Double-Hard Debias: Tailoring Word Embeddings for Gender Bias Mitigation

Abstract

Word embeddings derived from human-generated corpora inherit strong gender bias which can be further amplified by downstream models. Some commonly adopted debiasing approaches, including the seminal Hard Debias algorithm, apply post-processing procedures that project pre-trained word embeddings into a subspace orthogonal to an inferred gender subspace. We discover that semantic-agnostic corpus regularities such as word frequency captured by the word embeddings negatively impact the performance of these algorithms. We propose a simple but effective technique, Double Hard Debias, which purifies the word embeddings against such corpus regularities prior to inferring and removing the gender subspace. Experiments on three bias mitigation benchmarks show that our approach preserves the distributional semantics of the pre-trained word embeddings while reducing gender bias to a significantly larger degree than prior approaches.

arXiv commentAccepted to ACL 2020

References in corpus (7)

• Efficient Estimation of Word Representations in Vector Space
• Distributed Representations of Words and Phrases and their Compositionality
• Semantics derived automatically from language corpora contain human-like biases
• Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
• FRAGE: Frequency-Agnostic Word Representation
• Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias
• Understanding Undesirable Word Embedding Associations

Cited by in corpus (3)

• VERB: Visualizing and Interpreting Bias Mitigation Techniques for Word Representations
• Does Robustness Improve Fairness? Approaching Fairness with Word Substitution Robustness Methods for Text Classification
• Adversarial Examples Generation for Reducing Implicit Gender Bias in Pre-trained Models