Causalogue 和 Causaction
收藏arXiv2024-01-16 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2401.08221v1
下载链接
链接失效反馈官方服务：
资源简介：
Causalogue 是一个包含因果关系标签的对话数据集，由GPT-4根据预定义的因果规则生成，旨在分析话语间的因果关系。Causaction 是对Breakfast数据集重新标注后的视频数据集，包含1118个视频，记录了10种不同的早餐准备过程，用于分析视频中不同动作之间的因果关系。这两个数据集均用于研究不确定性数据中的因果关系，旨在解决多结构数据和多值表示共存时的因果发现挑战。
Causalogue is a dialogue dataset with causal relation labels, generated by GPT-4 based on predefined causal rules, aiming to analyze the causal relationships between utterances. Causaction is a re-annotated video dataset based on the Breakfast dataset, which contains 1118 videos documenting 10 distinct breakfast preparation processes, and is used to analyze the causal relationships between different actions in videos. Both of these datasets are employed for researching causal relations in uncertain data, with the goal of addressing the challenge of causal discovery when multi-structure data and multi-value representation coexist.
提供机构：
西安交通大学计算机科学与技术系
创建时间：
2024-01-16
