Web service discovery using semi-supervised Block Value Decomposition

Service communities help improve the service discovery process by targeting user queries at highly relevant sub-spaces. In this paper, we propose a semi-supervised web service community learning approach using Block Value Decomposition Co-clustering (SS-BVD). Our approach incorporates domain knowledge in the form of must-link and cannot-link constraints and leverages the duality between web services and their operations to significantly improve the homogeneity of communities. By employing BVD, our approach not only supports different numbers of communities for services and operations, but also forms communities for both the services and their operations simultaneously. Through experiments performed on real world web service data, we demonstrate the performance of SS-BVD for service community learning.