Abstract:
Web services have been used widely in modern software applications since they, as networked software units, provide certain functionality that can be incorporated into building software applications in a flexible manner. Like other software, Web services may experience changes and failures which make them inaccessible to service consuming applications. In this case, it is then necessary for those applications to find other alternative services. One of the effective approaches is to evaluate both structural similarity and semantic similarity between the description of the service in use and those of other candidate services in order to identify an alternative. This thesis follows an approach called URBE to determine structural and semantic similarity between Web services. In particular, we enhance the evaluation on data type similarity, by also considering family of data types and covariance/contravariance principle, and on name similarity, by also considering text similarity. The enhanced algorithm is called M-URBE. An experiment shows that, in comparison with URBE, M-URBE can improve the performance of Web service retrieval. In addition, a web service retrieval system is developed. It supports the M-URBE algorithm and analyzes the difference between the service is use and potential substitute services so that the service consumer can prepare for the difference before using any substitute service.