Model Theoretic Semantics for Information Integration

Due to the increasing necessity and availability of information from different sources, information integration is becoming one of the challenging issues in artificial intelligence and computer science. A successful methodology for information integration is based on federated databases. Differently form databases, a completely satisfactory formal treatment of federated databases is still missing. The goal of this paper is to fill this gap by providing a model theoretic semantics, called Local Models Semantics for federated databases. Our basic intuition is that a federated database can be formalized by representing each database as a set of local models. We argue that this perspective is a promising one, as many relevant problems in information integration, such as semantic heterogeneity, interschema dependencies, query distribution, local control over data and processing, and transparency, can be successfully represented by Local Models Semantics