基于LDA和Word2Vec模型的学位论文评阅意见主题挖掘与分析  

Mining and Analysis of Thesis Review Topics Based on LDA and Word2Vec Models

作　　者：王孟[1] 苏进城 陈志德[2] WANG Meng;SU Jincheng;CHEN Zhide(Graduate School,Fujian Normal University,Fuzhou 350117,China;School of Computer and Cyberspace Security,Fujian Normal University,Fuzhou 350117,China)

机构地区：[1]福建师范大学研究生院,福建福州350117 [2]福建师范大学计算机与网络空间安全学院,福建福州350117

出　　处：《福建师范大学学报（自然科学版）》2024年第5期41-51,共11页Journal of Fujian Normal University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61841701、62277010)。

摘　　要：选取某高校部分硕士学位论文评阅意见为研究对象,使用自然语言处理和机器学习技术进行自动化的硕士学位论文评阅意见主题挖掘与分析。首先,采用LDA(latent dirichlet allocation)模型对评阅数据进行主题建模,提取文本中的潜在主题,并将评阅意见转化为主题分布向量;其次,结合Word2Vec模型将评阅意见的关键词转化为向量表达;最后,采用TextRank方法提取关键词,以揭示评阅专家的关注核心主题。实验结果表明,所提方法能为高校管理人员提供切实有效的分析工具,有助于他们更好地分析总结评阅意见,同时也为硕士研究生撰写高质量学位论文提供有益借鉴。It was conducted to analyze the reviews of selected master s theses of a certain university.By leveraging natural language processing and machine learning technologies,we aim to automate the topic mining and analysis process for master s thesis reviews.Initially,the LDA(Latent Dirichlet Allocation)model is employed to perform topic modeling on the review data,extracting latent topics from the text and converting review comments into topic distribution vectors.Subsequently,the Word2Vec model is used to represent the keywords of review comments as numerical vectors.Finally,the TextRank method is utilized to extract keywords,revealing the core topics that are of prime interest to the review experts.Experimental results demonstrate that our approach offers a practical and effective analysis tool for university administrators,enabling them to better analyze and summarize review comments.Furthermore,it provides valuable insights and guidance for students aiming to compose high-quality master s theses,thus contributing significantly to the enhancement of academic quality in postgraduate education.

关 键 词：硕士学位论文 自然语言处理 LDA模型 Word2Vec模型 TextRank方法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]