Mitigating Gender Bias Amplification in Distribution by Posterior Regularization

05/13/2020
by   Shengyu Jia, et al.
1

Advanced machine learning techniques have boosted the performance of natural language processing. Nevertheless, recent studies, e.g., Zhao et al. (2017) show that these techniques inadvertently capture the societal bias hidden in the corpus and further amplify it. However, their analysis is conducted only on models' top predictions. In this paper, we investigate the gender bias amplification issue from the distribution perspective and demonstrate that the bias is amplified in the view of predicted probability distribution over labels. We further propose a bias mitigation approach based on posterior regularization. With little performance loss, our method can almost remove the bias amplification in the distribution. Our study sheds the light on understanding the bias amplification.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
06/21/2019

Mitigating Gender Bias in Natural Language Processing: Literature Review

As Natural Language Processing (NLP) and Machine Learning (ML) tools ris...
research
01/01/2023

CORGI-PM: A Chinese Corpus For Gender Bias Probing and Mitigation

As natural language processing (NLP) for gender bias becomes a significa...
research
09/16/2021

Balancing out Bias: Achieving Fairness Through Training Reweighting

Bias in natural language processing arises primarily from models learnin...
research
06/04/2023

Taught by the Internet, Exploring Bias in OpenAIs GPT3

This research delves into the current literature on bias in Natural Lang...
research
09/23/2019

NLVR2 Visual Bias Analysis

NLVR2 (Suhr et al., 2019) was designed to be robust for language bias th...
research
11/21/2022

Identifying gender bias in blockbuster movies through the lens of machine learning

The problem of gender bias is highly prevalent and well known. In this p...
research
03/17/2022

Entropy-based Attention Regularization Frees Unintended Bias Mitigation from Lists

Natural Language Processing (NLP) models risk overfitting to specific te...

Please sign up or login with your details

Forgot password? Click here to reset