Heterogeneous Graph Neural Networks for Concept Prerequisite Relation Learning in Educational Data

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2036–2047 June 6–11, 2021. ©2021 Association for Computational Linguistics 2036 Heterogeneous Graph Neural Networks for Concept Prerequisite Relation Learning in Educational Data Chenghao Jia, Yongliang Shen, Yechun Tang, Lu Sun, Weiming Lu∗ College of Computer Science and Technology, Zhejiang University, Hangzhou, China {chjia, syl, tangyechun, sunlu98, luwm}@zju.edu.cn Abstract Prerequisite relations among concepts are cru- cial for educational applications, such as cur- riculum planning and intelligent tutoring. In this paper, we propose a novel concept pre- requisite relation learning approach, named CPRL, which combines both concept represen- tation learned from a heterogeneous graph and concept pairwise features. Furthermore, we ex- tend CPRL under weakly supervised settings to make our method more practical, includ- ing learning prerequisite relations from learn- ing object dependencies and generating train- ing data with data programming. Our experi- ments on four datasets show that the proposed approach achieves the state-of-the-art results comparing with existing methods. 1 Introduction With the increasing availability of learning re- sources and the requirement of self-regulated learn- ing, there is a rising need to organize knowledge in a reasonable order. Concept prerequisite rela- tions are essentially considered as the dependen- cy among concepts, and they are crucial for peo- ple to learn, organize, apply and generate knowl- edge (Margolis and Laurence, 1999). For example, if someone wants to learn the knowledge about Conditional Random Fields, the knowledge about Hidden Markov Model should be learned first. Con- sequently, the concept Hidden Markov Model is a prerequisite concept of the concept Condition- al Random Fields. Nowadays, prerequisite rela- tions among concepts have played a crucial role in educational applications, such as curriculum planning (Liu et al., 2016) and intelligent tutor- ing (Wang and Liu, 2016; Chen et al., 2018). Recently, several attempts have been made to extract prerequisite relations among concepts from textbooks (Wang et al., 2016; Liang et al., 2018), MOOCs (Massive Open Online Courses) (Pan ∗ corresponding author et al., 2017), courses (Liang et al., 2015a; Liu et al., 2016; Liang et al., 2017; Li et al., 2019a; Roy et al., 2019) and scientific papers (Gordon et al., 2016). They either proposed a local statistical information, such as reference distance (Liang et al., 2015a) and cross-entropy (Gordon et al., 2016) to mea- sure the prerequisite relations between concepts, or proposed handcrafted features to learn a prereq- uisite relation classifier (Pan et al., 2017). Liang et al. (2017) proposed CPR-Recover to recover concept prerequisite relations from course depen- dencies. More recently, Li et al. (2019a) applied variational graph autoencoders to learn concept prerequisite relations from courses. While Roy et al. (2019) developed a supervised learning ap- proach called PREREQ. However, there are still several challenges to learn the prerequisite relations among concepts. Firstly, there are multiple and complex relations among concepts and learning resources, but they were not fully utilized before. Secondly, labeling training data is enormously expensive and time consuming, especially when domain …

Content truncated to 500 words