CAJ | 학술논문

随着互联网的发展,微博已成为人们获取信息的主要平台,为从海量微博中挖掘出有价值的主题信息,结合微博中的会话、转发和话题标签,将微博划分为用户兴趣、用户互动和话题微博3类,提出基于作者主题模型( ATM)的话题标签主题模型HC-ATM,使用Gibbs抽样法对模型进行推导,获取微博主题结构。在Twitter数据集上的实验结果表明,与ATM模型和基于潜在狄利克雷分布的微博生成模型相比, HC-ATM模型的主题困惑度更小、差异度更大,并且能有效挖掘出不同微博类型的主题分布。
수착호련망적발전,미박이성위인문획취신식적주요평태,위종해량미박중알굴출유개치적주제신식,결합미박중적회화、전발화화제표첨,장미박화분위용호흥취、용호호동화화제미박3류,제출기우작자주제모형( ATM)적화제표첨주제모형HC-ATM,사용Gibbs추양법대모형진행추도,획취미박주제결구。재Twitter수거집상적실험결과표명,여ATM모형화기우잠재적리극뢰분포적미박생성모형상비, HC-ATM모형적주제곤혹도경소、차이도경대,병차능유효알굴출불동미박류형적주제분포。
With the development of the Internet, microblog has become a major platform for people to obtain the information. In order to mine useful topic from microblog, based on the futures of microblog that having conversation tags,retweet tags and hashtags,this paper divides microblog into three kinds. They are microblogs about users’ interest, users interaction and hashtag-related. It designs a novel hashtag topic model named Hashtag Conversation Author Topic Model( HC-ATM) based on Author Topic Model( ATM) ,and uses Gibbs sampling implementation for inference of this model. Experiments on Twitter dataset show that HC-ATM outperforms the ATM and MicroBlog Latent Dirichlet Allocation( MB-LDA) in terms of both perplexity and KL-divergence. Besides,HC-ATM can mine topic distribution of different kinds of microblog effectively.