A text classification method integrating ontology and SVM
Aiming at processing the lack of conceptual semantics in the existing SVM-based classification, a text classification method integrated ontology and SVM is proposed. In this method, firstly, word features are mapped to concept features according to domain ontology, and then concept features and their weights will be sent into SVM for training and classification. The classification method integrated ontology and SVM reduces the dimension of classification space, thus saves the training time, and also saves the time for similarity calculation in the period of classification. The conceptional upper and lower extensions can solve the defect that the closely relationship between father and son concept will be treated as different word features in the classification. The text classification experiments in the domain of Bamboo and Rattan proved that compared to traditional SVM-based classification, the accuracy of classification has been greatly improved based on this method.
NETL
NSTL
万方数据
维普
