An Analysis of Roles Towards Ontology-Based Modelling Master’s Thesis

For decades, the concept of roles has frequently been discussed in the literature of various fields of computer science, such as conceptual modelling and knowledge representation, but an integrative definition of the concept has not yet been accomplished. On the basis of an extensive literature review, a general approach to roles is elaborated, which includes a system of classification. The motivation of this analysis comprises the integration of a well-founded concept of roles into the ontology-based knowledge representation language GOL (General Ontological Language). The evaluation of the literature reveals two recurring concepts in connection with roles, called player and context. For instance, a student could be a role whose player is some human and whose context is determined by a relation to some university. The basis of the role approach being developed is formed by the hypothesis that roles characterise their player, while they themselves are determined by contexts. The GOL categories relator, process and substance can be identified as those ontological categories which cover all contexts in the literature reviewed. A distinct role type is postulated for each of these categories, which is then examined and formalised in particular with regard to the relation to the corresponding contexts. Discussions of problems as well as of properties of the representation based on these role types supplement these examinations. The extraction of common properties of all roles from the particular role types completes the approach. In addition, the proposed role concept is compared to other ontological notions. A reinterpretation of some works selected from the reviewed literature shall demonstrate the expressiveness of the role approach. Referring to the motivation of this thesis, the axioms presented are proposed for integration into GOL. Further, an adaptation of the GOL representation of material facts is suggested in order to avoid problems which were previously shown to be solvable in terms of roles. Finally, a sample model from the domain of clinical trials shall clarify the application of the concepts developed. Moreover, it illustrates the potential precision of modelling based on these concepts. Altogether, this thesis is intended as a contribution to knowledge representation on a clear ontological basis, paving the way for application-independent reuse of knowledge and other benefits. 1In May 2003, this report was submitted as a Master’s Thesis to the Department of Computer Science at the University of Leipzig, Faculty of Mathematics and Computer Science. University of Leipzig 1 / 123