Graph Fusion Network for Text Classification

作者：

Highlights：

• We transform external knowledge into structural information to build better graphs.

• Our model can perform inference for new documents without rebuilding the whole graph.

• We propose a unified Graph Fusion Network (GFN) for text classification.

• Our model achieves the best performance.

摘要

•We transform external knowledge into structural information to build better graphs.•Our model can perform inference for new documents without rebuilding the whole graph.•We propose a unified Graph Fusion Network (GFN) for text classification.•Our model achieves the best performance.

论文关键词：Graph Neural Networks,Text classification,External knowledge,Graph fusion

论文评审过程：Received 22 June 2021, Revised 22 August 2021, Accepted 28 October 2021, Available online 10 November 2021, Version of Record 29 December 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.107659