Ontology-driven knowledge graph
You’ve processed a document with an ontology. Let’s see the difference it makes.
Compare the graphs
Go back to the Graph Explorer from the Workflows page.
Look at the knowledge graph from the ontology-driven extraction and compare it with the earlier ontology-free result. The differences should be clear:
- Entity types match the ontology — instead of generic nodes, you see the specific types you defined (or that the sample ontology defined)
- Relationships are consistent — the same relationship names appear across different entities, making the graph queryable with precision
- Less noise — the ontology filters out irrelevant extractions that the LLM might otherwise produce
Why this matters
Without an ontology, extraction is best-effort — the LLM extracts whatever it considers important, using whatever labels it chooses. Two documents about the same topic might produce completely different entity types and relationship names.
With an ontology, extraction is guided. The LLM maps document content to the entity types and relationships you defined. The result is a knowledge graph that’s consistent, predictable, and precise enough to query with SPARQL — which is exactly what we’ll do next.
Next
Query with SPARQL — run precise queries against the structured knowledge.