Online Reviews Helpfulness Analysis Based On Social Review Graph

With the rapid development of the Internet,the number of online reviews has grown exponentially.Online reviews are consumers' evaluations of the items or services they purchased,and have great reference value for potential consumers to make purchasing decisions.However,the reference value of online reviews is uneven.The helpfulness of online reviews measures the reference value of online reviews for consumers to purchase items or services.Analysis of the helpfulness of online reviews helps consumers to make purchasing decisions quickly.However,the current number of online reviews is huge,making it difficult for consumers to easily find helpful online reviews,and current mainstream e-commerce sites(such as Amazon,Taobao,etc.)rely on online reviews for ratings or posting time to rank online reviews.However,these ranking methods are relatively simple and it is difficult to effectively help consumers make purchasing decisions.Therefore,how to find valuable reviews from massive online reviews is a huge challenge.In order to extract helpful online reviews from a large number of online reviews,this paper analyzes the helpfulness of online reviews from the perspective of users' influence on the helpfulness of online reviews.In the review system of the e-commerce website,the user plays two roles,the author and the voter.This paper finds that the user is a key factor affecting the helpfulness of the online review by analyzing the author's reviews and the voter voting for the review.This paper constructs a social review graph model to describe the impact relationship between users and online reviews,and analyzes the impact of user quality on the helpfulness of online reviews.The graph model consists of two nodes,user and review,and constructs three types of edges:user-review edge,user-user edge,and review-review edge to describe users and reviews,different users,and different reviews.The relationship between the two,and then the user relationship matrix and the review relationship matrix are used to model the user quality and the helpfulness of online reviews,and iteratively calculate through the state transition formula of the relation matrix to obtain the helpfulness of online reviews based on social review graphs..Based on the social review graph model,this paper proposes a helpful review fusion ranking model for online reviews.The model combines the social review graph model with text analysis to analyze and calculate the helpfulness of online reviews,and ranks online reviews according to the helpfulness of online reviews.Online review helpfulness is measured in terms of both the helpfulness of text-based online reviews and the helpfulness of online reviews based on social review graphs.The helpfulness of text-based online review is calculated by combining the product attribute keyword calculation and the review text similarity analysis,and then the helpfulness of text-based online review and the helpfulness of online review based on social review graph are calculated by weighted calculation.Combine,calculate the helpfulness of online reviews,and rank online reviews accordingly.This paper uses the Amazon E-Commerce platform's product reviews as the experimental evaluation object,using three real online review data sets,including a public Amazon clothing review data set,two mobile power and book review data sets crawling from Amazon.Experimental analysis and evaluation of social review graph models and online review helpfulness fusion ranking methods.This paper evaluates the performance of the social review graph model.The results show that user quality has a positive correlation effect on the helpfulness of online reviews,and demonstrates the effectiveness of the social review graph model for analyzing the helpfulness of online reviews;online review helpfulness fusion ordering The method uses the artificial ranking result as the true value,and compares it with other popular algorithms such as Amazon top rated ranking method and review helpfulness analysis method based on complex network.The results show that compared with the traditional ranking method,this paper proposes the fusion ranking method has a high accuracy.