Typical Character Recognition And Tracking In Network Community

Users in the network platform have become the main content producers.A certain kind of user has a significant guiding effect on the network social thought and public opinion from the positive or negative,thus deriving a variety of typical characters.These typical characters play an important role in the governance of cyberspace security,and the technology of typical character recognition and tracking is the main means of governance implementation.Existing typical character recognition and tracking technology is mainly applied to single typical character,which is not suitable for recognizing and tracking multi-kinds of typical character in the network community.In order to solve this problem,research design and implement a multi-type typical character recognition model based on multi-dimensional feature analysis.In view of the sparse attributes of users in the network community,the recognition model designs 17 features in terms of publication content,user behavior and statistical data.Among them,content similarity based on word vectors and content aggregation based on Meanshift are key features to solve the problem of short and spoken content published by users in the network community.On the other hand,research also design and implement a multi-dimensional typical character tracking model of network community,which integrates interactive relationship,trajectory,user evaluation and other tracking technologies.The tracking model uses the relationship generation algorithm based on link detection to construct user interaction.At the same time,the influence index based on the influence logic and forgetting curve of the network community and the typical index based on the recognition feature analysis are designed to quantify the user characteristics of the network community.Combining the time traj ectory of the two indices and the user relationship graph,the multi-dimensional typical character tracking in the network community is realized.In order to better enable managers to use the recognition model and tracking model,a typical character recognition and tracking system in the network community is designed and implemented.It provides the function from data acquisition to visualization result presentation for typical character recognition and tracking.In order to verify the recognition effect of various typical character recognition models in network community,the subj ect selected two typical characters of Navy and sensitive characters in network community for comparative experiment.Experiments show that the recognition model are better than the reference recognition model.In order to validate the multi-dimensional typical characters tracking model in the network community,the model were compared and tested in four latitudes:relational,historic,predictive and evaluative.Experiments show that the tracking dimension of this tracking model is higher than that of other tracking technologies.