BiToD

该数据集是为多语言面向任务的对话系统设计的，采用了简约的自然语言设计理念。在数据集生成过程中，设计了一个管道以最小化自然语言输入，以便专注于核心的槽值提取任务。该数据集的主要任务是对话状态跟踪（Dialogue State Tracking）。

This dataset is designed for multilingual task-oriented dialogue systems, adopting the minimalist natural language design philosophy. During the dataset generation process, a processing pipeline was devised to minimize natural language input, so as to focus on the core slot value extraction task. The primary task of this dataset is Dialogue State Tracking (DST).