CookDial

CookDial数据集由IDLab, Ghent University – imec创建，包含260个面向任务的对话，每个对话都基于一个食谱文档。数据集旨在促进研究基于程序性知识理解的对话系统。CookDial对话具有两个独特特征：(i) 对话流程与支持文档之间的程序性对齐；(ii) 涉及长句分割、指令改写和对话上下文中指代消解的复杂代理决策。数据集的应用领域包括开发能够理解和执行特定顺序步骤的对话系统，如烹饪指导。

The CookDial dataset was created by IDLab, Ghent University – imec, and contains 260 task-oriented dialogues, each grounded in a corresponding recipe document. This dataset aims to facilitate research on dialogue systems focused on understanding procedural knowledge. CookDial dialogues feature two unique characteristics: (i) procedural alignment between the dialogue flow and the supporting reference documents; (ii) complex agent decision-making involving long sentence segmentation, instruction paraphrasing, and coreference resolution in dialogue contexts. Application scenarios of the dataset include developing dialogue systems that can understand and execute specific sequential steps, such as cooking guidance.