This dataset contains long-form feature articles from multiple U.S. online media outlets, covering a range of subject domains and writing styles.
Use Cases
- Pre-training for stylistic diversity: Articles from multiple outlets and domains can be used to expose models to a variety of editorial voices that a single-publisher corpus cannot provide.
- Long-form summarization training and evaluation: Feature-length articles can be used as source material to train and evaluate summarization models on texts longer than those in typical news-snippet benchmarks.