Called the Wikidata Embedding Project, the system applies a vector-based semantic search to the existing data on Wikipedia and its sister platforms, consisting of nearly 120 million entries.