Easy Adaptation to Mitigate Gender Bias in Multilingual Text Classification

TL;DR

This paper introduces a domain adaptation method to reduce gender bias in multilingual text classification tasks, demonstrating improved fairness and performance across hate speech detection and rating prediction.

Contribution

It presents a novel application of domain adaptation to mitigate gender bias in multilingual text classification, an area not previously explored.

Findings

Effective reduction of gender bias in multilingual settings

Improved classifier performance on hate speech detection and rating prediction

Outperforms three fair-aware baseline methods

Abstract

Existing approaches to mitigate demographic biases evaluate on monolingual data, however, multilingual data has not been examined. In this work, we treat the gender as domains (e.g., male vs. female) and present a standard domain adaptation model to reduce the gender bias and improve performance of text classifiers under multilingual settings. We evaluate our approach on two text classification tasks, hate speech detection and rating prediction, and demonstrate the effectiveness of our approach with three fair-aware baselines.