Novel top-down methods for Hierarchical Text Classification.docVIP

Novel top-down methods for Hierarchical Text Classification 　　（a. Modern Educational Technology and Information Center Jiangxi University of Science and Technology, Ganzhou,Jiangxi 341000 China; 　　b. Department of Computer Science and Technology GuanZhou University Sontan College, Zengcheng,GuangZhou, Guangdong,511370 China ） 　　Abstract: To classify large-scale text corpora, one common approach is using hierarchical text classification and classifying text documents in a top-down manner. Classification methods using top-down approach can scale well and cope with changes to the category trees. However, all these methods suffer from a common problem: a high level of misclassification document has unrecoverable. We define an virtual subclass for each non-leaf category to help the rejected documents go back to ancestor category ,thus improving the overall performance .Our experiments using Support Vector Machine (SVM) classifiers on the 20newsgroup collection have shown that they all could reduce blocking and improve the classification accuracy. Our experiments have also shown that the virtual category method delivered the best performance. 　　Key words: hierarchical classification , virtual category, top-down approach 　　1. Introduction 　　Text classification is a crucial and well-proven method for organizing the collection of large scale documents. The predefined categories are formed by different criterions, e.g. “Economic”, “Sports”and “Computer” in news classification, “Junk Email” and “Ordinary Email” in email classification. In the literature, many algorithms [1] have been proposed, such as Support Vector Machines (SVM), k-Nearest Neighbor (kNN), Naive Bayes (NB) and so on. Empirical evaluations have shown that most of these methods are quite effective in traditional text classification applications. In past several years, hierarchical text classification has become an active research topic in database area[2][3]. Different with traditional classification,