Researches On Semantic-based Automatic Service Composition

With the development of network technology, people's demand for network applications is moving to diversified, intelligent and personalized direction. With Web services technology matures, more and more business providers provide their services in the form of Web services. However, a single service is function limited, how to make use of Web services to provide users with comprehensive, complex services, particularly automatic service compositions has become today's research hot spots.The key points to implement automatic web service compositioin are the semantic technology and AI planning technology. Nowadays, academics have given several research results on automatic service composition technologies. Unfortunately, most researches still remain in the academic level, and the researches is too theoretical to lacking of corresponding industry practical instance. There is still a certain distance from theory to actual application.This thesis first describes concepts and features of Web services and service composition. In particular, the research situation and important technologies of service composition are also metioned. On this basis, we give a service composition framework. In this framework, message description and behavior description of service can be extracted from service description file. Then we find the behavior of a service can be represented by a finite state machine. Since the finite state machines can be changed to DPDL formulas, so we can use tableaux technique to check the existence of a composition, and use mealy algorithm to calculate the composition result. If such composition exists, the message semantic will be analyzed in order to check the message matching between composed services. Since the service interface definition can be represented by ontology concepts, and the service behavior can be represented using linear logic formulas. So we can use inference rules of linear logic to check the matchability and satisfiability of service message. Meanwhile, we extend the inference rules of linear logic for service composition. Finally, the thesis represents a multi-agent and mediator service based service compostion running framework, which supports the translation from composition result to BPEL description, in order to drive the running of service composition.