StereoBias
收藏资源简介:
StereoBias数据集是专为检测偏见和刻板印象而创建的,包含5012个句子,这些句子被标注为五个类别中的偏见和刻板印象:宗教、性别、社会经济地位、种族和职业,其余类型的偏见被标记为“其他”。数据集由 StereoSet 和 Crows-Pairs 数据集的句子组成,经过三个标注者的独立标注,最终通过多数投票确定标签。该数据集旨在帮助研究模型在偏见和刻板印象检测方面的性能,并为构建更公平和有效的AI系统提供支持。
The StereoBias dataset was specifically created for bias and stereotype detection. It contains 5,012 sentences annotated for biases and stereotypes across five categories: religion, gender, socioeconomic status, race, and occupation, while other types of biases are labeled as "Other". The dataset is compiled from sentences sourced from the StereoSet and Crows-Pairs datasets, which underwent independent annotation by three annotators, with final labels determined via majority voting. This dataset aims to assist research on model performance in bias and stereotype detection, and provide support for building more equitable and effective AI systems.
StereotypeAsCatalystForBias数据集概述
基本信息
- 数据集名称:StereotypeAsCatalystForBias
- 托管平台:GitHub
数据集描述
(注:根据提供的README内容,该数据集未包含具体描述信息)




