Measuring Bias in Contextualized Word Representations

Keita Kurita, Nidhi Vyas, Ayush Pareek, Alan W Black, Yulia Tsvetkov


Abstract
Contextual word embeddings such as BERT have achieved state of the art performance in numerous NLP tasks. Since they are optimized to capture the statistical properties of training data, they tend to pick up on and amplify social stereotypes present in the data as well. In this study, we (1) propose a template-based method to quantify bias in BERT; (2) show that this method obtains more consistent results in capturing social biases than the traditional cosine based method; and (3) conduct a case study, evaluating gender bias in a downstream task of Gender Pronoun Resolution. Although our case study focuses on gender bias, the proposed technique is generalizable to unveiling other biases, including in multiclass settings, such as racial and religious biases.
Anthology ID:
W19-3823
Volume:
Proceedings of the First Workshop on Gender Bias in Natural Language Processing
Month:
August
Year:
2019
Address:
Florence, Italy
Venues:
ACL | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
166–172
URL:
https://www.aclweb.org/anthology/W19-3823
DOI:
10.18653/v1/W19-3823
Bib Export formats:
BibTeX MODS XML EndNote
PDF:
https://www.aclweb.org/anthology/W19-3823.pdf