AMFB: Attention based multimodal Factorized Bilinear Pooling for multimodal Fake News Detection

作者：

Highlights：

• Multimodality increases the accuracy and efficiency of a fake news detection system.

• An Attention Based Stacked Bi-directional Long Short Term Memory (ABS-BiLSTM) network captures textual information.

• An Attention Based Multilevel Convolution Neural Network–Recurrent Neural Network (ABM-CNN–RNN)captures the visual features.

• Multimodal Factorized Bilinear (MFB) pooling fuses the textual and visual features.

• Multilayer Perceptron (MLP) classifies the multimedia news post as fake or real.

摘要

•Multimodality increases the accuracy and efficiency of a fake news detection system.•An Attention Based Stacked Bi-directional Long Short Term Memory (ABS-BiLSTM) network captures textual information.•An Attention Based Multilevel Convolution Neural Network–Recurrent Neural Network (ABM-CNN–RNN)captures the visual features.•Multimodal Factorized Bilinear (MFB) pooling fuses the textual and visual features.•Multilayer Perceptron (MLP) classifies the multimedia news post as fake or real.

论文关键词：Multimodal fake news detection,Deep learning,Attention mechanism,Multimodal Feature fusion,Multimodal Factorized Bilinear Pooling

论文评审过程：Received 12 August 2020, Revised 11 March 2021, Accepted 9 June 2021, Available online 29 June 2021, Version of Record 7 July 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115412