Missing Entity Type Instances Dataset

本研究构建了一个大规模数据集，用于推断知识库中缺失的实体类型实例。该数据集由马萨诸塞大学阿默斯特分校和微软研究院合作创建，旨在通过分析现有知识库内容及外部维基百科信息，自动评估和完善知识库的完整性。数据集的创建涉及使用两个不同时间点的知识库快照，以更真实地模拟知识库的更新过程。此数据集主要应用于知识库完成任务，特别是解决实体类型信息缺失的问题，提高下游任务如问答系统的性能。

This study develops a large-scale dataset for inferring missing entity type instances within knowledge bases. Collaboratively constructed by the University of Massachusetts Amherst and Microsoft Research, this dataset aims to automatically evaluate and enhance the completeness of knowledge bases by analyzing existing knowledge base content and external Wikipedia information. The dataset's creation utilizes two knowledge base snapshots from distinct timestamps to more realistically simulate the knowledge base update process. Primarily targeted at knowledge base completion tasks, this dataset specifically addresses the problem of missing entity type information, thereby improving the performance of downstream applications such as question answering systems.