LLM-Driven Ontology Construction for Enterprise Knowledge Graphs

TL;DR

Automating ontology construction for enterprise knowledge graphs from unstructured data remains challenging and manual. OntoEKG introduces a two-stage LLM-driven pipeline that first extracts $C_T$ and $P_T$ from text $T$, then applies Entailment to derive a taxonomy, followed by RDF serialization into Turtle; datatypes are reified as separate classes and classes map to $owl:Class$, properties to $owl:ObjectProperty$. The paper contributes a formal two-phase framework with explicit RDF output, a new end-to-end benchmark spanning Data, Finance, and Logistics, and empirical insights including a Data-domain fuzzy-match F1 of $0.724$ and notable limitations in scope control and hierarchical reasoning. These results demonstrate the potential of AI-assisted ontology engineering in real-world enterprises and motivate community-driven benchmarks for end-to-end ontology construction, with future work on provenance extraction and cross-document consistency.

Abstract

Enterprise Knowledge Graphs have become essential for unifying heterogeneous data and enforcing semantic governance. However, the construction of their underlying ontologies remains a resource-intensive, manual process that relies heavily on domain expertise. This paper introduces OntoEKG, a LLM-driven pipeline designed to accelerate the generation of domain-specific ontologies from unstructured enterprise data. Our approach decomposes the modelling task into two distinct phases: an extraction module that identifies core classes and properties, and an entailment module that logically structures these elements into a hierarchy before serialising them into standard RDF. Addressing the significant lack of comprehensive benchmarks for end-to-end ontology construction, we adopt a new evaluation dataset derived from documents across the Data, Finance, and Logistics sectors. Experimental results highlight both the potential and the challenges of this approach, achieving a fuzzy-match F1-score of 0.724 in the Data domain while revealing limitations in scope definition and hierarchical reasoning.

Figures (3)

• Figure 1: An example of Ontology Construction task.
• Figure 2: The full OntoEKG pipeline.
• Figure 3: The generated ontology for the Data use case.