计算机科学与探索 ›› 2014, Vol. 8 ›› Issue (9): 1146-1152.DOI: 10.3778/j.issn.1673-9418.1404011

• 人工智能与模式识别 • 上一篇    

基于耦合关系的情感词语义分析方法

王  伟1+,孟祥福2,肖春娇3   

  1. 1. 辽宁工程技术大学 科学技术处,辽宁 葫芦岛 125105
    2. 辽宁工程技术大学 电子与信息工程学院,辽宁 葫芦岛 125105
    3. 辽宁工程技术大学 基础教学部,辽宁 葫芦岛 125105
  • 出版日期:2014-09-01 发布日期:2014-09-03

Analysis Approach of Emotional Word Based on Coupling Relationship

WANG Wei1+, MENG Xiangfu2, XIAO Chunjiao3   

  1. 1. Office of Science and Technology, Liaoning Technical University, Huludao, Liaoning 125105, China
    2. School of Electronics and Information Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China
    3. Department of Basic Education, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Online:2014-09-01 Published:2014-09-03

摘要: 针对传统话题模型不能很好地获取文本情感信息并进行情感分类的问题,提出了情感LDA(latent Dirichlet allocation)模型,并通过对文本情感进行建模分析,提出了情感词耦合关系的LDA模型。该模型不但考虑了情感词的话题语境,而且考虑了词的情感耦合关系,并且通过引入情感变量对情感词的概率分布进行控制,采用隐马尔科夫模型对情感词耦合关系的转移进行建模分析。实验表明,该模型可以对情感词耦合关系和话题同时进行分析,不仅能有效地进行文本情感建模,而且提升了情感分类结果的准确度。

关键词: 潜在Dirichlet分配(LDA)模型, 情感词耦合, 隐马尔科夫模型(HMM), 文本情感建模

Abstract: Against to the problem that traditional topic model cannot obtain emotional information of text well for emotional classification, this paper proposes a latent Dirichlet allocation (LDA) model of emotion. By the text emotion modeling analysis, this paper also proposes an LDA model based on emotional word coupling relationship. The model considers not only the topic of emotional word context, but also the word coupling relationship. It controls the probability distribution of the emotional word by introducing emotional variables, and uses hidden Markov model to analyze the transformation of emotional word coupling relationship. The experiments show that, when analyzing the emotional word coupling relationship and topic at the same time, the model not only is effective for text sentiment modeling, but also can enhance the accuracy of sentiment classification results.

Key words: latent Dirichlet allocation (LDA) model, emotional word coupling, hidden Markov model (HMM), text sentiment modeling