To tie the individual modelling-construct descriptions together, UEML also comprises a common ontology into which the semantic construct descriptions are mapped. Specifically, all the classes, properties, states and transformations used to represent a construct are mirrored in the common ontology, which thereby comes to interrelate the constructs descriptions at the semantic level.
To organise the classes, properties, states and transformations used to describe modelling constructs, the common UEML ontology is organised as four taxonomies: The classes in the ontology are organised in a conventional generalisation hierarchy. Properties, on the other hand, are organised in a precedence hierarchy, resembling the generalisation hierarchy for classes. One thing precedes another if all the properties that possess the second property must also possess the first. For example, being a mammal precedes being a human, because everyone who is a human is also a mammal. There are similar hierarchies of states and sub states and of transformations and sub transformations.
Of course, the four taxonomies are also interrelated in several ways. Classes are related to the properties that characterise them. Properties are related to the states they define. States are in turn entered and exited by transformations. Certain types of properties are laws that restrict other properties. State laws restrict states, whereas transformation laws effect transformations. Finally, both classes, properties, states and transformations – including the state and transformation laws – are described by attributes. For example, they all have unique names and there are cardinality constraints and role names for the relationships between classes and properties. The idea is that the organisation of the UEML ontology as four distinct, but interrelated taxonomies makes it possible to evolve the ontology over time without increasing complexity more than necessary. New classes, properties, states and transformations will always have a clearly identifiable location where they can be added to the appropriate taxonomy.
By comparing their mappings onto the common ontology, exact relationships between any pair of constructs can be determined. All the modelling constructs in the UEML are thereby interrelated at the most detailed level possible via the common ontology: If two modelling constructs are identical, they will map onto the exact same ontology concepts. If two modelling constructs do not overlap at all, they will map onto completely distinct ontology concepts, i.e., onto concepts which are not even hierarchically related. The third case is likely to be most common, where two modelling constructs map onto some identical ontology concepts, some ontology concepts that are hierarchically related and some ontology concepts that are completely distinct. But in all cases, the hierarchically organised common ontology makes it possible to determine the exact representational (“semantic”) relationship between any pair or group of modelling constructs.
The figure (still missing) shows the meta-meta for the common ontology, also based on the earlier work of Opdahl & Henderson-Sellers (2004, 2005). Together, the representation mapping and the common ontology are the base of the web of languages. They are the starting point for identifying correspondences and for supporting integrated use of models in the future.
The common UEML ontology was first populated with a set of initial classes, properties, states and transformations that were derived directly from Bunge’s ontological model (Bunge 1977, 1979) and the Bunge-Wand-Weber representation model of information systems, the so-called BWW model (Wand & Weber 1988, 1993, 1995). Since then, it has grown incrementally through the analysis of incorporated constructs from a combination of academic and industrial modelling languages, such as ARIS, BMM, BPMN, coloured Petri nets, GRL, IDEF3?, ISO/DIS 19440, KAOS, UEML 1.0 and selected diagram types from UML 2.0. In consequence, the most general concepts in the common ontology are ontologically committed, in the sense that they have grown out of Bunge's ontology and the BWW model, whereas the more specific ones have emerged through analysis of languages.
References
- Bunge, M. (1977). Treatise on Basic Philosophy: Vol. 3: Ontology I: The Furniture of the World. Boston:Reidel.
- Bunge, M. (1979). Treatise on Basic Philosophy: Vol. 4: Ontology II: A World of Systems. Boston:Reidel.
- Opdahl, A.L. & Henderson-Sellers, B. (2004). A Template for Defining Enterprise Modelling Constructs. Journal of Database Management 15(2).
- Opdahl, A.L. and Henderson-Sellers, B. (2005). Template-Based Definition of Information Systems and Enterprise Modelling Constructs. In Ontologies and Business System Analysis, Peter Green and Michael Rosemann (eds.). Idea Group Publishing, 2005.
- Wand, Y. & Weber, R. (1988). An ontological analysis of some fundamental information systems concepts. In Proc. "Ninth International Conference on Information Systems", (DeGross, J.I. & Olson, M.H. (eds.), Minneapolis/USA, November 30–December 3, 1988, pp. 213–225.
- Wand, Y. & Weber, R. (1988). An Ontological Model of an Information System. IEEE Transactions of Software Engineering,16 (11):1282-1292, IEEE Press.
- Wand, Y. & Weber, R. (1993). On the ontological expressiveness of information systems analysis and design grammars. Journal of Information Systems, 3:217–237.
- Wand, Y. & Weber, R. (1995). On the deep structure of information systems. Information Systems Journal, 5:203–223.