A semantic web service classification system

Abstract

Semantic Web Services allows web services to be searchable through discovery,

composition and invocation and monitoring. Existing systems, such as MODiCo [1],

allows automatic semantic web service discovery and composition but at high

computational costs. By considering only web services in the domain of interest, the

effectiveness and efficiency of web service discovery and composition are expected to be

improved significantly. Hence, a Semantic Web Service Classification System is

proposed.

The classification system is designed with high speed performance and classification

accuracies in mind. With these two criteria in mind, pure textual descriptions approach

has been selected as the main approach in dealing with the task of semantic web service

classification. Supervised machine learning algorithms are used with this approach.

Experiments have shown that implementing a semantic web service classification system

for existing systems, such as MODiCo, is feasible. Our approach is able to achieve good

classification accuracies and speed performance using SVM as the machine learning

algorithm. Top three classification results can be used to further improve the

classification system. Further work such as multi-labeled classification methods and

optimization of machine learning algorithms are areas worth researching.