FairGraphBase
收藏资源简介:
FairGraphBase数据集是基于YAGO、DBpedia和Wikidata三个知识图谱生成的大型图数据集,包含人的属性、关系和个人属性等信息。数据集的目标标签设置为职业,敏感属性设置为国籍和性别,旨在评估公平感知图神经网络在知识图谱中的公平性。数据集的生成过程包括从知识图谱中提取实体、个人属性和关系,并生成属性图。FairGraphBase数据集为公平感知图神经网络的研究提供了一个新的基准,有助于理解和解决公平性问题。
FairGraphBase is a large-scale graph dataset generated from three knowledge graphs: YAGO, DBpedia and Wikidata. It contains information such as human attributes, relational facts and personal attributes. The dataset takes occupation as its target label, and sets nationality and gender as the sensitive attributes, aiming to evaluate the fairness of fairness-aware graph neural networks when applied to knowledge graphs. The dataset construction process includes extracting entities, personal attributes and relational facts from the three source knowledge graphs, and generating attributed graphs. FairGraphBase provides a new benchmark for research on fairness-aware graph neural networks, helping to advance the understanding and resolution of fairness-related issues.

- 1Benchmarking Fairness-aware Graph Neural Networks in Knowledge Graphs大阪大学 · 2025年



