Modularization is a key requirement to manage the size and complexity of large ontologies by replacing each one by a set of smaller ontologies. Two reasons for this requirement are that current ontology languages such as OWL do not allow partial reuse of ontologies and ontologies are ever growing to cover more knowledge in a specific domain. Many existing modularization methods focus on either semantics or structural aspects of ontologies while both of them are important. In this paper, we consider both semantic and structure by combining these aspects using random walk algorithms to achieve a balance between them. We also define weights for different relations to take semantic into account. The proposed method is designed using two algorithms: a greedy algorithm and a heuristic one to reduce run-time and time-complexity. Our goal is to produce reusable modules of high quality and support large ontologies. The results of the experiments show that our algorithms perform well in comparison with existing golden standard.

Keywords