中文摘要：

[目的／意义]在保障知识网络的整体性征的条件下，从原始知识网络中提取具有显著意义的层次知识网络，奠定基于关联频度提取层次知识网络的理论基础。[方法／过程]以无标度网络与分形几何作为基础理论支撑，在对关键词知识网络和标签知识网络关联频度分布进行分析的基础上，采用关联频度作为阈值提取层次知识网络。并对层次知识网络的无标度性和小世界效应两项网络整体性征进行验证。[结果／结论]知识网络的关联频度分布服从幂律分布。以关联频度为阈值提取的层次知识网络在节点度值分布和关联频度分布方面都保持了原始整体网络的无标度性。层次知识网络能够很好地保持原始网络所具有的小世界特征。基于关联频度提取的层次知识网络与原始知识网络等效。

英文摘要：

[ Purpose/significance ] Under the condition of guaranteeing the whole character of knowledge networks, this paper aims to extract a significant knowledge networks at level from the original knowledge networks, in order to estab- lish the theoretical basis of the knowledge network at level extracted by correlation frequency. [ Method/process] Under the basic theory support of scale-free network and fractal geometry, based on the analysis on the correlation frequency dis- tribution of keywords knowledge network and tags knowledge network, this article extracts the knowledge networks at level using the correlation frequency as the threshold. The scale-free and small-world effect of the knowledge networks at level are verified. [ Result/conclusion] The correlation frequency distributions of the knowledge networks match power-law dis- tribution. The knowledge networks at level, which take the correlation frequency as the threshold, keep the scale-free of whole original networks at nodes degree distribution and correlation frequency distribution. The knowledge networks at lev- el can keep the small-world characteristics of the original network well. The knowledge network at level based on correla- tion frequency is equivalent to the original knowledge network.