Unmasking Contextual Stereotypes: Measuring and Mitigating BERT's Gender Bias

TL;DR

This paper measures gender bias in BERT across English and German, compares it with real-world data, and proposes mitigation via fine-tuning and data substitution, highlighting cross-linguistic challenges.

Contribution

It introduces a method to measure gender bias in contextual embeddings and evaluates bias mitigation techniques across languages with different morphological features.

Findings

Bias measurement is effective for English but less so for German.

Mitigation reduces bias in English but faces challenges in German.

Cross-linguistic bias assessment is crucial for multilingual NLP models.

Abstract

Contextualized word embeddings have been replacing standard embeddings as the representational knowledge source of choice in NLP systems. Since a variety of biases have previously been found in standard word embeddings, it is crucial to assess biases encoded in their replacements as well. Focusing on BERT (Devlin et al., 2018), we measure gender bias by studying associations between gender-denoting target words and names of professions in English and German, comparing the findings with real-world workforce statistics. We mitigate bias by fine-tuning BERT on the GAP corpus (Webster et al., 2018), after applying Counterfactual Data Substitution (CDS) (Maudslay et al., 2019). We show that our method of measuring bias is appropriate for languages such as English, but not for languages with a rich morphology and gender-marking, such as German. Our results highlight the importance of investigating bias and mitigation techniques cross-linguistically, especially in view of the current emphasis on large-scale, multilingual language models.