Research Of Virtual Identity Association In Multi-social Network Based On Deep Learning

At present,social network has penetrated into all aspects of people's lives,and become an important social media.There are a variety of social networking platforms today,such as Sina-Blog,Twitter,Zhihu and so on.Different social networks offer different services.Sina-Blog can be used to record personal feelings at any time,sense the feeling of strangers and make friends.By Zhihu the user can solve their own problems,the platform includes expertise in areas of user interest.Users in order to better enjoy the different social network services,often have multiple social networks account.This phenomenon has led to the decentralization of user information,making the analysis of social networks have encountered a bottleneck.The significance of research on virtual identity association is to integrate the identity information of different social network platforms belonging to the same real user.The methods currently used in this area can be summarized in tow ways:based on username association and text-based association.However,most users do not usually register the same or similar username on different platforms,so the method based on username association is uncertain.The method of text association is often to extract a large number of features to construct user profile for analysis,but the traditional feature extraction is a cumbersome process,and the features extracted by the traditional method is not necessarily able to have a strong expressive force.The method of Deep Learning is not require annotation data,and through the self-learning method can obtain strong expression features.This paper,the method of Deep learning for text-based association in-depth study,the main work is as follows:1,By analyzing the commonly used techniques of text association,the drawbacks of traditional artificial feature extraction are obtained,thus giving the reason for using the Deep learning method.2,Deep Learning frameworks such as CNN,RNN and LSTM are introduced,which will lead to the problems that should be solved in Deep Learning Model.3,Taking into account the contents are published by users in social network need to convert to vector.In order to avoid ignoring the semantic information in the text,a model named CWSR is proposed to calculate the Chinese semantic word vector representation in this paper.4,The Chinese semantic word vector is obtained by CWSR model,and then combined with the user's personal information,this paper presents a model named Indetity2 Vec for computing user identity vector representation5.In this paper,the Keras Deep Learning package is used to build the model,and the model is validated and compared with the commonly used techniques.