Similarity Measures for Web Service Composition Models

A Web service composition is an interconnected set of multiple specialized Web service operations, which complement each other to offer an improved tool capable of solving more complex problems. Manual design and implementation of Web service compositions are among the most difficult and error prone tasks. To face this complexity and to reduce errors at design time, the developer can alternatively search and reuse existing compositions that have solved similar problems. Thus the problem of designing and implementing Web service compositions can be reduced to the problem of finding and selecting the composition closest to an initial specification. To achieve this goal, there is the need to define and use similarity measures to determine how close is a given composition with respect to any given specification. Comparison of Web service compositions can be done using two possible sources: composition designs (models), and execution logs of compositions. In particular, in this paper a set of similarity measures are described for Web service composition models. The main objective is to measure and assess the degree of closeness between two given compositions of Web services regardless of their modelling language.