TCSE
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/sonjbin/TCQA
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为TCSE，是为了训练和测试时间上下文相关的跨度提取任务而生成的一组数据。测试集是从训练数据未包含的维基百科页面中生成的，包含了118,104条用于训练的TCSE数据和9,323条用于测试的TCSE数据，这些数据是基于10,302个模板生成的。该任务的目标是时间上下文依赖的跨度提取。
This dataset is named TCSE. It is a collection of data generated for training and testing temporal context-aware span extraction tasks. The test set is constructed from Wikipedia pages excluded from the training corpus, and the dataset includes 118,104 TCSE training instances and 9,323 TCSE test instances, all of which are generated based on 10,302 templates. The core objective of this task is to extract spans dependent on temporal context.
