A Multi-modal Adversarial Framework for Fake News Detection under Uncertain Missing Modalities
The current fake news detection method have shifted from the single-modal data to the multi-modal data.To deal with the missing modality problem that may exist in real-world scenarios,this paper proposes a new fake news detection framework with a modality discriminator for the multi-modal feature learning.It learns the transfer-ring features between different modality combinations in the process of adversarial training with the feature genera-tor.Transferring features enable fake news detection under uncertain missing modalities.Experiments on real data-sets demonstrate that the proposed framework outperforms state-of-the-art multi-modal fake news detection methods under uncertain missing modalities.
万方数据
维普
