使用人口普查数据测量语言模型中的规范性和描述性偏见（英文版）

英文标题：Measuring Normative and Descriptive Biases in Language Models Using Census Data中文摘要：本文研究了在预训练语言模型中，关于职业性别分布如何反映在模型中，通过使用模板化句子结合特定职业性别分布数据，通过评估不同语言的模型得出结论。英文摘要：We investigate in this paper how distributions of occupations with respect togender is reflected in pre-trained language models. Such distributions are notalways aligned to normative ideals, nor do they necessarily reflect adescriptive assessment of reality. In this paper, we introduce an approach formeasuring to what degree pre-trai