Classification Framework for Public Web Services using Semantic Data Analysis

Abstract

Service oriented architecture is a popular trend in the current business world. There are thousands of public web services available over the service registries like Universal Description Discovery and Integration and web sites over the internet. Some of them are freely available, however consumers have to pay for most of the services. When the usage is increase, it becomes difficult for web service authors to manage them properly and for web service consumers to find desired services, which suits their needs. Web services use simple object access protocol. It is a XML based message passing protocol. It enables web services to work platform independently. This research is an attempt to implement a novel method or a framework to classify a given set of web services to a set of predefined classes or categories, so it helps web service authors to manage them easily and also, helps web service users to find desired services efficiently in the reduced search domain. Eventually it helps service discovery, integration and orchestration. There are lot of research that have been carried out in web service classification. They can be primarily divided into two categories. They are semantic based methods and machine learning techniques. Both these methods have their advantages and disadvantages. This novel classification framework use both text matching and semantic web technology. WordNet a free word corpora for english language was used for semantic similarity matching. Slightly changed version of the longest common subsequence and edit distance was used for the text classification process. Evaluating the system with other classification frameworks it can be showed that the proposed hybrid model improves the classification rate significantly. To evaluate the proposed model we used two existing classification methods. Web pages from BBC web site was used to prove the proposed classifier is an un biased and can be used for any other classification task. We hope to extend this research in future to improve the efficiency by using parallel programming techniques.