Compensatory Debiasing For Gender Imbalances In Language Models

Tae Jin Woo, Woo Jeoung Nam, Yeong Joon Ju, Seong Whan Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

Pre-trained language models (PLMs) learn gender bias from imbalances in human-written corpora. This bias leads to critical social issues when deploying PLMs in real-world scenarios. However, minimizing bias is limited by the trade-off due to the degradation of language modeling performance. It is particularly challenging to detach and remove biased representations in the embedding space because the learned linguistic knowledge entails bias. To address this problem, we propose a compensatory debiasing strategy to reduce gender bias while preserving linguistic knowledge. This strategy utilizes two types of sentences to distinguish biased knowledge: stereotype and non-stereotype sentences. We assign small angles and distances to pairs of representations of the two gender groups to mitigate bias for the stereotype sentences. At the same time, we maximize the agreement for the representations of the debiasing model and the original model to maintain linguistic knowledge for the non-stereotype sentences. To validate our approach, we measure the performance of the debiased model using the following evaluation metrics: SEAT, StereoSet, CrowS-Pairs, and GLUE. Our experimental results demonstrate that the model fine-tuned by our strategy has the lowest level of bias while retaining knowledge of PLMs.

Original languageEnglish
Title of host publicationICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728163277
DOIs
Publication statusPublished - 2023
Event48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023 - Rhodes Island, Greece
Duration: 2023 Jun 42023 Jun 10

Publication series

NameICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume2023-June
ISSN (Print)1520-6149

Conference

Conference48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023
Country/TerritoryGreece
CityRhodes Island
Period23/6/423/6/10

Bibliographical note

Publisher Copyright:
© 2023 IEEE.

Keywords

  • gender bias mitigation
  • Language model
  • social bias

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Compensatory Debiasing For Gender Imbalances In Language Models'. Together they form a unique fingerprint.

Cite this