Paper 1

EvoMatch: An Evolutionary Algorithm for Inferring Schematic Correspondences

Authors: Chenjuan Guo, Cornelia Hedeler Norman W. Paton, and Alvaro A.A. Fernandes

Volume 12 (2013)

Abstract

Schema matching provides an important foundation for both manual and semi-automatic derivation of mappings between sources. However, schema matchers typically return large numbers of potentially inconsistent matches that are neither conducive to automatic mapping generation nor readily digested by mapping developers. This paper pre- sents a method, EvoMatch, for automatically inferring schematic cor- respondences, from which mappings can be generated directly. It aims to oer a more expressive characterization of the relationships between sources than matches identied by existing schema matching methods. In particular, the paper contributes: i) an evolutionary search method for inferring schematic correspondences; ii) an objective function for cal- culating the tness value of a solution within the search space; and iii) an empirical evaluation assessing the eectiveness of EvoMatch for inferring schematic correspondences in comparison with well established existing techniques. In doing so, EvoMatch automatically identies correspon- dences that can be used directly to bootstrap information integration systems, or to inform the manual renement of mappings.