Semi-automatic matching of heterogeneous model-based specifications
IT-systems are often described by a variety of specifications such as UML, Java, BPMN, WSDL, etc. These heterogeneous specifications constitute different views on the same system, resulting in the challenge of matching. That is in context of Model Driven Engineering the discovery of semantic correspondences between model elements, which can be used for tasks such as transformation or trace link generation. Although support by semi-automatic matching has been proposed, current approaches show deficits. They leave room for improvement in matching quality, do not appropriately address scalability, and miss a thorough evaluation. They were proposed to specifically target differencing and versioning in contrast to matching. We tackle these issues by proposing a configurable combination of matchers that considers both: meta-models and models. We propose to adopt established schema matching techniques and to utilize information gained from meta-model matching for the task of model matching. Additionally, the graph qualities planarity and reducibility are used to apply graph isomorphism and clustering algorithms for enhanced structural matching. Finally, we build upon a generic model-based infrastructure (EMF) allowing for an easy integration of heterogeneous specifications and the realization of scenarios from the area of service engineering for our proposed evaluation.
Full Text: PDF