Text Classification Based on Nonlinear Dimensionality Reduction Techniques and Support Vector Machines

Abstract

Text classification is an important task in the field of natural language processing. The dimension of the text data is huge for the text documents are usually represented with the vector space model. Thus, it is greatly time-consuming to perform existed text categorization methods. Moreover, it is almost unimaginable to store and enquire high-dimensional text data. To improve the executing efficiency of classification methods, we present a classification algorithm based on nonlinear dimensionality reduction techniques and support vector machines. In the procedure, the ISOMAP algorithm is firstly executed to reduce the dimension of the high-dimensional text data. Then the low-dimensional data are classified with a multi-class classifier based single-class SVM. Experimental results demonstrate that the executing efficiency of categorization methods is greatly improved after decreasing the dimension of the text data without loss of the classification accuracy.