Topology and semantic based topic dependency structure discovery

Abstract

As an important enabler in achieving the maximum potential of text data analysis, topic relationship dependency structure discovery is employed to effectively support the advanced text data analysis intelligent application. The proposed framework combines an analysis approach of complex network and the Latent Dirichlet Allocation (LDA) model for topic relationship network discovery. The approach is to identify topics of the text data based on the LDA and to discover the graphical semantic structure of the intrinsic association dependency between topics. This not only exploits the association dependency between topics but also leverages a series of upper-level semantic topics covered by the text data. The results of evaluation and experimental analysis show that the proposed method is effective and feasible. The results of the proposed work imply that the topics and relationships between them can be detected by this approach. It also provides complete semantic interpretation.