Double-Hard Debias: Tailoring Word Embeddings for Gender Bias Mitigation
This addresses gender bias in NLP applications, offering an incremental improvement over existing debiasing methods.
The paper tackles gender bias in word embeddings by proposing Double-Hard Debias, a technique that purifies embeddings against corpus regularities before bias removal, resulting in significantly reduced bias while preserving semantics on three benchmarks.
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.