| The innovation of information technology has driven the rapid development of social networks.People can get thousands of information from all over the world without leaving their homes.However,the convenience of information transmission also leads to the rampant spread of rumors,which not only affects people’s normal life,but even disturbs social order.Therefore,rumors must be discovered and clarified as soon as possible to reduce their harm.However,the efficiency of manual rumor detection is low,and the identification work is often lag,so the automatic rumor detection research is of great significance.In this paper,a rumor detection model based on multimodal information is proposed on the basis of summarizing the shortcomings of the existing work,and its effectiveness and superiority are verified by experiments.Meanwhile,a rumor detection system is established according to the model.The main work of this paper is as follows:1)From the three perspectives of text content,image information and social context,the model uses deep learning method to extract text and image features.At the same time,the easily ignored hidden state is also used as the basis for rumor detection.Then combine these features with hand-extracted social features at multiple levels,finally,the fused features are used for rumor detection.In the process of feature extraction,attention mechanism is used to focus on key features.Meanwhile,multi-level feature fusion method is used to make full use of the information relation among multiple modes.In addition,experiments were carried out on public rumor data sets,and the experimental results show that the rumor detection model based on multi-modal information has higher recognition accuracy.2)A rumor detection system is designed and implemented.The main working modules of the system include crawler module,data preprocessing module,algorithm module,user level strategy and data storage.Crawler module is responsible for crawling the user’s personal information and historical microblog data according to the microblog account;The data preprocessing module is responsible for manually extracting social features;Algorithm module is responsible for switching and providing algorithm model services;The user level strategy is responsible for finding and specially managing microblog users who are highly likely to publish rumors;The data store is responsible for persisting the necessary data generated during system operation.The system can not only judge rumors on a piece of information,but also detect rumors in advance by monitoring micro-blogs of low-trust users,so as to play a better auxiliary role. |
