Research On Heterogeneous Data Sources Integration Oriented Semantic Modeling & Evolution Techniques

With the increase of the number of independent database applications, enterprises desire an effective solution of heterogeneous and distributed databases integration. To achieve the target of interoperability between these different databases, semantic integration is necessary. In this field, ontology model-driven approach shows a lot of advantages. However, this approach heavily depends on the ontology model which sits atop of the whole integration architecture. In Chinese aeronautical enterprises, there are two main obstacles for this approach's application. They are:1. An ontology model doesn't exist in most of these enterprises;2. It's too difficult to build practicable domain ontology without considerable consuming on both investment and time.This thesis tries to resolve these problems. We developed the concept of ontology model into semantic model, while the ontology model driven approach is developed into semantic model driven approach. We introduced an evolving method for semantic model to enable semantic model could be reused in the different integration environment.Three main research aspects as below are contained in this thesis: 1. Semantic model-driven approach for enterprise data integration. We proposed a new architecture of model-driven enterprise heterogeneous data sources integration, and introduce a new concept of Semantic Model, which take the place of Ontology Model in the integration architecture.2. Theory and practice of semantic model. Semantic model sits atop of the whole architecture. We suggested a model structure which is consists of concepts tree and mapping relations. Base on Protege contributed by Stanford SMI Group, we studied a modeling method.3. A method of semantic model evolution. The objective of integration is always in changing for enterprises. To minimize the consumption of investment and time in integration, semantic must be reused as possible. To achieve this purpose, semantic model must be modified somehow to face different task of integration. Therefore, we proposed a semantic model evolution method based on mapping data sources to the existed model.