Learning from Non-Assessed Resources: Deep Multi-Type Knowledge Tracing
Chunpai Wang, Siqian Zhao, Shaghayegh Sahebi
Jun 30, 2021 19:30 UTC+2
—
Session C3
—
Zoom link
Keywords: knowledge tracing, multiple learning resources, student performance prediction
Abstract:
The state of the art deep knowledge tracing methods only leverage the information of assessed learning materials (such as quizzes, assignments, and exercises) and predict student's future performance based on past historical records of these learning materials. However, most of students' activities are non-assessed, such as watching video lectures, participating a discussion forum, and reading a section of textbook, all likely could contribute to the students' knowledge growth. In this paper, we propose a novel deep learning based knowledge tracing model that explicitly model student's knowledge transition over assessed and non-assessed learning activities. By this method, we could discover the latent knowledge concepts of each non-assessed and assessed learning material, which could be useful for a better instructional policy design.In addition, we compare our propose method with various state of the art knowledge tracing method as well as their extended versions that implicitly model non-assessed learning activities. Extensive experiments are conducted on four real-world datasets to validate the effectiveness of our proposed model.