Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287666191> ?p ?o ?g. }
Showing items 1 to 69 of
69
with 100 items per page.
- W4287666191 abstract "Using a language model (LM) pretrained on two languages with large monolingual data in order to initialize an unsupervised neural machine translation (UNMT) system yields state-of-the-art results. When limited data is available for one language, however, this method leads to poor translations. We present an effective approach that reuses an LM that is pretrained only on the high-resource language. The monolingual LM is fine-tuned on both languages and is then used to initialize a UNMT model. To reuse the pretrained LM, we have to modify its predefined vocabulary, to account for the new language. We therefore propose a novel vocabulary extension method. Our approach, RE-LM, outperforms a competitive cross-lingual pretraining model (XLM) in English-Macedonian (En-Mk) and English-Albanian (En-Sq), yielding more than +8.3 BLEU points for all four translation directions." @default.
- W4287666191 created "2022-07-25" @default.
- W4287666191 creator A5061754269 @default.
- W4287666191 creator A5067446532 @default.
- W4287666191 creator A5068580049 @default.
- W4287666191 date "2020-09-16" @default.
- W4287666191 modified "2023-09-28" @default.
- W4287666191 title "Reusing a Pretrained Language Model on Languages with Limited Corpora for Unsupervised NMT" @default.
- W4287666191 doi "https://doi.org/10.48550/arxiv.2009.07610" @default.
- W4287666191 hasPublicationYear "2020" @default.
- W4287666191 type Work @default.
- W4287666191 citedByCount "0" @default.
- W4287666191 crossrefType "posted-content" @default.
- W4287666191 hasAuthorship W4287666191A5061754269 @default.
- W4287666191 hasAuthorship W4287666191A5067446532 @default.
- W4287666191 hasAuthorship W4287666191A5068580049 @default.
- W4287666191 hasBestOaLocation W42876661911 @default.
- W4287666191 hasConcept C104317684 @default.
- W4287666191 hasConcept C105580179 @default.
- W4287666191 hasConcept C137293760 @default.
- W4287666191 hasConcept C138885662 @default.
- W4287666191 hasConcept C149364088 @default.
- W4287666191 hasConcept C154945302 @default.
- W4287666191 hasConcept C185592680 @default.
- W4287666191 hasConcept C18903297 @default.
- W4287666191 hasConcept C203005215 @default.
- W4287666191 hasConcept C204321447 @default.
- W4287666191 hasConcept C206588197 @default.
- W4287666191 hasConcept C2777601683 @default.
- W4287666191 hasConcept C2985367798 @default.
- W4287666191 hasConcept C41008148 @default.
- W4287666191 hasConcept C41895202 @default.
- W4287666191 hasConcept C55493867 @default.
- W4287666191 hasConcept C622187 @default.
- W4287666191 hasConcept C86803240 @default.
- W4287666191 hasConceptScore W4287666191C104317684 @default.
- W4287666191 hasConceptScore W4287666191C105580179 @default.
- W4287666191 hasConceptScore W4287666191C137293760 @default.
- W4287666191 hasConceptScore W4287666191C138885662 @default.
- W4287666191 hasConceptScore W4287666191C149364088 @default.
- W4287666191 hasConceptScore W4287666191C154945302 @default.
- W4287666191 hasConceptScore W4287666191C185592680 @default.
- W4287666191 hasConceptScore W4287666191C18903297 @default.
- W4287666191 hasConceptScore W4287666191C203005215 @default.
- W4287666191 hasConceptScore W4287666191C204321447 @default.
- W4287666191 hasConceptScore W4287666191C206588197 @default.
- W4287666191 hasConceptScore W4287666191C2777601683 @default.
- W4287666191 hasConceptScore W4287666191C2985367798 @default.
- W4287666191 hasConceptScore W4287666191C41008148 @default.
- W4287666191 hasConceptScore W4287666191C41895202 @default.
- W4287666191 hasConceptScore W4287666191C55493867 @default.
- W4287666191 hasConceptScore W4287666191C622187 @default.
- W4287666191 hasConceptScore W4287666191C86803240 @default.
- W4287666191 hasLocation W42876661911 @default.
- W4287666191 hasOpenAccess W4287666191 @default.
- W4287666191 hasPrimaryLocation W42876661911 @default.
- W4287666191 hasRelatedWork W10667231 @default.
- W4287666191 hasRelatedWork W11012074 @default.
- W4287666191 hasRelatedWork W11890898 @default.
- W4287666191 hasRelatedWork W12732426 @default.
- W4287666191 hasRelatedWork W13343785 @default.
- W4287666191 hasRelatedWork W13780460 @default.
- W4287666191 hasRelatedWork W1619002 @default.
- W4287666191 hasRelatedWork W4629839 @default.
- W4287666191 hasRelatedWork W6659077 @default.
- W4287666191 hasRelatedWork W867563 @default.
- W4287666191 isParatext "false" @default.
- W4287666191 isRetracted "false" @default.
- W4287666191 workType "article" @default.