Significance of Chain of Thought in Gender Bias Mitigation for English-Dravidian Machine Translation

TL;DR

This study investigates gender bias in English-Dravidian machine translation, demonstrating that Chain of Thought prompting significantly reduces bias and highlighting the importance of language-specific strategies for fairer translations.

Contribution

The paper introduces the application of Chain of Thought prompting to mitigate gender bias in English-Dravidian machine translation, showing substantial bias reduction and emphasizing language-specific approaches.

Findings

Bias reduction from 80% to 4% in Telugu

Bias reduction from 40% to 0% in Kannada

Highlights need for language-specific bias mitigation strategies

Abstract

Gender bias in machine translation (MT) sys- tems poses a significant challenge to achieving accurate and inclusive translations. This paper examines gender bias in machine translation systems for languages such as Telugu and Kan- nada from the Dravidian family, analyzing how gender inflections affect translation accuracy and neutrality using Google Translate and Chat- GPT. It finds that while plural forms can reduce bias, individual-centric sentences often main- tain the bias due to historical stereotypes. The study evaluates the Chain of Thought process- ing, noting significant bias mitigation from 80% to 4% in Telugu and from 40% to 0% in Kan- nada. It also compares Telugu and Kannada translations, emphasizing the need for language specific strategies to address these challenges and suggesting directions for future research to enhance fairness in both data preparation and prompts during inference.