LegalCore
收藏资源简介:
LegalCore是一个针对法律领域的第一个事件及其共指信息注释的数据集。该数据集由德克萨斯A&M大学和Adobe Research共同创建,包含100个法律合同文档,总计约25万个tokens。数据集以法律合同文档为对象,这些文档的长度远超新闻文章,平均每个文档约2.5万个tokens。数据集不仅标注了个别事件提及,还标注了事件之间的共指关系,包括跨章节的远距离共指链接。该数据集旨在推动法律领域事件共指消解的研究,并可用于评估大型语言模型在事件识别和事件共指消解任务上的性能。
LegalCore is the first annotated dataset focused on events and their coreferential information in the legal domain. Co-created by Texas A&M University and Adobe Research, the dataset comprises 100 legal contract documents, with a total of approximately 250,000 tokens. The target documents of this dataset are legal contracts, which are considerably longer than news articles, with an average length of about 25,000 tokens per document. In addition to annotating individual event mentions, the dataset also labels coreferential relationships between events, including long-distance cross-chapter coreferential links. This dataset is designed to promote research on event coreference resolution in the legal field, and can be utilized to evaluate the performance of large language models (LLMs) on the tasks of event recognition and event coreference resolution.

- 1LegalCore: A Dataset for Legal Documents Event Coreference Resolution德克萨斯A&M大学计算机科学与工程学院, Adobe Research · 2025年



