Multi-task rumor detection model and method based on dynamic network

According to the multi-task rumor detection model and method based on the dynamic network, dynamic information can be captured, the detection effect can be improved, captured dynamic space features and time sequence text features which change along with time are organically combined by taking time steps as basic sharing units, and the detection efficiency is improved. And an attention mechanism is used to update output features, so that the multi-task rumor detection model and method based on the dynamic network are formed. The method comprises the following steps: firstly, constructing rumor event propagation diagrams at different moments, modeling the propagation diagrams at different moments by using a dynamic graph convolutional network, and capturing dynamic structure features of rumor events; the time sequence sharing layer is used as a sharing network to carry out feature interaction on event dynamic structure features and time sequence text features at each same time step, and finally an attention mec