Affiliations: The Institute of Scientific and Industrial Research, Osaka University, 8-1 Mihogaoka, Ibaraki, Osaka, 567-0047 Japan. Tel.: +81 6 6879 8416; Fax: +81 6 6879 2123; E-mail: {miz,sunagawa,kozaki,kita}@ei.sanken.osaka-u.ac.jp
Abstract: One of the major goals of ontology is to represent properly the underlying conceptual structure of the messy world reflecting the reality as much as possible. Ontology building tools should be designed to help developers create good ontologies. However, few of them can deal with roles adequately. Needless to say, the world is full of roles (e.g., wife, teacher, president, fuel, food, product, output, height) although there has been extensive theoretical research on roles, we do not yet have a comprehensive and usable theory yet. Our group has already constructed an ontology-development tool known as Hozo which has the ability to deal with roles. However, although Hozo allows users to represent roles better than other existing tools, the underlying theoretical foundations are still unclear and there is some room for improvement concerning the generality of how to deal with roles. In this paper, as an extension of the Hozo framework for roles, we present a framework for organizing role concepts according to their context dependencies. We also focus on the clarification of role properties and requirements on the model of roles and on the feedback obtained from Hozo-based ontology building experiences. Establishment of a computational model of roles contributes to building good ontologies because such a theory would provide useful guidelines for dealing with view-related and context-dependent distinctions related to roles.
