华中科技大学学报（自然科学版）Issue(z1)：66-70,5.DOI:10.13245/j.hust.15S1016
基于主题情感混合模型的细粒度观点挖掘
Fine-grained opinion mining based on topic and sentiment hybrid maximum entropy model
摘要
Abstract
On the basis of improving standard latent Dirichlet allocation (LDA)model,a topic and sentiment hybrid maximum entropy LDA model was proposed for fine-grained opinion mining of online reviews.Firstly,a maximum entropy component was added to the traditional LDA model to distin-guish background words,aspect words and opinion words.Both the local and global division of aspect words and opinion words can be further realized.Secondly,a sentiment layer was inserted between topic layer and word layer.The proposed model was extended from three layers to four layers.Final-ly,sentiment polarity analysis was done to simultaneously acquire the sentiment polarity of the whole review and each topic.Under this case,fine-grained topic-sentiment abstract can be concluded.The related experimental results verify the validity of the proposed model and theory.关键词
观点挖掘/潜在狄利克雷分布模型/主题情感混合模型/最大熵/细粒度Key words
review mining/latent Dirichlet allocation (LDA)model/topic and sentiment hybrid mod-el/maximum entropy/fine-grained分类
信息技术与安全科学引用本文复制引用
马长林,谢罗迪,王梦,司琪..基于主题情感混合模型的细粒度观点挖掘[J].华中科技大学学报（自然科学版）,2015,(z1):66-70,5.基金项目
国家自然科学基金资助项目（61003192）． （）
