Research On Distributed Service Discovery For Semantic Web Services

With the widespread use of Web services, the number of services grows rapidly. Services show multi-type, multi-shape and several degrees of complexity. The key issues of web services research is that how to find the needed service quickly and accurately from a large scale of web services.After in-depth analysis of local features of requestors and providers'behavior in SOA, diversity of services register center and status of web services descriptions, this thesis puts forward an idea which using the P2P technology to solve distributional discovery of semantic web services from the perspective of the behavior features. Fully using the descriptions of ServiceProfile and ServiceModel in OWL-S, a distributional semantic discovery system is designed, including three modules—Indexing, categorical register of service functions and multi-level service discovery. The Catalog module includes DHT indexing between domains and conceptual group indexing in each group. The DHT Catalog is used to maintain the topology between domains, which can reduce the conflicts between DHT and semantic discovery by designing hash function based on semantic similarity, improve the rate of domain selection, and reduce the number of times of transmitting request among domains. The conceptual group indexing realizes to pre-classify the registers into groups, which is based on the modularized features of domain ontology, partitions the big ontology into some conceptual groups with similar meaning, and then establishes a mapping between the registered services and these concept groups. The categorical register of service functions tries to realize the services register according to their functions which are divided into groups according to the similarity of service functions, in this way it can reduce the comparing area of service discovery. The concept mapping to 0-1 optimized storage and "and operation" are used for comparing concept group indexing, which can improve the performance of discovery efficiency. The multi-level service discovery divides into service function match and service process match which designs based on multi-level need of requester. Using Petri net to formalize the ServiceModel and nets comparing technology is used to match the processes, which also improve the precision of discovery.The results of tests show that, registering with pre-classification by service functions which reduce the average response time of service discovery about 60%, and increase the average precision of multi-level services matching about 10%. The model permit service discovery between different domains and multi-level service discovery can find more accurate services.