A Novel Feature Based Review score to Classify True Reviewer
Kunada Dhana Sree Devi, Rakesh Nayak, L. Lakshmi, Ravisankar Malladi · 2023
With the enormous growth of online business, customers are enjoying a prioritized role, as a vital source for product information, since the review given by a customer is the major influencer to the business. Not all reviews are reliable as causality may also be a part of review writer. Many of the ecommerce websites are facing a tough challenge of identifying customers equinity in giving a reliable review for the products. Reviews do mislead at a times. Genuine customer review really affects the true opinions about the product and can even have a great impact on user recommendations which truly enhances the business. Review rewards are true biases to classify the opinions in the customer reviews. Many of customers give focused reviews to bag these rewards. NLP is todays technology handy tool in analyzing the reviews to hold any true context. In this paper, we present a feature-based review score, generated by mining the opinions of customer, to penalize customers for whom giving review is a child’s play. Our model emphasis on true review writers are real deservers of user benefits and serving them can raise the business. Our approach used contextual sense of the words used in text reviews to generate a feature based for that reviewer. The proposed method introduced three types of feature scores: the contextual score, rating score and the upvotes score for a single user, which are generated based on the context and consistency of short and long reviews given by him for various products. This score for the reviewer is used to classify the reviewer to be genuine. Our approach used a stack, to pile various scores generated from features that have high influence on text reviews. Experimental results showed the proposed feature based reviewer classification is more accurate and appropriate, in classifying true reviewers who really matter for the business.