Research On Personalized Knowledge Recommendation System For Virtual Experiment Teaching In Middle School

AR experiments is a typical E-learning mode.In essence,it is a process of putting theory into practice,which needs to be closely connected with the experimental knowledge.Therefore,the knowledge graph in the field of artificial intelligence is used to effectively organize experimental knowledge,and named it middle school experimental knowledge graph.In addition,using the rich semantic information in the knowledge graph to assist students to master knowledge through active recommendation is a work of research significance and application value.The AR experiment platform based on cloud provides teachers with production tools,supports crowdsourcing to generate experimental teaching resources.For AR experiments,this thesis studies the completion and application of domain knowledge graph,and realizes a personalized knowledge recommendation system to assist students to learn experimental knowledge.The research contents of this thesis are as follows:(1)In terms of the completion and integration of the middle school experimental knowledge graph,a method based on representation learning is proposed.The knowledge subgraph is obtained through the knowledge extraction of crowdsourcing resources,and the quality evaluation of the subgraph is completed.On this basis,the knowledge completion task is completed based on the graph neural network and translation model,Through four groups of experiments,the effectiveness of the construction method of negative case triples and the confidence evaluation method of synthesizing the most similar semantic relationship are verified.(2)In terms of personalized recommendation based on the middle school experimental knowledge graph,on the basis of the existing personalized knowledge recommendation model KGPB(knowledge graph path Bayesian),the auxiliary recommendation of students’ historical experiment information is added,and an improved KGAB(knowledge graph attention Bayesian)recommendation model is proposed.Through the student history experiment evaluation sequence modeling to capture the potential knowledge needs,so as to strengthen the effect and interpretability of knowledge point recommendation.Compared with the existing methods,the NDCG(Normalized Discounted Cummulative Gain)index is significantly improved.(3)In the construction of personalized knowledge recommendation system for middle school AR experiment,combined with knowledge completion model and personalized recommendation model,the system is divided into KG expansion module,knowledge recommendation module and KG query module.The detailed structure and process of each module of the system are introduced.Finally,combined with a specific case,it is proved that the system can complete the effective recommendation of personalized knowledge.