CS&E Colloquium: Mitigating Language-Dependent Ethnic Bias in BERT

The computer science colloquium takes place on Mondays from 11:15 a.m. - 12:15 p.m.

This week's speaker, Alice Oh (KAIST), will be giving a talk titled "Mitigating Language-Dependent Ethnic Bias in BERT."

Abstract

BERT and other large-scale language models (LMs) contain gender and racial bias. They also exhibit other dimensions of social bias, most of which have not been studied in depth, and some of which vary depending on the language. In this talk, I present a study of ethnic bias and how it varies across languages by analyzing and mitigating ethnic bias in monolingual BERT for English, German, Spanish, Korean, Turkish, and Chinese. To observe and quantify ethnic bias, we develop a novel metric called Categorical Bias score. Then we propose two methods for mitigation; first using a multilingual model, and second using contextual word alignment of two monolingual models. We compare our proposed methods with monolingual BERT and show that these methods effectively alleviate the ethnic bias. Which of the two methods works better depends on the amount of NLP resources available for that language. We additionally experiment with Arabic and Greek to verify that our proposed methods work for a wider variety of languages.

Biography

Alice Oh is a Professor in the School of Computing at KAIST. She received her PhD in 2008 from MIT and joined KAIST in the same year. Her major research area is at the intersection of machine learning and computational social science. Within machine learning, she studies various models designed for analyzing written text including social media posts, news articles, and personal conversations. She also looks at non-textual data such as social network friendship and logs from online games for which she interacts closely with social scientists for an interdisciplinary approach to computational social science. She has served as Tutorial Chair for NeurIPS 2019, Diversity & Inclusion Chair for ICLR 2019, and Program Chair for ICLR 2021. She is serving as Program Chair for NeurIPS 2022 and General Chair for ACM FAccT 2022.