The Design And Implementation Of A Question Recommendation System Based On Knowledge Graph And Knowledge Tracing

With the vigorous development of the Internet,more and more online education systems have begun to become popular,providing a question bank for learners to check and fill in the vacancies.However,many question bank resources are redundant.They simply list knowledge points and corresponding test questions,and cannot push appropriate practice questions according to the students’ understanding of knowledge concepts and the situation of answering questions.Therefore,in response to these needs,a test question recommendation system based on knowledge graphs and Bayesian knowledge tracking is proposed.The system uses knowledge graphs to store course knowledge points and uses Bayesian knowledge tracking to evaluate the user’s mastery of each knowledge point.Master the status and answering situation to recommend test questions to users.The system uses the knowledge graph to associate the concepts and knowledge points of the junior high school mathematics curriculum.Obtain a resource library by crawling textbooks and teaching resources on the Internet,and then extract the text information of the resource library to obtain knowledge point entities,and use neo4 j to manage the knowledge graph nodes.The system performs word segmentation on the crawled test question text,extracts characteristic words to traverse the knowledge graph,and uses gensim to calculate the semantic similarity of the characteristic words and node attributes to match the test questions with the knowledge graph nodes.And the system uses Bayesian knowledge tracking technology to model the students.By setting the initial parameters of the model,and then according to the student’s answer sequence,the EM algorithm is used to train the Bayesian tracking model.After the Bayesian tracking model is obtained,it is used to decode the hidden state sequence and evaluate the learning situation of the learner.After the user’s mastery of a certain knowledge point reaches the threshold of the Bayesian tracking model,the next knowledge point that needs to be practiced is searched for according to the knowledge graph,otherwise,the recommended test questions of the knowledge point are regenerated for the user to practice.According to the user’s current knowledge points to find similar users,extract other users’ wrong questions to practice,and analyze the user’s wrong questions,use the weight of the word vector to calculate the cosine similarity,and select similar The test questions are pushed to the user.The test question recommendation system has been tested and has good stability without crashes and abnormalities.At the same time,the system can correctly assess the learner’s mastery and recommend appropriate test questions as needed.And provide various learning resources such as videos and materials to meet the needs of users.