MULTIOPED

MULTIOPED是一个开放领域的新闻社论语料库，专注于自动视角发现。该数据集由宾夕法尼亚大学创建，包含2794条数据，每条数据包含一个争议话题的查询和两个从不同角度回应的新闻社论。每个社论都附有一个单句视角，抽象地总结了社论的关键论点。MULTIOPED旨在支持与新闻社论中论证结构相关的多种任务，特别是在自动视角发现方面。数据集的应用领域包括自动视角发现、论证结构分析和多任务学习，旨在解决新闻社论中隐含论证结构的自动识别和分析问题。

MULTIOPED is an open-domain news editorial corpus dedicated to automatic perspective discovery. Developed by the University of Pennsylvania, this dataset comprises 2,794 instances. Each instance includes a query regarding a controversial topic, paired with two news editorials that respond from distinct perspectives. Each editorial is accompanied by a single-sentence perspective statement that abstractly summarizes the core argument of the editorial. MULTIOPED is designed to support a variety of tasks related to the argumentative structure of news editorials, with a particular focus on automatic perspective discovery. Its application areas include automatic perspective discovery, argumentative structure analysis, and multi-task learning, aiming to address the automatic identification and analysis of implicit argumentative structures in news editorials.