Deploying ontologies in software design

Abstract

In this thesis we will be concerned with the relation between ontologies and software design. Ontologies are studied in the artificial intelligence community as a means to explicitly represent standardised domain knowledge in order to enable knowledge shar¬ ing and reuse. We deploy ontologies in software design with emphasis on a traditional software engineering theme: error detection. In particular, we identify a type of error that is often difficult to detect: conceptual errors. These are related to the description of the domain whom which the system will operate. They require subjective knowledge about correct forms of domain description to detect them. Ontologies provide these forms of domain description and we are interested in applying them and verify their correctness(chapter 1). After presenting an in depth analysis of the field of ontologies and software testing as conceived and implemented by the software engineering and artificial intelligence communities(chapter 2), we discuss an approach which enabled us to deploy ontologies in the early phases of software development (i.e., specifications) in order to detect conceptual errors (chapter 3). This is based on the provision of ontological axioms which are used to verify conformance of specification constructs to the underpinning ontology. To facilitate the integration of ontology with applications that adopt it we developed an architecture and built tools to implement this form of conceptual error check(chapter 4). We apply and evaluate the architecture in a variety of contexts to identify potential uses (chapter 5). An implication of this method for de¬ ploying ontologies to reason about the correctness of applications is to raise our trust in the given ontologies. However, when the ontologies themselves are erroneous we might fail to reveal pernicious discrepancies. To cope with this problem we extended the architecture to a multi-layer form(chapter 4) which gives us the ability to check the ontologies themselves for correctness. We apply this multi-layer architecture to cap¬ ture errors found in a complex ontologies lattice(chapter 6). We further elaborate on the weaknesses in ontology evaluation methods and employ a technique stemming from software engineering, that of experience management, to facilitate ontology testing and deployment(chapter 7). The work presented in this thesis aims to improve practice in ontology use and identify areas to which ontologies could be of benefits other than the advocated ones of knowledge sharing and reuse(chapter 8).