generate_reference_files.py

import os
from datasets import load_dataset

os.makedirs('./reference_files', exist_ok=True)


for split in ['validation', 'test']:
    dataset = load_dataset('asahi417/qg_newsqa', '', split=split)
    for data in ['question', 'answer', 'sentence', 'paragraph']:
        with open('./reference_files/{}-{}.txt'.format(data, split), 'w') as f:
            if data == 'paragraph':
                f.write('\n'.join(dataset['paragraph_id']))
            else:
                f.write('\n'.join(dataset[data]))