Wikimedia Deutschland has announced the launch of the Wikidata Embedding Project, a new database designed to make Wikipedia’s extensive knowledge base more accessible to AI systems.
The project applies vector-based semantic search, a technique that helps machines understand meaning and relationships between words, to nearly 120 million entries across Wikipedia and its sister platforms. It also incorporates support for the Model Context Protocol (MCP), a standard that enables AI systems to communicate directly with data sources, making the information more usable for natural language queries.
Developed in partnership with Jina.AI and DataStax (an IBM-owned real-time training data company), the system builds on Wikidata’s history of providing machine-readable information. Unlike the older tools, which relied on keyword searches and SPARQL queries, the new framework integrates smoothly with retrieval-augmented generation (RAG) systems, helping AI models ground responses in verified Wikipedia content.
The enriched structure provides semantic depth. For instance, a query for “scientist” would return not only prominent nuclear scientists and Bell Labs researchers but also translations, related terms like “researcher” or “scholar,” and Wikimedia-cleared images.
The database is now publicly accessible on Toolforge, and a developer webinar will be held on October 9.
Related: Mastodon, the Open Source X Competitor, Launches New In-App Donation Drive
The launch comes amid increasing demand for high-quality, reliable datasets in AI development. While some critics question Wikipedia’s role in training models, experts note that its curated data is far more dependable than broad-scraped datasets like Common Crawl.
In a statement, Wikidata AI project manager Philippe Saadé emphasized the project’s open approach:
“This Embedding Project launch shows that powerful AI doesn’t have to be controlled by a handful of companies. It can be open, collaborative, and built to serve everyone.”
