Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey on Gender Bias in Natural Language Processing

Karolina Stanczak, Isabelle Augenstein

TL;DR

The paper systematically surveys 304 NLP studies on gender bias, mapping how gender is defined in society and linguistics to formal bias concepts, and examining resources, measurement, detection, and mitigation. It identifies four core limitations hindering progress: binary gender treatment, monolingual focus, lack of bias testing in new methods, and methodological flaws with narrow definitions and baselines. It synthesizes diverse datasets, lexica, and evaluation metrics (e.g., PMI, WEAT/SEAT, WinoBias, WinoMT, StereoSet, CrowS-Pairs) and outlines data- and algorithmic-based debiasing strategies, emphasizing language-specific and inclusive approaches. The authors advocate for transparent gender definitions, expanded multilingual resources, formal bias testing in model development, and robust, standardized benchmarks to drive responsible progress in gender bias mitigation for NLP.

Abstract

Language can be used as a means of reproducing and enforcing harmful stereotypes and biases and has been analysed as such in numerous research. In this paper, we present a survey of 304 papers on gender bias in natural language processing. We analyse definitions of gender and its categories within social sciences and connect them to formal definitions of gender bias in NLP research. We survey lexica and datasets applied in research on gender bias and then compare and contrast approaches to detecting and mitigating gender bias. We find that research on gender bias suffers from four core limitations. 1) Most research treats gender as a binary variable neglecting its fluidity and continuity. 2) Most of the work has been conducted in monolingual setups for English or other high-resource languages. 3) Despite a myriad of papers on gender bias in NLP methods, we find that most of the newly developed algorithms do not test their models for bias and disregard possible ethical considerations of their work. 4) Finally, methodologies developed in this line of research are fundamentally flawed covering very limited definitions of gender bias and lacking evaluation baselines and pipelines. We suggest recommendations towards overcoming these limitations as a guide for future research.

A Survey on Gender Bias in Natural Language Processing

TL;DR

The paper systematically surveys 304 NLP studies on gender bias, mapping how gender is defined in society and linguistics to formal bias concepts, and examining resources, measurement, detection, and mitigation. It identifies four core limitations hindering progress: binary gender treatment, monolingual focus, lack of bias testing in new methods, and methodological flaws with narrow definitions and baselines. It synthesizes diverse datasets, lexica, and evaluation metrics (e.g., PMI, WEAT/SEAT, WinoBias, WinoMT, StereoSet, CrowS-Pairs) and outlines data- and algorithmic-based debiasing strategies, emphasizing language-specific and inclusive approaches. The authors advocate for transparent gender definitions, expanded multilingual resources, formal bias testing in model development, and robust, standardized benchmarks to drive responsible progress in gender bias mitigation for NLP.

Abstract

Language can be used as a means of reproducing and enforcing harmful stereotypes and biases and has been analysed as such in numerous research. In this paper, we present a survey of 304 papers on gender bias in natural language processing. We analyse definitions of gender and its categories within social sciences and connect them to formal definitions of gender bias in NLP research. We survey lexica and datasets applied in research on gender bias and then compare and contrast approaches to detecting and mitigating gender bias. We find that research on gender bias suffers from four core limitations. 1) Most research treats gender as a binary variable neglecting its fluidity and continuity. 2) Most of the work has been conducted in monolingual setups for English or other high-resource languages. 3) Despite a myriad of papers on gender bias in NLP methods, we find that most of the newly developed algorithms do not test their models for bias and disregard possible ethical considerations of their work. 4) Finally, methodologies developed in this line of research are fundamentally flawed covering very limited definitions of gender bias and lacking evaluation baselines and pipelines. We suggest recommendations towards overcoming these limitations as a guide for future research.
Paper Structure (63 sections, 8 equations, 1 figure, 3 tables)

This paper contains 63 sections, 8 equations, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Gender in Society and Linguistics
  4. Non-binary gender
  5. Determining gender
  6. Gender Bias, Sexism and Harms they make
  7. Gender Bias
  8. Sexism
  9. Harms
  10. Bias in NLP
  11. Resources
  12. Gender lexica
  13. Sentiment
  14. Gender Ladenness
  15. Valence, Arousal, and Dominance
  16. ...and 48 more sections

Figures (1)

  • Figure 1: Cumulative number of papers published on gender bias prior to June 2021.