Bringing Together Structured and Unstructured Sources: The OUMSUIS Approach

摘要

Integration of heterogeneous sources is a means to offer the user an access to multiple information sources in a unified way through queries submitted to a global schema. We propose a semantic web-based mediator model, to provide unified access to various sources which may be both structured (database systems) and unstructured (textual medical reports, scientific publications, etc.). The mediator level is composed of the global ontology and a set of ontologies which make it possible to characterize sources (one or several ontologies by source). Unstructured sources are seen through their Semantic Document Representation obtained by a semantic characterization process. A reverse engineering process is then applied on each Semantic Document Representation schema and each structured source schema in order to provide semi-automatically a set of local ontologies. These local ontologies are articulated around the global schema following the global centric approach. A service called Terminology Server (ServO) is used to perform queries and manage the ontologies. The ontology-based query model combines databases and information retrieval techniques. We illustrate this approach with a case study in the brain disease field but it is sufficiently generic to be used in other domains.