基于异构网络的学科交叉主题发现方法

Interdisciplinary Topics Discovery Method Based on Heterogeneous Networks

【目的/意义】通过异构网络的多模关系分析可以整合更多的主题关联关系,从而提高学科交叉主题识别的准确度。【方法/过程】系统调研了已有的2-模异构网络的社区识别方法,将2-模网络社区直接识别分为投影方法、非投影方法、扩展的多模网络分析和超网络分析,并对比这些已有方法在学科交叉主题识别中的优劣。在此基础上选择对应分析作为2-模网络社区识别方法,对情报学的学科交叉主题进行识别和分析。【结果/结论】对应分析作为一种2-模关系分析方法可直接识别异构网络的学科交叉主题,并保证更少的信息遗漏。

[ Purpose/significance] Multi-mode analysis to the heterogeneous network can integrate the relationship between more topics, so as to improve the accuracy of interdisciplinary topics discovery. [Method/process] This article systemly invested the current method of topic community recognition based on heterogeneous networks of 2 mode, and divided the community direct identification of 2-mode network into projection method, non-projection method, extended and super muhi-mode networks. Then we compared the pros and cons of these existing methods in interdisciplinary topic recognition. On this basis, we selected the corresponding analysis as interdisciplinary topics recognition method of 2-mode network and conducted an empirical analysis on and Library and Information Science. [Result/conclusion] Correspondence analysis as a muhimodal analysis method can directly discover the interdisciplinary topics of heterogeneous network, so to ensure less information missed.