EventNarrative
收藏资源简介:
EventNarrative是一个大规模的事件中心知识图谱到文本数据集,由佛罗里达大学计算机科学系创建。该数据集包含约220,000个图及其对应的自然语言文本,利用丰富的本体论,所有KG实体都与文本链接,并通过手动注释确认高质量数据。数据集旨在推动事件中心研究,并为研究人员提供一个定义良好、大规模的数据集,以更好地评估现有的和未来的知识图谱到文本模型。数据集内容涵盖多种事件类型,从体育赛季到社交媒体活动,关系包括位置、事件类型和开始/结束时间。创建过程中,首先从EventKG提取事件,然后为每个事件增加额外的对应Wikidata信息。数据集的应用领域包括事件描述和知识图谱到文本生成模型的评估。
EventNarrative is a large-scale event-centric knowledge graph-to-text dataset created by the Department of Computer Science at the University of Florida. This dataset contains approximately 220,000 graphs and their corresponding natural language texts. Leveraging a rich ontology, all knowledge graph (KG) entities are linked to their respective texts, and high-quality data is validated via manual annotations. This dataset aims to advance event-centric research and provide researchers with a well-defined, large-scale resource to better evaluate existing and future knowledge graph-to-text models. The dataset covers a diverse range of event types, spanning from sports seasons to social media campaigns, with relations including location, event type, and start/end times. In its creation process, events were first extracted from EventKG, and additional corresponding Wikidata information was then added to each event. Application scenarios of this dataset include event description and the evaluation of knowledge graph-to-text generation models.




