Berro, Alain, Megdiche-Bousarsar, Imen and Teste, Olivier (2015) A Linear Program For Holistic Matching : Assessment on Schema Matching Benchmarks. In: 26th International Conference on Database and Expert Systems Applications, 1 September 2015 - 4 September 2015, Valencia, Spain.

Preview

Text Download (1MB) | Preview

Abstract

Schema matching is a key task in several applications such as data integration and ontology engineering. All application fields require the matching of several schemes also known as "holistic matching", but the difficulty of the problem spawned much more attention to pairwise schema matching rather than the latter. In this paper, we propose a new approach for holistic matching. We suggest modelling the problem with some techniques borrowed from the combinatorial optimization field. We propose a linear program, named LP4HM, which extends the maximum-weighted graph matching problem with different linear constraints. The latter encompass matching setup constraints, especially cardinality and threshold constraints; and schema structural constraints, especially superclass/subclass and coherence constraints. The matching quality of LP4HM is evaluated on a recent benchmark dedicated to assessing schema matching tools. Experimentations show competitive results compared to other tools, in particular for recall and HSR quality measures