Research And Implementation Of Personality Prediction Based On Social Network Text Information

As an important human psychological structure,personality characteristics play an important role in the prediction of human behavior and mental state,making personality analysis could be applied in many fields such as mental health,recommendation,and public opinion monitoring.However,traditional methods of personality analysis mostly rely on scale evaluation methods.Those methods consume a lot of manpower and time and lack timeliness,which greatly limits the applicability of personality analysis.With the rapid development of online social network in recent years,people generate a large amount of content containing their personality information on social networks every day.This provides a huge amount of effective data for personality analysis and makes personality analysis based on social network text information gain more and more academic attention.Although studies in this field have made great progress,there are some shortcomings such as data label sparsity,semantic features cannot fully reflect personality characteristics,and it is difficult to effectively use knowledge of other disciplines,which greatly limits the effect of personality analysis.In view of these problems,this paper studies and implements a semi-supervised personality analysis algorithm that can integrate computer science and psychology knowledge,so that the algorithm can obtain better analysis results with lower label rate.This subject has conducted the following research:1.Propose a network representation learning method personality2vec which is specifically for personality analysis.The method consists of a new network construction,a biased walk strategy and a modified skip-gram algorithm.The method could embed users into personality vectors and achieve good performance in predicting personality;2.Propose a personality prediction method PerKG which introduced knowledge graph into personality analysis.With a personality knowledge graph constructed based on the existing psychological knowledge and an improved graph embedding method,the method can obtain the knowledge of other disciplines and generate personality vectors that contain abundant personality information for users without labels and achieve good performance in predicting personality;3.In order to reduce the user's difficulty of using the algorithm and improve the efficiency of personality analysis,this paper design and implement a personality analysis prototype system to provide users with a visual operation platform and implement functions of data storage,personality analysis and results display.