On Web Services Matchmaking With Semantic Relatedness Measure

Web Service matchmaking has attracted great research attention recently, with a noteworthy focus on semantics-based matchmaking. However, it is our observation in this thesis that the current mainstream semantic Web Service matchmaking techniques stress overmuch on the rigid "matching" between requests and service advertisement, using logic-based reasoning to facilitate automatic service discovery and composition. However, in real-world applications, such accurate matching is not always feasible, nor is sometimes desirable for users. It is often the case that a user states some vague request and expects to manually select a few from a list of services that are likely to fit the user request. To cater for the vagueness of the user requests, there is a need to persue more flexible matching technologies.The major aim of the present thesis is to address the above-mentioned problem. To do so, a two-tier semantic matchmaking model is proposed, which combines the ontology-based relatedness measure between concepts and the matchmaking of Web Services. First, the OWL-S-based service and request descriptions are investigated by taking Web Services in the transportation field as an example. The upper ontology SUMO and the attached transportation ontology are used in conjunction with OWL-S to annotate transportation-related Web services and requests. Based on this semantic description scheme, the proposed matchmaking model is comprised of two tiers. The first tier is called service Category matchmaking, which compares the relatedness between a reqest and a service advertisement in terms of the "ServiceCategory" description in the OWL-S ServiceProfile sub-model. The comparison result is to filter out the irrelevant Web services so as to descrease the computational load of the second tier matchmaking. The second tier is the functional matchmaking beween requests and advertisements; and the functional matchmaking is further realized by a weighted summalization of the Inputs matchmaking and the Outputs matchmaking.The matchmaking algorithms used in ServiceCategory, Inputs, and Outputs are all based on the semantic relatedness measure between concepts. Hence, the present thesis furthermore studies the semantic relatedness algorithm between two concepts. The existing semantic relatedness measures are analyzed and a hybrid similarity measure is proposed by extending the semantic-distance-based measure.