Structured Wikipedia
Structured Contents, also published as the structured-wikipedia dataset, is Wikimedia Enterprise's machine-readable rendering of Wikipedia. Rather than raw wikitext or rendered HTML, it delivers each article as structured JSON: the abstract, the short description, the infobox, the individual sections, parsed tables, lists, images and references, plus a link to the matching Wikidata entity. It is built by the Wikimedia Foundation and is currently a beta, so the schema and coverage are still settling.
You can reach it two ways. The On-demand API returns a single article for any Wikipedia language, while the Snapshot API and the Hugging Face mirror (wikimedia/structured-wikipedia) provide bulk files covering nine languages so far: English, Spanish, Portuguese, French, German, Italian, Dutch, Indonesian and Welsh. The Hugging Face release ships Parquet and carries the English and French editions at roughly 44 GB, readable directly from DuckDB, Pandas, Polars or Spark.
For RAG this is the shape you actually want. Because the parsing is already done, you skip the fiddly wikitext handling that the raw Wikipedia dumps force on you, and the section segmentation maps almost directly onto retrieval passages: one chunk per section, with the abstract as a ready-made summary and the references as provenance. It suits grounded question answering and entity-aware pipelines, where the Wikidata identifiers let you disambiguate and link the things your documents mention.
The licence follows Wikipedia itself. The text is Creative Commons Attribution-ShareAlike 4.0, with some content also under the GFDL. Commercial use is permitted, but you must attribute Wikipedia and release any derived text under the same share-alike terms, and embedded media can carry its own licences, so check before redistributing. Compared with the plain Wikipedia dumps we list, this trades completeness and every-language coverage for a cleaner, pre-structured form; pair it with Wikidata when you need the underlying facts as statements rather than prose.
Related sources
DBpedia
A knowledge graph built by pulling the structured parts of Wikipedia, mainly the infoboxes, into machine-readable data. It holds billions of facts about people, places, organisations, and more as RDF triples, small subject, predicate, object statements, which you can search with the SPARQL query language.
Freebase
A collaborative knowledge base once run by Google, now retired but still available as downloadable data dumps. It holds structured facts about millions of entities, and much of its content has since moved into Wikidata.
OpenCyc
The open release of Cyc, one of the oldest attempts to hand-build common-sense knowledge for machines. It holds hundreds of thousands of concepts and millions of assertions about how the everyday world fits together. Now archived, but the data is still available.
SYNTH
A fully open synthetic corpus of amplified multilingual encyclopaedic text with built-in reasoning traces and exercises covering RAG, information extraction and QA. Released by PleIAs with the AI Alliance, it targets training and evaluating small, grounded, citeable reasoning models.