Explainable exercise recommendation with knowledge graph

Area of Science:

• Educational Technology
• Artificial Intelligence in Education
• Knowledge Representation

Background:

• Recommending exercises to students is crucial for learning efficiency but challenging due to resource variety and diverse student needs.
• Existing methods like collaborative filtering struggle with suitable recommendations, while deep learning lacks explainability, limiting practical application.
• There is a need for an explainable exercise recommendation system that can effectively match students with appropriate exercises and provide rationale.

Purpose of the Study:

• To propose and evaluate KG4EER, a novel knowledge graph-based system for explainable exercise recommendation.
• To address the limitations of existing recommendation approaches in educational settings by incorporating explainability.
• To improve student learning efficiency through personalized and justified exercise suggestions.

Main Methods:

• Developed KG4EER, an explainable exercise recommendation system utilizing a knowledge graph.
• Implemented a feature extraction module to represent student learning characteristics.
• Constructed a knowledge graph integrating knowledge concepts, students, and exercises, along with their interrelationships.

Main Results:

• KG4EER demonstrated superior performance compared to existing baseline methods across three real-world datasets.
• The system effectively matches diverse students with suitable exercises, providing clear explanations for recommendations.
• Expert interviews corroborated the system's effectiveness and robust explainability.

Conclusions:

• KG4EER offers a significant advancement in explainable exercise recommendation systems.
• The knowledge graph approach effectively bridges the gap between student needs and exercise resources.
• The system's explainability enhances user trust and practical utility in educational contexts.