新型智能导学系统构建及其关键技术  

Novel Intelligent Tutoring System Design and Its Key Technologies

作　　者：卢宇 骈扬 陈鹏鹤 Yu Lu;Yang Pian;Penghe Chen

机构地区：[1]北京师范大学教育学部未来教育高精尖创新中心 [2]北京师范大学教育技术学院

出　　处：《中国远程教育》2023年第7期30-38,46,共10页Chinese Journal of Distance Education

基　　金：北京市教育科学“十四五”规划2021年度重点课题“人工智能驱动的新一代智能导学系统构建研究”(项目编号:CHAA21036)的研究成果。

摘　　要：智能导学系统能够提供个性化的自主学习服务,但仍然面临诸多挑战,如对学习者理解不足、用户体验较差、自动化程度有限等。针对这些挑战与问题,本研究设计并实现了涵盖领域知识层、基础模型层和用户接口层的三层架构新型导学系统,并以“雷达数学”系统为例进行介绍。领域知识层借助知识空间等理论,构建符合教学和认知规律的学科知识图谱和教学策略库;基础模型层利用深度学习、自然语言处理等技术,实现更精准的学习者模型、个性化教学策略模型和自动批阅模型,提升系统对学习者的理解和自动化程度;用户接口层基于可解释人工智能技术与游戏化设计理论,构建可解释的认知地图和个性化导航等服务,改善系统用户的体验。在此基础上,本研究针对学习者、教师、家长等不同角色,阐述了新型智能导学系统的三类典型应用场景并对其发展趋势进行了展望。Intelligent tutoring system(ITS)provides personalized and autonomous learning services,but it still faces many challenges,such as insufficient understanding of learners,poor user experience,and limited automation.To address these challenges and problems,this study designs and implements a new three-layer ITS,consisting of domain knowledge layer,basic model layer and user interface layer,and introduces the“Radar Math”system for mathematics as an example.The domain knowledge layer constructs a subject knowledge base and a teaching strategy base,which conforms to the laws of teaching and cognition by leveraging knowledge space and zone of proxi-mal development theories.The basic model layer adopts deep learning and natural language processing techniques to implement more accurate learner models,personalized teaching strategy models and automatic scoring models,which improves the system's understanding on learners and the capabilities of automation.The user interface layer uses interpretable Al techniques and gamification design theory to build interpretable cognitive maps and personalized navigation services to improve the user experience of the system.On this basis,the study illustrates three typical application scenarios of the new ITS for different roles of education,typically including students,teachers,and parents,and provides an outlook on its future trend.

关 键 词：智能导学系统 深度学习 自然语言处理 可解释人工智能 认知地图 

分 类 号：G40-057[文化科学—教育学原理]