NEXUS: Interpretable Knowledge-Graph Recommendation for User-Adaptive Exploration of Digital Natural Heritage

Natural heritage comprises biological, ecological, historical, and digital records whose values emerge through their interrelations, yet these records are often managed separately. This study develops NEXUS, an interpretable knowledge-graph framework that links a Natural Heritage Ontology (NHO) with a User Context Ontology (UCO) to support context-sensitive information prioritization. Expert requirements were structured through Kano–QFD analysis, and rule-based profile–relation affinities were used to adjust relation weights across four exploration profiles. The framework was implemented in Neo4j using Korean crane-related resources and evaluated through semantic validation, ranking analysis, sensitivity testing, and a formative user study. Across 12 focal nodes, profiles produced different result orderings (Overlap@5 = 0.303; Overlap@10 = 0.625; Kendall’s $\tau =0.180$), while uniform affinities yielded identical rankings. Rankings remained stable under local affinity perturbations. A formative evaluation ($n=20$) showed high ratings in both adaptive and static conditions, but no significant difference between them. These findings show that user context can be incorporated into knowledge-graph ranking in an explicit and inspectable manner, while the relevance of the resulting rankings and their effects on user understanding require further validation.