from data to knowledge: a unified relational-semantic architecture with context-aware term modeling for next-generation knowledge systems

Database design for knowledge-centric systems suffers from fragmentation: relational models provide integrity but implicit semantics; knowledge graphs offer explicit relationships but weak transactions; ontology-based approaches (obda) suffer mapping complexity; multi-model systems lack conceptual coherence. this paper proposes a unified relational-semantic architecture that bridges these gaps. the model introduces five principles: term-entity independence (context-neutral terms), triadic contextualization (term–domain–module tuples for polysemy resolution), relationships as first-class citizens (explicit relations with typed relationtypes), governance through typed schemas (module-relation compatibility mrt and declarative relation constraints), and service-oriented access control. evaluation on a realistic dataset (5,000 terms, 18,547 relations) in islamic sciences compares the proposed model against pure relational, neo4j, and obda baselines. results indicate that the model provides native support for contextualized term definitions across domains and modules—a feature absent from all baselines. low overhead (45–99%) guarantees 99.4% constraint accuracy. for deep transitive closure, neo4j is 4–5× faster (mitigable by materialized paths). the model outperforms obda by 3–8× and achieves a conceptual coherence score of 4.7/5, compared to 2.2–3.1/5 for baselines. the architecture offers a practical, empirically validated alternative for digital libraries, encyclopedias, and scholarly knowledge systems where contextual meaning and semantic relationships are paramount.