Financial Fake News Detection via Context-Aware Embedding and Sequential Representation using Cross-Joint Networks

Fake news is a pervasive phenomenon over online social media platforms. The computation detection of fake news is a well-known problem. Financial fake news has immense potential to mislead the financial domain, but it has drawn less attention from researchers. In this paper, we introduce a new approach for financial fake news detection. Starting with preparing financial fake news datasets using topic modelling and Transformer-based techniques. We introduce two cross-joint networks. The first cross-joint network - CAEN is used to give context-aware linguistic and financial embeddings. The combined embedding outcome is passed to the second cross-joint network - CSRN which is employed to produce contextual sequential representation to detect financial fake news. We evaluate proposed approach on two benchmark datasets and it shows impressive results. We have compared the evaluation results of the proposed approach with relevant state-of-the-art and numerous neural network-based baseline methods, and it is noticed that it also shows remarkably better.