基于共现网络的用户评论聚类分析与语义识别研究  

A Study on User Comment Clustering Analysis and Semantic Recognition Based on Co-occurrence Network

作　　者：李森涛 Li Sentao

机构地区：[1]郑州大学信息管理学院,河南郑州450001

出　　处：《图书馆研究与工作》2023年第9期31-39,共9页Library Science Research & Work

摘　　要：挖掘网络知识问答社区用户评论的语义关系,有利于识别用户信息需求特征,为用户提供更具有价值的评论,提升问答社区的信息服务质量。文章以“百度贴吧图书馆吧”中的评论文本数据进行实证研究,通过LDA主题模型进行评论主题聚类分析,利用Word2vec模型将主题关键词转换为词向量,并以此构建评论语义共现网络,最后进行评论排序对比。通过这种方法不仅能够正确聚类不同评论主题,还能够筛选出易被用户所忽视的高质量评论。研究发现,用户评论共分为6类,用户主要关注于学科前景发展,而对于基础设施服务和泛在化服务关注度较低。Mining the semantic relationships of user comments in a network knowledge Q&A community can help identify user information needs and provide more valuable comments,thereby improving the information service quality of the Q&A community.This paper conducts empirical research using comment text data from the"Library Bar"under"Baidu Post Bar".By using LDA topic models for comment clustering analysis,the study transforms the topic keywords into word vectors using the Word2vec model,constructs a comment semantic co-occurrence network,and finally compares comment ranking results.This method can not only correctly cluster different comment topics but also screen out high-quality comments that are easily overlooked by users.The study found that community library user comments can be divided into six categories,and users are mainly concerned about the development of disciplinary prospects,with less attention paid to infrastructure services and ubiquitous services.

关 键 词：网络知识问答社区 主题聚类 共现网络 语义识别 用户评论 

分 类 号：G252.0[文化科学—图书馆学]