摘要
针对当前基于LDA模型扩展的文本情感分析方法未能考虑同一词语在不同语境下其情感极性的差异及非特征情感词对微博文本情感极性的影响这两个问题,提出一种基于语境分类和遗传算法的微博情感分析方法。该方法首先利用LDA模型构造微博主题集及微博主题词集,然后用微博标签数据逐一对各微博主题词集应用遗传算法自动迭代计算得出词集中词语的情感值,最后利用词集词语的情感值计算微博文本情感极性。实验结果表明,该方法精确度比LDA提升3.12%,召回率达87.32%,F1达73.79%,能够从语境和非特征情感词获取微博情感信息,有效提高情感分类准确率。
In view of the fact that the current text sentiment analysis methods based on LDA model extension fail to consider the differ⁃ences of emotional polarity of the same word in different contexts and the influence of non characteristic emotional words on the emo⁃tional polarity of Weibo text,this paper proposes a method of Weibo sentiment analysis based on context classification and genetic algo⁃rithm.Firstly,Weibo topic sets and Weibo subject word sets have been constructed by using LDA model.Then based on the Weibo tag data,genetic algorithm is applied to each Weibo subject word set one by one,and the emotional value of the words in the word set is calculated automatically.Finally,the emotion value of the word sets are used to calculate the emotional polarity of Weibo.Experimen⁃tal results show that compared with other methods,the accuracy is 3.12%higher than LDA,the recall rate is 87.32%,and F1 is 73.79%.This method can obtain emotional information from context and non-featured emotional words in Weibo,and thus it effective⁃ly improves the accuracy of sentimental classification.
作者
邓凯凯
陆向艳
阮开栋
许欣
刘峻
DENG Kai-kai;LU Xiang-yan;RUAN Kai-dong;XU Xin;LIU Jun(School of Computer and Electronic Information,Guangxi University,Nanning 530004,China)
出处
《软件导刊》
2021年第1期178-184,共7页
Software Guide
基金
广西软科学研究计划项目(桂科AB17205002,2019ZL0046)
广西高校科研项目(KY2015YB008)
广西大学科研基金项目(XJZ130355)
广西研究生教育创新计划项目(JGY2015004)。
关键词
微博情感分析
语境
LDA
非特征情感词
遗传算法
sentiment analysis of Weibo
context
LDA
non-emotional feature words
genetic algorithm
