摘要
分析传统LDA模型在进行微博热点挖掘时所得概率结果抽象且难以结合实际解释的缺点;考虑到微博本身的数据特点和信息论中信息量的观点,提出微博热度的概念,并将其引入到LDA模型的热点挖掘研究中,构建基于微博热度的LDA模型;通过API采集微博数据上的实验,证明新方法与旧方法具有相同的性能,而且能得到更直观的微博热度表,并得出更具有说服力的挖掘结论。
This paper analyses shortcomings in the traditional LDA (Latent Dirichlet Allocation) model when per- forming microblog hotspot mining, which include that excavated probability results is abstract and is difficult to interpret. Taking into account the characteristics of the microblog and the viewpoint of the information quantity in information theory, it proposes the concept of microblog heat, introduces it into the hotspots mining research of the LDA model, and frams the LDA model based on microblog heat. With experiments on microblog data collected through API, this paper proves that the new method has the same performance compared to the old one, furthermore, it can express a more intuitive table of microblog heatand draw a more convincible conclusion.
出处
《图书情报工作》
CSSCI
北大核心
2014年第5期58-63,共6页
Library and Information Service
基金
国家自然科学基金项目"社会化媒体集成检索与语义分析方法研究"(项目编号:71273194)研究成果之一