 Abstract | Conceptual modelling presented as a framework for database design is a discipline
of great importance in many areas in computer science that seeks to represent
real-world phenomenon using semantic primitives. To date, traditional (static)
conceptual models have been successfully used and extended to deal with the
semantics of relatively stable real world applications. However, the capturing
of semantics is a seemingly endless task as it involves various dimensions and
categories.
It is argued in this thesis that the incorporation of complex domain structures
in conceptual modelling to represent the semantic domains of an attribute and
the relationships within a concept in ontologies would provide more expressive
and richer semantics. Additionally, it is argued that basic relationships in the
entity-relationship model may need to be modified or extended to handle a broad
spectrum of situations that arise from differing perspectives of the real world.
Furthermore, it is argued that a conceptual model should allow rapid and simultaneous
storage of data and data modelling as unexpected and sudden events
require data to be modelled rapidly. This thesis begins with an extensive review
of the field of conceptual modelling and an exploration of the concepts of mesodata,
ontologies and semantic relationships in conceptual modelling as well as
various aspects of extensions to the entity-relationship model.
Using these foundations, a classification of ER modelling extensions (CERME)
framework is introduced that forms the basis of common aspects and comparative
criteria which can be used to categorise and compare various proposals. In addition,
the Mesodata Entity-Relationship (MDER) model, Mesodata Object Role
Model (MDORM), Ontological Entity-Relationship (OntoER) model, Ontological
Object Role Modelling (OntoORM) and Ontological Unified Modeling Language
(OntoUML) class diagrams are presented that allow advanced semantics to be
associated with the domains of an attribute. It is also demonstrated that these
proposed models can be mapped into the commonly accepted standard relational
model. Furthermore, for some of the modelling issues that are not easily accommodated
into the ER model, this thesis introduces a new relationship construct,
polymorphic relationships, to handle this situation. To this end, a novel approach
to conceptual modelling, the LItER model, is presented that incorporates the previously
proposed concepts of mesodata, ontologies and polymorphic relationships
into the model which allows data to be modelled rapidly. |
