A Web Service Similarity Refinement Framework Using Automata Comparison

A key issue in web service based technology is finding an effective web service similarity comparison method. Current solution approaches include similarities comparisons both on signature level and behavior level. The accuracy of comparison is generally satisfactory if the web services are atomic. However, for composite web services with complicated internal structures, the precision of similarity results depends on the comparison approaches. To enhance this precision for composite services, the paper proposes a two-stage framework to integrate the signature level and behavior level similarities. The first stage adopts the traditional WordNet-based interface measuring method to generate a set of possible similar services. Then in the second stage, an automaton is built for each service. By characterizing and comparing the languages generated from those automata, we can enhance the precision of similarity measures of similar services from the previous stage.