Linked OpenStreetMap, Wikidata, Wikipedia and Wikivoyage dataset

A dataset published by Wikimedia Enterprise in October 2026 that links map features to encyclopaedic records. A central links table holds about 24 million links, each tying a Wikidata item to an OpenStreetMap element, a Wikipedia article or a Wikivoyage article, with a column saying how the link was established. Around it sit the OpenStreetMap elements themselves with tags and geometry, the Wikidata items and their locations, 1.5 million geolocated English Wikipedia articles and 31,000 Wikivoyage articles. Michal Migurski, an OpenStreetMap contributor, built it.

It saves twin builders the join they would otherwise do by hand: asking what a place is, what Wikipedia says about it and where its geometry lies in one query. Every table carries the Wikidata identifier, so any two can be joined on it. The files are Parquet and GeoParquet, meant for DuckDB, pandas, Polars or Spark; the whole set is about 59 GB, of which the OpenStreetMap table is 36 GB.

Licensing follows the sources: the links and Wikidata tables are CC0, the OpenStreetMap table is ODbL, and the Wikipedia and Wikivoyage tables are CC BY-SA 4.0. The snapshot reflects the sources as of 15 September 2026; no refresh schedule is given, and anyone wanting newer data is told to re-extract from the source snapshots.