Text classification based on semi-supervised learning

Abstract

In this paper, we present our solution and experimental results of the application of semi-supervised machine learning techniques and the improvement of SVM algorithm to build text classification applications. Firstly, we create a features model which is based on labeled data, and then we will be improved it by the unlabeled data. The technique that is to be added a label into new data is based on binary classification. Our experiment is implemented on three data layers which are extracted from papers in three topics sports, entertainment and education on VNEXPRESS.NET. We experimented and compared the accuracy of the classification results between before and after improve features model through semi-supervised machine learning method and classification algorithm based on SVM model. Experiments show that classification quality is enhanced after improvement features model.