MINDS-14

该数据集是一个多语言数据集，主要用于评估不同语言环境下的语言模型，特别关注口语理解能力。该数据集以英语数据为主，同时也包含了法语、波兰语和韩语的数据，但这三种语言的代表性相对较低。该数据集涵盖了118种语言，各种语言的比例各不相同，其任务旨在提升跨多种语言的语意理解能力。

This multilingual dataset is primarily utilized for evaluating language models across diverse linguistic contexts, with particular emphasis on spoken language understanding. Composed mainly of English data, the dataset also includes data in French, Polish, and Korean, albeit with relatively low representativeness for these three languages. It covers 118 languages with varying proportional distributions, and its tasks are designed to enhance cross-lingual semantic understanding capabilities.