Metamodel Matching Based on Planar Graph Edit
p.245-259 in: Theory and Practice of Model
Transformations; Springer LNCS 6142/2010;
ACM Digital Library
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.