Research On Knowledge Recommendation Method Based On Educational Knowledge Graph

Under the development trend of online and offline blended teaching,a wide variety of learning resources are emerging,and various learning platforms are also increasing.How to effectively use technical means to integrate and optimize this knowledge and information to make it play a full role in the teaching process,provide targeted help for learners,and improve teaching has become the key to current research.The research of knowledge graph and recommendation technology shows important practical application value in this respect.Using knowledge graph can help to sort out the subject system and integrate fragmented knowledge;using recommendation technology can help learners quickly and automatically choose teaching resources that are suitable for their own cognitive level and learning purpose from massive resources.In previous studies on educational knowledge graphs,researchers regarded subject terms or teaching resources as nodes in the graph,and regarded the associations between subject terms or teaching resources as edges in the graph,which can effectively connect subject knowledge into a whole.However,in these completed studies,there is often a lack of complete information about the knowledge points themselves,and the type of teaching resources covered are single,which is limited in the degree of help to learners;in terms of personalized recommendation,most of the recommendations in the teaching field are only for the knowledge points or a certain type of teaching resources,which have not taken into account the cognitive needs of learners.In view of the above-mentioned shortcomings,this thesis makes a series of research and improvement on the construction of multi-modal educational knowledge graph and the knowledge recommendation based on educational knowledge graph.The details are as follows:(1)Aiming at the deficiencies of neglecting the importance of terminology information for teaching and the single resource type,this thesis constructs a small-scale multi-modal educational knowledge graph of database course.Firstly,the data source suitable for the construction of knowledge graph is found from a large amount of database course related information;secondly,the appropriate named entity recognition combination model Dict-BERT-BILSTM-CRF is constructed to extract database course terms from the text information,and extract the attribute and relationship of terminology,at the same time,the multi-modal teaching resources are fused;finally,the obtained information is used to a multi-modal educational knowledge graph of the database curriculum suitable for knowledge recommendation.(2)Aiming at the deficiency that most recommendation methods only focus on knowledge points or a certain type of teaching resources,while considering the cognitive level of learners and how to greatly improve the learning effect in a limited time,this thesis proposes a knowledge recommendation method ALKPT-KI based on educational knowledge graph,and recommend multi-modal teaching resources based on this method in the next step.This method integrates the learner's knowledge mastery and the importance of knowledge points,and generates Top-N knowledge recommendations that conform to the learners' cognitive level and behavior habits.In particular,ALKPT is an improvement of KPT model,which adapts to the target task and reduces part of the calculation.Finally,through comparative experiments on the SQL data set,with Recall@N and MAP@N as indicators,it is proved that the recommended method has better performance.(3)In order to better serve learners and promote the continuous development of online and offline hybrid education,a knowledge recommendation prototype system based on educational knowledge graph is built.This thesis analyzes the function of the prototype system,gives the architecture of the prototype system,realizes three functions of knowledge visualization,knowledge card and knowledge recommendation based on knowledge points(including video,PPT,audio,etc.),and gives the function implementation interface.