创新创业理论研究与实践 ›› 2023, Vol. 6 ›› Issue (16): 9-14.

• 理论研究 • 上一篇    下一篇

基于GPT模式的LLM模型构建高职学校人才培养工作状态数据采集与管理平台问答系统研究

林德力, 范玉婷, 廖心妍, 林叶, 林志远, 陈致州   

  1. 闽江师范高等专科学校,福建福州 350108
  • 出版日期:2023-08-25 发布日期:2023-09-15
  • 作者简介:林德力(1984—),男,福建福州人,硕士,实验师,研究方向:智慧校园建设、数据统计与治理。
  • 基金资助:
    2018年福建省电化教育馆教育信息技术研究课题“高等院校智慧教室建设的研究”(FJDJ1857)

LLM Model Based on GPT Mode to Construct a Question and Answer System for the Data Collection and Management Platform of Talent Training in Higher Vocational Schools

LIN Deli, FAN Yuting, LIAO Xinyan, LIN Ye, LIN Zhiyuan, CHEN Zhizhou   

  1. Minjiang Teachers College, Fuzhou Fujian, 350108, China
  • Online:2023-08-25 Published:2023-09-15

摘要: 该文针对数据统计过程中繁杂的问答体系,探讨基于GPT模式的LLM模型构建高等职业学校人才培养工作状态数据采集与管理平台问答系统。基于GPT模式的LLM模型问答系统构建的三个学习过程包含训练监督策略模型、训练奖励模型和利用PPO算法微调SFT模型。问答系统构建涉及学校综合数据信息指标、师资队伍数据指标、学生情况数据指标、专业办学数据指标、产教融合数据指标五个方面的思维链建设,为准确构建逻辑体系搭建基础框架。

关键词: LLM模型, 高等职业学校, 人才培养, 数据采集, ChatGPT, 问答系统

Abstract: In view of the complicated question and answer system in the process of data statistics, this paper discusses the question and answer system of the data collection and management platform for talent training in higher vocational schools based on the LLM model of GPT mode. The three learning processes of the LLM model question answering system based on the GPT mode include the training supervision strategy model, the training reward model and the use of PPO algorithm to fine-tune the SFT model. The construction of the question and answer system involves the construction of the thinking chain in five aspects: the school's comprehensive data and information index, the faculty's data index, the student's situation data index, the professional school-running data index, and the industry-teaching integration data index, which builds the basic framework for the accurate construction of the logical system.

Key words: LLM model, Higher vocational schools, Talent training, Data acquisition, ChatGPT, Q&A system

中图分类号: