A study on exercise recommendation method using Knowledge Graph for computer network course
Lei Zhu, Yaolin Liu, Xinhong Hei, Yichuan Wang, Haining Meng, Jiuyuan Jiao, Long Pan · 2020
With the vast applications of massive online learning platforms during the coVID-19 outbreak, the personalized exercise recommendation methods play an import role on computer aided instruction(CAI). Most existing methods generates the exercises according to the contents and knowledge system structure, lacking semantic relationships between exercises and its knowledge. Knowledge graph is widely used to represent the semi-structured and schemaless information (nodes) and their relation (edges), and indicate the sentence embedding grammatical structure and semantic relations, thus it can be applied on computer aided instruction to automatically generate the personalized exercises. Aiming to improve the efficiency of exercise recommendation, this paper studies the feature information of computer network course, and proposes a content and knowledge graph based personalized exercise recommendation method. More specifically, knowledge graph is firstly constructed from entities and relations of computer network course, and the information vectors of exercises are generated by combining the knowledge with the exercises content. And then the learner's historical log data is analyzed, and the semantic similarity between exercises and their knowledge are generated for the wrong answers. According the semantic similarity of knowledge, the final exercises are recommended for the learners. Experimental results show that the proposed method can improve the efficiency of exercises recommendation.