Investigating Issues and Challenges of Sentiment Analysis-Based Recommender System

Abstract

Now in a day millions of products and customers are adding to the market. In current era, a wide range of customers are putting their interest to purchase products from the internet. So it becomes necessary that quality products can be purchased by customer with greater satisfaction. In recommender systems, items are recommended to users based on their search query, as well as their context information including history of purchased items, taste, location, gender and age group of the user and the sentiment about the products of inventory by analysing product reviews. We can solve this problem by considering cooperative environment, increase the performance of the learners compared to the best recommendation method that comprises the complete realisation of user arrivals and the inventory of items, as well as the context-dependent effects on purchase of items, and evaluate results on a dataset. The main idea behind to choosing sentiment analysis to recommend items according the popularity of the items and review of other users about the same items to convince user needs as per elicited user preferences expressed in textual reviews because now in a day's wide range of users of social networking sites passing their sentiments in reviews about different products. The review datasets consist of millions of reviews which plays an important role to improve the prediction accuracy. This technique is named as sentiment analysis and maps such preferences onto some predicted rating scales that can be understood by existing recommendation algorithms mainly done by extracting opinion words from review and then aggregating the ratings of such words to determine the dominant or average sentiment implied by the user for particular item.