UBC Theses and Dissertations
Using design languages for conceptual modelling : the UML case Evermann, Joerg Magnus
Information systems are representations of and situated in the business and organization. In order to develop effective information systems (IS), the first step must be to understand and describe this real world domain. System analysis is this first step, the final result of which is the conceptual model, a formal description of the business and organizational domain. It serves as the communication medium to further common domain understanding. System design, the second step in IS development, builds on this understanding, together with other requirements regarding e.g. functionality, performance, quality, usability, to design the software system. The conceptual model serves as input to this phase. Every model must be expressed in a language. However, there exists no widely accepted language for conceptual modelling of the business or the organization. On the other hand, recent years have seen the emergence and wide acceptance of object-oriented languages in general, and the Unified Modelling Language (UML) specifically, for IS design. This study examines the suitability of using such design languages for conceptual modelling. In order for a language to be usable for modelling business and organizational domains, the language constructs must possess real-world semantics, i.e. it must be clear what they refer to in the real world, not only in the software domain. Based on ontology, the branch of philosophy that deals with what exists in the real world, this study assigns such meaning to UML constructs. Based on these semantics, ontological assumptions are used to derive modelling rules for UML when UML is used for conceptual modelling. These rules are formalized using the UML meta-model. A case study is conducted which applies the proposed rules in a medium size IS development project and notes their beneficial effects on the analysis process and the final conceptual model. An experimental study is conducted to show specific benefits to domain understanding, induced by models which conform to the proposed rules. The chosen method of analysis of languages is applicable not only to UML, chosen as an example here, but is generalizable to other languages as well. The results derived in this study, other than the formalization by means of the specific language meta-model, are therefore generalizable to other object-oriented languages.
Item Citations and Data