Tracing Knowledge State with Individualized Ability and Question Difficulty

Abstract

Knowledge tracing(KT) refers to the task of modeling students’ evolving knowledge state according to their historical learning trajectories. Although many methods have been proposed to solve KT task, most of them ignore the difference of students and questions, i.e., the students’ learning ability are different from each other. To this end, in this paper, a learning ability estimation module is proposed to extract students’ learning ability according to their learning history and a novel method to obtain questions’ representation is designed. Besides, a knowledge state estimation module is proposed to estimation students’ knowledge state which takes both students’ learning ability and their learning interaction into consideration when modeling. Extensive experiments demonstrate that the proposed model could improve thr results of knowledge tracing through modeling individualized students’ learning ability and questions’ difficulty in learning process.