Establishing a knowledge base to assist integration of heterogeneous databases

In this paper the establishment of a supporting knowledge base for a looselycoupled federation of heterogeneous databases available over a computer network is described. In such a federation, user information requirements are not static. The changes and evolution in user information needs reflect both the changing information available in the federation as databases join and leave it and their contents change, and the varied requirements the users have of the information available in the federation. As a user's awareness of the contents of the federation improves, their requirements from it often evolve. This means that users need different integrating views if they are to realise its full potential. The knowledge base will be used to help users of the federated databases to create views of the federation and to determine the available information in the federation. The knowledge base (KB) is created and evolved incrementally by analysing both meta-data of databases as they join the federation and views as they arc created over the databases in the federation. The KB is organised as a semantic network of concept clusters, and can be viewed as a conceptual model describing the semantics of all databases in the federation. It records not only concepts and their inter-relationships defined in schemas but also their frequency of occurrence within the federation, and is used both as a semantic dictionary and as a basis to enrich schemas Romantically during the detection of semantic relationships among schema components when creating user views. We semantically enrich schemas prior to integration by generating the best subnetwork that spans the concepts common to the schema and the knowledge base and then by unifying this sub-network with the schema. Hence the terms in schemas arc given interpretations not by considering them in isolation but by taking into account contexts, derived from the knowledge base with respect to those schema terms. An architecture to facilitate the establishment of this knowledge base and to exploit its functionality is described. The novelty of our research is that we create a knowledge base for a federation by gathering and organising information from it to assist groups of users to create dynamic and flexible views over the databases in the federation which meet their dynamic information requirements. This knowledge base enables reuse of the semantics when creating views.