Résumé en anglais

The importance gained by the process models in modern information systems led to the increasing proliferation of process model repositories. Matching process models and assessing their similarity are critical functionalities required for the management of these collections. In this work1 we present an efficient graph-based technique for matching and evaluating the similarity of semantically annotated process models. Approximate graph matching algorithms which are used in literature (e.g., error correcting sub-graph isomorphism detection), are exponential in size of the graphs. In order to reduce the execution time and improve the applicability of the algorithm for matching and retrieval of process models, we propose a graph summarization technique which reduces the size of the graphs to be compared. Moreover, while most of the related works detect only 1-1 activity mappings, our matching approach is able to detect complex mappings (m-n) between activities based on their input/output sets. Experiments showed that the summarization technique reduces considerably the execution time, maintaining at the same time a good quality of the matching.