Codes for my bachelor thesis: Multi-Modal Fusion for Fake News Detection.
The model was tested on two different datasets Weibo and Biendata and achieved quite good performance. In terms of the Weibo dataset, it surpassed att-RNN, EANN and MVAE, which were previously proposed, and achieved the accuracy of 89.3%, and achieved the accuracy of 98.3% offline in the new Biendata dataset. This paper also makes a deep analysis of the misclassified samples and puts forward some feasible plans for the future.
Biendata https://www.biendata.xyz/competition/falsenews_3/
Weibo dataset: https://pan.baidu.com/s/1q5z63vqMtoWZSmq-YFv57A password:jk0t