Bengali Fake News Detection: A Multi-Layered LSTM Ensemble Approach

In the contemporary digital landscape, combatting the The fight with the spread of fake news remains an utmost challenge in the contemporary digital industry that affects everything from political forces to public opinion. Since most available studies tend to focus on detecting misinformation through obscure newspapers or rumors on Facebook, our study diverges from this trend, primarily focusing on mainstream Bangladeshi journalism. Taking a critical review of credible sources to include top TVs and newspapers, our work propounds a new direction in the field of fake news detection. As we tap into their long-standing credibility and high readership counts, we have a very strong database to back our ensemble model. A good model would use a special type of recurrent neural network, known as Long Short-Term Memory (LSTM) networks, to be able to take care of the sequence data, which is inherent in news content. Our rigorous experiments and model iterations result in an accuracy rate of 82.43 %, which stands remarkably high vis-à-vis previous benchmarks. These give a high motivation for our approach toward accurately discerning which piece of news is authentic and which one is fabricated.