Modeling Content Relationship In Social Networks And Its Visualization System

With the continuous development of the Internet,social networking platforms have also flourished.Social networks are the main platform for information acquisition and dissemination.Users can rely on the platform to freely express their opinions and exchanges on various events.However,these massive amounts of information are intricately intertwined and cannot be intuitively obtained for potential value.Therefore,study the content of social networks.has great significance.Through research content and then analysis of user behavior,public opinion monitoring can be effectively carried out,and personalized recommendations can be provided to users.The emergence of hot events is accompanied by a large number of participants,and the opinions expressed by users are complex and diverse.If public opinion guidance cannot be made in time,serious losses may result.The research in this thesis is based on the content of social networks.Aiming at the problem that massive tweet information is difficult to process and hot events are not easy to find,firstly,we analyze the propagation law of tweets in the time dimension based on statistical methods,and analyze hot topics in the mass tweets content,mining hot events that interest users,and analyze The duration of hot events and the reverse verification of the propagation law of tweets;for the difficulty in discovering the hidden relationship between the content of tweets,a content relationship network model based on hot events is generated and displayed in the form of a relationship graph,which can be more intuitive Mining the influential tweets and users;for the problem of failing to effectively classify the opinions expressed by users’ tweets,propose an opinion classification model for hot topics,which can effectively classify the content of tweets.Finally,the above-mentioned work content is designed and realized a visualization system,through which public opinion response and management and control measures can be made to hot events in a timely manner.The main research work of this thesis is as follows:(1)A hot event discovery method based on the maximum keyword popularity of the day is proposed.Use the unsupervised word segmentation algorithm based on the prefix dictionary to process the word segmentation of the content of the tweets,and calculate the word segmentation results to further improve the stop word dictionary.The stop word dictionary and the largest keyword popularity of the day are used to double filter the word segmentation results,and the experimental results are in time The event heat map is generated in the dimension,and the experimental results are consistent with the actual events through verification.(2)Propose a BERT-CNN model for the classification of social data text opinions.Due to the weak ability of the BERT model to extract local feature information of the text,the multi-convolution kernel structure of the CNN network can well make up for the shortcomings of BERT.Therefore,the CNN network is introduced at the end of the word embedding layer BERT to build a BERT-CNN model,and compare experiments with multiple classification models under the real data set to verify that the BERT-CNN model performs best under this data set.(3)A generative model of content relational network is proposed.Use tweets as visual nodes,tweet relationships and keyword relationships as node connections,calculate text content similarity between tweets based on keyword relationships,and use content similarity to measure the strong content relationship between tweets It is weak,and it is displayed on the visual image,which provides an idea for the visualization of the content relationship network.(4)Independently designed and developed a B/S structure front-end separation visualization system.Designed a database table for Twitter social data,and optimized the data table to greatly reduce retrieval time.The front-end provides a responsive page,and secondary development is carried out on the basis of Echarts to increase the visualization results to support human-computer interaction.The system provides many visual interfaces,integrates the research results of many students in the laboratory,and provides visual services.