Researchers build a knowledge graph that stays organized — the schema doesn't dissolve when you reuse it

A team built a 34-million-entity knowledge graph from Wikidata with a schema designed to be reusable, not buried inside construction code. Instead of creating a schema as a side effect of building the graph, they designed the schema first — organized around entity types and property relationships — so downstream users can apply it independently to new problems without rebuilding everything.

Until now, knowledge graphs got locked into their construction pipelines. You'd build a graph for one job, and the schema decisions lived in the code that created it — if you wanted to use that schema for something else, you had to reverse-engineer or rebuild it. This approach separates the schema from the building process, which means a domain expert can reuse the same organizational structure across different tasks without consulting the original builders. That reduces friction for anyone building on existing data.