GUARD Project

Ensuring Interoperable and Trustworthy Knowledge Graphs for Defence and National Security AI

Project funded by The Turing Defence and National Security Grand Challenge.

Knowledge graphs (KGs) and ontologies are vital for trustworthy AI, but their effectiveness is often limited by interoperability gaps and quality issues. Interoperability of ontologies at the upper- and mid-level has been advanced through initiatives such as the Basic Formal Ontology (BFO) and the Common Core Ontologies (CCO). However, achieving the same interoperability at the application level remains a major challenge, particularly in defence and national security, where broad conceptual coverage must be reconciled with specialised subdomains. An ontology foundry for this domain, such as that proposed by the US Department of Defense, must ensure interoperability, comprehensive coverage, and high quality. Without these guarantees, the deployment of ontologies and knowledge graphs (KGs) risks limiting their effectiveness in downstream applications.

Two key challenges must be addressed. The first challenge is interoperability and coverage. Existing ontologies and KGs relevant to defence model overlapping domains, but their interoperability remains limited, and greater integration with linked data resources is required to extend coverage across geolocations, organisations, diseases, and environmental hazards. The second challenge is quality. Although OWL, SHACL, and ShEx enable the detection of inconsistencies and definition of integrity constraints, their practical use is often limited by scalability issues, particularly when integrating large or multiple KGs.

Team: Ernesto Jimenez-Ruiz (PI) and Dave Herron (RA)

Advisory board: Catia Pequita (University of Lisbon), Paul Cripps (Dstl),

Relevant resources:


References

2025

  1. LLM OA Oracles
    Large Language Models as Oracles for Ontology Alignment
    Sviatoslav Lushnei, Dmytro Shumskyi, Severyn Shykula, Ernesto Jiménez-Ruiz, and Artur Garcez
    CoRR, 2025