In this paper, we build a large language model about counseling conversations by using automatic conversation generation technology to simulate an automatic conversation between a counselor and a visitor. In order to optimize the dialogue recognition and mental health counseling service and construct a cognitive event map, a dialogue generation model that can introduce external knowledge is proposed to improve the accuracy of mental health counseling responses. Analyze the performance of the dialog generation model with the introduction of external knowledge on BLEU, ROUGE-L, FEQA, QuestEval, P, R and F1 metrics. To carry out experiential mental health counseling services for college freshmen and analyze the changes in positive psychological qualities of college freshmen before and after the intervention. After the implementation of experiential mental health counseling services, the positive psychological quality scores of the experimental class were higher than those of the control class, and reached statistically significant in the dimensions of “self-management and humility” and “spirituality and transcendence” and the total score. This suggests that the mental health counseling service using deep learning technology can be used as a way to assist the training of positive mental qualities of college students.
