Reasoning about Change with Domain-specific Process Ontologies

Abstract

The expanding use of information systems in industrial and commercial settings has increased the need for interoperation between software systems. This is, however, inhibited because different systems may use distinct terminologies or assume different meanings for the same terms. A common solution to this problem is to develop ontologies which act as an intermediate language between different parties. For systems which communicate process information within a specific domain, the semantic integration requires an intermediate language between theories which axiomatize possible processes within the domain. In this thesis, we take the first steps toward integrating dynamic information systems. We present a methodology for design and verification of process ontologies which are specific to a particular domain, such as biology, or manufacturing. Furthermore, we show how ontology modularization techniques can be exploited as a means of reusing existing domain-specific process ontologies in developing new domain-specific process ontologies. We also discuss the applications of domain-specific process ontologies in axiomatizing qualitative spatial change. We then describe cases in which domain-specific process ontologies are not sufficient for semantic integration, and characterize a class of theories of actions, called canonical action theories, which can act as intermediate languages between different parties in those cases. In particular, we show that for each domain with finite number of elements there exists a canonical action theory such that all correct plans and process descriptions of the domain can be described in terms of activities axiomatized by the canonical action theory. Both in designing domain-specific process ontologies and characterizing canonical action theories we use an algebraic structure for representing change which enables us to abstract away domain-dependent properties of processes and activities, and demonstrate general properties of formalisms required for semantic integration of dynamic information systems.