metadata
license: apache-2.0
This is the Japanese subset of TyDi QA.
number of examples
primary task
- train: 16288 (original: 166916)
- validation: 1709 (original: 18670)
secondary task ... no japanese data
filtering script
from datasets import load_dataset
dataset = load_dataset("tydiqa", "secondary_task")
ja_dataset = dataset.filter(lambda example: example['id'].startswith('jp'))