How to Split: the Effect of Word Segmentation on Gender Bias in Speech Translation

TL;DR

This paper investigates how different word segmentation methods affect gender bias in speech translation, revealing that sub-word methods like BPE increase bias, while character-based segmentation can reduce it without harming translation quality.

Contribution

It introduces a comparative analysis of segmentation strategies on gender bias in speech translation and proposes a combined approach to mitigate bias while maintaining quality.

Findings

BPE segmentation increases gender bias in translation.

Character-based segmentation better captures gendered linguistic features.

A combined segmentation approach balances bias reduction and translation quality.

Abstract

Having recognized gender bias as a major issue affecting current translation technologies, researchers have primarily attempted to mitigate it by working on the data front. However, whether algorithmic aspects concur to exacerbate unwanted outputs remains so far under-investigated. In this work, we bring the analysis on gender bias in automatic translation onto a seemingly neutral yet critical component: word segmentation. Can segmenting methods influence the ability to translate gender? Do certain segmentation approaches penalize the representation of feminine linguistic markings? We address these questions by comparing 5 existing segmentation strategies on the target side of speech translation systems. Our results on two language pairs (English-Italian/French) show that state-of-the-art sub-word splitting (BPE) comes at the cost of higher gender bias. In light of this finding, we propose a combined approach that preserves BPE overall translation quality, while leveraging the higher ability of character-based segmentation to properly translate gender.