Research On News Recommendation System Based On Keywords Expansion

With the rapid development of the Internet and large data,people increasingly rely on the network to obtain information.News,as an important medium for people to obtain information,is more and more popular with users.How to discover news that users are interested in in in the mass of information has become a hot research topic.Starting from the basic text classification,because of the short content of news text,traditional text processing methods often cause the lack of semantic information when analyzing news text,which is one of the bottlenecks restricting the performance of short text classification.This paper uses external corpus to train Word2 Vec model,expands the keywords extracted by traditional keyword extraction algorithm based on external semantic information,and studies the feasibility of expanding short text keywords based on external semantic information according to different expansion ?methods.Finally,it uses K-Nearest neighbor classification algorithm(K-Nearest Neighbor,KNN),a comprehensive short text classification algorithm based on Keyword Extension-K-Nearest Neighbor is proposed(CAKE-KNN).Finally,the accuracy and efficiency of the method proposed in this paper are verified on the experimental dataset,which is close to the current mainstream text classification algorithm.Further,this paper designs a news recommendation system(Recommendation Systems based on Keyword Expansion,KERS),based on the analysis of users' news reading behavior and insight into the characteristics of news communication,proposes star rating methods of users' interest level and filtering methods of comment violation content(advertisement,yellow,violence,reactionary speech,etc.).Combining with the text classification algorithm proposed,this paper formulates the recommended strategies,and makes experimental determination of the main parameters,and other,The comparison results of the recommendation system show that the recommendation strategies proposed in this paper can meet the user personalized news recommendation task to a certain extent and alleviate the cold start problem.