MD-SEE
收藏资源简介:
MD-SEE数据集是一个多维度事件抽取数据集,旨在解决实际应用中事件抽取的挑战,如选择合适的模式和执行抽取过程。数据集由12个数据集组成,涵盖不同领域、复杂性和语言设置。数据集构建过程中,通过模式改写和检索增强生成,将事件抽取任务分解为模式检索和模式感知抽取。数据集的应用领域包括知识图谱构建、问答系统、信息检索和事件预测等。
The MD-SEE dataset is a multi-dimensional event extraction dataset that aims to address the challenges of event extraction in real-world applications, such as selecting appropriate extraction schemas and executing extraction procedures. The dataset consists of 12 constituent datasets covering diverse domains, complexity levels, and language settings. During the dataset construction process, the event extraction task is decomposed into schema retrieval and schema-aware extraction via schema rewriting and retrieval-augmented generation. The applicable domains of this dataset include knowledge graph construction, question answering systems, information retrieval, event prediction, and other related fields.




