多模态课程知识图谱与大模型协同的智慧教育实践  

Practice of smart education driven by collaboration of multimodal curriculum knowledge graph and large language model

作　　者：王利琴 陈博文 董永峰 许智宏 王旭 WANG Liqin;CHEN Bowen;DONG Yongfeng;XU Zhihong;WANG Xu(School of Big Data and Artificial Intelligence,Hebei University of Technology,Tianjin 300401,China)

机构地区：[1]河北工业大学人工智能与数据科学学院,天津300401

出　　处：《计算机应用文摘》2025年第1期43-47,共5页

基　　金：河北省高等教育教学改革研究与实践项目(2022GJJG049);河北省研究生教育教学改革研究项目(YJG2023023)。

摘　　要：随着智慧教育的推进,在线学习平台面临知识展示形式单一、在线答疑功能不完善、学习个性化需求无法满足等问题。为解决这些问题,首先,借助大模型和半自动化方法,构建了包含视频、图片、习题等内容的多模态课程知识图谱,实现了网状知识点关联的直观呈现。其次,基于多模态知识图谱和大语言模型,开发了2种知识问答服务,能够有效满足用户在在线学习中的实时问答需求。此外,采用认知诊断模型分析用户的答题情况,并根据分析结果有针对性地推荐薄弱知识点相关的习题和学习资源,从而更好地满足个性化学习需求。上述方法为智慧教育驱动的育人模式提供了新的途径,为用户提供了个性化、多样化和终身化的学习体验。With the advancement of smart education,online learning platforms are facing problems such as a single form of knowledge display,lack of online Q&A functions,and inability to meet personalized learning needs.To address these issues,firstly,a multimodal course knowledge graph containing videos,images,exercises,and other content was constructed using large-scale models and semi automated methods,thus achieving an intuitive presentation of networked knowledge point associations.Secondly,based on multimodal knowledge graphs and large language models,two knowledge question answering services have been developed,which can effectively meet users̓real-time question answering needs in online learning.In addition,a cognitive diagnostic model is used to analyze users̓answering situations,and targeted exercises and learning resources related to weak knowledge points are recommended based on the analysis results,in order to better meet personalized learning needs.The above methods provide a new approach for the education model driven by smart education,providing users with personalized,diversified,and lifelong learning experiences.

关 键 词：知识图谱 多模态 智能问答 习题推荐 大模型 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]