基于作者引用文献关系的潜在研究兴趣主题发现简

扩展LDA（latent dirichlet allocation）模型，提出基于作者引用文献关系的作者-兴趣主题-文献模型。每个作者被分配一个在所有主题上的多项概率分布，每个主题被分配一个在所有文献上的多项概率分布。在DBLP（digital bibliography ＆library project ）文献引用关系数据集上的实验表明，所提模型能有效地提取一个研究领域的主要潜在研究兴趣主题及其所包含的代表性文献，并能挖掘每个作者属于每个研究兴趣主题的分布。

Detecting potential research interest topics using relationship between authors and cited papers

Literatures cited in academic papers can reflect the authors’research interests to some extent.Thus,the citations,not papers themelves,can be used to discover authors’potential research interest topics.The extending LDA （latent dirichlet alloca-tion）model,an author-interest topic-literature model focusing on citation relationship is proposed in this paper.In this model, each author is associated with multinomial distribution over all the topics,while each topic is associated with multinomial distribu-tion over all the cited literatures.Experiment on citation relationships from DBLP dataset shows that it can detect potential inter-est topics and representative cited literatures effectively,and each author’s interest distribution on different topics can also be de-scribed.