简体中文
EnglishWhat Is The GDELT Project and How Can You Use Its Global News Data?
Updated
The GDELT Project is a free, open platform that monitors news media worldwide — print, broadcast, and web — in over 100 languages, and turns that coverage into a structured, continuously updating database of global human society. It is best suited for researchers, analysts, journalists, and developers who need machine-readable data on events, themes, emotions, and networks across countries and languages, rather than a single curated news feed. Its archives reach back to January 1, 1979, and update every 15 minutes.
What GDELT actually does
GDELT watches the world's news media from nearly every country, in more than 100 languages, every moment of the day. Rather than summarizing articles, it codifies them: it identifies the people, locations, organizations, themes, sources, emotions, counts, quotes, images, and events that appear in the coverage.
The result is described as "a realtime network diagram and database of global human society for open research." In practice, that means you get quantitative, queryable records of what is being reported, where, by whom, and with what emotional tone — not just a stream of headlines.
Two features distinguish it from Western-centric news aggregators:
- Global reach — it deliberately moves beyond a Western media focus toward a broader global perspective on what is happening and how the world feels about it.
- Language coverage — over 100 languages are monitored, which makes cross-language media analysis possible without you building your own translation and extraction pipeline.
Main data products and coverage
GDELT organizes its output around several layers, reflected in the site's own structure: Watching The Entire World, Computing on The Entire World: Events & Networks, and Querying, Analyzing and Downloading.
| Dimension | What it captures |
|---|---|
| Events | Who did what to whom, where, and when |
| Networks | Relationships among actors, sources, and themes |
| Themes | Recurring topics and narratives in coverage |
| Emotions | Sentiment and emotional tone of reporting |
| Sources | Which outlets and media types are driving coverage |
| Quotes, images, counts | Supporting extracted detail from articles |
Coverage spans print, broadcast, and web formats. The historical archive begins January 1, 1979, and the dataset refreshes every 15 minutes, so it supports both long-run historical analysis and near-realtime monitoring.
How to get started
The official site frames the entry point as "The World Right Now, This Moment," and provides dedicated sections for querying, analyzing, and downloading data. A practical path looks like this:
- Define your question. Decide whether you need events (e.g., protests, cooperation, conflict), themes, emotional tone, or network relationships. This determines which data layer you pull.
- Choose your access mode. GDELT offers querying, analysis, and download options. For one-off exploration, start with query tools; for reproducible research, download the data and process it locally.
- Set your scope. Filter by time range (back to 1979), geography, language, or actor. Narrowing scope early keeps result sets manageable.
- Verify against the source. Because GDELT is machine-extracted from news, spot-check records against the original articles before drawing conclusions.
- Iterate. Use the 15-minute update cycle for monitoring tasks, or freeze a historical slice for a fixed study.
The site also maintains a blog documenting updates and use cases, which is worth checking when a query returns unexpected results.
What kinds of work this supports
GDELT's design supports several concrete tasks:
- Tracking global events — following breaking developments anywhere on the planet as they are reported.
- Conflict forecasting and de-escalation research — the project explicitly frames its vision around mapping happiness and conflict and potentially forecasting global conflict.
- Cross-language media analysis — comparing how different societies and languages frame the same event.
- Emotion and discourse studies — quantifying the emotional tone of coverage at scale.
- Network analysis — examining relationships among people, organizations, and media sources.
Practical cautions
Because GDELT is built from automated extraction across enormous volumes of multilingual media, treat it as a signal source rather than ground truth. Coverage density varies by country and language, so absence of records is not proof that an event did not occur. For any high-stakes conclusion, triangulate GDELT data with primary sources and domain expertise.