PRUS: Product Recommender System Based on User Specifications and Customers Reviews

Abstract

The rising popularity of online shopping has led to a steady stream of new product evaluations. Consumers benefit from these evaluations as they make purchasing decisions. Many research projects rank products using these reviews, however, most of these methodologies have ignored negative polarity while evaluating products for client needs. The main contribution of this research is the inclusion of negative polarity in the analysis of product rankings alongside positive polarity. To account for reviews that contain many sentiments and different elements, the suggested method first breaks them down into sentences. This process aids in determining the polarity of products at the phrase level by extracting elements from product evaluations. The next step is to link the polarity to the review’s sentence-level features. Products are prioritized following user needs by assigning relative importance to each of the polarities. The Amazon review dataset has been used in the experimental assessments so that the efficacy of the suggested approach can be estimated. Experimental evaluation of PRUS utilizes rank score ( $RS$ ) and normalized discounted cumulative gain ( $nDCG$ ) score. Results indicate that PRUS gives independence to the user to select recommended list based on specific features with respect to positive or negative aspects of the products.

Keywords

• Customers reviews
• product ranking
• sentiment analysis
• user specification
• feature extraction