Peer-Review Dataset with Annotated Arguments

该数据集包含了来自多个计算机科学会议的同行评审，为了便于进行论点挖掘研究，这些评审已用论点类型进行了注释。数据集提供了词级别和句子级别的注释，总计包含28,502个标注词元和1,401个句子。为了评估注释质量，我们使用了Krippendorff's alpha方法。该数据集规模为12,135篇收集到的评审，其中77篇进行了样本标注，用于论点挖掘任务。

This dataset consists of peer reviews sourced from multiple computer science conferences. To facilitate argument mining research, these reviews have been annotated with argument types. It offers both word-level and sentence-level annotations, containing a total of 28,502 annotated word tokens and 1,401 sentences. To evaluate the quality of the annotations, Krippendorff's alpha coefficient was employed. The dataset encompasses 12,135 collected peer reviews, with 77 of them being annotated as samples for the argument mining task.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于论点挖掘的同行评审数据集，名为AMSR，包含从计算机科学会议中收集的评审文本，并标注了非论点、支持性和攻击性论点类型。数据集基于OpenReview平台爬取，采样了77份评审进行人工标注，适用于自然语言处理中的论点分析研究。