FICLE

该数据集名为FICLE（事实不一致性分类与解释），包含约8000个样本，每个样本由一个主张和带有不一致性类型及范围的上下文句子组成。该数据集旨在帮助检测文本中的事实不一致性，并提供解释。此外，该数据集源自FEVER数据集，利用带有“反驳”标签的样本，并包括对语法和语义领域的详细注释。规模上，共有8055个样本，任务是对事实不一致性进行检测和分类。

This dataset is named FICLE (Factuality Inconsistency Classification and Explanation). It comprises approximately 8,000 samples, each consisting of a claim and a contextual sentence annotated with inconsistency type and scope. This dataset aims to facilitate the detection of factual inconsistencies in text and provide corresponding explanations. Additionally, it is derived from the FEVER dataset, utilizing samples labeled with "refuted" and including detailed annotations across grammatical and semantic domains. In total, it contains 8,055 samples, and its core task is the detection and classification of factual inconsistencies.