A Neural Network Based Approach for Semantic Service Annotation

Nowadays, a large number of business owners provide advertising for their services on the web. Semantically annotating those services, which assists machines to understand their purpose, is a significant factor for improving the performance of automated service retrieval, selection, and composition. Unfortunately, most of the existing research into semantic service annotation focuses on annotating web services, not on business service information. Moreover, all are semi-automated approaches that require service providers to select proper annotations. As a result, those approaches are unsuitable for annotating very large numbers of services that have accrued or been updated over time. This paper outlines our proposal for a Neural Network (NN)-based approach to annotate business services. Its aim is to link a given service to a relevant service concept. In this case, we treat the task as a service classification problem. We apply a feed-forward neural network and a radial basis function network to determine relevance scores between service information and service concepts. A service is then linked to a service concept if its relevance score reaches the threshold. To evaluate the performance of this approach, it is compared with the ECBR algorithm. The experimental results demonstrate that the NN-based approach performs significantly better than the ECBR approach.