Generative Ontology Induction: Domain-Agnostic Schema Discovery from Document Corpora Using Large Language Models
Quick Answer
This paper shows that Generative Ontology Induction (GOI) is a domain-agnostic framework that generates structured ontologies from document corpora, achieving 95-100% coverage of structural nodes across various ontologies.
Quick Take
It introduces the Node Coverage Score to evaluate the completeness of generated outputs, outperforming generic templates significantly in specific domains.
Key Points
- GOI generates entities, properties, and relationships as a typed graph in YAML/JSON.
- Node Coverage Score measures the structural ontology nodes in generated outputs.
- Validation across four ontologies shows 95-100% structural coverage with GOI.
- Generic templates achieve only 52.2% to 78.3% coverage in specific cases.
- GOI's performance is consistent regardless of document type familiarity.
DeepSignal Analysis
What happened
Generative Ontology Induction (GOI) is introduced as a framework for creating structured ontologies from document corpora. It achieves high coverage of structural nodes, with results showing 95-100% coverage across various ontologies. The framework uses a new evaluation metric called the Node Coverage Score.
Key evidence
- GOI generates structured ontologies as typed graphs in YAML/JSON format, covering 95-100% of structural nodes in four tested ontologies.
- A generic three-field template achieved 97.8% coverage on an invoice schema but dropped significantly to 52.2% on a Job Description Ontology.
- The Node Coverage Score measures the fraction of structural ontology nodes appearing in generated outputs, indicating the completeness of GOI's results.
Why it matters
The development of GOI addresses a significant challenge in ontology engineering, which has been a bottleneck for knowledge-intensive AI systems. By providing a domain-agnostic solution, it allows for more efficient and comprehensive ontology creation, which can enhance the performance of AI systems that rely on structured knowledge. The high coverage rates suggest that GOI could be a valuable tool for various applications across different domains.
Paper Resources
Source Excerpt
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. We introduce Generative Ontology Induction (GOI), a domain-agnostic framework that induces a generative blueprint - entities, dimensions, properties, relationships, and constraints - from a corpus of examples and exports it as a typed graph (six
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →RAIL Guard: Closing the Evaluation-to-Remediation Gap in Responsible AI for Agents
RAIL Guard introduces a closed-loop AI pipeline for large language models (LLMs) that evaluates outputs across eight dimensions and iteratively remediates failures, achieving 96.9% convergence compared to 49.1% for traditional block-and-retry methods. The system reduces unsafe agent executions by 33% without impacting task completion and is available as open-source SDKs.