Improving Zero-Shot Translation by Disentangling Positional Information

Danni Liu, Jan Niehues, James Cross, Francisco Guzmán, Xian Li

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingAcademicpeer-review

Abstract

Multilingual neural machine translation has shown the capability of directly translating between language pairs unseen in training, i.e. zero-shot translation. Despite being conceptually attractive, it often suffers from low output quality. The difficulty of generalizing to new translation directions suggests the model representations are highly specific to those language pairs seen in training. We demonstrate that a main factor causing the language-specific representations is the positional correspondence to input tokens. We show that this can be easily alleviated by removing residual connections in an encoder layer. With this modification, we gain up to 18.5 BLEU points on zero-shot translation while retaining quality on supervised directions. The improvements are particularly prominent between related languages, where our proposed model outperforms pivot-based translation. Moreover, our approach allows easy integration of new languages, which substantially expands translation coverage. By thorough inspections of the hidden layer outputs, we show that our approach indeed leads to more language-independent representations.
Original languageEnglish
Title of host publicationProceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)
PublisherAssociation for Computational Linguistics
Pages1259–1273
Number of pages15
Volume1
EditionAugust 2021
DOIs
Publication statusPublished - 2021
EventThe Joint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing - Online, Unknown
Duration: 1 Aug 20216 Aug 2021
https://2021.aclweb.org/

Conference

ConferenceThe Joint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing
Abbreviated titleACL-IJCNLP 2021
Country/TerritoryUnknown
Period1/08/216/08/21
Internet address

Cite this