概念图在文献集合研究中的一个新模型

A New Model For Conceptmap's Application in the Research About Collection of Documents

概念图是一种知识以及知识之间关系的网络图形化表征。本研究中将图结构的非参数先验与潜在狄里克来分布结合在一起,构建了一个新的概率模型GLDA,用来学习文档中的概念图。并将此模型应用于维基百科中关于机器学习的文献集合加以验证,同时与高阶弹球分布模型(hPAM)与高阶潜在狄里克来分布模型(hLDA)进行了比较,对模型的优缺点进行了分析。

Conceptmap is a graphical representation of knowledge, and the relationship between knowledge. In the study, we combine the non- parametric prior with latent diriehlet allocation to create a new probability model called GraphLDA to learn the eonceptmap from the documents. We also apply the model to the collection of Machine Learning in Wikipedia, and compare the model with hierarchical Pachinko allocation model mad hierarchical latent Diriehlet allocation, then we analyse the model＇s advantages and disadvantages.