Dynamic dual graph networks for textbook question answering

Yaxian Wang, Jun Liu, Jie Ma, Hongwei Zeng, Lingling Zhang, Junjun Li

Abstract: Highlights•We propose a novel model Dynamic Dual Graph Networks for textbook question answering task, which includes a Diagram Graph Network (DGN) and a Textual Graph Network (TGN) to help understand the complex diagram and capture useful textual information respectively.•We propose a question-guided multi-step graph reasoning method to update the two graph networks dynamically through the top-K information propagation mechanism in every step.•Experimental results show that our proposed model outperforms the baselines on the TQA dataset, and extensive ablation studies are also conducted to validate the effectiveness of our proposed model.