Encompassing 120,000 examples in JSONL format pairing raw text from the C4 English dataset with corresponding website descriptions. Each entry maps unstructured web content to metadata fields sourced from Wikipedia to provide context about the source domain.
Use Cases
- Train text classification models to categorize web content using the Wikipedia-sourced metadata as ground truth labels
- Fine-tune language models to improve domain-specific context awareness by associating the 'text' field with its website description
- Conduct semantic alignment studies to measure the consistency between a website's raw 'text' content and its official Wikipedia description
Strengths
- 120,000 examples in JSONL format
- Includes a 'text' column containing raw content extracted from the C4 English dataset
- Features metadata fields containing website descriptions extracted from Wikipedia
- Pairs unstructured web data with authoritative domain-level context