Research On Rumor Feature Extraction And Recognition Algorithms On Microblog

With the rapid development of social media information,the emerging media platforms represented by Sina Weibo,Vibrato,and Fast Hands have become the main platforms for people to obtain information,share information,and disseminate information.However,when these social media platforms bring the convenience and richness of information access,it also brings convenience and extensiveness to the dissemination of false information.The spread of rumors in social media has severely hampered access to reliable information and may lead to huge economic losses or serious public panic in some emergencies.The traditional rumor detection method regards rumors as a classification problem and is devoted to extracting the social features of rumors,but these methods ignore the semantic features of the current huge official rumor data sets.In recent years,deep learning has achieved certain achievements in various fields,and word vector representation has shown good performance in applications,which has inspired the related research in the field of rumor detection.In order to solve the problem of the single feature in the current rumor detection,this paper starts with the traditional rumor detection method and the method based on deep learning,and uses the officially rumor dataset to extract the social features of the rumor to realize the early rumors detection.At the same time,the combination of word embedding representation and neural network model is used to improve the semantic feature detection of rumors,and to make false news recognition on social media in all aspects and levels.The main research work includes:1.Design a rumor detection algorithm based on the topic and prevention model.Firstly,the topics of the official rumor data set are extracted,and the statistical features of the special symbols of the rumor subset are extracted by subject classification.The sample is then compared to the Weibo official rumor subset,and its value is entered as a statistical feature into supervised machine learning.Finally,the experimental results show that the proposed classification features effectively improve the accuracy of detecting rumors.At the same time,it also has a good effect in the early detection of rumors.2.Design a deep neural network model based on multi-feature fusion to learn rumor detection of event-level news.Event-level news refers to the original Weibo and related comment with forwarding content.Event-level news can capture more semantic information and show the way that news transmit.The event-level news are divided into sub-events,each sub-event extracts semantic information,and the vector representation of the sub-event is input as an input vector into the neural network to learn the semantic features and potential time-series features of the sub-event.The neural network output and social features are combined to make classification results to improve the expression ability of the model.The experimental results show that the correct rate of the model is about 9%higher than that of the traditional rumor detection algorithm,which is about 4%higher than that of the message-level news rumor detection algorithm,and more effective than the single-semantic feature rumor detection mode.