Abstract
With the popularity of e-commerce platforms, customer product reviews tend to have a significant impact on markets. Nowadays, more and more reviews are available on products and usually, good contributions are produced by a relatively small set of reputable users and consumed by a large user population. Therefore, to identify the most helpful reviews is very essential to improve the product sales. Several researches have been done to predict the helpfulness of product reviews. However, it is found out that the quality of a review is not correlated to the rating of the product. Sentiments, however, have a direct significant impact on sales. Therefore, emotions can play a vital role to determine the helpfulness of reviews. Very few researches have been done on using emotion for prediction of helpfulness. Therefore, we intend to mine emotional patterns contained in the text of the reviews and use these patterns to predict its helpfulness. Our objective is divided into two parts: firstly, to predict helpfulness in the same product category (within domain) and secondly, across different product categories (cross domain). We identified several features based on the degree of emotions that can be used to mine the patterns. Our experiment results show that the proposed approach can outperform the existing baseline method upto 6.77 % accuracy for within domain prediction and upto 10.13 % for cross-domain predictions.