Multimodal fusion Mongolian rumor detection method based on knowledge perception attention network

The invention discloses a multi-modal fusion Mongolian rumor detection method based on a knowledge perception attention network, and the method comprises the steps: extracting image features and text features of news contents, and carrying out the fusion of the image features and the text features to obtain a multi-modal fusion feature map; a graph attention network is introduced, and different weights are distributed to each node of the feature graph; extracting entities in the news content, aligning the entities with corresponding entities in the knowledge graph to obtain an entity sequence, and further obtaining external knowledge of the entities, namely an entity context sequence; taking the multi-modal fused news content, the entity and the entity context as input, and extracting semantic features of the entity and the entity context; constructing an attention network to obtain representation information of different subspaces at different positions; and connecting the multi-modal news content representa