Matches in SemOpenAlex for { <https://semopenalex.org/work/W2996987694> ?p ?o ?g. }
- W2996987694 endingPage "7846" @default.
- W2996987694 startingPage "7839" @default.
- W2996987694 abstract "Non-autoregressive translation (NAT) models remove the dependence on previous target tokens and generate all target tokens in parallel, resulting in significant inference speedup but at the cost of inferior translation accuracy compared to autoregressive translation (AT) models. Considering that AT models have higher accuracy and are easier to train than NAT models, and both of them share the same model configurations, a natural idea to improve the accuracy of NAT models is to transfer a well-trained AT model to an NAT model through fine-tuning. However, since AT and NAT models differ greatly in training strategy, straightforward fine-tuning does not work well. In this work, we introduce curriculum learning into fine-tuning for NAT. Specifically, we design a curriculum in the fine-tuning process to progressively switch the training from autoregressive generation to non-autoregressive generation. Experiments on four benchmark translation datasets show that the proposed method achieves good improvement (more than 1 BLEU score) over previous NAT baselines in terms of translation accuracy, and greatly speed up (more than 10 times) the inference process over AT baselines." @default.
- W2996987694 created "2020-01-10" @default.
- W2996987694 creator A5009732907 @default.
- W2996987694 creator A5018286848 @default.
- W2996987694 creator A5020025718 @default.
- W2996987694 creator A5048237545 @default.
- W2996987694 creator A5055122985 @default.
- W2996987694 creator A5070990160 @default.
- W2996987694 date "2020-04-03" @default.
- W2996987694 modified "2023-10-16" @default.
- W2996987694 title "Fine-Tuning by Curriculum Learning for Non-Autoregressive Neural Machine Translation" @default.
- W2996987694 doi "https://doi.org/10.1609/aaai.v34i05.6289" @default.
- W2996987694 hasPublicationYear "2020" @default.
- W2996987694 type Work @default.
- W2996987694 sameAs 2996987694 @default.
- W2996987694 citedByCount "34" @default.
- W2996987694 countsByYear W29969876942020 @default.
- W2996987694 countsByYear W29969876942021 @default.
- W2996987694 countsByYear W29969876942022 @default.
- W2996987694 countsByYear W29969876942023 @default.
- W2996987694 crossrefType "journal-article" @default.
- W2996987694 hasAuthorship W2996987694A5009732907 @default.
- W2996987694 hasAuthorship W2996987694A5018286848 @default.
- W2996987694 hasAuthorship W2996987694A5020025718 @default.
- W2996987694 hasAuthorship W2996987694A5048237545 @default.
- W2996987694 hasAuthorship W2996987694A5055122985 @default.
- W2996987694 hasAuthorship W2996987694A5070990160 @default.
- W2996987694 hasBestOaLocation W29969876941 @default.
- W2996987694 hasConcept C104317684 @default.
- W2996987694 hasConcept C105580179 @default.
- W2996987694 hasConcept C119857082 @default.
- W2996987694 hasConcept C121332964 @default.
- W2996987694 hasConcept C13280743 @default.
- W2996987694 hasConcept C149364088 @default.
- W2996987694 hasConcept C149782125 @default.
- W2996987694 hasConcept C154945302 @default.
- W2996987694 hasConcept C157524613 @default.
- W2996987694 hasConcept C159877910 @default.
- W2996987694 hasConcept C173608175 @default.
- W2996987694 hasConcept C182516595 @default.
- W2996987694 hasConcept C185592680 @default.
- W2996987694 hasConcept C185798385 @default.
- W2996987694 hasConcept C199360897 @default.
- W2996987694 hasConcept C203005215 @default.
- W2996987694 hasConcept C205649164 @default.
- W2996987694 hasConcept C2776214188 @default.
- W2996987694 hasConcept C31258907 @default.
- W2996987694 hasConcept C33923547 @default.
- W2996987694 hasConcept C41008148 @default.
- W2996987694 hasConcept C50644808 @default.
- W2996987694 hasConcept C55493867 @default.
- W2996987694 hasConcept C62520636 @default.
- W2996987694 hasConcept C68339613 @default.
- W2996987694 hasConcept C98045186 @default.
- W2996987694 hasConceptScore W2996987694C104317684 @default.
- W2996987694 hasConceptScore W2996987694C105580179 @default.
- W2996987694 hasConceptScore W2996987694C119857082 @default.
- W2996987694 hasConceptScore W2996987694C121332964 @default.
- W2996987694 hasConceptScore W2996987694C13280743 @default.
- W2996987694 hasConceptScore W2996987694C149364088 @default.
- W2996987694 hasConceptScore W2996987694C149782125 @default.
- W2996987694 hasConceptScore W2996987694C154945302 @default.
- W2996987694 hasConceptScore W2996987694C157524613 @default.
- W2996987694 hasConceptScore W2996987694C159877910 @default.
- W2996987694 hasConceptScore W2996987694C173608175 @default.
- W2996987694 hasConceptScore W2996987694C182516595 @default.
- W2996987694 hasConceptScore W2996987694C185592680 @default.
- W2996987694 hasConceptScore W2996987694C185798385 @default.
- W2996987694 hasConceptScore W2996987694C199360897 @default.
- W2996987694 hasConceptScore W2996987694C203005215 @default.
- W2996987694 hasConceptScore W2996987694C205649164 @default.
- W2996987694 hasConceptScore W2996987694C2776214188 @default.
- W2996987694 hasConceptScore W2996987694C31258907 @default.
- W2996987694 hasConceptScore W2996987694C33923547 @default.
- W2996987694 hasConceptScore W2996987694C41008148 @default.
- W2996987694 hasConceptScore W2996987694C50644808 @default.
- W2996987694 hasConceptScore W2996987694C55493867 @default.
- W2996987694 hasConceptScore W2996987694C62520636 @default.
- W2996987694 hasConceptScore W2996987694C68339613 @default.
- W2996987694 hasConceptScore W2996987694C98045186 @default.
- W2996987694 hasIssue "05" @default.
- W2996987694 hasLocation W29969876941 @default.
- W2996987694 hasLocation W29969876942 @default.
- W2996987694 hasOpenAccess W2996987694 @default.
- W2996987694 hasPrimaryLocation W29969876941 @default.
- W2996987694 hasRelatedWork W2969565141 @default.
- W2996987694 hasRelatedWork W2990488589 @default.
- W2996987694 hasRelatedWork W2996843693 @default.
- W2996987694 hasRelatedWork W2996987694 @default.
- W2996987694 hasRelatedWork W3054488230 @default.
- W2996987694 hasRelatedWork W3092111461 @default.
- W2996987694 hasRelatedWork W3106147182 @default.
- W2996987694 hasRelatedWork W3206689743 @default.
- W2996987694 hasRelatedWork W4287646082 @default.
- W2996987694 hasRelatedWork W4303874710 @default.
- W2996987694 hasVolume "34" @default.
- W2996987694 isParatext "false" @default.
- W2996987694 isRetracted "false" @default.