User personality prediction based on topic preference and sentiment analysis using LSTM model

Abstract

Abstract Based on the original text information, this paper converts the users' theme preferences and text sentiment features into attention information and combines different forms with the LSTM (Long Short-Term Memory) model to predict the personality characteristics of social network users. Finally, the experimental results of multiple groups’ show that the Attention-based LSTM model proposed in the paper can achieve better results than the currently popular methods in the recognition of user personality traits and that the model has good generalization ability.