Wikimedia Deutschland launches AI-friendly Wikidata Embedding Project

Wikimedia Deutschland has launched the Wikidata Embedding Project, converting approximately 120 million structured data points from Wikidata into vector embeddings to make the knowledge base more accessible and usable for AI systems, particularly large language models. This vectorization allows AI to understand contextual relationships between terms, enabling more accurate and reliable information retrieval through semantic search and retrieval-augmented generation (RAG) techniques. Developed in collaboration with Jina AI and IBM’s DataStax, the project aims to democratize AI development by providing smaller companies and developers open access to high-quality, verified, multilingual data, helping them compete with larger tech firms. The project supports modern data protocols like the Model Context Protocol (MCP) to improve natural language query processing and prevent hallucination by grounding AI responses in fact-checked content. While Wikipedia’s user experience remains unchanged, the backend improvements facilitate the creation of smarter apps and chatbots by leveraging this AI-friendly dataset. The initiative demonstrates that powerful AI resources can be developed openly and collaboratively, expanding opportunities beyond the major tech companies.