Research On Electronic Word Of Mouth Mining Based On Online Reviews

With the development of Web2.0 and e-commerce,the mode of e-commerce has changed from the traditional business-oriented e-commerce mode to the consumer-oriented social e-commerce model.Under the consumer-oriented social e-commerce model,consumers no longer passively receive information provided by merchants,they can post customer reviews after purchasing products.These customer reviews as an important carrier of electronic word-of-mouth(eWOM)for products are very important for both merchants and potential consumers.However,these abundant and heterogeneous customer reviews data have resulted in information overload.It is impossible for merchants and consumers to make rational judgments in a short period of time by using complicated data.If we can mine the eWOM of product from a large amount of reviews data,it will improve their decision making efficiency and accuracy.In order to solve the mining problem,this paper studies two aspects of mining methods and mining data.At the method level,in order to make up for the limitation of existing mining methods,a product eWOM mining method integrating multiple heterogeneous review data is proposed to improve the accuracy of eWOM.At the data level,in order to make the eWOM more reliable,this paper studies the management strategy of increasing the number of reviews.Firstly,the density of reviews is analyzed.The research results reflect the participation or activity of consumers' feedback about product or service information.Merchants can make corresponding reward strategies to encourage more consumers to write reviews after purchasing products.Furthermore,through the study of the relationship bet\veen the retail price of products and the number of reviews,the law of change in the number of reviews under different retail prices is found.This research provides a scientific basis for the merchants to make price adjustment strategies and stimulate consumers to post their reviews.The generation of more reviews data also can provide sufficient data support for the mining of eWOM.The research conclusions of the paper are summarized as follows:(1)Research on eWOM mining method based on heterogeneous reviews data.A product eWOM mining method that can integrate multiple heterogeneous reviews data(including numeric rating,text reviews,and comparative voting)is proposed,which improves the accuracy of eWOM.Based on the in-depth study of the intrinsic characteristics of online review data,the heterogeneous review data is divided into txwo types:descriptive information and comparative information.Both descriptive information and comparative information are integrated into a digraph structure,from which an overall eWOM score for each product and a ranking of all products can be derived.We collect the data from a third-party review website.The results demonstrate that our method can provide improved performance compared with those of existing product ranking methods.In addition,we also design the implementation framework of the product ranking system based on eWOM score and develops the prototype system,which demonstrates the operability of the method in practical applications,and provides quick and accurate decision-making for the merchants and consumers.(2)The density of product review and its trend.Considering that the larger number of reviews is the guarantee for reliable eWOM,this paper conducts a detailed study on the density of reviews on purchased products based on the relationship model between the number of-reviews and the number of purchases.In order to model the relationship between the number of reviews and the number of purchases,we proposed a data-driven ideal to find the relationship model.We use symbolic regression method to intelligently learn the relationship model between the number of reviews and the number of purchases from real data.According to the discovered relationship model,two model selection strategies are proposed according to the different performance of the model.Finally,based on the selected feasible model,the density of reviews distribution for different products and the trend of the density of reviews are analyzed.The results show that for 60%of the products in the sample,the density of reviews is not stable.When the number of products purchased is relatively small,the density of reviews is relatively high,and when the number of purchases is large,the density of reviews will be relatively low.The results provide a mathematical model and decision-making basis for merchants to make a reasonable reward mechanism to stimulate more consumers to post reviews after purchase products.A large amount of review data provides a guarantee for tapping the real and obj ective eWOM of product.(3)Research on the relationship between the volume of reviews and the price.In order to find out the impact of product price changes on consumer review behavior,this paper uses data analysis method to establish a relationship model between product retail price and volume of review.Based on the experimental results of real data,it is f-ound that the most likely relationship between the volume of reviews and the price is monotonously decreasing,asymmetric U-shaped,and asymmetric inverted U-shaped relationship.We also found the changes for different product categories is different.The experimental results show that the rise in retail prices does not always lead to a decline in the volume of reviews or the average numeric rating.For 38%of the products in the sample,when the price increases to a certain level,the number of reviews will increase.These findings provide an objective basis for merchants to change consumer review behavior by price adjustment mechanisms.More reviews are the data foundation for reliable eWOM.(4)An eWOM generation,mining and diffuse model was designed.Based on the research of product eWOM mining based on reviews big data,this paper designs an eWOM generation,mining and diffuse model based on the existing social e-commerce model from the perspective of design science,and improves the effectiveness of the current online review system.This model can not only promote the generation of more online reviews,but also overcome the decision-making obstacles caused by the overload of review data.Users do not spend a lot of time reading reviews,but based on the results of product eWOM mining,quickly evaluate product or service quality to improve decision making efficiency.In addition,this model will help the platform to spread and push the product's eWOM,help consumers and merchants to quickly make the right decisions,increase user stickiness,and promote the healthy and sustainable development of the e-commerce industry.