The anonymity veil and virality related to the social platforms have been tactfully used to deceive and mislead society through creating or disseminating fake news. The content and meta-data associated with the information posted can serve as evidence to debunk its truth value. A hybrid deep neural model is proposed which combines a high-level representation of textual features learned using hierarchical attention network and meta-data features learnt using convolutional neural network for the multi-label classification of fake news in benchmark LIAR dataset. The experiments on this model-level fusion network achieves a performance accuracy of 43.8% and outperforms the existing state-of-the-art.
Volume 12 | Issue 6
Pages: 2132-2145
DOI: 10.5373/JARDCS/V12I6/S20201175