Exploring Word Sense Disambiguation Abilities of Neural Machine Translation Systems 

Rebecca Marvin, Philipp Koehn

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Neural machine translation systems have been shown to achieve state-of-the-art translation performance for many language pairs. In order to produce a correct translation, MT systems must learn how to disambiguate words with multiple senses and pick the correct translation. We explore the extent to which the word embeddings for ambiguous words are able to disambiguate senses at deeper layers of the NMT encoder, which are thought to represent words with surrounding context. Consistent with previous research, we find that the NMT system fails to translate many ambiguous words correctly. We provide an evaluation framework to use for proposed improvements to word sense disambiguation abilities of NMT systems.
Original languageEnglish
Title of host publicationProceedings of the 13th Conference of the Association for Machine Translation in the Americas (Volume 1: Research Papers)
Place of PublicationBoston, MA, USA
PublisherAssociation for Machine Translation in the Americas, AMTA
Pages125-131
Number of pages7
Publication statusPublished - 21 Mar 2018
Event13th Conference of the Association for Machine Translation in the Americas - Boston, United States
Duration: 17 Mar 201821 Mar 2018
https://amtaweb.org/

Conference

Conference13th Conference of the Association for Machine Translation in the Americas
Abbreviated titleAMTA 2018
Country/TerritoryUnited States
CityBoston
Period17/03/1821/03/18
Internet address

Cite this