Research On Intelligent Dialogue System For Psychological Counseling

The intense competition and enormous pressure in modern society pose serious challenges to people’s mental health,leading to many psychological problems.People urgently need psychological counseling and therapy services.However,China is currently facing a serious shortage of psychological counselors,and many visitors often need to make appointments and wait in line to receive psychological counseling services.At the same time,traditional psychological counseling services are also limited by time and geography,and can no longer meet the psychological health needs of the general public.Based on the humancomputer interaction model and combining psychological counseling theory with task-oriented dialogue system technology,this project explores the construction of an intelligent dialogue system for psychological counseling.It provides visitors with 24/7 psychological counseling and therapy services through online chat.Compared to open-ended dialogue systems and general task-oriented dialogue systems,dialogue systems for psychological counseling require stricter topic control ability and empathy for specific psychological issues.In order to meet the requirements of topic control,this article proposes a psychological counseling dialogue system architecture based on cognitive behavioral therapy theory and the Rasa task-oriented dialogue system framework to guide the conversation according to a specific process.In order to endow the conversational system with empathy,this study explores the technological implementation of empathy perception and empathy response in human-computer interaction.Specifically,a text paraphrasing generation model,4SPG,with semantic and style enhancers is proposed for empathy response technology to address the problems of relying heavily on training samples and difficultly simultaneously learning semantic similarity and style diversity in the paraphrasing generation task.Experimental results on QQP-Pos,Para NMT,and Para SCI-ar Xiv datasets show that the introduction of semantic enhancers significantly improves the BLEU4 score from 23.00 to 23.59,and the addition of style enhancers decreases the self-BLEU value to 25.95.These results indicate that the 4SPG model can balance semantic accuracy and style diversity in generating paraphrased content.Additionally,this study proposes a facial expression recognition model,DAN,aimed at achieving empathy perception in psychological counseling tasks and improving the problem of category overlap in facial expression recognition tasks.Experimental results show that the DAN model achieves an accuracy of 62.09% on the Affect Net-8 dataset,65.69% on the Affect Net-7 dataset,and 89.70% on the RAF-DB dataset,indicating that the DAN model has good facial expression recognition performance.Based on the above research,this work integrates and builds an intelligent dialogue system for psychological counseling.Intents and entity categories for psychological counseling were formulated based on CBT theory.With the use of offline speech recognition technology,data desensitization techniques,and data labeling tool doccano,a Chinese psychological counseling dataset with intent and entity labeling was created.Furthermore,the Rasa dialogue system framework was used to sequentially build a natural language understanding module,dialogue strategy module,and dialogue action module based on the psychological counseling dataset.Empathy responses and an empathy perception model were embedded into the dialogue system.The results of the system demonstration indicate that the dialogue system constructed in this work has preliminarily achieved empathy-based psychological counseling.