[RE] Double-Hard Debias: Tailoring Word Embeddings for Gender Bias Mitigation
arXiv:2104.06973v13 citations
Originality Synthesis-oriented
AI Analysis
This tackles the problem of unintended gender bias in NLP systems, which can affect fairness and accuracy in applications, and appears to be an incremental improvement over existing debiasing techniques.
The paper addresses gender bias in word embeddings, which can propagate to downstream NLP tasks, by proposing a method to mitigate this bias.
Despite widespread use in natural language processing (NLP) tasks, word embeddings have been criticized for inheriting unintended gender bias from training corpora. programmer is more closely associated with man and homemaker is more closely associated with woman. Such gender bias has also been shown to propagate in downstream tasks.