Towards a Knowledge-based Approach to Semantic Service Composition
1 Department of Electronics and Computer Science
University of Southampton, Highfield, Southampton, SO17 1BJ, U.K.
{lc, nrs, ft}@ecs.soton.ac.uk
2 Department of Computer Science
University of Manchester, Oxford Road, Manchester, M13 9PL, U.K.
{carole, colin.puleston}@cs.man.ac.uk
3 School of Engineering Sciences
University of Southampton, Highfield, Southampton, SO17 1BJ, U.K.
sc@ecs.soton.ac.uk
4 Epistemics Ltd., Strelley Hall, Nottingham NG8 6PE , U.K.
paul.smart@epistemics.co.uk
Abstract. The successful application of Grid and Web Service technologies to real-world problems, such as e-Science [1], requires not only the development of a common vocabulary and meta-data framework as the basis for inter-agent communication and service integration but also the access and use of a rich repository of domain-specific knowledge for problem solving. Both requirements are met by the respective outcomes of ontological and knowledge engineering initiatives. In this paper we discuss a novel, knowledge-based approach to resource synthesis (service composition), which draws on functionality of semantic web services to represent and expose available resources. The approach we use exploits domain knowledge to guide the service composition process and provide advice on service selection and instantiation. The approach has been implemented in a prototype workflow construction environment that supports runtime recommendation of a service solution, service discovery via semantic service descriptions, and knowledge-based configuration of selected services. The use of knowledge provides a basis for full automation of service composition via conventional planning algorithms. Workflows produced by this system can be executed through a domain-specific direct mapping mechanism or via a more fluid approach such as WSDL-based service grounding. The approach and prototype have been used to demonstrate practical benefits in the context of the Geodise initiative [2].
Research Paper at 2nd International Semantic Web Conference (ISWC2003), 20-23, October, 2003, Sanibel Island, Florida, USA.