Multiple-Debias: A Full-process Debiasing Method for Multilingual Pre-trained Language Models

TL;DR

This paper presents Multiple-Debias, a comprehensive method for reducing gender, racial, and religious biases in multilingual pre-trained language models across four languages, using data augmentation, self-debiasing, and parameter-efficient fine-tuning.

Contribution

The paper introduces a full-process multilingual debiasing framework that outperforms monolingual methods and enhances fairness by leveraging multiple languages.

Findings

Multilingual debiasing surpasses monolingual approaches in bias mitigation.

Integrating multilingual debiasing improves model fairness.

Extended CrowS-Pairs to four additional languages for validation.

Abstract

Multilingual Pre-trained Language Models (MPLMs) have become essential tools for natural language processing. However, they often exhibit biases related to sensitive attributes such as gender, race, and religion. In this paper, we introduce a comprehensive multilingual debiasing method named Multiple-Debias to address these issues across multiple languages. By incorporating multilingual counterfactual data augmentation and multilingual Self-Debias across both pre-processing and post-processing stages, alongside parameter-efficient fine-tuning, we significantly reduced biases in MPLMs across three sensitive attributes in four languages. We also extended CrowS-Pairs to German, Spanish, Chinese, and Japanese, validating our full-process multilingual debiasing method for gender, racial, and religious bias. Our experiments show that (i) multilingual debiasing methods surpass monolingual approaches in effectively mitigating biases, and (ii) integrating debiasing information from different languages notably improves the fairness of MPLMs.