Research On News Recommendation Method Based On Social Network

With the rapid growth of network information,because of information overload,people cannot obtain satisfactory or interesting news information from a large amount of news information.Personalized recommendation is the main method to solve the problem of information overload.By analyzing the preferences of different users and considering the mutual influence of each other,the corresponding users are recommended to the content they are interested in.Scholars at home and abroad have conducted extensive research on recommendation algorithms.Among them,the entity representation of converged news texts,the dynamic changes of user interest points,and the quantitative influence of influence in social networks are the main factors affecting the recommendation effect.In view of the above problems,this paper studies the news recommendation method from two aspects:feature quantification based on knowledge graph and social network structure information.In the aspect of feature quantification based on knowledge graph,this paper proposes a feature quantification method of hybrid knowledge graph embedding and topic word embedding for the problem that news contains a large number of entity relationships and user interest dynamic changes,aiming at the characteristics of different perspectives of news text.The method can not only integrate the information between the entities in the news text,but also enhance the emphasis of the text representation in the subject direction,and can quantify the influence of different browsing records on the user's interest in the time series through the attention mechanism,thereby obtaining more rich features.The text,the vector representation of the user.Experimental results based on our proposed algorithm show that we could get well entity representation of news texts,dynamic capture of user points of interest,and improve the accuracy of prediction.In terms of the social network structure information,aiming at the interaction between users groups,an improved sampling method for social network structure information is proposed to quantify the degree of influence between users.By designing the sampling strategy based on the number of interactions,the method obtains the neighbor users in the social network,and obtains the influence weight of the neighbor users on the target users through the interaction relationship and the content information.The experimental results show that the method can obtain the structural information between users in the social network,effectively solve the problem of the degree of influence in the social network and improve the accuracy of the recommendation.