Proceedings of the 3rd International Conference on Internet Technology and Educational Informatization, ITEI 2023, November 24–26, 2023, Zhengzhou, China

Research Article

A Prompt-based Approach for Discovering Prerequisite Relations Among Concepts

Download85 downloads
  • @INPROCEEDINGS{10.4108/eai.24-11-2023.2343577,
        author={Fangrui  Du and Wenxin  Hu and Changjiu  Qin and Zheng  Xiao},
        title={A Prompt-based Approach for Discovering Prerequisite Relations Among Concepts},
        proceedings={Proceedings of the 3rd International Conference on Internet Technology and Educational Informatization, ITEI 2023, November 24--26, 2023, Zhengzhou, China},
        publisher={EAI},
        proceedings_a={ITEI},
        year={2024},
        month={4},
        keywords={concept prerequisite learning prompt learning smart education},
        doi={10.4108/eai.24-11-2023.2343577}
    }
    
  • Fangrui Du
    Wenxin Hu
    Changjiu Qin
    Zheng Xiao
    Year: 2024
    A Prompt-based Approach for Discovering Prerequisite Relations Among Concepts
    ITEI
    EAI
    DOI: 10.4108/eai.24-11-2023.2343577
Fangrui Du1, Wenxin Hu1,*, Changjiu Qin1, Zheng Xiao1
  • 1: East China Normal University Shanghai
*Contact email: wxhu@cc.ecnu.edu.cn

Abstract

The concept prerequisite learning task is a crucial study in smart education and plays an important role in online course design, course guidance, and learning material recommendation systems. In general, concept prerequisite relations are the determination of whether two concepts have a prerequisite relationship, usually as a binary task. Nowadays, learning concept representations from pre-trained models has become a new trend in concept prerequisite learning (CPL) tasks. However, too many handicraft features are still required to discover the concept features in previous studies. In this paper, we propose a concept prerequisite relationship discovery method based on prompt learning, in which we design four prompt functions, mapping the predicted labels to the existing labels through answer engineering after the model training. Conducting thorough experiments across three publicly available benchmarks reveals enhancements of up to 13% in F1 score. It shows that the prompt learning approach can effectively improve the prediction of prerequisite relations and provide a new idea for the study of concept prerequisite relations tasks.