Classification of compressed and uncompressed text documents

Abstract

Computing the degree of closeness (similarity) between two sets of text documents is one of the core operations in many text mining applications like text classification, clustering and sentiment analysis. The efficiency of such applications mainly depends on the factors like selection of representation model, selection of the similarity metric and selection of learning algorithms. Among these three factors, selection of similarity measure is important since it contributes to the efficiency of most of the text mining applications. In this research article, an efficient similarity measure is proposed for computing the closeness between two sets of text documents. The proposed measure has the capacity of considering different real time situations like presence of a feature or absence of features for computing the degree of similarity between the documents. Furthermore, a compression modeling similarity measure is also proposed for text documents. Two different sets of experiments are conducted to validate the efficacy of the proposed similarity measures. Experimental results demonstrate that thef-measure score obtained from proposed similarity metric is better than thef-measure score of the existing state of the art techniques.