Matches in SemOpenAlex for { <https://semopenalex.org/work/W2936969148> ?p ?o ?g. }
Showing items 1 to 92 of
92
with 100 items per page.
- W2936969148 endingPage "325" @default.
- W2936969148 startingPage "313" @default.
- W2936969148 abstract "Speech translation has traditionally been approached through cascaded models consisting of a speech recognizer trained on a corpus of transcribed speech, and a machine translation system trained on parallel texts. Several recent works have shown the feasibility of collapsing the cascade into a single, direct model that can be trained in an end-to-end fashion on a corpus of translated speech. However, experiments are inconclusive on whether the cascade or the direct model is stronger, and have only been conducted under the unrealistic assumption that both are trained on equal amounts of data, ignoring other available speech recognition and machine translation corpora. In this paper, we demonstrate that direct speech translation models require more data to perform well than cascaded models, and although they allow including auxiliary data through multi-task training, they are poor at exploiting such data, putting them at a severe disadvantage. As a remedy, we propose the use of end- to-end trainable models with two attention mechanisms, the first establishing source speech to source text alignments, the second modeling source to target text alignment. We show that such models naturally decompose into multi-task–trainable recognition and translation tasks and propose an attention-passing technique that alleviates error propagation issues in a previous formulation of a model with two attention stages. Our proposed model outperforms all examined baselines and is able to exploit auxiliary training data much more effectively than direct attentional models." @default.
- W2936969148 created "2019-04-25" @default.
- W2936969148 creator A5007526409 @default.
- W2936969148 creator A5024736091 @default.
- W2936969148 creator A5046084081 @default.
- W2936969148 creator A5068811427 @default.
- W2936969148 date "2019-11-01" @default.
- W2936969148 modified "2023-10-11" @default.
- W2936969148 title "Attention-Passing Models for Robust and Data-Efficient End-to-End Speech Translation" @default.
- W2936969148 cites W2962717763 @default.
- W2936969148 cites W2964084097 @default.
- W2936969148 doi "https://doi.org/10.1162/tacl_a_00270" @default.
- W2936969148 hasPublicationYear "2019" @default.
- W2936969148 type Work @default.
- W2936969148 sameAs 2936969148 @default.
- W2936969148 citedByCount "78" @default.
- W2936969148 countsByYear W29369691482019 @default.
- W2936969148 countsByYear W29369691482020 @default.
- W2936969148 countsByYear W29369691482021 @default.
- W2936969148 countsByYear W29369691482022 @default.
- W2936969148 countsByYear W29369691482023 @default.
- W2936969148 crossrefType "journal-article" @default.
- W2936969148 hasAuthorship W2936969148A5007526409 @default.
- W2936969148 hasAuthorship W2936969148A5024736091 @default.
- W2936969148 hasAuthorship W2936969148A5046084081 @default.
- W2936969148 hasAuthorship W2936969148A5068811427 @default.
- W2936969148 hasBestOaLocation W29369691481 @default.
- W2936969148 hasConcept C104317684 @default.
- W2936969148 hasConcept C105580179 @default.
- W2936969148 hasConcept C137293760 @default.
- W2936969148 hasConcept C149364088 @default.
- W2936969148 hasConcept C154945302 @default.
- W2936969148 hasConcept C162324750 @default.
- W2936969148 hasConcept C165696696 @default.
- W2936969148 hasConcept C185592680 @default.
- W2936969148 hasConcept C187736073 @default.
- W2936969148 hasConcept C203005215 @default.
- W2936969148 hasConcept C204321447 @default.
- W2936969148 hasConcept C2780366754 @default.
- W2936969148 hasConcept C2780451532 @default.
- W2936969148 hasConcept C28490314 @default.
- W2936969148 hasConcept C34146451 @default.
- W2936969148 hasConcept C38652104 @default.
- W2936969148 hasConcept C41008148 @default.
- W2936969148 hasConcept C43617362 @default.
- W2936969148 hasConcept C51632099 @default.
- W2936969148 hasConcept C55493867 @default.
- W2936969148 hasConcept C74296488 @default.
- W2936969148 hasConceptScore W2936969148C104317684 @default.
- W2936969148 hasConceptScore W2936969148C105580179 @default.
- W2936969148 hasConceptScore W2936969148C137293760 @default.
- W2936969148 hasConceptScore W2936969148C149364088 @default.
- W2936969148 hasConceptScore W2936969148C154945302 @default.
- W2936969148 hasConceptScore W2936969148C162324750 @default.
- W2936969148 hasConceptScore W2936969148C165696696 @default.
- W2936969148 hasConceptScore W2936969148C185592680 @default.
- W2936969148 hasConceptScore W2936969148C187736073 @default.
- W2936969148 hasConceptScore W2936969148C203005215 @default.
- W2936969148 hasConceptScore W2936969148C204321447 @default.
- W2936969148 hasConceptScore W2936969148C2780366754 @default.
- W2936969148 hasConceptScore W2936969148C2780451532 @default.
- W2936969148 hasConceptScore W2936969148C28490314 @default.
- W2936969148 hasConceptScore W2936969148C34146451 @default.
- W2936969148 hasConceptScore W2936969148C38652104 @default.
- W2936969148 hasConceptScore W2936969148C41008148 @default.
- W2936969148 hasConceptScore W2936969148C43617362 @default.
- W2936969148 hasConceptScore W2936969148C51632099 @default.
- W2936969148 hasConceptScore W2936969148C55493867 @default.
- W2936969148 hasConceptScore W2936969148C74296488 @default.
- W2936969148 hasLocation W29369691481 @default.
- W2936969148 hasLocation W29369691482 @default.
- W2936969148 hasLocation W29369691483 @default.
- W2936969148 hasLocation W29369691484 @default.
- W2936969148 hasOpenAccess W2936969148 @default.
- W2936969148 hasPrimaryLocation W29369691481 @default.
- W2936969148 hasRelatedWork W2046935727 @default.
- W2936969148 hasRelatedWork W2066913438 @default.
- W2936969148 hasRelatedWork W2132187905 @default.
- W2936969148 hasRelatedWork W2140830074 @default.
- W2936969148 hasRelatedWork W2274408985 @default.
- W2936969148 hasRelatedWork W2405158679 @default.
- W2936969148 hasRelatedWork W2510503866 @default.
- W2936969148 hasRelatedWork W3120880016 @default.
- W2936969148 hasRelatedWork W3204668696 @default.
- W2936969148 hasRelatedWork W3204941900 @default.
- W2936969148 hasVolume "7" @default.
- W2936969148 isParatext "false" @default.
- W2936969148 isRetracted "false" @default.
- W2936969148 magId "2936969148" @default.
- W2936969148 workType "article" @default.