DuConv知识对话数据集

DuConv旨在考察模型是否可以在对话过程中充分利用外部知识(既包括结构化知识，也包括非结构化知识)，并且在生成对话回复的过程中引入外部知识，是首个bot主动的中文知识对话数据集。

DuConv is designed to evaluate whether models can fully leverage external knowledge, including both structured and unstructured knowledge, during dialogue interactions, and integrate such external knowledge when generating dialogue responses. It is the first Chinese knowledge-grounded dialogue dataset with proactive bot involvement.