Improving Performance of Product Recommendations Using User Reviews

Recommendation systems have gained importance with the rapid growth in e-commerce industry. Recommendation system utilizes user feedbacks to suggest products that might be useful to the user and also help in accessing the long tail products. Traditional recommendation systems rely on ratings provided by users. However, with advancement in data acquisition, most e-commerce websites today capture other useful feedbacks such as review and review helpfulness, etc. This paper proposes an approach to improve the performance of recommendation systems using user reviews. The experiments are performed on Amazon product dataset which consists of product ratings and reviews. A comparison between traditional rating-based and the proposed recommendation system shows improvement in the recall and root mean square error (RMSE) scores of recommendation system.