Performance Evaluation of Upper‐Level Ontologies in Developing Materials Science Ontologies and Knowledge Graphs

This study tackles a significant challenge in ontology development for materials science: selecting the most appropriate upper‐level ontologies for creating application‐level ontologies and knowledge graphs. Focusing on the use case of Brinell hardness testing, the research assesses the performance of various top‐level ontologies (TLOs)—basic formal ontology (BFO), elementary multiperspective material ontology (EMMO), and provenance ontology (PROVO)—in developing Brinell testing ontologies (BTOs). Consequently, three versions of BTOs are created using combinations of these TLOs along with their integrated mid‐ and domain‐level ontologies. The performance of these ontologies is evaluated based on ten parameters: semantic richness, domain coverage, extensibility, complexity, mapping efficiency, query efficiency, integration with other ontologies, adaptability to different data contexts, community acceptance, and documentation and maintainability. The results show that all candidate TLOs can effectively develop BTOs, each with its distinct advantages. BFO provides a well‐structured, understandable hierarchy, and excellent query efficiency, making it suitable for integration across various ontologies and applications. PROVO demonstrates balanced performance with strong integration capabilities. Meanwhile, EMMO offers high semantic richness and domain coverage, though its complex structure impacts query efficiency and integration with other ontologies.