Exploring Unclassified Texts Using Multiview Semisupervised Learning

Abstract

This chapter presents semi-supervised multi-view learning in the context of text mining. Semi-supervised learning uses both labelled and unlabelled data to improve classification performance. It also presents several multi-view semi-supervised algorithms, such as CO-TRAINING, CO-EM, CO-TESTING and CO-EMT, as well as reporting some experimental results using CO-TRAINING in text classification domains. Semi-supervised learning could be very useful whenever there is much more unlabelled than labelled data. This is likely to occur in several text mining applications, where obtaining unlabelled data is inexpensive, although manual labelling the data is