DISAPERE

DISAPERE数据集专注于学术同行评审中的话语结构，特别是审稿人与作者之间的互动。该数据集由马萨诸塞大学阿默斯特分校创建，包含506对审稿-反驳文本，总计超过20,000个句子，所有句子均由专家进行标注。数据集不仅关注核心论点，还对审稿和反驳中的每个句子进行标注，提供细粒度的非论点类型标签。DISAPERE旨在为研究者提供一个全面且高质量的测试集合，用于训练和微调模型，以帮助决策者如区域主席等更有效地理解和吸收他们必须阅读的讨论内容。

The DISAPERE dataset focuses on the discourse structure of academic peer review, particularly the interactions between reviewers and authors. Developed by the University of Massachusetts Amherst, it contains 506 pairs of reviewer-rebuttal texts, totaling over 20,000 sentences, all of which were annotated by domain experts. In addition to targeting core arguments, the dataset annotates every sentence in both reviews and rebuttals, providing fine-grained non-argument type labels. The DISAPERE dataset aims to provide researchers with a comprehensive and high-quality test collection for model training and fine-tuning, to help decision-makers such as area chairs more effectively understand and assimilate the discussion content they are required to review.