An application study on multimodal fake news detection based on Albert-ResNet50 Model

In today's interconnected world, where individuals can create and receive information freely, the proliferation of fake news has become a significant issue. This type of false information frequently appears in areas such as business or politics, and its widespread dissemination on the internet can disrupt the normal social order and create a biased net- work atmosphere, ultimately leading to the destruction of the normal network environment. The evolution of fake news, from early plain text to complex images and texts, has made its detection more difficult. To address this, we propose an Albert ResNet50 hybrid deep neural net- work model that combines implicit features of both text and images for detecting multimodal fake news. We tested our model on three fake news datasets, and the results showed an accuracy rate of 90.51%, 79.87%, and 92.93%, respectively. Compared to traditional models that only use text data, our multimodal model can better identify fake news.