The eNOVATION CBRN Domain-Specific Ontology is a structured semantic knowledge base that harmonises and interconnects information about CBRN Training Centres (TCs), their capabilities, technologies and operational scenarios across the European preparedness community. Built on RDF/OWL Semantic Web standards and aligned with established frameworks including the EU Civil Security Taxonomy and CWA 18019:2023, it transforms heterogeneous TC data into a unified, machine-interpretable knowledge graph that supports structured search, OWL reasoning and cross-institutional interoperability.

The ontology is organised into three interconnected sub-ontologies:
- Training Centres and Operational Capabilities: TC profiles, training courses, services, competences, facilities, certifications and staff roles.
- Scenarios, Incidents and Exercises: threat scenarios, incident types, exercise structures, evaluation methods and response actions.
- Equipment, Technologies and Procedures: detection and decontamination equipment, technologies and standard operating procedures.

The ontology is populated with real-world data collected from eNOVATION partner TCs through structured questionnaires. Object properties with inverse relations support bidirectional graph traversal, while controlled vocabularies (threat categories, TRL levels, domain classifications) ensure consistency and extensibility as the TC network grows.

The ontology is delivered with a full-stack platform that makes it accessible without requiring expertise in semantic technologies:
- SPARQL Endpoint: hosted on Apache Jena Fuseki with OWL rule-based reasoning, enabling expressive queries over both asserted and inferred knowledge.
- Application Backend: a FastAPI (Python) service layer orchestrating SPARQL query execution, result transformation and caching, containerised with Docker and PostgreSQL.
- Web Exploration Interface: a React/Tailwind CSS frontend secured with Keycloak authentication, supporting hierarchical and list-based class browsing, instance inspection and relationship navigation, no query language knowledge required.
- AI-Enabled Natural Language Query Engine: an LLM-based NL-to-SPARQL pipeline with a syntax validation layer and automated self-correction loop, translating free-text user questions into SPARQL queries against the ontology.

Validation was conducted in two phases. Phase 1 involved qualitative assessment with TC representatives through structured online sessions, confirming conceptual clarity and operational relevance. Phase 2, which is in-practice validation through eNOVATION joint activities and field exercises, is still ongoing and evaluates the ontology under realistic training coordination conditions.

Contact: Institute of Communication and Computer Systems (vkarakolis@epu.ntua.gr)
Available for JA: UCSC - Italy-Rome 05/2026