PFDial

该数据集名为PFDial，包含了12,705条高质量的中文对话指令，这些指令源自440个包含5,055个流程节点的UML流程图。数据集不仅包括顺序样本，还包含决策样本，并针对域外测试的准确性制定了特定的评估指标。该数据集的规模为12,705个训练样本，其任务旨在为流程驱动的对话系统提供训练和评估。

This dataset, named PFDial, contains 12,705 high-quality Chinese conversational instructions derived from 440 UML flowcharts that incorporate a total of 5,055 process nodes. In addition to sequential samples, the dataset also includes decision-making samples, and features customized evaluation metrics specifically designed for assessing the accuracy of out-of-domain tests. Comprising 12,705 training samples, this dataset is developed to provide training and evaluation resources for process-driven conversational systems.