File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
  • Find via Find It@HKUL

Conference Paper: Lexical-constrained neural machine translation with explicit alignment guidance

TitleLexical-constrained neural machine translation with explicit alignment guidance
Authors
KeywordsMachine Translation & Multilinguality
Issue Date2021
PublisherAAAI Press. The Journal's web site is located at https://aaai.org/Library/AAAI/aaai-library.php
Citation
Proceedings of the 35th Association for the Advancement of Artificial Intelligence (AAAI) Conference on Artificial Intelligence (AAAI-21), Virtual Conference, USA, 2-9 February 2021, v. 35 n. 14, p. 12630-12638 How to Cite?
AbstractLexically constrained neural machine translation (NMT), which leverages pre-specified translation to constrain NMT, has practical significance in interactive translation and NMT domain adaption. Previous work either modify the decoding algorithm or train the model on augmented dataset. These methods suffer from either high computational overheads or low copying success rates. In this paper, we investigate Att-Input and Att-Output, two alignment-based constrained decoding methods. These two methods revise the target tokens during decoding based on word alignments derived from encoder-decoder attention weights. Our study shows that Att-Input translates better while Att-Output is more computationally efficient. Capitalizing on both strengths, we further propose EAM-Output by introducing an explicit alignment module (EAM) to a pretrained Transformer. It decodes similarly as EAM-Output, except using alignments derived from the EAM. We leverage the word alignments induced from Att-Input as labels and train the EAM while keeping the parameters of the Transformer frozen. Experiments on WMT16 De-En and WMT16 Ro-En show the effectiveness of our approaches on constrained NMT. In particular, the proposed EAM-Output method consistently outperforms previous approaches in translation quality, with light computational overheads over unconstrained baseline.
DescriptionAAAI-21 Technical Tracks 14 / AAAI Technical Track on Speech and Natural Language Processing I
Persistent Identifierhttp://hdl.handle.net/10722/305948
ISSN

 

DC FieldValueLanguage
dc.contributor.authorChen, G-
dc.contributor.authorChen, Y-
dc.contributor.authorLi, VOK-
dc.date.accessioned2021-10-20T10:16:38Z-
dc.date.available2021-10-20T10:16:38Z-
dc.date.issued2021-
dc.identifier.citationProceedings of the 35th Association for the Advancement of Artificial Intelligence (AAAI) Conference on Artificial Intelligence (AAAI-21), Virtual Conference, USA, 2-9 February 2021, v. 35 n. 14, p. 12630-12638-
dc.identifier.issn2159-5399-
dc.identifier.urihttp://hdl.handle.net/10722/305948-
dc.descriptionAAAI-21 Technical Tracks 14 / AAAI Technical Track on Speech and Natural Language Processing I-
dc.description.abstractLexically constrained neural machine translation (NMT), which leverages pre-specified translation to constrain NMT, has practical significance in interactive translation and NMT domain adaption. Previous work either modify the decoding algorithm or train the model on augmented dataset. These methods suffer from either high computational overheads or low copying success rates. In this paper, we investigate Att-Input and Att-Output, two alignment-based constrained decoding methods. These two methods revise the target tokens during decoding based on word alignments derived from encoder-decoder attention weights. Our study shows that Att-Input translates better while Att-Output is more computationally efficient. Capitalizing on both strengths, we further propose EAM-Output by introducing an explicit alignment module (EAM) to a pretrained Transformer. It decodes similarly as EAM-Output, except using alignments derived from the EAM. We leverage the word alignments induced from Att-Input as labels and train the EAM while keeping the parameters of the Transformer frozen. Experiments on WMT16 De-En and WMT16 Ro-En show the effectiveness of our approaches on constrained NMT. In particular, the proposed EAM-Output method consistently outperforms previous approaches in translation quality, with light computational overheads over unconstrained baseline.-
dc.languageeng-
dc.publisherAAAI Press. The Journal's web site is located at https://aaai.org/Library/AAAI/aaai-library.php-
dc.relation.ispartofProceedings of the AAAI Conference on Artificial Intelligence-
dc.subjectMachine Translation & Multilinguality-
dc.titleLexical-constrained neural machine translation with explicit alignment guidance-
dc.typeConference_Paper-
dc.identifier.emailLi, VOK: vli@eee.hku.hk-
dc.identifier.authorityLi, VOK=rp00150-
dc.identifier.hkuros327673-
dc.identifier.volume35-
dc.identifier.issue14-
dc.identifier.spage12630-
dc.identifier.epage12638-
dc.publisher.placeUnited States-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats