Taskmaster-1

该数据集名为Taskmaster-1，涵盖了六个领域的13,215个基于任务的对话。这些对话是通过两种方法创建的：一种是采用代理和众包工人的“奥兹巫师”方法，另一种是工人独立编写对话的自对话方法。该数据集旨在为对话系统提供更真实、更多样化的会话数据。此外，该数据集旨在为面向目标的对话系统提供一个现实的基础线，并包含了每个对话的API调用和参数信息。规模上，该数据集包含了13,215个对话，其任务是面向目标对话系统的发展。

Taskmaster-1 is a dataset consisting of 13,215 task-oriented dialogues spanning six distinct domains. These dialogues were constructed via two methodologies: the "Oz Wizard" approach, which leverages both agents and crowd workers, and the self-dialogue approach, where crowd workers independently draft the full dialogues. This dataset aims to provide more realistic and diverse conversational data for dialogue systems. Additionally, it is designed to offer a realistic baseline for task-oriented dialogue systems and includes API call records and parameter information for each dialogue. In terms of scale, this dataset contains 13,215 dialogues and is intended for the development of task-oriented dialogue systems.