AG's corpus of news articles
- Type
- corpus
- Venue
- University of Pisa (groups.di.unipi.it)
- Year
- 2026
- Source
- unipi
- Access
- free
- Language
- English
- Added
- 2026-07-17T20:18:03.645808+00:00
- Verified
- 2026-07-17T20:18:03.645808+00:00
Summary
AG's corpus is a collection of more than 1 million news articles gathered from over 2,000 news sources by ComeToMyHead, an academic news search engine running since July 2004. It is provided by the academic community for non-commercial research purposes. The widely-used 'AG News' topic-classification subset takes the four largest classes (World, Sports, Business, Sci/Tech) yielding 120,000 training and 7,600 testing samples using title and description fields.
Keywords
news classification text nlp benchmark ag-news corpus non-commercial
Topics
NLP / Text Classification / News
Research notes
- The original unipi page was fetchable via search. Available in DB and XML versions. The AG News 4-class subset is one of the most widely used text classification benchmarks. Citation: Del Corso, Gulli, Romani 'Ranking a stream of news' WWW 2005; Gulli 'The anatomy of a news search engine' WWW 2005.