Unifying MDA and Knowledge Representation Technologies

As model-driven approaches to software engineering and ontology-based approaches to knowledge representation have continued to expand their influence and importance in recent years they have inevitably started to encroach on each others’ territories. There is therefore growing interest in understanding how precisely the two technologies relate to one another and how they can best be used together. Several papers have been published describing strategies for interchanging model and ontology information, and dedicated workshops on this topic have started to appear. This short position paper discusses the issues involved in integrating MDA and ontology representation languages and then argues the case for a core-level unification of the two technologies. INTRODUCTION Model Driven Development as embodied by the MDA, and ontology-based knowledge representation as embodied by the semantic web, are the IT industry’s primary visions for the future of software engineering and the World Wide Web respectively. However, as software applications become increasingly web-based, and web applications require increasing “intelligence”, it is clear that these technologies will need to work more and more closely together in the future. There is consequently a growing interest in finding practical ways of using them effectively together. Over the last couple of years researchers have put forward a number of different proposals for integrating Ontology Representation Languages (RDLs), exemplified by languages such as OWL [Bechhofer et. al. 2004] and DAML-OIL [McGuinness et. al. 2002], and MDA [Cranefield 2001, Baclawski et. al. 2002, Djuric et. al. 2003, Falkovych et. al. 2003]. Last year (2003) the OMG also issued an official RFP for an “Ontology Definition Metamodel” to define an ontology definition language within the framework of the MDA [OMG 2003]. The drive to integrate the two technologies is definitely picking up momentum therefore. However, all the existing initiatives to date share two fundamental premises:– Premise 1. MDA and ontology description technologies are inherently distinct technology spaces with at least some fundamentally distinct concepts and mechanisms. Premise 2. the two technologies should be integrated by using “official” OMG extension mechanisms to define an ontology representation language within the OMG’s MDA framework. The term “official” OMG extension mechanism is used here to refer to the extension mechanisms explicitly recognized by the OMG as ways of introducing new language capabilities into the overall MDA framework, namely:– 1. the definition of a new “metamodel” or profile as an instance of (or effective instance of) the MOF, 2. direct extension of the UML metamodel through metaclass specialization, 3. indirect extension of the UML using the profile mechanism (i.e. stereotypes, tags etc.). The basic idea behind current proposals is that the key features of the leading ontology representation languages such as OWL and DAML-OIL should be added to the MDA suite of standards by extending the UML and/or the predefined set of metamodels in one of these three ways. We therefore group these existing proposals together under the label of “extension-based” approaches. To the extent that MDA and ORLs have been developed by different communities, and employ a superficially distinct set of concepts, it is undeniably the case that they represent distinct technology spaces. However, it has never been unequivocally established whether these differences are truly fundamental consequences of the requirements of model and ontology representation, or whether they are merely reflections of their different development histories and backgrounds. If the latter, the case for extension-based integration approaches is significantly weakened, and the case for a more fundamental merging of the two technologies at