Graph-based partitioning of large-scale ontologies

With the growing utilization of ontologies in almost all branches of science and industry, not only the number of available ontologies has increased considerably but also many widely used ontologies has reached a size that cannot be handled by the available reasoners. Due to the size and the monolithic nature of large-scale ontologies, problems with large monolithic ontologies in terms of reusability, scalability and maintenance have led to the increasing modularization techniques for ontologies. In this paper a novel two-phase partitioning approach is proposed which partitions the large ontologies into smaller modules based on the weighted graph constructed from the ontologies. In the first phase it clusters the sparse weighted graph into several sub clusters, and in the second phase it iteratively selects two clusters for which both RI and RC between them are high and merges them into one module until it has reached the original requirements. And the experiments have demonstrated that this method performs quite well.