TyDiQA

该数据集包含了204K个问答对，覆盖了11种在类型学上各异的语言。它专注于那些在仅包含英语的语料库中找不到的语言现象。这些问题是由不知道答案的人编写的，并在每种语言中直接收集，未经翻译。该数据集的任务是问答。

This dataset contains 204K question-answer pairs, covering 11 typologically diverse languages. It focuses on linguistic phenomena that are absent from English-only corpora. The questions were written by people who did not know the answers, and were collected directly in each language without translation. The core task of this dataset is question answering.