Text Classification with Intra-layer and Inter-layer Graph Attention Networks

Anonymous

Abstract: Text classification is a classical task in the field of natural language processing. Recently, graph neural networks (GNNs) have received considerable attention and made great breakthroughs on this task. However, current GNN-based methods neither fully utilize edge information nor obtain higher-order interactions of words. To address these problems, we propose the Intra-layer and Inter-layer Graph ATtention networks (IIGAT) to obtain the higher-order interactions of word nodes and construct multi-dimensional edges between word nodes in the intra-layer GAT to enrich the semantic information of words. Extensive experiments on four benchmark datasets demonstrate the effectiveness of our methods on the text classification task.

Paper Type: long

Research Area: Information Retrieval and Text Mining