Matches in SemOpenAlex for { <https://semopenalex.org/work/W2981648103> ?p ?o ?g. }
- W2981648103 abstract "Autoregressive sequence models achieve state-of-the-art performance in domains like machine translation. However, due to the autoregressive factorization nature, these models suffer from heavy latency during inference. Recently, non-autoregressive sequence models were proposed to reduce the inference time. However, these models assume that the decoding process of each token is conditionally independent of others. Such a generation process sometimes makes the output sentence inconsistent, and thus the learned non-autoregressive models could only achieve inferior accuracy compared to their autoregressive counterparts. To improve then decoding consistency and reduce the inference cost at the same time, we propose to incorporate a structured inference module into the non-autoregressive models. Specifically, we design an efficient approximation for Conditional Random Fields (CRF) for non-autoregressive sequence models, and further propose a dynamic transition technique to model positional contexts in the CRF. Experiments in machine translation show that while increasing little latency (8~14ms), our model could achieve significantly better translation performance than previous non-autoregressive models on different translation datasets. In particular, for the WMT14 En-De dataset, our model obtains a BLEU score of 26.80, which largely outperforms the previous non-autoregressive baselines and is only 0.61 lower in BLEU than purely autoregressive models." @default.
- W2981648103 created "2019-11-01" @default.
- W2981648103 creator A5009362721 @default.
- W2981648103 creator A5020049212 @default.
- W2981648103 creator A5060171279 @default.
- W2981648103 creator A5063684082 @default.
- W2981648103 creator A5068285627 @default.
- W2981648103 creator A5091046771 @default.
- W2981648103 date "2019-10-25" @default.
- W2981648103 modified "2023-09-23" @default.
- W2981648103 title "Fast Structured Decoding for Sequence Models" @default.
- W2981648103 cites W1821462560 @default.
- W2981648103 cites W1934019294 @default.
- W2981648103 cites W2006969979 @default.
- W2981648103 cites W2129999749 @default.
- W2981648103 cites W2147880316 @default.
- W2981648103 cites W2157331557 @default.
- W2981648103 cites W2158899491 @default.
- W2981648103 cites W2183341477 @default.
- W2981648103 cites W2487501366 @default.
- W2981648103 cites W2525778437 @default.
- W2981648103 cites W2789543585 @default.
- W2981648103 cites W2804145368 @default.
- W2981648103 cites W2904441038 @default.
- W2981648103 cites W2907945666 @default.
- W2981648103 cites W2949847915 @default.
- W2981648103 cites W2962969034 @default.
- W2981648103 cites W2963403868 @default.
- W2981648103 cites W2963418779 @default.
- W2981648103 cites W2963434219 @default.
- W2981648103 cites W2963536265 @default.
- W2981648103 cites W2963572611 @default.
- W2981648103 cites W2963736842 @default.
- W2981648103 cites W2964121744 @default.
- W2981648103 cites W2964140243 @default.
- W2981648103 cites W2964190861 @default.
- W2981648103 cites W2964265128 @default.
- W2981648103 cites W2964308564 @default.
- W2981648103 hasPublicationYear "2019" @default.
- W2981648103 type Work @default.
- W2981648103 sameAs 2981648103 @default.
- W2981648103 citedByCount "0" @default.
- W2981648103 crossrefType "posted-content" @default.
- W2981648103 hasAuthorship W2981648103A5009362721 @default.
- W2981648103 hasAuthorship W2981648103A5020049212 @default.
- W2981648103 hasAuthorship W2981648103A5060171279 @default.
- W2981648103 hasAuthorship W2981648103A5063684082 @default.
- W2981648103 hasAuthorship W2981648103A5068285627 @default.
- W2981648103 hasAuthorship W2981648103A5091046771 @default.
- W2981648103 hasConcept C11413529 @default.
- W2981648103 hasConcept C119857082 @default.
- W2981648103 hasConcept C149782125 @default.
- W2981648103 hasConcept C151406439 @default.
- W2981648103 hasConcept C152565575 @default.
- W2981648103 hasConcept C154945302 @default.
- W2981648103 hasConcept C159877910 @default.
- W2981648103 hasConcept C194657046 @default.
- W2981648103 hasConcept C203005215 @default.
- W2981648103 hasConcept C24338571 @default.
- W2981648103 hasConcept C2776214188 @default.
- W2981648103 hasConcept C2778112365 @default.
- W2981648103 hasConcept C28490314 @default.
- W2981648103 hasConcept C33923547 @default.
- W2981648103 hasConcept C41008148 @default.
- W2981648103 hasConcept C54355233 @default.
- W2981648103 hasConcept C57273362 @default.
- W2981648103 hasConcept C86803240 @default.
- W2981648103 hasConceptScore W2981648103C11413529 @default.
- W2981648103 hasConceptScore W2981648103C119857082 @default.
- W2981648103 hasConceptScore W2981648103C149782125 @default.
- W2981648103 hasConceptScore W2981648103C151406439 @default.
- W2981648103 hasConceptScore W2981648103C152565575 @default.
- W2981648103 hasConceptScore W2981648103C154945302 @default.
- W2981648103 hasConceptScore W2981648103C159877910 @default.
- W2981648103 hasConceptScore W2981648103C194657046 @default.
- W2981648103 hasConceptScore W2981648103C203005215 @default.
- W2981648103 hasConceptScore W2981648103C24338571 @default.
- W2981648103 hasConceptScore W2981648103C2776214188 @default.
- W2981648103 hasConceptScore W2981648103C2778112365 @default.
- W2981648103 hasConceptScore W2981648103C28490314 @default.
- W2981648103 hasConceptScore W2981648103C33923547 @default.
- W2981648103 hasConceptScore W2981648103C41008148 @default.
- W2981648103 hasConceptScore W2981648103C54355233 @default.
- W2981648103 hasConceptScore W2981648103C57273362 @default.
- W2981648103 hasConceptScore W2981648103C86803240 @default.
- W2981648103 hasOpenAccess W2981648103 @default.
- W2981648103 hasRelatedWork W2098152933 @default.
- W2981648103 hasRelatedWork W2185923243 @default.
- W2981648103 hasRelatedWork W2789163545 @default.
- W2981648103 hasRelatedWork W2789543585 @default.
- W2981648103 hasRelatedWork W2964089333 @default.
- W2981648103 hasRelatedWork W2969565141 @default.
- W2981648103 hasRelatedWork W2972034672 @default.
- W2981648103 hasRelatedWork W2990004017 @default.
- W2981648103 hasRelatedWork W2990389671 @default.
- W2981648103 hasRelatedWork W3000840023 @default.
- W2981648103 hasRelatedWork W3015162217 @default.
- W2981648103 hasRelatedWork W3022372506 @default.
- W2981648103 hasRelatedWork W3030343076 @default.
- W2981648103 hasRelatedWork W3034216012 @default.