Attenuating Bias in Word Vectors

TL;DR

This paper presents simple methods to detect and reduce gender, race, ethnicity, and age biases in word embeddings, improving fairness in NLP applications.

Contribution

It introduces new techniques using names to identify and attenuate various biases in word vectors, extending beyond gender to other social biases.

Findings

Names effectively reveal gender bias in embeddings.

Bias attenuation methods reduce stereotypical associations.

Names can detect multiple social biases in word vectors.

Abstract

Word vector representations are well developed tools for various NLP and Machine Learning tasks and are known to retain significant semantic and syntactic structure of languages. But they are prone to carrying and amplifying bias which can perpetrate discrimination in various applications. In this work, we explore new simple ways to detect the most stereotypically gendered words in an embedding and remove the bias from them. We verify how names are masked carriers of gender bias and then use that as a tool to attenuate bias in embeddings. Further, we extend this property of names to show how names can be used to detect other types of bias in the embeddings such as bias based on race, ethnicity, and age.