面向新浪微博的情感社区检测算法

面向社交网络的情感社区检测，可应用于公共健康、舆情监测等领域.以新浪微博为平台建立一种情感社区检测框架，首先融合微博情感表情特征和情感词典，提出基于朴素贝叶斯算法的半词典半表情(naive Bayes based semi-lexicon and semi-emoji，SL-SE-NB)分类模型以实现对文本的情感极性预测；提出一种基于潜在狄利克雷分配(latent Dirichlet allocation，LDA)话题模型的用户-超话题-关键词(user-topic-keywords，UTK) 模型抽取用户话题；基于标签传播算法(label propagation algorithm，LPA)并加入话题概念，提出基于种子集与最小边介数的标签传播情感社区发现算法(label propagation algorithm based seeds and min-edge betweenness，SMB-LPA).最后通过实验验证了所提出算法的有效性和高效性.

Sentiment Community Detection Algorithm for Sina Weibo

Sentiment community detection in social networks would be very valuable in many areas， such as public health， public opinion monitoring and so on. A framework of sentiment community detection was established on Sina Weibo. Firstly， the sentiment expression features and lexicon of Weibo were combined， and the classification model SL-SE-NB (naive Bayes algorithm based semi-lexicon and semi-emoji) was proposed to predict the sentiment polarity of texts. And then， the UTK (user-topic-keywords) model based on LDA (latent Dirichlet allocation) was proposed to extract user topics. Based on LPA(label propagation algorithm) and adding topic concepts， SMB-LPA (label propagation algorithm based seeds and min-edge betweenness) was proposed to discovery sentiment community. Finally， the experimental results proved the effectiveness and efficiency of the proposed algorithms.