Curiosity

Curiosity数据集由马里兰大学计算机科学系和Facebook的研究人员共同创建，旨在研究用户与数字助手之间的信息寻求对话。该数据集包含14,048个英语对话，每个对话围绕地理主题进行，如地缘政治实体和地点，并标注了用户的先验知识、消息级的对话行为以及用户对消息的反应。数据集的创建过程涉及两步：首先通过定制界面收集对话，然后进行对话行为的事后标注。Curiosity数据集的应用领域主要集中在改善对话系统，特别是那些需要理解用户知识背景并据此提供相关信息的系统，以提高用户参与度和对话质量。

The Curiosity Dataset was co-created by researchers from the Department of Computer Science at the University of Maryland and Facebook, aiming to study information-seeking conversations between users and digital assistants. This dataset includes 14,048 English conversations, each centered on geographic topics such as geopolitical entities and locations, and is annotated with users' prior knowledge, message-level dialogue acts, and users' reactions to messages. The dataset was developed through a two-step process: first, conversations were collected via a custom-built interface, followed by post-hoc annotation of dialogue acts. The main application scenarios of the Curiosity Dataset focus on improving dialogue systems, especially those that need to understand users' knowledge backgrounds to provide relevant information, so as to enhance user engagement and conversation quality.