A Global Database of Society (www.gdeltproject.org)

GDELT is a massive, free, open-data platform that continuously ingests print, broadcast, web and social media from nearly every country in over 100 languages to create a computable, near‑realtime “global graph” of human society. Backed by Google Jigsaw and partnered with the Internet Archive, GDELT produces three 15‑minute‑updated data streams: an Event Database (300+ coded event types, ~60 attributes per event, georeferenced, records back to 1979 and extending toward 1800), a Global Knowledge Graph (millions of themes, thousands of emotions, named entities and relationships), and a visual‑narrative stream that samples up to ~1M images/day processed with Google Vision API. Its Translingual pipeline machine‑translates 65 languages (covering ~98.4% of non‑English volume) to enable uniform analysis. For the AI/ML community GDELT is significant both as a research corpus and an operational monitoring platform: it supplies trillions of datapoints for tasks like event extraction, cross‑lingual NLP, relation extraction, temporal and graph modeling, multimodal vision+text work, and conflict/early‑warning forecasting. The dataset’s scale, multilingual coverage, historical depth, and frequent updates make it ideal for training and benchmarking large models, studying bias and representativeness in global media, and building real‑time analytic pipelines using BigQuery, GDELT’s Analysis Service, or raw CSV exports.