A new transductive support vector machine approach to text categorization

Transductive inference is well suited for text categorization tasks with an enormous amount of unlabeled data in addition to a small number of labeled data. We present a new transductive support vector machine approach for text categorization in order to make use of the large amount of unlabeled data efficiently. In our experiments the performance of transductive methods greatly exceeds that of conventional SVM. Experimental results also show that our algorithm outperforms Joachims' TSVM, especially when there is a significant deviation between the distribution of training and test data.