Automated Web Service Composition

Abstract

Web content has traditionally been dominated by documents intended for human interpretation. However, the recent advent of Web Services has played a key role in evolving the Web from a mere document repository to a provider of services. The range of services available may be further expanded by composing Web Services together to create new services which provide novel functionality. The ability of Web Service composition to create new services which are capable of fulfilling a need that no single service can satisfy is its major benefit. Web Service composition also has the potential to make Web Service invocation more efficient by aggregrating disparate processes into a single service.

The full potential of Web Service composition can only be realised if the process is automated and performed on demand. Specific advantages of automated service composition over static service composition include its ability to: take advantage of new services as they become available; take account of changing user preferences; cope with service unavailability.

In order to automate service composition, services must describe their own characteristics in an unambiguous machine readable format. In other words, it must be possible for a machine to determine what those services actually do. The current standard for encoding such descriptions is OWL-S. In order to solve a Web Service composition problem a machine must also be able to determine which services to execute, and in what order. This problem can be solved by AI planning.

The Web Services composition solution presented in this report reasons over Web Service described in OWL-S and uses a type of planning known as HTN planning in order to specify the composition.