Gender Bias in Nepali-English Machine Translation: A Comparison of LLMs and Existing MT Systems
Abstract: Bias in Nepali NLP is seldom addressed due to its classification as a low-resource language, perpetuating biases in subsequent systems. Our work addresses gender bias in Nepali-English machine translation. With the advent of Large Language Models, there is an opportunity to mitigate this bias. We quantify and evaluate gender bias by building an occupation corpus and contextualizing three gender-bias challenge sets for Nepali. While gender bias is prominent in existing translation systems, LLMs perform better in both gender-neutral and gender-specific contexts. Despite their quirks, LLMs can be a valuable alternative to traditional machine learning systems for culture-rich languages like Nepali.
Paper Type: Short
Research Area: Ethics, Bias, and Fairness
Research Area Keywords: Ethics, Bias, and Fairness, Machine Translation, Computational Social Science and Cultural Analytics
Contribution Types: Model analysis & interpretability, Data resources, Data analysis
Languages Studied: Nepali, English
Submission Number: 2835
Loading