Diving Deep into Context-Aware Neural Machine Translation

TL;DR

This paper thoroughly evaluates various document-level context-aware neural machine translation models across multiple domains, revealing that different architectures excel in different tasks and that back-translation enhances performance when document-level data is scarce.

Contribution

It provides a comprehensive analysis of the effectiveness of different context-aware NMT architectures across diverse domains and tasks, highlighting the importance of task-specific approaches and data augmentation techniques.

Findings

No single best architecture for all tasks

Context-aware models improve pronoun and headline translation

Document-level back-translation compensates for limited data

Abstract

Context-aware neural machine translation (NMT) is a promising direction to improve the translation quality by making use of the additional context, e.g., document-level translation, or having meta-information. Although there exist various architectures and analyses, the effectiveness of different context-aware NMT models is not well explored yet. This paper analyzes the performance of document-level NMT models on four diverse domains with a varied amount of parallel document-level bilingual data. We conduct a comprehensive set of experiments to investigate the impact of document-level NMT. We find that there is no single best approach to document-level NMT, but rather that different architectures come out on top on different tasks. Looking at task-specific problems, such as pronoun resolution or headline translation, we find improvements in the context-aware systems, even in cases where the corpus-level metrics like BLEU show no significant improvement. We also show that document-level back-translation significantly helps to compensate for the lack of document-level bi-texts.