Does Context Help Mitigate Gender Bias in Neural Machine Translation?

TL;DR

This paper investigates whether context-aware neural machine translation models effectively reduce gender bias, finding that while they improve feminine term translation, they may still sustain or increase bias, indicating the need for more targeted solutions.

Contribution

The study provides a detailed analysis of gender bias in context-aware NMT models, revealing limitations and the necessity for more precise bias mitigation methods.

Findings

Context-aware models improve feminine term translation accuracy.

Models can still maintain or amplify gender bias.

Highlights need for fine-grained bias mitigation approaches.

Abstract

Neural Machine Translation models tend to perpetuate gender bias present in their training data distribution. Context-aware models have been previously suggested as a means to mitigate this type of bias. In this work, we examine this claim by analysing in detail the translation of stereotypical professions in English to German, and translation with non-informative context in Basque to Spanish. Our results show that, although context-aware models can significantly enhance translation accuracy for feminine terms, they can still maintain or even amplify gender bias. These results highlight the need for more fine-grained approaches to bias mitigation in Neural Machine Translation.