Matches in SemOpenAlex for { <https://semopenalex.org/work/W3104273515> ?p ?o ?g. }
- W3104273515 abstract "The scarcity of large parallel corpora is an important obstacle for neural machine translation. A common solution is to exploit the knowledge of language models (LM) trained on abundant monolingual data. In this work, we propose a novel approach to incorporate a LM as prior in a neural translation model (TM). Specifically, we add a regularization term, which pushes the output distributions of the TM to be probable under the LM prior, while avoiding wrong predictions when the TM “disagrees” with the LM. This objective relates to knowledge distillation, where the LM can be viewed as teaching the TM about the target language. The proposed approach does not compromise decoding speed, because the LM is used only at training time, unlike previous work that requires it during inference. We present an analysis of the effects that different methods have on the distributions of the TM. Results on two low-resource machine translation datasets show clear improvements even with limited monolingual data." @default.
- W3104273515 created "2020-11-23" @default.
- W3104273515 creator A5035665045 @default.
- W3104273515 creator A5036521865 @default.
- W3104273515 creator A5038456766 @default.
- W3104273515 date "2020-01-01" @default.
- W3104273515 modified "2023-09-29" @default.
- W3104273515 title "Language Model Prior for Low-Resource Neural Machine Translation" @default.
- W3104273515 cites W1533861849 @default.
- W3104273515 cites W1821462560 @default.
- W3104273515 cites W1902237438 @default.
- W3104273515 cites W1915251500 @default.
- W3104273515 cites W2006969979 @default.
- W3104273515 cites W2041404167 @default.
- W3104273515 cites W2064675550 @default.
- W3104273515 cites W2130942839 @default.
- W3104273515 cites W2154368244 @default.
- W3104273515 cites W2183341477 @default.
- W3104273515 cites W2294370754 @default.
- W3104273515 cites W2512924740 @default.
- W3104273515 cites W2526471240 @default.
- W3104273515 cites W2555428947 @default.
- W3104273515 cites W2886540570 @default.
- W3104273515 cites W2888779557 @default.
- W3104273515 cites W2927431361 @default.
- W3104273515 cites W2948210185 @default.
- W3104273515 cites W2962964385 @default.
- W3104273515 cites W2963174344 @default.
- W3104273515 cites W2963216553 @default.
- W3104273515 cites W2963250244 @default.
- W3104273515 cites W2963310665 @default.
- W3104273515 cites W2963341956 @default.
- W3104273515 cites W2963347649 @default.
- W3104273515 cites W2963403868 @default.
- W3104273515 cites W2963413917 @default.
- W3104273515 cites W2963506925 @default.
- W3104273515 cites W2963532001 @default.
- W3104273515 cites W2963593215 @default.
- W3104273515 cites W2963681240 @default.
- W3104273515 cites W2964121744 @default.
- W3104273515 cites W2964308564 @default.
- W3104273515 cites W2967985939 @default.
- W3104273515 cites W2970694516 @default.
- W3104273515 cites W2970756316 @default.
- W3104273515 cites W2970971581 @default.
- W3104273515 cites W2979047515 @default.
- W3104273515 cites W2986367395 @default.
- W3104273515 cites W2986562961 @default.
- W3104273515 cites W2989276524 @default.
- W3104273515 cites W2994928925 @default.
- W3104273515 cites W3004127093 @default.
- W3104273515 cites W3034772996 @default.
- W3104273515 doi "https://doi.org/10.18653/v1/2020.emnlp-main.615" @default.
- W3104273515 hasPublicationYear "2020" @default.
- W3104273515 type Work @default.
- W3104273515 sameAs 3104273515 @default.
- W3104273515 citedByCount "16" @default.
- W3104273515 countsByYear W31042735152021 @default.
- W3104273515 countsByYear W31042735152022 @default.
- W3104273515 countsByYear W31042735152023 @default.
- W3104273515 crossrefType "proceedings-article" @default.
- W3104273515 hasAuthorship W3104273515A5035665045 @default.
- W3104273515 hasAuthorship W3104273515A5036521865 @default.
- W3104273515 hasAuthorship W3104273515A5038456766 @default.
- W3104273515 hasBestOaLocation W31042735151 @default.
- W3104273515 hasConcept C104317684 @default.
- W3104273515 hasConcept C105580179 @default.
- W3104273515 hasConcept C11413529 @default.
- W3104273515 hasConcept C119857082 @default.
- W3104273515 hasConcept C137293760 @default.
- W3104273515 hasConcept C149364088 @default.
- W3104273515 hasConcept C154945302 @default.
- W3104273515 hasConcept C165696696 @default.
- W3104273515 hasConcept C17744445 @default.
- W3104273515 hasConcept C185592680 @default.
- W3104273515 hasConcept C199539241 @default.
- W3104273515 hasConcept C203005215 @default.
- W3104273515 hasConcept C204321447 @default.
- W3104273515 hasConcept C2776135515 @default.
- W3104273515 hasConcept C2776214188 @default.
- W3104273515 hasConcept C2776650193 @default.
- W3104273515 hasConcept C38652104 @default.
- W3104273515 hasConcept C41008148 @default.
- W3104273515 hasConcept C50644808 @default.
- W3104273515 hasConcept C55493867 @default.
- W3104273515 hasConcept C57273362 @default.
- W3104273515 hasConceptScore W3104273515C104317684 @default.
- W3104273515 hasConceptScore W3104273515C105580179 @default.
- W3104273515 hasConceptScore W3104273515C11413529 @default.
- W3104273515 hasConceptScore W3104273515C119857082 @default.
- W3104273515 hasConceptScore W3104273515C137293760 @default.
- W3104273515 hasConceptScore W3104273515C149364088 @default.
- W3104273515 hasConceptScore W3104273515C154945302 @default.
- W3104273515 hasConceptScore W3104273515C165696696 @default.
- W3104273515 hasConceptScore W3104273515C17744445 @default.
- W3104273515 hasConceptScore W3104273515C185592680 @default.
- W3104273515 hasConceptScore W3104273515C199539241 @default.
- W3104273515 hasConceptScore W3104273515C203005215 @default.
- W3104273515 hasConceptScore W3104273515C204321447 @default.
- W3104273515 hasConceptScore W3104273515C2776135515 @default.