Towards Automated Semantics-Driven Web Service Composition: Case Study on Question Answering Systems

When following the Component-based Architecture (CBA), Web Services (WSs), as individual components, are composed to meet the functional and non-functional requirements of users. As the number of available WSs increases, this process becomes more complex. Therefore, designers must be able to easily (1) define, (2) select, (3) verify, and (4) execute composite WSs. While many papers focus on the execution phase, our work specifically aims at the definition and verification phases of combining WSs. To achieve this, we introduce a WS composition method that uses semantic descriptions to define functional information and select relevant WSs. A formal mathematical model is then used to verify the composite WSs. In our approach, we map OWL-based semantic service descriptions to Coloured Petri Nets model and use its algorithms to automatically compute reachable combinations. We apply our method to Question Answering (QA) systems, where WSs serve as functional services that are combined to answer natural-language questions. The results demonstrate that the proposed method helps to significantly reduce the search space of composite WSs during the selection phase.