Doc2Dial

该数据集名为Doc2Dial，包含了3,474个对话和44,149个回合用于训练，以及661个对话和8,539个回合用于评估，它专门为面向目标的、基于文档的对话任务而设计。该数据集被用于DialDoc21共享任务，重点关注知识识别和响应生成。在规模上，训练集包含3,474个对话和44,149个回合，而评估集包含661个对话和8,539个回合，任务类型为面向目标的文档基础对话。

This dataset, named Doc2Dial, consists of 3,474 dialogues with 44,149 turns for training, and 661 dialogues with 8,539 turns for evaluation. It is specifically developed for goal-oriented document-grounded dialogue tasks. This dataset was employed in the DialDoc21 shared task, which focuses on knowledge identification and response generation. Regarding its scale, the training set contains 3,474 dialogues and 44,149 turns, while the evaluation set comprises 661 dialogues and 8,539 turns, and the corresponding task type is goal-oriented document-grounded dialogue.