Iterative cross-training: An algorithm for learning from unlabeled Web pages

Abstract

The article presents a new learning method, called iterative cross-training (ICT), for classifying Web pages in three classification problems, i.e., (1) classification of Thai/non-Thai Web pages, (2) classification of course/non-course home pages, and (3) classification of university-related Web pages. Given domain knowledge or a small set of labeled data, our method combines two classifiers that are able to use effectively unlabeled examples to iteratively train each other. We compare ICT against the other learning methods: a supervised word segmentation classifier, a supervised naïve Bayes classifier, and a co–training-style classifier. The experimental results on three classification problems show that ICT gives better performance than those of the other classifiers. One of the advantages of ICT is that it needs only a small set of prelabeled data or no prelabeled data in the case that domain knowledge is available. © 2004 Wiley Periodicals, Inc.