A Multimodal Framework For Robust Propaganda Detection in Digital News Media

The proliferation of propaganda in multimedia content has grown to be a serious issue in the current digital landscape, as false information quickly shapes the public opinion. This study uses the N24News dataset, which blends text and image data, to present a novel method for identifying propaganda. Using convolutional neural networks (CNNs) for visual feature extraction and natural language processing (NLP) techniques for text analysis, the suggested structure allows for a comprehensive multimodel detection system. The technology efficiently combines textual and visual elements using sophisticated fusion techniques, making it possible to accurately identify a wide range of propaganda tactics across categories. The framework performs better on assessment metrics including F1-score, precision, and recall demonstrating its effectiveness in experimental results. These results demonstrate how complex propaganda patterns may be found using multimodal approaches in greater dependability and resilience for identifying inaccurate information. Future work will concentrate on extending this framework to include media forms such as video and audio to adapt to evolving propaganda techniques.