A Graph-Structure-Based Method for Chinese Document Representation towards Clustering Application

In this paper, we propose a graph-structure-based method to represent knowledge for Chinese document clustering. First, we introduce a new knowledge representation method called Graph Space Model (GSM) to convert each document to a graph structure, and then we adopt Maximum Common Subgraph (MCS) to compute the similarities between any two graph structures, which can be further used for document clustering. The results show that the GSM approach can outperform VSM method in representing capability of Chinese documents.