Discovering Linkage Patterns among Web Services Using Business Process Knowledge

Discovering relations among web services has been a fruitful research topic in services computing. Such relations are useful for service discovery, selection, composition, etc. The much information about web services is available, the better/more relations can be discovered. Based on the technical information provided in WSDL files, simple relations can be discovered. Finding more comprehensive relations requires additional information, such as semantic descriptions, information about service consumers, or service compositions. Rich semantic service descriptions are not so common in practice. Experiments have shown that similar users do not necessarily use the same web services. Using service compositions gives good results, however, researchers assume that they have access to such compositions, which is not the typical case in practice. Additionally, using service compositions is not sufficient to determine types and strengths of such relations. In this work, we propose a novel approach to discover relations among web services in the form of linkage patterns based on the configurations of business processes that use them. We specify types and weights of the discovered linkage patterns based on control flow patterns in business processes. We have implemented this approach using Oryx, a process modeling tool and repository, and Depot, a service registry, and validated it through real-world examples, as we show in this paper.