Research On Learning Path Recommendation Based On Educational Knowledge Graph

With the advent of artificial intelligence and information age,the education industry has also developed rapidly.Learners’ learning methods are no longer limited to traditional offline classroom learning,and more and more learners choose to use the Internet for learning.Online learning focus on the initiative of learners.Although there are massive learning resources,learners still need to explore and organize knowledge from different forms and sources.Therefore,in the face of a large amount of fragmented knowledge,learners are prone to information overload and learning lost.In recent years,knowledge graph technology has received more and more attention,and as an effective means of organizing and managing knowledge,it has also been well applied in the education field.Taking computer courses as an example,this paper integrates multi-source heterogeneous educational data,constructs a highquality educational knowledge graph by extracting educational entities and relationships in the data,and conducts learning path recommendation based on knowledge graphs.On the one hand,the structure of knowledge points is displayed to learners in a more intuitive way.On the other hand,the curriculum knowledge point model is designed by combining the entities and relationships in the knowledge graph,and learning path recommendation strategies are proposed for different relationships,so as to realize the integration of educational knowledge and the recommendation of learning path for underlying knowledge.The main research work of this paper is as follows:1.According to the characteristics of multi-source heterogeneous educational data,the data schema of educational knowledge graph is defined,including educational entity types and relation types.This paper contains three types of entities: structure,algorithm,and basic terminology,and defines five types of educational relationships between entities.2.Considering the professionalism and domain of educational entity recognition,this paper proposes a pre-trained language model that integrates educational domain vocabulary,which can adaptively capture effective domain information,extract contextual semantics with Bi-directional Long Short-Term Memory network,and use Conditional Random Fields for sequence labeling.Experimental results show that the model proposed in this paper shows better performance in entity recognition.3.In relation extraction task,this paper combines educational entities with the pre-trained language model to realize automatic labeling of semantic relationships between entities in educational resources.Compared with other baseline methods,the model proposed in the paper performs the best in relation classification.4.Based on the structure of knowledge points,this paper designs a course knowledge point model.According to the established educational knowledge graph,a deep recommendation strategy based on inclusion relationship and a hybrid recommendation strategy based on other relationships are proposed.The course knowledge model and recommendation strategy have good reference significance for learners.Experiments show that the entity recognition algorithm and relation extraction algorithm proposed in this paper have good performance in the field of education,which lays a foundation for building a high-quality knowledge graph.At the same time,according to the course knowledge learning model,different learning path recommendation strategies are proposed for different educational relationships in the knowledge graph.Starting from the structure of knowledge points,it can not only provide learners with a reasonable learning sequence,but also help learners fully learn the implicit information between knowledge points and achieve efficient learning.