Generalized Uncertain Databases: First Steps

Abstract

Existing uncertain databases have difficulty managing data when exact confidence values or probabilities are not available. Confidence values may be known imprecisely or coarsely, or even be missing altogether. We propose a generalized uncertain database that can manage data with such incomplete knowledge of uncertainty. We develop a semantics for generalized uncertain databases based on Dempster-Shafer theory. We propose a representation scheme for generalized uncertain databases that generalizes the Trio representation. Our approach builds upon Trio's query processing techniques to extend them to operate on generalized uncertain databases.