Abstract: We constructed a psychological assessment model based on text emotional characteristics. Firstly, acording to the emotional polarity and part of speech of words, word features were designed, and each word in the text was mapped into an emotional word vector, which was then used as the input of convolutional neural network. Attention mechanism was added to optimize the output results, the text vector representation containing emotional features was obtained. Secondly, the Bayesian regularization algorithm was used to optimize the weights, control and balance the fitting degree of the neural network, and improve the generalization ability of the BP neural network algorithm. Finally, the text vector was used as the input of Bayesian regularization neural network to predict the students’ psychological state. Compared with the results of psychological assessment, the experimental results show that the effect of the model is ideal.

Key words: emotional characteristic, convolutional neural network,  , Bayesian regularization