1 Introduction

The goal of text categorization is to classify a collection of documents and assign each document to the applicable categories. However, building text classification ”by hand” is time-consuming, thus it is more practical to use machine learning approach to classify from examples - to perform the category assignments automatically.

Many research in text classification focuses on the binary classification using Support Vector Machine (usually abbreviated as SVM for short) learning approach. However, SVM is typically used to perform binary classification, and binary classification is always not enough in practice. Hence, many scientist has started working on finding the best combination of SVM with other means to perform multi-class classification. One approach is to develop a alternative implementation of Support Vector Machine with information gain as evaluation method.

1.1 Text Categorization Tasks

There are numerous tasks that can be automated with text categorization methods. Here are some of them [1]:

• Document Organization: organizing tons of documents into categories for browsing.
• Text Filtering: Text filtering is the activity of classifying a stream of incoming documents dispatched in an asynchronous way.
• Spam filtering: documents are to be assigned to one of two categories: spam and not spam. Note that spam filtering is a cops and robbers game.
• webpage categorization build the classification based on the nature of the hyptertext documents, and the hierarchical structure of the category set
• Word sense disambiguation: depending on the context of an ambiguous word, decide in which sense was it used.
• Information retrieval: search engines may classify documents to be relevant or non-relevant for a given query, based on user feedback.