Task2Dial

该数据集名为Task2Dial，是一个创新的数据集，包含基于文档的任务驱动对话，其中信息提供者向信息接受者提供指令以完成特定任务。这些对话需要使用常识知识，并且具有丰富的词汇量和变化性。由于对话中涉及常识知识的运用以及从基础文档中进行释义的需求，该数据集具有一定的挑战性。它包含平均18.15轮对话和每轮19.79个词汇。该数据集的任务是增强常识知识的任务型对话。

This dataset, named Task2Dial, is an innovative corpus consisting of document-grounded task-oriented dialogues, where information providers deliver instructions to information recipients to complete specific tasks. These dialogues require the application of commonsense knowledge, and feature rich vocabulary and high conversational variability. Given the requirements of leveraging commonsense knowledge and paraphrasing from the source documents, this dataset presents notable challenges. On average, each dialogue comprises 18.15 conversational turns and 19.79 tokens per turn. The core objective of this dataset is to advance commonsense-aware task-oriented dialogues.