ChineseSquad
ChineseSquad (中文机器阅读理解数据集) is a dataset specifically designed for Chinese machine reading comprehension. It is created by translating and manually correcting the original SQuAD (Stanford Question Answering Dataset) into Chinese. The dataset includes both V1.1 and V2.0 versions of SQuAD. However, due to some translation challenges (especially with short answers and document translations), the Chinese version has slightly fewer examples compared to the original English SQuAD¹.
Here are some key details about the ChineseSquad dataset:
- Data Sources:
- ChineseSquad is derived from the original SQuAD dataset through machine translation and subsequent manual corrections.
- It includes both the V1.1 and V2.0 versions.
- Data Size:
- The dataset contains both questions with answers and questions without answers.
- Here's a breakdown of the data:
- squad-zen 1.0 train: 68,213 examples with answers, 43,498 examples without answers (total: 110k)
- squad-zen 1.0 dev: 8,326 examples with answers, 5,954 examples without answers (total: 14k)
- squad 2.0 train: 46,530 examples with answers, 43,498 examples without answers (total: 90k)
- squad 2.0 dev: 3,391 examples with answers, 5,945 examples without answers (total: 9k)
- squad 1.1 dev: 7,679 examples (no answers provided)
- squad 1.1 train: 55,526 examples (no answers provided)