Semantic Nearest Neighbor Search in OWL Ontologies

A nearest neighbor search procedure is presented, for retrieving resources in knowledge bases expressed in OWL. The procedure exploits a semidistance for annotated resources, that is based on a number of dimensions corresponding to a committee of features represented by OWL concept descriptions. The procedure can retrieve resources belonging to query concepts expressed in OWL, by analogy with other training instances, on the grounds of the classification of the nearest ones w.r.t. the dissimilarity measure. Besides, it may also be able to suggest new assertions that are not logically entailed by the knowledge base due to open world semantics. In the experimentation, where we compare the performance of the procedure to running a reasoner, we show that it can be quite accurate and augment the scope of its applicability, improving w.r.t. previous prototypes that adopted other semantic measures.