One size does not fit all: Customizing ontology alignment using user feedback

Abstract

A key problem in ontology alignment is that different ontological features (e.g., lexical, structural or semantic) vary widely in their importance for different ontology comparisons. In this paper, we present a set of principled techniques that exploit user feedback to customize the alignment process for a given pair of ontologies. Specifically, we propose an iterative supervised-learning approach to (i) determine the weights assigned to each alignment strategy and use these weights to combine them for matching ontology entities; and (ii) determine the degree to which the information from such matches should be propagated to their neighbors along different relationships for collective matching. We demonstrate the utility of these techniques with standard benchmark datasets and large, real-world ontologies, showing improvements in F-scores of up to 70% from the weighting mechanism and up to 40% from collective matching, compared to an unweighted linear combination of matching strategies without information propagation. © 2010 Springer-Verlag.