Using seed words to learn to categorize Chinese text

In this paper, we focus on text categorization model by unsupervised learning techniques that do not require labeled data. We propose a feature learning bootstrapping algorithm (FLB) using a small number of seed words, in that features for each of categories could be automatically learned from a large amount of unlabeled documents. Using these learned features we develop a new Naive Bayes classifier named NB_FLB. Experimental results show that the NB_FLB classifier performs better than other Naive Bayes classifiers by supervised learning in small number of features cases.