Research On Network Rumor Detection Based On Deep Learning

With the rapid development of the Internet,social media has greatly improved the speed and scope of information dissemination.Social media,however,the lack of regulation characteristics of not restricted also fueled rumors spread rapidly,Internet rumors spread to the public for real and effective information caused serious interference,light person influence perception judgment and reality,people have a negative impact on personal life,the person that weigh can cause serious social panic,affect the harmonious and stable development of the society.Therefore,accurate detection of network rumors has very important research value and social significance for preventing the widespread spread of rumors and preventing the harm brought by network rumors.Faced with massive social media information,traditional Internet rumor detection methods rely on manual construction of feature training discriminant classifier.This method is not only time-consuming and laborious,but also has problems such as lag in recognition and incomplete coverage.With the development of artificial intelligence technology,deep learning relies on the advantage of automatically learning deep features in sample data without relying on artificial construction of features,and has achieved good research results in various fields.So rumors detection method based on traditional network shortcomings and the advantages of deep learning,this paper mainly revolves around a variety of deep learning model of Internet rumors detection technology research,through the use of different structure of deep learning model to in-depth study of Internet rumors,further improve the recognition accuracy of Internet rumors detection system.The main work and innovation of this paper are as follows:(1)The background and significance of rumor detection research are studied,and the history and current situation of relevant research are summarized from three aspects of rumor research,rumor propagation and rumor detection.(2)Study the characteristics and text representation of network rumors,and learn the basic theory of network rumors.Firstly,this paper studies the characteristics of Internet rumors in detail from three aspects: text characteristics,user characteristics and transmission characteristics.Then it summarizes the text representation methods of Internet rumors,including Boolean model,vector space model and word embedding model,to provide theoretical basis for subsequent experiments.(3)The related technologies of rumor detection are studied,including traditional machine learning algorithms,including support vector machine and artificial neural network.Then,deep learning algorithms,including convolutional neural network and cyclic neural network,are introduced.Finally,the attention mechanism is introduced to provide theoretical support for the subsequent experimental comparison.(4)Designed a kind of Internet rumors detection model based on hierarchical attention,the model adopts the structure of hierarchical attention,could be divided into words Internet rumors event hierarchy,hierarchy and event hierarchy,and integrating the statistical characteristics of Internet rumors,the fusion of the hierarchical temporal characteristics of the event and deep semantic information,obtained the accurate representation of Internet rumors.Through rumor detection experiments and ablation experiments designed on sina weibo and Twitter rumor data sets,the model is verified to have a good rumor detection effect,and the validity of each part of the model is also verified.Finally,an early rumor detection experiment was designed to prove the effectiveness of the proposed model in early rumor detection.(5)This paper proposes a network rumor detection model based on generative adversarial network.The model uses a serial-to-sequence model to build a generator to learn the distribution characteristics of real samples,and generates pseudo-samples close to the real samples through reverse generation to suppress the generation distortion.The model adopts the idea of adversarial learning,and uses the network rumor detection model based on hierarchical attention as the discriminator,learns the deep abstract features of network rumor through adversarial training,so as to improve the recognition accuracy and robustness of the model.Through a large number of comparative experiments on rumor data sets,the validity and reliability of this model in rumor detection task and early rumor detection task are verified.