News Recommendation System Based On Emotion Analysis And Knowledge Map

With the popularization of the Internet,huge user groups have generated massive amounts of data,but it is difficult for the public to effectively receive and process massive amounts of data,resulting in the problem of information overload.The recommendation system is an important means to solve this problem.At present,most recommendation systems rely on user portraits,but many news sites do not include user systems,so user portraits cannot be generated.In the end,these recommendation systems are not well used in the field of news.In order to solve this problem,based on the content-based text similarity recommendation algorithm,this thesis has carried out in-depth research work,proposed a series of improved algorithms,and developed a set of recommendation systems in the news domain based on these improved algorithms.The main work is as follows:(1)An emotion classification algorithm of news text based on multi model fusion is proposed.Firstly,this thesis analyzes the existing sentiment classification algorithms,at the same time expands the existing Chinese sentiment dictionary,and finally transforms the differential evolution algorithm to form a multi model fusion emotion classification algorithm.Through experimental analysis,the multi model fusion sentiment classification algorithm on a specific data set is superior to the fused classification algorithm in the classification accuracy and F1 evaluation indicators,and in the sentiment dictionary-based sentiment classification method,the expanded sentiment dictionary is also better than the existing Chinese sentiment dictionary in the classification accuracy and F1 evaluation indicators.(2)A recommendation algorithm based on text similarity based on knowledge graphs is proposed.In this thesis,based on the requirements of news reports,named entity recognition technology is used to obtain entity information,and then the knowledge map is used to expand the entity information.Finally,the expanded knowledge is used as a text feature to apply to the recommendation algorithm based on text similarity.A text similarity recommendation algorithm based on knowledge graphs is formed.Through experimental analysis,the text similarity based recommendation algorithm based on knowledge map is higher than the original text similarity based recommendation algorithm in terms of coverage and diversity.(3)Design and implement a recommendation system for the news domain.Using the emotion classification algorithm of news text based on multi model fusion and the text similarity recommendation algorithm based on knowledge graphs,on this basis,combined with the engineering background to build a recommendation system,through the micro-service technology,the recommendation system has the characteristics of high cohesion and low coupling.At the same time,the system provides a variety of solutions for data synchronization between systems.Finally,to facilitate system operation and maintenance,components such as tracking and system performance monitoring are added.