---
license: apache-2.0
---

This is the Japanese subset of [TyDi QA](https://github.com/google-research-datasets/tydiqa).

**number of examples**

- primary task
  - train: 16288 (original: 166916)
  - validation: 1709 (original: 18670)

- secondary task ... no japanese data


**filtering script**

```python
from datasets import load_dataset

dataset = load_dataset("tydiqa", "secondary_task")
ja_dataset = dataset.filter(lambda example: example['id'].startswith('jp'))
```