An information content based partitioning method for the anatomical ontology matching task

Abstract

Anatomy ontology matching has been attracting a lot of interest and attention of researchers, especially, biologists, medics and geneticists. This is a very difficult task due to the huge size of anatomy ontologies. Despite the fact that many ontology matching tools have been proposed so far, most of them achieve good results only for small size ontologies. In a recent survey [22], the authors pointed out that the large scale ontology matching problem still presents a real challenge because it is a time consuming and memory intensive process. According to state of the art works, the authors also state that partitioning large scale ontology is a promising solution to deal with this issue. Therefore, in this paper, we propose a partitioning approach to break up the large matching problem into smaller matching subproblems. At first, we propose a method to semantically split anatomy ontology into groups called clusters. It relies on a specific method for computing semantic similarities between concepts based on both their information content on anatomy ontology, and a scalable agglomerative hierarchical clustering algorithm. We then propose a filtering method to select the possible similar partitions in order to reduce the computation time. The experimental analysis demonstrates that our approach is capable of solving the scalability ontology matching problem and encourages us to the future works.