An Analysis of Multimodal Approaches for Fake News Detection

Fake news refers to factually incorrect reports that are spreading at an increasing rate, negatively affecting public opinion, media credibility, and social stability. If not screened and prevented promptly, fake news can influence the publics judgment of the facts. More seriously, it may affect the direction of international public opinion and the credibility of the national media. Thus, detecting and managing fake news is the focus of public opinion issues. To solve the problem, the multimodal fake news detection method came into being, which is a detection method that extracts and fuses features from multiple modalities of news (such as text, pictures, and audio), and then evaluates the results. This paper provides an overview of current relevant news detection techniques and analyzes the advantages of multimodal approaches in the current field of fake news detection, as well as the problems that may be encountered. The results indicate that multimodal fake news detection can improve detection accuracy by fusing multiple data modalities, but still faces a series of challenges such as dataset diversity and model interpretability, with great potential for future development.