Grep-BiasIR

Grep-BiasIR是由因斯布鲁克大学和林茨理工大学联合创建的数据集，旨在研究信息检索结果中的性别表征偏见。该数据集包含118个偏见敏感的中性搜索查询，覆盖了7个与性别相关的主题，如体能、育儿等。每个查询附带一个相关和一个不相关的文档，这些文档分别以女性、男性和中性三种形式呈现。数据集通过两位博士后研究人员的严格审核，确保高质量。Grep-BiasIR的应用领域包括分析搜索引擎行为、研究性别偏见对人类判断的影响以及用户查询生成分析，旨在揭示和解决信息检索中的性别偏见问题。

Grep-BiasIR is a dataset jointly developed by the University of Innsbruck and Linz Technical University, aimed at investigating gender representation bias in information retrieval results. The dataset includes 118 bias-sensitive neutral search queries spanning seven gender-related topics such as physical fitness and parenting. Each query is paired with one relevant document and one irrelevant document, which are presented in three distinct variants: female, male, and neutral. To ensure high data quality, the dataset has been rigorously reviewed by two postdoctoral researchers. Application scenarios of Grep-BiasIR cover analyzing search engine behaviors, studying the impact of gender bias on human judgment, and conducting user query generation analysis, with the ultimate goal of revealing and addressing gender bias issues in information retrieval.