Knowledge Graph Matching Track Gold Standard

该数据集是为了评估知识图谱匹配工具而创建的部分黄金标准，特别用于2018年和2019年的OAEI知识图谱赛道。这些黄金标准包含了来自多个维基的高主题重叠的实体和架构映射，以及一个涉及歌词维基的隐藏评估任务。这些黄金标准既包括简单匹配也包括非简单匹配，其中大量映射是通过亚马逊MTurk的众包方式创建的。实例级映射的标注者间一致性达到了0.87。数据集规模包括385,000个独立维基，总文章数量超过5千万篇；其中还包含了来自歌词维基的2百万个实例。该数据集的任务是本体匹配和知识图谱对齐。

This dataset is a collection of partial gold standards developed to evaluate knowledge graph matching tools, specifically for the Knowledge Graph Tracks of OAEI 2018 and 2019. These gold standards include entity and schema mappings with high thematic overlap from multiple wikis, as well as a hidden evaluation task involving Lyrics Wiki. These gold standards cover both simple and complex mappings, with a large portion of the mappings generated via crowdsourcing on Amazon MTurk. The inter-annotator agreement for instance-level mappings reaches 0.87. In terms of scale, the dataset encompasses 385,000 independent wikis with a total of over 50 million articles, and it also contains 2 million instances from Lyrics Wiki. The tasks addressed by this dataset are ontology matching and knowledge graph alignment.