News articles and human-written summaries from newser.com categorized for multi-document summarization tasks. Each record pairs a professionally edited `summary` with a `document` field containing multiple source articles concatenated using a '|||||' delimiter.
Use Cases
- Train transformer-based models for multi-document summarization using the `document` and `summary` features.
- Evaluate the factual consistency of generated text against the editor-provided `summary` labels.
- Develop algorithms to parse and process multiple distinct articles from the single `document` string using the '|||||' delimiter.
Strengths
- Uses the '|||||' special token to separate multiple source articles within the `document` column.
- Contains summaries professionally written by editors from newser.com.
- Provides a multi-to-one mapping between source news articles and a single human-written summary.