A toward framework for generic uncertainty management

The need for an automatic inference process able to deal with information coming from unreliable sources is becoming a relevant issue both on corporate networks and on the open Web. Mathematical theories to reason with uncertain information have been successfully applied in several situations, but each one of these models is tailored to deal with a specific semantics of uncertainty. In this paper, we put forward the idea of using explicit representations of the different types of uncertainty for partitioning the inference process into parts. By coordinating multiple independent reasoning processes, we are sometimes able to apply a specific model to each type of uncertain information, and recombine the final results via a suitable reconciliation process. We validated our approach applying it to the classic schema matching problem, and using the Ontology Alignment Evaluation Initiative, (OAEI) tests to assess the results.