MULTIMODAL FAKE NEWS DETECTION

Abstract

Fake news detection is a critical task in today's information landscape, with the proliferation of misinformation across various media platforms. This study explores the use of multi-modal data to improve fake news detection, particularly focusing on the relationship between image and text modalities. We hypothesize that leveraging the similarity between image and text representations can enhance the accuracy of fake news detection systems. Using a multi-modal framework, we jointly learn representations from both image and text data, while simultaneously optimizing for fake news detection. Our approach aims to capture the complementary information present in both modalities, thereby improving the overall performance of the detection system. We investigate various architectures and loss functions tailored to the multi-modal nature of the problem. To evaluate the effectiveness of our proposed approach, we conduct experiments on benchmark datasets for fake news detection. Our results demonstrate that incorporating the relationship between image and text similarity leads to significant improvements in detection accuracy compared to single-modal approaches. Furthermore, we analyze the learned representations to gain insights into the underlying factors driving fake news propagation across different media formats. Overall, our study sheds light on the potential of multi-task learning for enhancing fake news detection systems, particularly by exploiting the relationship between image and text modalities. By leveraging multi-modal information, we can develop more robust and effective tools for combating misinformation in online platforms.