Freebase

Freebase was a community-built knowledge base of entities and the facts joining them: people, places, films, companies, books, and far more, reaching tens of millions of entities and billions of facts at its height. Google bought it in 2010, drew on it to seed its own Knowledge Graph, then closed the public service in 2015 and handed the data to the Wikidata community.

The live site and its API are long gone, but the final data dumps are still downloadable, published as RDF triples: compact statements shaped as subject, predicate, and object (for example, "Paris, capital-of, France"). The full dump is bulky and leans on Freebase's own machine identifiers and schema, so budget time to load it into a triple store or graph database and to map those opaque identifiers to readable labels before it is any use.

For a brand-new project, Freebase is rarely the natural starting point. Its value today is historical and academic: reproducing or benchmarking against research that was built on it (a good number of entity-linking and knowledge-graph question-answering datasets still quote Freebase identifiers), studying how the facts were modelled, or recovering a specific 2015-era snapshot. If you want a living knowledge graph for RAG, Wikidata absorbed much of Freebase and continues to grow.

The obvious drawback is staleness: updates stopped in 2015, so anything time-sensitive is now wrong or missing, and there is no support or correction path. The dumps are released under CC BY 2.5, which allows commercial reuse as long as you credit Freebase, with no share-alike condition attached. Mapping Freebase identifiers onto current Wikidata ones takes real effort, though published crosswalks exist to smooth the way.

Treat it as the archived ancestor of Wikidata and DBpedia: use those for anything current, and reach for Freebase mainly to bridge back to older work that still speaks its identifiers.