西安石油大学学报（社会科学版）2025，Vol.34Issue(4)：69-78,10.
基于LDA-RoBERTa模型的在线评论主题及情感分析:以京东商城手机为例
Online Review Topic and Sentiment Analysis Based on the LDA-RoBERTa Model:A Case Study of JD Mobile Phones
摘要
Abstract
Accurately identifying the"pain points"of consumers in online reviews has significant practi-cal value for optimizing products,improving services,and increasing the repurchase rate.This paper proposed a topic sentiment recognition model based on the fusion of LDA-RoBERTa dual models.Taking the smart phones of JD.com as an example,this paper explored the correlation between different sentiment polarities and the topics that consumers were concerned about.Research shows that:The Chinese-RoBERTa-www-ext pre-training model builds an emotion classifier to achieve fine-grained emotion recognition(with an accura-cy rate of 98％),and identifies six major themes that consumers are concerned about(brand reputation,per-ceived value,product performance,user experience,appearance design,and product functions).It is found that product performance,user experience and product functions are the areas where negative comments are most concentrated,revealing the core"pain points"of mobile phone products.It is proposed to strengthen user-ori-ented R&D and optimize products,strictly adhere to the bottom line of quality,and build a whole-cycle shop-ping service system.关键词
电子商务/预训练模型/LDA主题模型/文本挖掘/在线评论/情感分析Key words
e-commerce/pre-training model/LDA topic model/text mining/online review/sentiment analysis分类
管理科学引用本文复制引用
黄珣,段金利..基于LDA-RoBERTa模型的在线评论主题及情感分析:以京东商城手机为例[J].西安石油大学学报（社会科学版）,2025,34(4):69-78,10.基金项目
福建省社会科学基金博士扶持项目"DRGs支付改革背景下医疗服务定价方法及补偿标准研究"(FJ2024BF007). （FJ2024BF007）
