摘要
智能教辅系统的个性化及便捷性有助于充分发挥其教学作用,其中知识追踪(KT)则是一项重要但棘手的任务,它随着时间来追踪学习者不断变化的关键词掌握程度,并预测学生在下一次测试中的表现。很多研究者已经关注该领域并提出了一些策略,如贝叶斯知识追踪(BKT)及深度知识追踪(DKT)。关键词(又称概念)之间的传播影响已被教育理论证明是学习的关键因素之一,然而却未能得到充分探索。提出了一种新框架,称为基于关键词结构的知识追踪(Keywords Structure based Knowledge Tracing,KKT)模型,利用关键词结构中的多重关系来模拟关键词间的相互影响。KKT框架应用图神经网络(GNN)将关键词结构映射为图,同时考虑对练习序列的时间影响和对关键词间结构的空间影响。通过在开放数据集上实验,结果证明提出的KKT模型具有很好的预测性和可解释性。
The individuation and convenience of the intelligent teaching assistant system help to give full play to its teaching role, and knowledge tracing(KT) is an important but difficult task. It tracks the learners’ changing keywords mastery over time and predicts the students’ performance in the next test. Many researchers have paid attention to this field and proposed some strategies, such as bayesian knowledge tracing(BKT) and deep knowledge tracing(DKT). The influence of communication between keywords(or concepts) has been proved by educational theory to be one of the key factors of learning, but it has not been fully explored. This paper proposes a new framework called keywords structure based knowledge tracing(KKT) model, which uses multiple relationships in keyword structure to simulate the interaction between concepts. KKT framework apply graph neural network(GNN) to map the keyword structure into a graph, taking into account the time impact on the practice sequence and the spatial impact on the structure between keywords.
出处
《工业控制计算机》
2023年第2期104-106,共3页
Industrial Control Computer
关键词
知识追踪
关键词结构
图神经网络
knowledge tracing
keyword structure
graph neural network
