学习者知识追踪研究进展综述

Learner modeling is one of the supporting techniques of adaptive learning systems, among which learner knowledge state modeling represented by knowledge tracing is the most widely studied.Three representative knowledge tracing techniques are Bayesian Knowledge Tracing(BKT) based on hidden Markov model, Additive Factor Model(AFM) based on logistic regression model and Deep Knowledge Tracing(DKT) based on recurrent neural network.It is found that the BKT is suitable for know-ledge tracing in learning tasks that only contain single knowledge skill, AFM and DKT can be used for tracing students’ know-ledge state in learning tasks that have more than one knowledge skills.However, the DKT is hard to be interpreted from the perspective of pedagogy.Based on reviewing the existing research and inspired by the knowledge space theory, this paper argues that the integration of the prerequisite relationship among knowledge skills and the knowledge tracing is a promising research direction.Finally, this paper proposes a protype of additive factor model integrating knowledge prerequisite relationship.