Marked Attribute Bias in Natural Language Inference

TL;DR

This paper investigates gender bias in natural language inference, focusing on how biased word embeddings contribute to marked attribute bias, and proposes a new debiasing method that improves bias mitigation in static embeddings.

Contribution

It identifies limitations of current debiasing techniques for static word embeddings and introduces a novel intrinsic bias measure and postprocessing scheme to better mitigate marked attribute bias.

Findings

Current debiasing methods do not mitigate marked attribute bias effectively.

A new intrinsic bias measure correlates with the marked attribute effect.

The proposed debiasing method achieves state-of-the-art results on bias test sets.

Abstract

Reporting and providing test sets for harmful bias in NLP applications is essential for building a robust understanding of the current problem. We present a new observation of gender bias in a downstream NLP application: marked attribute bias in natural language inference. Bias in downstream applications can stem from training data, word embeddings, or be amplified by the model in use. However, focusing on biased word embeddings is potentially the most impactful first step due to their universal nature. Here we seek to understand how the intrinsic properties of word embeddings contribute to this observed marked attribute effect, and whether current post-processing methods address the bias successfully. An investigation of the current debiasing landscape reveals two open problems: none of the current debiased embeddings mitigate the marked attribute error, and none of the intrinsic bias measures are predictive of the marked attribute effect. By noticing that a new type of intrinsic bias measure correlates meaningfully with the marked attribute effect, we propose a new postprocessing debiasing scheme for static word embeddings. The proposed method applied to existing embeddings achieves new best results on the marked attribute bias test set. See https://github.com/hillary-dawkins/MAB.