731008 hasConcept C205649164 @default.
- W4311731008 hasConcept C2776135515 @default.
- W4311731008 hasConcept C2776182073 @default.
- W4311731008 hasConcept C2776214188 @default.
- W4311731008 hasConcept C2780366754 @default.
- W4311731008 hasConcept C28490314 @default.
- W4311731008 hasConcept C2994044699 @default.
- W4311731008 hasConcept C33923547 @default.
- W4311731008 hasConcept C41008148 @default.
- W4311731008 hasConcept C43521106 @default.
- W4311731008 hasConcept C45273575 @default.
- W4311731008 hasConcept C57273362 @default.
- W4311731008 hasConcept C60048801 @default.
- W4311731008 hasConcept C94375191 @default.
- W4311731008 hasConceptScore W4311731008C111472728 @default.
- W4311731008 hasConceptScore W4311731008C11413529 @default.
- W4311731008 hasConceptScore W4311731008C13280743 @default.
- W4311731008 hasConceptScore W4311731008C138885662 @default.
- W4311731008 hasConceptScore W4311731008C153083717 @default.
- W4311731008 hasConceptScore W4311731008C154945302 @default.
- W4311731008 hasConceptScore W4311731008C163294075 @default.
- W4311731008 hasConceptScore W4311731008C185798385 @default.
- W4311731008 hasConceptScore W4311731008C199360897 @default.
- W4311731008 hasConceptScore W4311731008C203005215 @default.
- W4311731008 hasConceptScore W4311731008C205649164 @default.
- W4311731008 hasConceptScore W4311731008C2776135515 @default.
- W4311731008 hasConceptScore W4311731008C2776182073 @default.
- W4311731008 hasConceptScore W4311731008C2776214188 @default.
- W4311731008 hasConceptScore W4311731008C2780366754 @default.
- W4311731008 hasConceptScore W4311731008C28490314 @default.
- W4311731008 hasConceptScore W4311731008C2994044699 @default.
- W4311731008 hasConceptScore W4311731008C33923547 @default.
- W4311731008 hasConceptScore W4311731008C41008148 @default.
- W4311731008 hasConceptScore W4311731008C43521106 @default.
- W4311731008 hasConceptScore W4311731008C45273575 @default.
- W4311731008 hasConceptScore W4311731008C57273362 @default.
- W4311731008 hasConceptScore W4311731008C60048801 @default.
- W4311731008 hasConceptScore W4311731008C94375191 @default.
- W4311731008 hasLocation W43117310081 @default.
- W4311731008 hasOpenAccess W4311731008 @default.
- W4311731008 hasPrimaryLocation W43117310081 @default.
- W4311731008 hasRelatedWork W2098486943 @default.
- W4311731008 hasRelatedWork W2154135679 @default.
- W4311731008 hasRelatedWork W2977183928 @default.
- W4311731008 hasRelatedWork W3000153094 @default.
- W4311731008 hasRelatedWork W3011858470 @default.
- W4311731008 hasRelatedWork W3043902818 @default.
- W4311731008 hasRelatedWork W3129072390 @default.
- W4311731008 hasRelatedWork W4221152531 @default.
- W4311731008 hasRelatedWork W4311731008 @default.
- W4311731008 hasRelatedWork W4375869276 @default.
- W4311731008 isParatext "false" @default.
- W4311731008 isRetracted "false" @default.
- W4311731008 workType "article" @default.