Hybridizing genetic algorithms and hill climbing for similarity aggregation in ontology matching

Ontology Matching aims at finding correspondences between two different ontologies with overlapping parts in order to bring them into a mutual agreement. The set of correspondences, called alignment, is obtained by computing an aggregated similarity value for all pairs of ontology entities through a weighted approach. Unfortunately, the similarity aggregation task is a very complex optimization process, above all, when no information is known about ontology characteristics. This work presents a hybrid approach which aims at efficiently optimizing the weights for the similarity aggregation task without knowing a priori the ontology features. The effectiveness of our approach is shown by aligning ontologies belonging to the well-known OAEI benchmark dataset and by executing a comparison based on the Wilcoxon's signed rank test which highlights that our proposal statistically outperforms both its genetic counterpart and a traditional no evolutionary approach.