Semantics of an extended relational model for managing uncertain information

Abstract

The Information Source Tracking (IST) method has been recently developed for the modeling and manipulation of uncertain and inaccurate data. In this paper we extend the IST method to model uncertain and inaccurate information at the finest granularity of data in the relational model, namely at the attribute value level. An extended relational model is proposed in which each attribute value in a tuple in an extended relation is associated with an information source vector showing the sources (observers) that contribute to that attribute value, and the nature of their contribution. We will discuss how the relational algebra operations can be extended and implemented to trace the information sources that correspond to each attribute/tuple in the answer to a query. We also present a semantic interpretation of our extended relational model, and prove that the extended relations are 'correct' with respect to the semantic model.