Cohesion and coupling metrics for ontology modules Sunju Oh • Heon Y. Yeom • Joongho Ahn Published online: 3 April 2011 Ó Springer Science+Business Media, LLC 2011 Abstract In recent years, an increasing number of ontologies and semantic web applications have been devel- oped and used. A conscious effort has been made to develop methods to modularize ontologies. These methods contribute to building a new ontology. However, few studies have focused on the evaluative methods for ontol- ogy modules. In this study, we propose novel metrics to measure ontology modularity. To evaluate the ontology modules, we introduce cohesion and coupling based on the theory of software metrics. A cohesion metric and two coupling metrics were used to measure cohesion and cou- pling for ontology modules. The proposed metrics provide more detailed support in considering the different types of relationships between classes in ontology modules. In addition, the new coupling metrics contribute to checking the consistency between the ontology modules and their original ontology. The proposed metrics were validated using well-known verification frameworks and empirical experiments to complement the previous investigations. The results of this study offer ontology engineers valuable criteria with which to evaluate ontology modules and help ontology users select qualifying ontology modules. Keywords Ontology  Modularity  Metrics  Cohesion  Coupling 1 Introduction Ontologies are the backbone of the Semantic Web, a semantic-aware version of the World Wide Web [14, 23, 30, 36, 54, 57, 60]. They provide an essential knowledge- sharing infrastructure of semantic-driven applications. Due to the increased information and knowledge available on the Internet, ontologies have become considerably large, making it difficult to reuse, maintain, and understand a single large-scale ontology. This obstacle with large monolithical ontologies has led to the increasing interest in techniques to carry out ontology modularization [15, 16]. Small and modular ontologies can effectively contribute to building a new ontology. This can reduce development time and effort while improving the quality of the ontol- ogy. However, these approaches often rely on individual assumptions of modularity and are based on a particular application scenario (e.g., reasoning, visualization, evolu- tion, and ontology reuse). Therefore, it is difficult to compare modules produced by these varying techniques [12, 13]. Previous research has proposed metrics for ontologies and has provided some principal work to study the nature of measures for ontology in general [22, 46, 58]. How- ever, most ontology metrics have been designed to eval- uate ontologies themselves [22, 46]. Therefore, existing metrics cannot be applied to ontology modules and determine how well modularized an ontology is because modules have characteristics different from those of ontology itself. Furthermore, most metrics are based on the hierarchical structure of ontology and fail to consider A prior version of this paper was presented at the 2009 IEEE International Conference on e-Business Engineering (ICEBE2009). S. Oh (&)  H. Y. Yeom School of Computer Science and Engineering, Seoul National University, 599 Gwanak-no, Gwanak-gu, Seoul 151-744, Korea e-mail: ohsunju7@snu.ac.kr J. Ahn Graduate School of Business, Seoul National University, 599 Gwanak-no, Gwanak-gu, Seoul 151-916, Korea 123 Inf Technol Manag (2011) 12:81–96 DOI 10.1007/s10799-011-0094-5