Product review classification using LSA with improved feature weight model

Abstract

Product review classification plays a vital role in understanding the likes and dislikes of users of the product. This analysis can be utilized to improve the quality of product as expected by users. In this paper, classification of reviews is performed using Naïve Bayes and K-nearest neighbor algorithms. The key factors contributing to improving classification performance are proper feature weights and less number of dimensions. To improve feature weights, a novel feature weight modification technique is proposed which is based on sentiment scores of the Synset words of input set of words. And to reduce the number of dimensions, we used Latent Semantic Analysis (LSA) technique. From the results, it is proved that the proposed method of modifying weights gives significant improvement in classification performance.