Text Classification Based on Latent Semantic Indexing and Graph Embedding

Abstract

Text classification has become a key technology for processing and organizing large amounts of documents. In order to improve the performance of text classification, we propose a method based on latent semantic indexing and graph embedding technology. The method transforms original document space into semantic feature space by means of latent semantic indexing, and then graph embedding in this space is employed to construct an adjacent graph that reflects semantic similarity between documents and its complementary graph which reflects non-neighboring relationship. The proposed method tries to obtain the global structure and the local structure of documents. Using the $k$ nearest neighbor classifier on 20-Newsgroups dataset, the experimental results show that this method can improve the performance of text classification.