Time Interval Aware Self-Attention approach for Knowledge Tracing. (September 2022)

• Record Type: Journal Article
• Authors: Wei, Liting; Li, Bin; Li, Yun; Zhu, Yi
• Is Part Of: Computers & electrical engineering. Volume 102(2022)
• Publication Date: 2022-09
• Subjects: Knowledge Tracing -- Attention mechanism -- Time interval -- Temporal convolutional network; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.108179
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: Knowledge Tracing aims to discover the students' states of mastery for different knowledge or concepts on interaction sequences. It has been widely applied to intelligent educational systems. In recent years, deep neural network-based methods have achieved sound performance on this task because they have a powerful feature representation learning ability. However, two main problems hinder the further development of the existing deep learning-based Knowledge Tracing methods. The first problem is that the interaction history is treated as an ordered sequence, which ignores the individual time intervals between each interaction. The second problem is that higher-order temporal features for knowledge tracing have not been extensively explored. In this paper, to address these problems, we propose a method called the Time Interval Aware Self-Attention approach for Knowledge Tracing (TISAKT), which takes the relationship between exercises into consideration. More specifically, firstly, the relationship between exercises is modeled from the interactions introduced into the absolute positions and the time intervals of answered exercises at the same time. Then, temporal features are exploited implicitly by a temporal convolutional network. Extensive experiments on the two real-world datasets show that our model consistently outperforms state-of-the-art knowledge tracing methods.