图谱波纹特征驱动的知识追踪算法  

Knowledge Tracking Algorithm Driven by Graph Ripple Feature

作　　者：李俊华 王聪 李敏[1] LI Jun-hua;WANG Cong;LI Min(College of Computer Science,Sichuan Normal University,Chengdu 610101,China;Department of Computer Science and Technology,Sichuan Police College,Luzhou 646000,China)

机构地区：[1]四川师范大学计算机科学学院,成都610101 [2]四川警察学院计算机科学与技术系,四川泸州646000

出　　处：《小型微型计算机系统》2023年第7期1419-1427,共9页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61602331)资助;四川省重点实验室开放课题项目(NDSMS201606)资助;四川省教育厅重点项目(17ZA0322)资助;四川省教育厅科研项目(17ZB0361)资助。

摘　　要：知识追踪是智能教学系统中用于建模学生知识状态的关键技术.尽管研究者们提出了多种知识追踪模型,但这些模型没有充分挖掘知识点之间的关联性,限制了知识追踪的性能和应用.为此,本文提出了RKT(RippleNet Knowledge Tracing),一种利用知识状态传播将知识图谱融入到知识追踪的模型.首先以学生掌握的题目为起点,将题目-知识点的对应关系及知识点之间的关联形成一个知识图谱,然后将学生的知识状态沿着该知识图谱不断传播,产生向外扩散的多个“波纹”,再叠加起来,形成学生对未知题目的多阶响应,自动探索学生潜在的学习能力,最后预测回答正确的概率.通过对Assistment2009_Skill_builder、Junyi Academy等经典数据集的对比实验,结果表明RKT在AUC和ACC性能上有着显著提升.Knowledge tracking is a key technology for modeling students′knowledge status in Intelligent Tutoring Systems.Although researchers have proposed a variety of knowledge tracking models,these models do not fully explore the correlation between knowledge points,which limits the performance and application of knowledge tracking.To this end,this paper proposes RKT(RippleNet Knowledge Tracing),a model that uses knowledge state propagation to integrate knowledge graphs into knowledge tracing.Firstly,taking the questions mastered by students as the starting point,the corresponding relationship between questions and knowledge points and the correlation between knowledge points form a knowledge graphs,and then the students′knowledge state is continuously spread along the knowledge graphs,resulting in multiple"ripples"spreading outward,and then superimposed to form a student′s multi-level response to unknown questions and automatically explore students′potential learning ability,finally predict the probability of correct answers.Through comparative experiments on classic data sets such as Assistment2009_Skill_builder and Junyi Academy,the results show that the performance of RKT on AUC and ACC has been significantly improved.

关 键 词：知识追踪 知识图谱 神经网络 知识状态 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]