Computational Cost of Querying for Related Entities in Different Ontologies

The computational cost of querying for similar entities across ontologies is high since, in the worst case, every pair of entities will have to be considered. Therefore, links discovered during ontology alignment have been used to speed up querying across ontologies by following relatedness links to discover similar entities. We derive the computational complexity of querying across ontologies using the ontology alignment links discovered using the Unified Fuzzy Ontology Matching (UFOM) framework. We consider querying for related entities by following either a single alignment link or by following multiple alignment links. These methods have different computational complexity and produce different query results. We also study the impact of the specific implementation approach on query time. We consider implementations based on multiple accesses of the triplestore using a high-level procedural language and by execution of a single SPARQL graph query on the ontology server. These approaches were evaluated using ontologies derived from an enterprise-scale dataset. Experimental results show that an implementation using nested for-loops in a procedural language outperformed by nearly 2× an implementation based on a single SPARQL query.