Using UML for conceptual modeling: towards an ontological core

Before developing an information system, the business and organizational domain in which the information system is to be used must be examined and understood. Creating conceptual models in the system analysis phase to faithfully represent the domain is critical for successful information system development. Although it is widely accepted that UML could be used both for modeling software, and for modeling the problem domain that is supported by a system, i.e. conceptual modeling, its suitability for the latter in the early development phase has been questioned. In fact, the semantics of its constructs (such as object, class, attribute, link, association, and association class) are clear with respect to software design and coding, but vague with respect to conceptual modeling. -- To endow UML with semantics for conceptual modeling, in this thesis, an ontological framework of UML based on Bunge's ontology is proposed, focusing on static aspects (class/object diagrams and links in collaborations). The framework assigns precise ontological semantics for a core set of UML constructs (object, attribute, class/type, link, association, state, state transition, operation, and role) in class, object, and collaboration diagrams and is used to resolve a number of confusions in the UML literature. -- Furthermore, in the thesis, a novel ontological metamodel of classifiers based on Bunge's ontology, OntoClean methodology, and Guizzardi et al.’s ontological profile is proposed, focusing on discussing the definition, properties, and representation of the notion of role in the object-oriented literature. The metamodel conforms to the fundamental role features identified in the literature and handles the counting problem and related role identity problem. -- The study also compares conceptual models created using the metamodel to those created using ER approach with respect to conceptual database modeling and describes how to map a conceptual model based on the metamodel into relational database schema. Using examples, the study demonstrates that relational database schemata generated using our approach are more stable with respect to requirements change, and moreover a number of real world semantics and rules can be implemented as integrity constraints of a relational database schema