Advanced Attention-Enhanced BiLSTM-GRU Model for Real vs. Fake News Detection

INTRODUCTION: Social media has become one of the primary platforms for the rapid dissemination of both real and fake information. In today's digital society, individuals are often easily influenced by misleading content, making fake news a powerful tool for manipulation. A common strategy employed on social media is the intentional spread of false information with the aim of deceiving and misleading the public. OBJECTIVES: In some cases, this misinformation is so convincing that it significantly influences public opinion and behaviour, with long-term consequences. According to research, only 54% of people are capable of detecting deception without external assistance. To address this growing concern, the proposed model leverages Deep Learning techniques to identify fake news content in both Tamil and English languages. METHODS: The dataset for this study was sourced from an open-access repository on GitHub. After thorough pre-processing, the dataset was divided into training (80%) and testing (20%) subsets. Additionally, the proposed modified attention-based revamping weighted BiLSTM-GRU model is introduced and evaluated. RESULTS: The effectiveness of each model is measured using various performance metrics, including accuracy, precision, recall, F1-score, Bilingual Evaluation Understudy, Area under the ROC Curve, and the Receiver Operating Characteristic curve. Comparative analysis shows that the proposed model outperforms the existing methods across all evaluation parameters. CONCLUSION: A key advantage of the proposed approach is its capability to accurately detect fake text in both Tamil and English. Furthermore, the model demonstrates strong performance when tested on datasets generated using AI tools like ChatGPT, effectively identifying real and fake content with high precision.