Cloze Coherence (CC)
收藏arXiv2025-09-30 收录
下载链接：
https://www.cs.rochester.edu/nlp/rocstories/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个大规模的、包含连贯与不连贯例子的集合。其中，连贯的例子是由完整且书写良好的文本组成，而不连贯的例子则源于句法或语义的扰动。该数据集重点关注由两个短句构成的连贯例子，这些短句之间存在共指关系并展现出修辞关系。数据集的规模包括97,903个训练例子、1,871个开发例子以及1,871个测试例子，其任务是进行语篇连贯性评估。
This dataset is a large-scale collection encompassing both coherent and incoherent examples. Specifically, coherent examples consist of complete and well-written texts, while incoherent ones are derived from syntactic or semantic perturbations. This dataset focuses on coherent examples constructed from two short sentences, which bear coreferential relations and demonstrate rhetorical relationships between one another. The dataset comprises 97,903 training examples, 1,871 development examples, and 1,871 test examples, with its target task being discourse coherence evaluation.
提供机构：
Authors of the paper
