README.md

metadata

license: mit
language:
  - en
tags:
  - scrape

Scrape Content Dataset v1

A human-curated benchmark dataset for evaluating web scraping engines on content quality.

Overview

This dataset contains 1,000 web pages with human-annotated ground truth for evaluating how well web scraping engines capture core content while avoiding noise (navigation, ads, footers, etc.). The dataset was created in 2025-10-21 and may become outdated over time.

Dataset Structure

CSV format with columns:

• id: Sequential identifier
• url: Full URL of the web page
• truth_text: ~100-word core snippet (main content)
• lie_text: ~10-word non-core snippet (navigation/footer/ads)
• error: Optional error message if page retrieval failed

Limitations

Created: 2025-10-21

This dataset contains real-world web pages that may become outdated over time. Known issues include:

• URLs may become unavailable
• Content may be updated or removed
• Some pages may require JavaScript to load properly

For reproducible research, consider using this as a snapshot benchmark for the time period.

Contributing

We welcome contributions to improve this dataset:

• Bug Reports: Report issues with specific entries
• Updated Annotations: Improve ground truth for existing pages
• New Content: Suggest additional web pages for evaluation
• Dataset Maintenance: Help keep URLs and content current

License

MIT License - This dataset is provided for research and benchmarking purposes. Please respect website terms of service and use appropriate rate limiting.