Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods
This addresses bias in NLP systems for fairness, though it is incremental as it builds on existing debiasing techniques.
The paper tackles gender bias in coreference resolution by introducing the WinoBias benchmark, showing that existing systems link gendered pronouns to pro-stereotypical entities with 21.1 higher F1 score on average, and proposes a data-augmentation method combined with word-embedding debiasing to remove this bias without harming performance on standard benchmarks.
We introduce a new benchmark, WinoBias, for coreference resolution focused on gender bias. Our corpus contains Winograd-schema style sentences with entities corresponding to people referred by their occupation (e.g. the nurse, the doctor, the carpenter). We demonstrate that a rule-based, a feature-rich, and a neural coreference system all link gendered pronouns to pro-stereotypical entities with higher accuracy than anti-stereotypical entities, by an average difference of 21.1 in F1 score. Finally, we demonstrate a data-augmentation approach that, in combination with existing word-embedding debiasing techniques, removes the bias demonstrated by these systems in WinoBias without significantly affecting their performance on existing coreference benchmark datasets. Our dataset and code are available at http://winobias.org.