MTOP

MTOP数据集是由Facebook创建的一个全面的多语言任务导向语义解析基准。该数据集包含100,000个标注的语句，涵盖6种语言和11个领域，旨在解决现有数据集在语言多样性和标注量上的不足。MTOP不仅包含简单的意图和槽位检测，还支持复杂的嵌套查询，适用于更系统的多语言实验和评估。数据集的创建过程涉及合成语句的生成、翻译和质量控制，确保了数据的高质量和多样性。MTOP的应用领域广泛，特别是在多语言任务导向对话系统的研究中，有助于提升模型在不同语言环境下的性能。

The MTOP dataset is a comprehensive multilingual task-oriented semantic parsing benchmark developed by Facebook. It includes 100,000 annotated utterances spanning 6 languages and 11 domains, and is constructed to address the limitations of existing datasets in terms of linguistic diversity and annotation volume. Beyond basic intent and slot detection, MTOP also supports complex nested queries, making it applicable for more rigorous multilingual experiments and evaluations. The dataset creation process encompasses synthetic utterance generation, translation, and quality control procedures, which guarantee the high quality and diversity of the data. MTOP has a wide range of application scenarios, especially in the research of multilingual task-oriented dialogue systems, where it assists in enhancing the performance of models across different language contexts.