Abstract
A mechanism for determining the probabilities for the possible values in a partial value is proposed. Partial values with probabilities are employed in a heterogeneous database system to resolve schema and data conflicts, and to provide a more informative query answer. In this paper, we extend the definition of the similarity function in Jaccard$s similarity coefficients to estimate the probabilities. This extension is based on our categorization of attribute semantics. We consider a model in which sample tuples contain only definite values, followed by considering more general cases.