| With years of rapid development of Web services technology, new requirementsare gradually raised far Web service discovery technology, which also becomes aresearch hotspot in the field, especially intelligent Web services technology based onsemantic that attracts people's interest. Bringing Semantic Web technology into Webservices, people make the Web service description have the semantic information.Compared UDDI and other traditional discovery technology it improves the discoveryalgorithm accuracy. However, every area are required to build their own ontology andmatching both sides need to use the same ontology, which increases the complexity ofthe whole discovery mechanism and reduce the efficiency of the discovery algorithm.4n the other hand, the user's requestion needs too much information in thesemantic-based matching method which is often hard for users. For the above twoissues, we propose a Kernel-based methods and WordNet Web service discoverymechanisrra, which is divided into two parts.The first part introduces the classification of Web services mechanisms based onKernel methods and WordNet. With the increase of services, we believe that the firststep of service discovery process should classify all services according to thepredetermined in the register, which can improve accuracy the matching algorithm.Referencing to the text retrieval technology, first of all, we parse and extracte WSDLto feature vector, then WordNet HypernymslHypornyms concept helpe us reduce thedimension of feature vectors. Finally, we propose a method of calculation based onKernel Vector Cosine similarity function instead of the traditional functions, whichcould improve the accuracy of the fnal classification.The second part introduces the Web Service Matching method based on theWordNet Concept Tree. In this paper, we use the WordNet Concept Tree of PrincetonUniversity as a semantic ontology dictionary. Firstly, we design a more practical user'sQuery Information. Secondly, refer to the method of the first part, we generate servicefeature vector. At last, we propose matching algorithm between the query andadvVector based on Nuno Seco `s WordNet semantic similarity.Finally, two experiments verify the above two parts. The first experiment showsthat the accuracy of classification based on Kernel methods and WordNet be higherthan the traditional classification based on IR. Through the second experiment, on theone hand,it demonstrate effectiveness of Query Information based Double Option. On the other hand, it verity the semantic similarity computation based on WordNetconcept tree can reach the semantic level. which improve the intelligence of themechanism of Web Service discovery. |
