EXPERT

EXPERT数据集由太平洋西北国家实验室创建，包含七个动态异构学术图数据集，涵盖人工智能和核不扩散两个领域。数据集通过提取科学出版物的上下文和内容信息构建，包含3.5M节点和34M时间戳边缘，用于测试多步图预测模型的有效性。数据集的创建涉及从顶级会议和数据库收集论文，使用GROBID和CERMINE解析PDF以提取元数据和内容，并通过专家筛选关键词进行过滤。该数据集旨在解决动态异构图学习中的挑战，特别是在科学合作、伙伴关系和能力发展方面的应用。

The EXPERT dataset was created by the Pacific Northwest National Laboratory. It comprises seven dynamically heterogeneous academic graph datasets covering two domains: artificial intelligence and nuclear non-proliferation. Constructed by extracting contextual and content information from scientific publications, the dataset contains 3.5 million nodes and 34 million timestamped edges, and is used to test the effectiveness of multi-step graph prediction models. The creation of the dataset involved collecting papers from top conferences and databases, parsing PDF documents using GROBID and CERMINE to extract metadata and content, and filtering the collected materials via expert-curated keywords. This dataset aims to address the challenges in dynamic heterogeneous graph learning, particularly for applications in scientific collaboration, partnerships, and capability development.