Comprehensive Information Model

A comprehensive information model needed to be provided, which incorporates static and dynamic aspects of information integration, and supports advanced “human-like” reasoning, based on a rich semantics. Current information integration systems are rather poor in this respect, and provide only limited support (if any) for expressing constraint relationships between the local sources and a global view of the data. The source data are integrated in such systems under implicit assumptions such as soundness and completeness; arising inconsistencies are handled at low levels in a procedural manner, without a clear understanding to the user about the effects on the semantics of the overall system. What we barely needed was a much richer information model in which knowledge about the sources, their semantics and relationships can be declaratively expressed, such that on the basis of a clear semantics, reasoning about the sources was possible and can be exploited for meaningful integration. Furthermore, the information model should be capable of expressing criteria such as source preference, or strategies for data integration that the user might select. From the declarative specification, the integration process may then provide results which are transparently obtained by exploiting all available knowledge in a meaningful way. Of course, increasing the expressiveness of the Information Model has an impact on the complexity of reasoning about the information specified in such a model, which in turn influences the complexity of
the information integration activity, in particular the task of answering queries posed to the Information Model. Therefore, a crucial aspect that needs a deep analysis is the trade-off between the expressiveness requirements above illustrated and the issue of providing efficient information integration algorithms.