README.md

WikiQA Filtered Dataset

Source

This dataset is derived from WikiQA "snap-stanford-lms/simpleQA-wikipedia-subset", filtered to ensure the articles text from clean_wikiqa_train_dataset.py contains the answer.

Cleaning Process

1. Answer Verification: Excluded 580 items where the article doesn't contain the answer
2. Final Dataset: Retained 1952 high-quality question-answer pairs from 2532 original items

Statistics

• Total Items: 1952
• Unique Articles: 1518
• Questions per Article: 1.29
• Exclusion Rate: 22.9%

Features

• metadata: Question metadata including topic and answer type
• problem: The question text
• answer: The correct answer
• doc_id: Wikipedia document ID
• doc_title: Wikipedia article title