File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Go from the general to the particular: Multi-domain translation with Domain Transformation Networks

TitleGo from the general to the particular: Multi-domain translation with Domain Transformation Networks
Authors
Issue Date2020
PublisherAAAI Press. The Journal's web site is located at https://aaai.org/Library/AAAI/aaai-library.php
Citation
Proceedings of the 34th Association for the Advancement of Artificial Intelligence (AAAI) Conference on Artificial Intelligence (AAAI-20), New York, NY, USA, 7-12 February 2020, v. 34 n. 5, p. 9233-9241 How to Cite?
AbstractThe key challenge of multi-domain translation lies in simultaneously encoding both the general knowledge shared across domains and the particular knowledge distinctive to each domain in a unified model. Previous work shows that the standard neural machine translation (NMT) model, trained on mixed-domain data, generally captures the general knowledge, but misses the domain-specific knowledge. In response to this problem, we augment NMT model with additional domain transformation networks to transform the general representations to domain-specific representations, which are subsequently fed to the NMT decoder. To guarantee the knowledge transformation, we also propose two complementary supervision signals by leveraging the power of knowledge distillation and adversarial learning. Experimental results on several language pairs, covering both balanced and unbalanced multi-domain translation, demonstrate the effectiveness and universality of the proposed approach. Encouragingly, the proposed unified model achieves comparable results with the fine-tuning approach that requires multiple models to preserve the particular knowledge. Further analyses reveal that the domain transformation networks successfully capture the domain-specific knowledge as expected.1
DescriptionAAAI-20 Technical Tracks 5 / Section: AAAI Technical Track: Natural Language Processing
Persistent Identifierhttp://hdl.handle.net/10722/287897
ISSN

 

DC FieldValueLanguage
dc.contributor.authorWang, Y-
dc.contributor.authorWang, L-
dc.contributor.authorShi, S-
dc.contributor.authorLi, VOK-
dc.contributor.authorTu, Z-
dc.date.accessioned2020-10-05T12:04:50Z-
dc.date.available2020-10-05T12:04:50Z-
dc.date.issued2020-
dc.identifier.citationProceedings of the 34th Association for the Advancement of Artificial Intelligence (AAAI) Conference on Artificial Intelligence (AAAI-20), New York, NY, USA, 7-12 February 2020, v. 34 n. 5, p. 9233-9241-
dc.identifier.issn2159-5399-
dc.identifier.urihttp://hdl.handle.net/10722/287897-
dc.descriptionAAAI-20 Technical Tracks 5 / Section: AAAI Technical Track: Natural Language Processing-
dc.description.abstractThe key challenge of multi-domain translation lies in simultaneously encoding both the general knowledge shared across domains and the particular knowledge distinctive to each domain in a unified model. Previous work shows that the standard neural machine translation (NMT) model, trained on mixed-domain data, generally captures the general knowledge, but misses the domain-specific knowledge. In response to this problem, we augment NMT model with additional domain transformation networks to transform the general representations to domain-specific representations, which are subsequently fed to the NMT decoder. To guarantee the knowledge transformation, we also propose two complementary supervision signals by leveraging the power of knowledge distillation and adversarial learning. Experimental results on several language pairs, covering both balanced and unbalanced multi-domain translation, demonstrate the effectiveness and universality of the proposed approach. Encouragingly, the proposed unified model achieves comparable results with the fine-tuning approach that requires multiple models to preserve the particular knowledge. Further analyses reveal that the domain transformation networks successfully capture the domain-specific knowledge as expected.1-
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.titleGo from the general to the particular: Multi-domain translation with Domain Transformation Networks-
dc.typeConference_Paper-
dc.identifier.emailLi, VOK: vli@eee.hku.hk-
dc.identifier.authorityLi, VOK=rp00150-
dc.identifier.doi10.1609/aaai.v34i05.6461-
dc.identifier.hkuros315146-
dc.identifier.volume34-
dc.identifier.issue5-
dc.identifier.spage9233-
dc.identifier.epage9241-
dc.publisher.placeUnited States-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats