Journal
KNOWLEDGE-BASED SYSTEMS
Volume 253, Issue -, Pages -Publisher
ELSEVIER
DOI: 10.1016/j.knosys.2022.109547
Keywords
Cognitive diagnosis; Graph neural networks; Interpretable machine learning
Categories
Funding
- National Natural Science Foundation of China [2020B444]
- [61976001]
- [61922073]
- [61672483]
Ask authors/readers for more resources
This paper proposes a graph-based Cognitive Diagnosis model that directly discovers the interactions among students, skills, and questions through a heterogeneous cognitive graph. The model designs a performance-relative propagator and an attentive knowledge aggregator to handle this task. Extensive experimental results demonstrate the effectiveness and extendibility of the model.
For intelligent tutoring systems, Cognitive Diagnosis (CD) is a fundamental task that aims to estimate the mastery degree of a student on each skill according to the exercise record. The CD task is considered rather challenging since we need to model inner-relations and inter-relations among students, skills, and questions to obtain more abundant information. Most existing methods attempt to solve this problem through two-way interactions between students and questions (or between students and skills), ignoring potential high-order relations among entities. Furthermore, how to construct an end -to-end framework that can model the complex interactions among different types of entities at the same time remains unexplored. Therefore, in this paper, we propose a graph-based Cognitive Diagnosis model (GCDM) that directly discovers the interactions among students, skills, and questions through a heterogeneous cognitive graph. Specifically, we design two graph-based layers: a performance-relative propagator and an attentive knowledge aggregator. The former is applied to propagate a student's cognitive state through different types of graph edges, while the latter selectively gathers messages from neighboring graph nodes. Extensive experimental results on two real-world datasets clearly show the effectiveness and extendibility of our proposed model. (C) 2022 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available