New LLM framework GOI achieves 95-100% structural ontology coverage from document corpora

Ontology engineering remains a critical bottleneck in knowledge-intensive AI systems. Existing automated approaches either depend on predefined schemas, operate within narrow domains, or produce unstructured outputs unsuitable for downstream pipelines. In a new preprint (arXiv:2607.16201), researcher Sergei Sergienko introduces Generative Ontology Induction (GOI), a domain-agnostic framework that induces a generative blueprint — encompassing entities, dimensions, properties, relationships, and constraints — from a corpus of examples. The output is a typed graph with six node types and seven edge types, exported as YAML or JSON, making it directly usable in knowledge graphs and AI pipelines.

The framework introduces a novel evaluation metric called Node Coverage Score, which measures the fraction of structural ontology nodes (classes, properties, and dimensions) appearing in generated outputs. In controlled generative validation on four contrasting ontologies — a familiar Software Services Invoice schema, a custom Job Description Ontology, a confidential Pain-Management Clinical Visit Record Ontology, and a Professional Services Contract & Statement of Work Ontology — GOI-prompted generation covered 95-100% of the structural backbone in every case. By comparison, a generic three-field template held at 97.8% on the invoice schema but plummeted to 52.2% on the Job Description Ontology, 62.2% on the Pain-Management ontology, and 78.3% on the Professional Services Contract ontology. Importantly, structural coverage remained high regardless of how familiar the document type was to the model, demonstrating GOI's robustness and domain agnosticism.