Research And Implementation On Matchmaking Strategies In Semantic Web Service Discovery

As a new computing paradigm, Service-Oriented Computing (SOC) has becomea hot topic in the ?eld of software. Web service has been becoming the main streamimplementation of SOC, with the development of Web service standard and tech-nique. The task of Web service discovery is to locate Web services in repository thatsatisfy user's requirements, according to the functional and non-functional descrip-tions. As a primary task in SOC, Web service discovery has become a challengingproblem which will ensure the successful application of SOC in industry.The standard Web service description can not exactly describe the functionaland non-functional semantics of Web services, due to the lack of semantic annota-tions. Hence, the semantic ambiguity of Web service description brings di?culties tothe automation of service discovery. Fortunately, Semantic Web service technologyprovides a way to describe Web service with formal and explicit semantics and afoundation for automatic service discovery, composition and invocation. However,the current methods of Semantic Web service discovery can not solve the discoveryproblems in many practical applications very well, and the e?ectiveness of discoverystill has plenty of room for improvement.This thesis focuses on Web service discovery methods, including the query mech-anism for heterogeneous Semantic Web services and the e?ective matchmaking algo-rithms that make use of the characteristics of Web service descriptions. The detailsare as follows.1) The research on query mechanism for heterogeneous Semantic Web servicesA query framework, named iXQuery, is proposed to support the query on het-erogeneous Semantic Web services. This framework extends XQuery with similarityjoins for heterogeneous Semantic Web service matchmaking. iXQuery also de?nes auni?ed abstract Semantic Web service ontology, which can be mapped to heteroge-nous Semantic Web service ontologies. The query in terms of this ontology is madetransparent to heterogeneous Semantic Web service ontologies.2) A Web service discovery method based on SAWSDL description character-istics Current information retrieval based service discovery methods ignore the char-acteristics of Web service description, and the service discovery problem is consideredas the normal text retrieval. Thus the e?ectiveness of service discovery is not sat-is?ed. To solve this problem, a method is proposed to integrate two main kindsof features of Web service description language (WSDL) documents, i.e., the struc-ture features and the reference features. Then a novel multi-vector model for Webservices is introduced, which is distinguished from the general text representationmodel. This model exhibits explicitly the features of Web services, and helps toimprove the e?ectiveness of Web service discovery.3) A novel matchmaking method based on ?ne-grained data semanticsThe service discovery methods based on structure matching has high compu-tational cost and can not distinguish the Web services with similar semantics butin di?erent structure de?nitions, due to the ?exibility of XML Schema. To solvethis problem, a so-called ?ne-grained data semantics based method is proposed,which can be obtained by a transformation algorithm that decomposes parametersat message level into a set of ?ne-grained data parameters. The ?ne-grained datasemantics is transparent to the speci?c data structure of message-level parameters,therefore, it can help to match successfully similar Web services with di?erent datastructures of parameters. Experiments show that the proposed method can improvethe e?ectiveness of service discovery.4) A method to enrich the semantic description of Web serviceMost of current Semantic Web services have annotations only for input andoutput parameters, which is not enough to describe the exact semantics of Webservices. A method that extracts semantic constraints from Web service contextis proposed to enhance the semantic representation of Web service. Experimentalresults show that this method can enrich the semantic description of Web serviceand reduce the semantic bias between an interface parameter and an annotation,thus the e?ectiveness of service discovery is improved.5) The matchmaking method by integrating several matchmaking results intoan overall similarityCurrent research demonstrates several kinds of matchmaking on di?erent ser-vice descriptions should be integrated to get an overall similarity, which can help to improve the e?ectiveness of service discovery. A statistic model based servicematchmaking method is proposed to exploit machine learning technique to learnthe weights for di?erent matchmaking strategies from the queries, whose relevantservices are known. Experimental results show that the statistical model basedmethod can e?ectively integrate several matching results from di?erent service de-scriptions, and then improve the e?ectiveness of service discovery.In summary, we have proposed a query mechanism for heterogeneous SemanticWeb services and several service matchmaking strategies, according to some keyissues in automatic service discovery, which provide suggestions for future researchand engineering on automatic service discovery both in theory and in practice. Aheterogeneous Semantic Web service matchmaker, called H-iMatcher, has also beenimplemented that integrates the methods mentioned above as well as several basicmatchmaking algorithms.