Probabilistic Bias Mitigation in Word Embeddings

TL;DR

This paper introduces a probabilistic framework for bias mitigation in word embeddings, proposing a new method that more effectively reduces bias while preserving semantic quality, addressing limitations of previous approaches.

Contribution

It presents a novel probabilistic bias mitigation technique that improves bias reduction in word embeddings without compromising their semantic utility.

Findings

Significantly reduces bias according to multiple metrics

Maintains embedding quality across benchmark tasks

Outperforms existing bias mitigation methods

Abstract

It has been shown that word embeddings derived from large corpora tend to incorporate biases present in their training data. Various methods for mitigating these biases have been proposed, but recent work has demonstrated that these methods hide but fail to truly remove the biases, which can still be observed in word nearest-neighbor statistics. In this work we propose a probabilistic view of word embedding bias. We leverage this framework to present a novel method for mitigating bias which relies on probabilistic observations to yield a more robust bias mitigation algorithm. We demonstrate that this method effectively reduces bias according to three separate measures of bias while maintaining embedding quality across various popular benchmark semantic tasks