The Promotion Of Constructing Ontology Automatization With The Clustering Method Of Neural Network

The ontology theory has been widely approved in the information science domain. It becomes more and more important to construct complete and accurate domain ontology. At present, most of researches in the world require the participation of experts to construct ontology. However, it is very difficult to deal with large numbers of data with this method by manual. The object of this thesis is to make use of the Self-Organizing Map(SOM) neural network to deal with the data of constructing ontology, and to find terms, properties and relations among properties which are necessary to construct ontology.At first, we performed SOM Neural Network clustering experiments. According to the advantage of SOM Neural Network, we used this method to cluster ontology data based on large-scale corpus. Secondly, we improved the clustering algorithm through the combination of SOM Neural Network, which compensates the disadvantage that the borderline of clustering by SOM Neural Network is not clear due to the small semantic distance among data.The key point of cluster is to construct the input vector of SOM Neural Network in the process of clustering operation. This thesis showed the ways how to construct the context windows and how to calculate increasing information of character word and TFIDF( Term Frequency Inverted by Document Frequency). According to the difference between the method of...