心理学
领域(数学分析)
计算机科学
社会心理学
认知心理学
透视图(图形)
人工智能
认知科学
感知
集合(抽象数据类型)
发展心理学
认识论
作者
Di Wu,Tenghao Zhang,Jiasen Wang
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-06-01
卷期号:: 1-7
标识
DOI:10.1109/ijcnn69854.2026.11728528
摘要
To address the limited adaptability of existing psychological dialogue data in the field of adolescent mental health support, and the insufficient guidance ability of current models, the Adolescent Psychological Support Dialogue Generation Model of Domain Data-Driven (APS-DM) is proposed. By extracting multi-level adolescent psychological topics, using topic-related keywords to filter and integrate data corpus, a context-based role-playing prompt is designed according to the characteristics of adolescent psychological development stages. It is intended to guide GPT-4 in performing multi-turn semantic reconstruction, thereby constructing an adolescent psychological support dialogue dataset (AdoPsyDialogue). Besides, a structured thinking-guided prompt is designed to constrain the response generation scenarios. A parameter-efficient fine-tuning method is used to optimize the base model, enhancing its ability to provide adolescent psychological support responses and guidance. Experimental results on the QsnTest dataset demonstrate that the APS-DM model significantly outperforms models such as ChatGLM2-6B and ChatGPT-3.5 in both automatic and manual metrics.
科研通智能强力驱动
Strongly Powered by AbleSci AI