面向B2C电商网站的消费者评论有用性评价模型研究

Research on the Model for Evaluating the‘Helpfulness’of Consumer Reviews towards B2C E-Commerce Website

[目的/意义]目前各大电子商务网站产生了海量的评论信息,对于消费者而言,查阅和分析这些信息将面临巨大的挑战。因此,有必要对评论的有用性进行综合评价,为消费者过滤出真正有价值的内容。[方法/过程]为此,本文提出并研究了一种在线消费者评论的有用性评价模型,为消费者的网购决策提供支持。该模型主要基于分类算法,识别在线消费者评论的有用性,并按其概率值大小进行排序。根据在线消费者评论的特点,提取了一系列分类特征用于其有用性评价,然后利用支持向量机对评论进行分类并从中识别有用的记录。利用来自B2C电子商务网站的3个在线消费者评论数据集(手机、女鞋、糖果巧克力)对提出的模型进行实证分析。[结果/结论]研究结果显示,该模型能够量化地评价在线消费者评论的有用性并对其进行有效的分类排序。该模型主要依赖语义特征进行排序,而对非语义特征的依赖较少。通过选择合适的概率阈值,能够缩小验证空间,并显著提升分类精确度。

[Purpose/Significance]There is a huge number of consumer reviews in those large-scale e- commerce websites,which hence poses a challenge for a customer to go through all of them.It is necessary to evaluate the helpfulness of reviews and highlight valuable ones for potential consumers.[ Method/Process]Therefore,this paper tried to find out an approach to evaluate and rank online consumer reviews according to their helpfulness.In this study,it proposed a classification-based ranking model for evaluating the helpfulness and importance of online consumer reviews.It extracted and identified a series of features from reviews in order to evaluate their helpfulness.SVM model was used to classify and identify helpful reviews.This paper had carried out an empirical analysis on the proposed model by using three collections of online consumer reviews(mobile phones,women s shoes,and candy/chocolate).[Result/Conclusion]The empirical results showed that our model could be used to evaluate and rank the helpfulness of online consumer reviews quantatively.The model relied mainly on semantic features rather than non-semantic ones.Moreover,it could reduce the verification space and significantly increase the classification precision by choosing a suitable probability threshold value.