基于PLSA模型的在线评论量化研究  

Research on Online Comment Quantification Based on PLSA Model

作　　者：王青芸[1] 周靖 李艳青 尧志毅 WANG Qingyun;ZHOU Jing;LI Yanqing;YAO Zhiyi(School of Mathematics and Computer Science,Gannan Normal University,Ganzhou 341000,China)

机构地区：[1]赣南师范大学数学与计算机科学学院,江西赣州341000

出　　处：《赣南师范大学学报》2021年第3期20-23,共4页Journal of Gannan Normal University

基　　金：江西省教育厅科技项目(GJJ201405,GJJ202015);江西省自然科学基金项目(20202BABL203031)。

摘　　要：随着大数据时代的到来,网络购物的快速发展,越来越多的网民可以跨地域界线进行方便、快捷的购物交流.因此,由互联网用户创造的海量数据使得在线评论成为一种重要的网络口碑.本文以十几年来亚马逊在线市场几种产品的评价为例,首先依据评论的帮助等级确定评论使用价值.排除使用价值过低的评论后,再对剩余数据使用PLSA模型进行潜在语义分析,得出每篇在线评论的情感因素.将情感因素量化后,作为每条评论的情感评分,并以此将评论分为积极评论和消极评论.最后建立“时间-星级-评论”模型,研究星级对客户评论的影响.该模型从评论的来源着手,研究影响评论的因素,提早对产品的评论情感趋势进行预测,使公司能在销售前期便对产品的销售策略进行优化调整.With the advent of the era of big data and the rapid development of online shopping,more and more netizens are able to conduct convenient and fast shopping exchanges across geographical boundaries.As a result,the vast amount of data created by Internet users makes online reviews an important form of Internet word of mouth.This paper takes the evaluation of several products in amazon's online market over the past decade as an example:Firstly,determine the use value of the reviews based on its helpfulness ratings.After excluding the reviews with low usage value,the potential semantic analysis of the remaining data using the PLSA model was carried out to obtain the emotional factors of each online reviews.The emotional factors were quantified as the emotional rating of each review,and the reviews were then divided into positive and negative reviews.Finally,a"time-star rating-reviews"model is established to study the influence of star-rating on customer reviews.Starting from the sources of reviews,this model studied the factors affecting reviews and predicted the emotional trend of product reviews in advance,so that the company can optimize and adjust the sales strategy of products in the early stage of sales.

关 键 词：在线评论 PLSA 量化分析 情感极性分析 NLTK 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]