Multi-Level Modeling with M-Objects and M-Relationships

B. Neumayr
PT1002 (2010)

Kurzfassung (Englisch)

Information Systems maintain a representation of the state of a domain in order to provide information about that domain. They are often employed in a cross-organizational setting and have to represent different sub-domains. These sub-domains and different organizations are homogeneous as well as heterogeneous with regard to the domain itself, as well as with regard to information needs. Emphasis on homogeneity results in inflexible information systems. On the other hand, emphasis on heterogeneity leads to `information islands', which are difficult to integrate.

As a prerequisite for information system development, conceptual modeling provides a high-level representation of a domain, independent of a specific implementation. For this purpose, conceptual modeling makes use of the classical abstraction principles: classification, generalization, and aggregation. In order to represent complex domains, being both homogeneous as well as heterogeneous, these principles need to be combined. Using traditional modeling techniques, this results in multiple representation of single objects in different abstraction hierarchies. This, in turn, results in models and information systems that are hard to comprehend and to maintain.

Multi-Level Objects (M-Objects) and Multi-Level Relationships (M-Relationships) allow to represent objects and relationships at higher abstraction levels, which encapsulate descriptions of sets of objects and relationships at multiple lower levels of abstraction. A single concretization hierarchy of m-objects (or m-relationships) combines and differentiates multiple classification-, generalization-, and aggregation-hierarchies. In this manner, m-objects and m-relationships facilitate stepwise modeling of complex domains at multiple abstraction levels. Concretization hierarchies guarantee a certain level of homogeneity between sub-hierarchies and, on the other hand, they also allow for heterogeneities between different sub-hierarchies.

In this thesis, we provide formal definitions and examples of m-objects and m-relationships. The m-object and m-relationship approach is complemented by formally defined operations for creating, manipulating and querying multi-level models.

Applications of m-objects and m-relationships are manifold, we especially discuss web ontologies and data warehousing. Applied to ontology engineering with OWL DL, m-objects and m-relationships provide a more powerful alternative to current multi-level modeling techniques, such as punning. In data warehousing, m-objects and m-relationships provide a solution to the well discussed problem of modeling heterogeneous dimension hierarchies and also allow heterogeneities within cubes. This solution has been implemented as an extension package for Oracle DB.