在线产品评论文本情感倾向与评分相关性研究  

Correlation between Text Emotional Tendencies in Online Product Review and Rating Scores

作　　者：徐德华[1] 杨芷凌 XU Dehua;YANG Zhiling

机构地区：[1]同济大学经济与管理学院,上海201804

出　　处：《图书情报导刊》2022年第2期59-65,共7页Journal of Library and Information Science

基　　金：国家自然科学基金项目“项目描述对网络借贷投资决策的影响:多维语言特征分析视角”(项目编号:71771177)。

摘　　要：探究了评分和文本评论之间的关系,检验了两者是否表达出一致的情感倾向。从携程网采集酒店评论和评分数据,利用情感词典,计算了评论文本的特征情感评分及总体得分,通过线性回归和二元回归方法,研究了其与用户实际评分的相关性。实验结果显示,文本情感倾向与评分的相关系数是0.522,为中等程度相关。鲁棒性检验发现,采用不同的情感强度计算方法和不同平台的评价数据,相关系数变化不大。评分与文本评论表达的情感倾向并非完全一致。This paper explores the relationship between user ratings and text comments,and tests whether the two ex⁃press consistent emotional tendencies,collecting hotel reviews and rating data from Ctrip.com and using emotion lexicon to calculate the characteristic emotional scores and overall scores of review texts,and uses linear regression and binary re⁃gression methods to study their relevance to actual user ratings.Experimental results show that the correlation coefficient between text emotion tendency and user ratings is 0.522,which is moderately related.The robustness test finds that the correlation coefficients do not change much when using different emotional strength calculation methods and evaluation da⁃ta from different platforms.User ratings are not completely consistent with the emotional tendencies expressed in text re⁃views.

关 键 词：在线评分 情感词典 文本情感倾向 相关性分析 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]