文章摘要
肖 吴1,2 ,王 曙3,4 ,刘雨平5 ,叶 鹏5*.基于大语言模型的高校学生焦虑心理分析[J].海南师范大学学报自科版,2025,38(3):289-295
基于大语言模型的高校学生焦虑心理分析
Psychological Analysis of College Students' Anxiety Based on Large Language Model
  
DOI:10.12051/j.issn.1674-4942.2025.03.005
中文关键词: 焦虑心理  高校学生  大语言模型  GPT模型  BERT模型
英文关键词: anxiety psychology  college student  large language model  GPT model  BERT model
基金项目:江苏高校哲学社会科学研究一般项目(2022SJYB2125);教育部产学合作协同育人项目(230804691081731)
作者单位
肖 吴1,2 ,王 曙3,4 ,刘雨平5 ,叶 鹏5* 1. 扬州大学 人力资源处江苏 扬州 225009 2. 东北财经大学 国民经济工程实验室辽宁 大连 116025 3. 中国科学院 地理科学与资源研究所资源与环境信息系统国家重点实验室北京 100101 4. 江苏省地理信息资源开发与利用协同创新中心江苏 南京 210023 5. 扬州大学 土木与交通学院江苏 扬州 225127 
摘要点击次数: 654
全文下载次数: 192
中文摘要:
      针对社交媒体平台上的高校学生评论开展焦虑心理分析有助于及时探测高校学生的 心理健康问题。然而,由于庞大的数据规模以及较高的发布频率,基于社交媒体数据分析高校学 生的焦虑心理仍然面临挑战。大语言模型正在引领人工智能领域发展进入新纪元,并且在自然语 言的对话、理解和推理方面展现出优异性能。基于大语言模型进行高校学生焦虑心理分析,并比 较GPT家族和BERT家族中不同大语言模型微调后适用于高校学生焦虑心理分析的有效性。结果 表明,GPT-3.5 Turbo 0125 和 RoBERTa-base 分别是其模型家族中性能最优的 2 个模型,GPT-3.5 Turbo 0125整体性能更佳,其精确率达到96.27%。总体上,GPT和BERT家族的大语言模型在高校 学生焦虑心理分析中都展现出强大潜力,为生成式人工智能助力高校学生心理健康教育提供了理 论借鉴和技术支撑。
英文摘要:
      The psychological analysis of college students' anxiety on the social media platform is helpful to detect the men⁃ tal health problems of college students in time. However, due to the large scale of data and the high frequency of release, it is still a challenge to analyze of college students' anxiety based on social media data. The large language model is leading the development of artificial intelligence into a new era, and has shown excellent performance in dialogue, understanding and reasoning of natural language. This study investigated the large language model-based approaches for psychological analysis in college students' anxiety, with comparative evaluation of fine-tuned GPT-family and BERT-family models. The results showed that GPT-3.5 Turbo 0125 and RoBERTa-base were the two models with the best performance in the two model families, respectively. The overall performance of GPT-3.5 Turbo 0125 was better, and its precision rate reached 96.27%. In general, the large language models of GPT and BERT families have shown great potential in the psychological analysis of college students' anxiety, which provides theoretical reference and technical support for generative artificial in⁃ telligence to help college students' mental health education.
查看全文   查看/发表评论  下载PDF阅读器
关闭