A Deep Learning Multimodal Framework for Fake News Detection

The swift spread of fake news on social media platforms presents significant challenges to the society, necessitating the development of a more efficient model for fake news detection. Existing fake news detection methods primarily focus on linguistic and compositional characteristics, which may not be able to differentiate between various forms of fake news and impede effective detection. This paper proposes an innovative approach to address such challenges. It introduces a comprehensive framework for detecting fake news, leveraging advanced multimodal techniques to analyze multilingual text and visual data. The proposed framework employs Natural Language Processing (NLP) for text preprocessing, the DeepL translator for language consistency, and vectorization for feature extraction. For the detection models, Long Short-Term Memory (LSTM) networks are used for sequential text analysis, while the Contrastive Language-Image Pretraining (CLIP) model is utilized for image analysis to be performed. The combined features are then processed through a decision-making layer for the news to be classified as real or fake. Results demonstrate the model's high efficacy, with an accuracy of 99.22% for text and 93.12% for text and images, outperforming the existing state-of-the-art techniques.