An Ontology-Driven Similarity Algorithm
This paper presents our similarity algorithm between relations in a user query written in FOL (first order logic) and ontological relations. Our similarity algorithm takes two graphs and produces a mapping between elements of the two graphs (i.e. graphs associated to the query, a subsection of ontology relevant to the query). The algorithm assesses structural similarity and concept similarity. An evaluation of our algorithm using the KMi Planet ontology is presented. We also carried out an experiment to test the human judgment about similarity using context and without context. Our similarity algorithm has been mainly used in AQUA, our question answering system, in the query reformulation process.