Constructing an Ontology Automatically by Projective ART Neural Network

Abstract

Ontology is playing an important role in Semantic Web, biomedical informatics and knowledge management. At the same time, constructing and maintaining ontology has become challenges in efficiency and accuracy. In this study, we present a novel ontology construction based on artificial neural network and Bayesian network. First, we collected the web pages related to the problem domain. Then utilize the labels from the HTML tags to selected keywords and utilize WordNet to determine the meaningful keywords called terms. Next, calculate Entropy value to determine the weight of terms. After above steps, using a projective adaptive resonance theory neural network(PART) clusters the terms. Finally, the system outputs an ontology using Bayesian network to express the hierarchical relation among the keywords.