基于知识采纳模型的在线评论有用性识别——以大众点评网为例  

Knowledge Adoption Model-based Online Reviews Helpfulness Identification:Case Study of Dianping Website

作　　者：唐艺楠 徐德华[1] 

机构地区：[1]同济大学经济与管理学院,上海200092

出　　处：《情报探索》2017年第6期8-14,共7页Information Research

摘　　要：[目的/意义]旨在自动识别高质量的在线评论。[方法/过程]以大众点评网某商家2398条评论为样本,提取其评论文本特征和评论者特征,采用监督学习的方法进行评论质量识别。[结果/结论]5种质量分类算法中,Gradient Boosting模型和Ada Boosting模型效果最好。这2种模型中,对分类结果贡献度最高的是评论词语总数,评论文本广度对评论有用性有正向影响;相对于消极情感值,积极情感值有更大的贡献度;评论者特征中,评论者的贡献值对评论有用性影响最大。[Purpose/significance]The paper is to automatically identify high quality online reviews. [Method/process]The paper takes a merchant＇s 2398 reviews on Dianping website as sample, extracts their text features and reviewer features, and adopts supervised learning method to do reviews helpfulness identification. [Result/conclusion]Among 5 quality classification algorithms, GradientBoosting model and Ada Boosting model work the best. These two models show that the highest contribution to classification results is total number of commentary words; the width of review text has positive influence on review helpfulness; positive emotion value has more contribution than passive emotion value; and reviewer＇s contribution value has the greatest impact on review helpfulness among the reviewer features.

关 键 词：在线餐饮评论 有用性 评论文本特征 评论者特征 文本分类 

分 类 号：F713.365.2[经济管理—产业经济]