Design And Implementation Of Weibo Rumor Detection System Based On Deep Learnin

With the continuous development of the Internet age,social networks have become one of the important channels for people to obtain information.However,there are a lot of false information on social media,which may mislead people and have bad influence,and even endanger social stability.Therefore,how to use computers to accurately detect false information in social media has become a hot research topic in recent years.At present,rumor detection on social media usually relies on manual methods,which requires a lot of manpower,material and financial resources,and requires experts to have a certain knowledge reserve.To solve this problem,scholars have devoted themselves to researching an automatic rumor detection model.Although the existing models improve the detection accuracy to a certain extent,they ignore the correlation between events.Therefore,this thesis proposes a detection method based on a deep learning,and designs and implements a Weibo rumor detection system.The main work of this thesis is as follows:(1)Aiming at the problem of single event source text information,a rumor detection model CROSS＿SC with cross-attention fusion of source text and comment text is proposed.First,pretrain the source text and comment text,and convert it into a data form that can be recognized and processed by a computer.Second,the source text uses the Text RNN model to extract feature vectors,while the comment text uses the Text CNN model to extract feature vectors.After that,the feature vectors of the two texts are fused with cross-attention to obtain two new feature vectors,and then the new feature vectors are spliced to obtain the representation vector of the event,which makes the event information more abundant and paves the way for the next step.Finally,the event representation vector is introduced into the fully connected layer to perform rumor detection as an independent whole.(2)Aiming at the fact that events in social networks do not exist in isolation,a rumor detection model CROSS＿SC＿GAT based on graph neural network is proposed.This method uses the abovementioned event representation vector as the node feature vector,so that the node feature vector has rich semantic information.At the same time,the edge of the event correlation graph,that is,the node adjacency matrix,is constructed according to whether there are the same commenters or forwarders between events,so as to improve the correlation between events and capture the connection between events more accurately.After that,the representation vector and adjacency matrix of the event are input into the graph attention network to output the classification result of the event node.(3)In order to facilitate user operation and result display,a prototype system was designed and implemented.Weibo rumor detection system adopts B/S architecture and uses Django framework.The system includes three modules: user login,rumor detection,and data expansion.The user login module uses the OAuth2 protocol to implement third-party login.After the user successfully logs in,the system will obtain their personal information and display it on the interface.The rumor detection module is the core module of this system,which embeds the above models into the system to provide users with the function of rumor detection.The data expansion module improves the accuracy of rumor detection by crawling Sina Weibo data and storing the data in the database for training models.