Service Discovery By Automated Structural And Semantic Matching

Pervasive computing systems need to locate and use services dynamically. Current models of service location and interaction rely on a fixed contract, or service description, located by name and/or a combination of keywords. This implies a priori agreement on the service description, leading to fragility and the inability for systems to interoperate unless they were built to match an existing standard. This thesis implements a technique for relaxing the fixed contract assumption, so that the nearest match amongst a set of services can be located. The system takes into account structural, data type and naming differences. The naming differences are handled by using an ontology, so that the program maps similar concepts. The search algorithm lays out a series of steps that could be used to form an adapter, in the cases where a match is close enough to be automatically resolved. Creation of these adapters, such as a set of XSLT transforms, is left as future work. Several researchers have proposed techniques for service integration and flexible composition, but there is a paucity of computational results in the literature. This work proposes a semantic matching algorithm and reports on the performance and accuracy, using a set of services created by a number of different programmers. The various types of incompatibilities that cannot be resolved automatically are characterized and this thesis concludes with a discussion of future work, as well as listing the incompatibilities that cannot be bridged programmatically.