FakeNewsCorpus: Millions of News Articles from Curated Sources

This corpus contains millions of news articles scraped from a curated list of sources, published by several27 and last updated in January 2020. It provides a large-scale text repository for natural language processing and fake news detection research.

Use Cases

• Classification of news veracity using the curated source labels
• Language model pre-training using the news article text
• Bias analysis across the curated list of news outlets

Strengths

• Scale of millions of news articles
• Apache-2.0 open-source license
• Curated source list for news veracity labeling

Limitations

• Temporal staleness with no updates since early 2020
• Potential for formatting noise inherent in scraped text
• Lack of detailed column documentation in metadata

Provenance

Source

Curated list of news data sources

Collection Method

scraped

Time Range

Up to January 2020

Freshness

Last updated January 2020; no recent updates recorded.