Unraveling the Tangle of Disinformation: A Multimodal Approach for Fake News Identification on Social Media

Abstract

The growth of interactive and multimedia content on the Internet has made it an essential news source for people worldwide. Social media is a platform for sharing information and facilitates the spread of fake news. The dissemination of disinformation on social media has a significant impact on society. Conventional methods used in the identification of fake news often struggle to analyze textual, visual, and combined aspects of news shared on social media. Therefore, we propose the Multimodal Approach for Fake News Identification (MuAFaNI), which uses a combined representation of text and images to assess news authenticity as fake or real. MuAFaNI uses the RoBERTa language model for text analysis and ResNet-50 for image analysis. Experiments on two prominent social media datasets, Twitter and Weibo, showed that MuAFaNI performed better than state-of-the-art fake news techniques in terms of accuracy, precision, recall and F1 score.