Metamodel matching based on planar graph edit distance

Abstract

A prerequisite for implementing a model transformation is a mapping between metamodel elements. A mapping consists of matches and requires the task of discovering semantic correspondences between elements. This task is called metamodel matching. Recently, semi-automatic matching has been proposed to support transformation development by mapping generation. However, current matching approaches utilize labels, types and similarity propagation approaches rather than graph isomorphism as structural matching. In constrast, we propose to apply an efficient approximate graph edit distance algorithm and present the necessary adjustments and extensions of the general algorithm as well as an optimization with ranked partial seed mappings. We evaluated the algorithm using 20 large-size mappings demonstrating effectively the improvements, especially regarding the correctness of matches found. © 2010 Springer-Verlag.