A method of information source selection for domain ontology construction

A method of information source selection is critical to building a domain ontology with regard to ontology quality and efficiency. A good method can, on the one hand, enhance quality and efficiency of building ontology, and, on the other hand, generalize and develop the field of ontology. Considerable progress has been achieved in this respect; yet, a traditional method only takes concepts into account, and falls short in solving practical problems as well. AVRN (Abstract method, VSM Vector Space Model, Relation distance method and neural network) method is proposed in this paper. Firstly, characteristics of information sources are addresses by abstract analysis method, such as conceptuality, relativity and document predictability. Then, in order to compute document weights, these characteristics' weights are determined by improved Vector Space Model, based on ontology relation distance method and neural network. Finally, the document weights are obtained from the neural network with training samples that is outputted by the OnMaker software. Combined with a real document data set of “Wetland Protection”, the method is tested and a good order effect on the document selection for ontology construction is attained.