A Causal Inference Method for Reducing Gender Bias in Word Embedding Relations
This work addresses gender bias in NLP systems, which can perpetuate stereotypes and discrimination, by improving debiasing methods for word embeddings, though it is incremental as it builds on prior debiasing techniques.
The paper tackles gender bias in word embedding relations, which existing debiasing methods fail to address, by proposing a causal inference approach that achieves state-of-the-art results on gender-debiasing, lexical- and sentence-level evaluations, and downstream coreference resolution tasks.
Word embedding has become essential for natural language processing as it boosts empirical performances of various tasks. However, recent research discovers that gender bias is incorporated in neural word embeddings, and downstream tasks that rely on these biased word vectors also produce gender-biased results. While some word-embedding gender-debiasing methods have been developed, these methods mainly focus on reducing gender bias associated with gender direction and fail to reduce the gender bias presented in word embedding relations. In this paper, we design a causal and simple approach for mitigating gender bias in word vector relation by utilizing the statistical dependency between gender-definition word embeddings and gender-biased word embeddings. Our method attains state-of-the-art results on gender-debiasing tasks, lexical- and sentence-level evaluation tasks, and downstream coreference resolution tasks.