ArgSciChat

ArgSciChat是一个包含41个科学家之间关于20篇NLP论文的辩论对话数据集。该数据集由达姆施塔特工业大学的UKP实验室创建，旨在通过对话形式探讨科学论文的论证内容。数据集中的对话涉及探索性和论证性交互，每条对话由NLP专家同步进行，涉及单篇论文的内容讨论。ArgSciChat数据集的应用领域包括评估特定领域对话代理的少样本迁移学习方法，以及用于训练和评估文档基础的对话代理。

ArgSciChat is a conversational dataset consisting of 41 debates between scientists about 20 NLP papers. This dataset was developed by the UKP Lab at Technische Universität Darmstadt, with the goal of exploring the argumentative content of scientific papers via conversational interactions. The conversations in this dataset involve both exploratory and argumentative interactions: each conversation is carried out synchronously by NLP experts and centers on the content discussion of one individual paper. The application domains of the ArgSciChat dataset include evaluating few-shot transfer learning methods for domain-specific conversational agents, as well as training and evaluating document-grounded conversational agents.