Measuring similarity of (annotated) data tables

Abstract

We propose a measure of similarity for classical relations or data tables. It is obtained as a generalization of the Egli-Milner ordering and the Hausdorff metric. This measure allows us to compare different data tables. We attempt to extend the measure from classical relations to D-relations, also called relations with similarities, which generalize a large group of annotated relations. The general form of such a measure now contains a function for comparing annotations. We expose some properties of the annotation-comparing function and find suitable candidates in the case of some special annotation domains.