Fusion of Global and Local Features for Text Classification

Abstract

Text classification is an important problem in natural language processing. In this work, we propose a feature fusion model for document classification, which has excellent text representation ability. We first build a shared graph from training data sets to bring global information. Then we build individual graph for each document which maintain the word order. Finally, we get the document representation vector through attention layer. Extensive experiments show that our method performs well on three standard datasets, which illustrates the effectiveness of our model.