PerSenT

Person SenTiment，一个用于新闻领域作者情绪预测的挑战数据集。我们介绍 PerSenT，一个众包数据集，它捕获作者对新闻文章中主要实体的情绪。该数据集包含 5.3k 个文档的注释和 3.2k 个独特实体的 38k 个段落。该数据集还包括段落级情感注释，以为任务提供更细粒度的监督。Person SenTiment: A challenge dataset for author sentiment prediction in the news domain. We introduce PerSenT, a crowdsourced dataset that captures authors' sentiments toward the main entities in news articles. The dataset comprises annotations for 5.3k documents, as well as 38k paragraphs associated with 3.2k unique entities. It also includes paragraph-level sentiment annotations to provide finer-grained supervision for the task.

搜集汇总

背景与挑战

背景概述

PerSenT是一个新闻领域作者情绪预测挑战数据集，通过众包方式捕获作者对新闻文章中主要实体的情绪。它包含5.3k个文档注释和3.2k个独特实体的38k个段落，并提供段落级情感注释以实现细粒度监督。