Research And Implementation Of User Interest Analysis Method Based On Emotional Cognition And Personalized Features

With the development of online social media,research on user interest and development of relating application on this basis are of great significance.Because of the personalized information contained in Weibo users' profiles,and emotional or sentiment information in the short texts of microblogs published by users,the microblog texts are analyzed and mined to complete their emotional cognition and personality analysis.It is meaningful to extract the interest and improve the personalized recommendation on this basis.This thesis constructs the user energy model,improves the LDA-based interest extraction model,and improves the personalized recommendation.The main tasks are as follows:1)According to the users' emotional semantic analysis,and employing the Big-Five psychological personality method,this thesis constructs the user energy model.Through the algorithm comparison,the model of the users' positive and negative energy is constructed,the users' text is converted into vector input to the SVM for training,the new corpus is analyzed,and the new user energy polarity is predicted.2)Construct a personalized interest extraction algorithm based on improved LDA to explore an effective method for user interest tag extraction.The word pairs are processed by the LTP dependency parsing tool,and the interest tag extraction is performed via the LDA algorithm.By comparing with BTM algorithm,multiple benchmarks are used for evaluation.Experimental results show that the interest tags extracted by the improved algorithm are more accurate.3)A recommendation method based on user energy model and interest tags is proposed,which can recommend users with similar energy in the group.The collaborative filtering algorithm is used to expand the interest tags,and Doc2 vec is used to vectorize the text to achieve similar recommendations among users.Experimental results show that the proposed model has certain value for user energy prediction;the improved LDA algorithm can reduce the confusion and time complexity,meanwhile it can improve the accuracy;the proposed recommendation algorithm can recommend similar users.