GDELT Event Database and Global Knowledge Graph
The GDELT Project
Machine-read news from around the world, coded into events every 15 minutes.
GDELT runs news articles in 65 languages through software that extracts who did what to whom, where, and in what tone, producing a coded event table back to 1979 and a Global Knowledge Graph of people, organizations, themes, and places mentioned. It is used for conflict and protest tracking, media analysis, and early-warning models.
Access
- Cost
- FreeRaw files are free; BigQuery queries are billed by Google after its free tier.
- License
- Provider's own termsFree and open per GDELT's terms of use (citation requested)
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadCSV, BigQuery, JSON API
What's in it
- One row is
- One row per coded event (Events) or per article (GKG), with separate mention records
- Covers
- Worldwide news coverage, geocoded to country and city
- Years
- 1979-present for events; 2013-present for the GKG
- Updated
- Real time
- Size
- Hundreds of millions of event records; the 2015 GKG alone exceeds 2.5 TB
Quality
Unmatched in scale and timeliness, but fully automated coding means many duplicate, misread, or misplaced events.
- Event counts track news volume, which grew enormously over time
- Automated coding misclassifies many events and duplicates others
- Pre-2013 data was backfilled from a different, smaller source set
Connects to
- ACLED (Armed Conflict Location & Event Data): Human-coded conflict events to validate GDELT's automated counts
- Common Crawl: Common Crawl's news crawl offers the full article text GDELT summarizes
Questions it could answer
- How did global media tone about a country shift in the months before an election?
- Where were protest events reported most often last year?
- Which themes co-occur most with coverage of a given company?