Place semantics into context: Service community discovery from the WSDL corpus

Abstract

We propose a novel framework to automatically discover service communities that group together related services in a diverse and large scale service space. Community discovery is a key enabler to address a set of fundamental issues in service computing, which include service discovery, service composition, and quality-based service selection. The standard Web service description language, WSDL, primarily describes a service from the syntactic perspective and rarely provides rich service descriptions. This hinders the direct application of traditional document clustering approaches. In order to attack this central challenge, the proposed framework applies Non-negative Matrix Factorization (NMF) to the WSDL corpus for service community discovery. NMF has demonstrated its effectiveness in clustering high-dimensional sparse data while offering intuitive interpretability of the clustering result. NMF-based community discovery is further augmented via semantic extensions of the WSDL descriptions. The extended semantics are first computed based on the information sources outside the WSDL corpus. They are then seamlessly integrated with NMF, which makes the semantic extensions fit in the context of the original services. The experiments on real world Web services are presented to show the effectiveness of the proposed framework. © 2011 Springer-Verlag.