Abstracts of Randomized Controlled Trials (AbstRCT)

该数据集包含了700篇不同疾病随机对照试验（RCTs）的摘要，每个摘要都按跨度级别标注了三种论证组件（AC）类型（主张、主要主张、证据）和三种论证关系（AR）类型（支持、攻击、部分攻击）。在实验中，将“证据”标签用作前提，数据集被划分为350个训练样本、50个开发样本和100个测试样本。该任务的主题是论证挖掘。

This dataset comprises 700 abstracts of randomized controlled trials (RCTs) across various diseases. Each abstract is annotated at the span level with three types of argumentative components (ACs): Claim, Main Claim, and Evidence, as well as three types of argumentative relations (ARs): Support, Attack, and Partial Attack. In the experiments, the "Evidence" label was used as the premise, and the dataset was split into 350 training samples, 50 development samples, and 100 test samples. The task focuses on argument mining.