False news detection method and system based on cross-modal enhancement

The invention relates to a false news detection method and system based on cross-modal enhancement, and the method comprises the steps: obtaining an image and a text in news, and obtaining corresponding image features, text features and features connected with the image features and the text features; obtaining an image-text cross attention feature and a text-image cross attention feature based on a cross attention mechanism; taking the image features, the text features, the features connecting the image features and the text features, the image-text cross attention features and the text-image cross attention features as corresponding modals, and utilizing a classifier to obtain inference probabilities corresponding to the modals; and connecting all the obtained inference probabilities, obtaining predicted news category probability distribution based on the classifier, and determining a final false news detection result by using a loss function. 本发明涉及基于跨模态增强的虚假新闻检测方法及系统，获取新闻中的图像和文本，得到相应的图像特征、文本特征和两者相连接的特征；基于交