Research On Distributed Service Discovery For Semantic Web Services

Using ontology based Semantic Web markup languages which create a computer-interpretable description of service, Semantic Web Services make it possible to discovery Web Services automatically. In real world, Semantic Web Services are usually described by a set of concepts belonging to large-scale domain ontologies. Along with the large quantities of the concept combinations, the number of Web Services also grows dramatically. How to efficiently manage the huge number of Web Services described by the large-scale ontologies becomes a critical issue in the large-scale service discovery area.As ontology graph has an intrinsic fine hierarchy and modularity, this paper partitions the ontology graph into several concept groups with semantic similiarity. Then cluster the Semantic Web Services by mapping their description concept sets to the concept group sets. Then a structured P2P network is proposed to manage the clusters.Based on the ROCK (A Robust Clustering Algorithm for Categorical Attributes) cluster algorithm, this paper introduces a large-scale ontology graph oriented partitioning algorithm, R0CK0n2. R0CK0n2 partitions the large-scale ontology graph into several concept groups. Then using the proposed ROCKOn2Cluster algorithm, the distributed Web Services belonging to the same concept groups are congregated together. In order to manage the clustered services, an ontology partition and prefix-routing based semantic web services discovery system, Spring, is proposed. Through ROCKOn2 algorithm, the contents and nodeId of peers are represented by varying length of concept groups; meanwhile, Spring is an structured integrated semantic P2P system with prefix routing mechanism supporting varying encoding length. The experiment and the practical usage in PISOMWare application shows that Spring system has stable routing hops and efficient discovery ability which is suitable for the large-scale ontology based distributed Web applications.