An Extensive Selection of Features as Combinations for Automatic Text Categorization

Abstract

The merits of modern web search engines that intend to access such pages limit relatively to users’ requirement relying highly on information retrieval techniques. For accessing most relevant user subject specific pages, building a categorization system that can analyse the content and present information precisely could be a good alternative. For Text Categorization, most of the researchers relied highly on trained dataset. Each trained dataset is usually large in size due to which most approximations, computations are time consuming. This makes the entire categorization system slow and inaccurate. The proposed method is novel and the number of features is used. This paper explores the effect of word and other values of word in the document, which express the features of a word in the document. The proposed features are exploited by tf-itf, position of the word and compactness. These features are combined and evaluated. The Experimental results showed a significant improvement in Text categorization process.