CaseHOLD (Case Holdings On Legal Decisions)

CaseHOLD (关于法律决定的案件持有) 是一个法律数据集，由53,000多个选择题组成，用于识别所引用案件的相关持有。该数据集向律师提出了一项基本任务，并且从NLP的角度来看 (具有BiLSTM基线的0.4 F1) 具有法律意义和困难。司法裁决中的引用背景是问题的提示。答案选择是在法律决定中保留从引用文本中得出的陈述。每个引用文本有五个答案选择。正确答案是与引用文本相对应的保留声明。四个错误的答案是其他持有声明。
要阅读有关数据集的更多信息，请参阅我们的论文或博客。

CaseHOLD (Case Holdings on Legal Decisions) is a legal dataset consisting of over 53,000 multiple-choice questions designed to identify the relevant holdings of cited cases. This dataset presents a core task for legal practitioners, and possesses both legal significance and inherent difficulty from an NLP perspective, with a BiLSTM baseline achieving an F1 score of 0.4. The citation context within judicial rulings serves as the prompt for each question. Each answer option is a holding statement extracted from the cited text in a legal decision. Every cited text is paired with five answer options: the correct answer is the holding statement corresponding to the cited text, while the four incorrect options are other unrelated holding statements.
For more information about this dataset, please refer to our associated paper or blog post.

搜集汇总

背景与挑战

背景概述

CaseHOLD是一个包含53,000多个选择题的法律数据集，用于识别法律案件中的相关持有声明，具有法律意义和NLP挑战性，每个问题提供五个答案选择。