Construction and Application Research of Intelligent Education Knowledge Graph Based on Multi-modal Learning

This study explores the methods and applications of constructing an intelligent education knowledge graph based on multi-modal learning materials (text, images, videos) Through deep learning algorithms for feature extraction and relationship modeling, a knowledge graph containing 120 entities and 100 relationships was built with an accuracy of $87 %$. The graph demonstrates excellent performance in personalized learning path recommendations, capable of recommending based on students’ prior knowledge and learning styles, leading to a $27 %$ improvement in learning efficiency. The study also considers the “uncanny valley” phenomenon, avoiding discomfort caused by AI mimicking human behavior excessively, ensuring a positive user experience.