BriefGPT.xyz
Jun, 2019
测量上下文化词表示中的偏见
Measuring Bias in Contextualized Word Representations
HTML
PDF
Keita Kurita, Nidhi Vyas, Ayush Pareek, Alan W Black, Yulia Tsvetkov
TL;DR
本研究基于模板方法提出了一种量化BERT中偏见的方法，并且通过性别代词解析的案例研究证明了该方法在捕捉社会偏见方面的优越性，同时也指出了该方法的普遍适用性，包括在多类别设置中使用的种族和宗教偏见。
Abstract
Contextual word embeddings such as
bert
have achieved state of the art performance in numerous
nlp
tasks. Since they are optimized to capture the statistical properties of training data, they tend to pick up on a
→