Research On The Method Of Extracting Course Knowledge Graph For Wisdom Education

In recent years,MOOC(Massive Open Online Course)and online education(eLearning)have been widely discussed in the field of smart education.With the injection of users,a large amount of teaching behavior data and knowledge resources have been accumulated on the smart education platform.The analyzing and mining of these two important types of educational data have injected new motivation into the development of smart education.When students are learning online,they will learn many course concepts.Most of the current online education platforms are organized with courses and are classified according to subject areas,universities,and years according to course information.The concept of the course is implicit in the course,which requires learners to organize,classify and organize during learning the course.How to extract the course concepts automatically in a course by using big data is one of the central issue and difficulties in recent smart education research.At the same time,there are many relationships between course concepts: prerequisites of course concept,subordinate relationship of course concept and inclusion relationship of course concept.However,the relationship between these course concepts in the MOOC online education platform also needs to be learned by the learners.How to use big data analysis to automatically extract the relationship between course concepts in a course is also one of the central issue and difficulties in today's smart education research.Extracting the relationship between course concepts and course concepts is actually extracting the knowledge graph of course concepts.Aiming at the problem of course concept extraction in course knowledge graph extraction,this paper proposes a model algorithm for course outline corpus and course caption corpus.Because density of course concepts are high density in the corpus of the course outline,most noun phrases can be used for course concepts.The extraction of course concepts in the outline corpus depends on the language template method to extract this part of the course concepts.For the corpus of course video caption,this article first uses the graph propagation of confidence algorithm to extract course concepts,and then changes the original candidate course concept extraction method.The entities extracted by distant supervision are used as candidate course concepts in the graph propagation model for iterative extraction.Experiments verify that the accuracy of the improved algorithm is about 10% higher than the propagation algorithm of confidence.Aiming at the problem of course relationship extraction in course knowledge graph,this article focuses on extracting the relationship between course concepts: the subordinate relationship of the course concept and the prerequisite of the course concept.For the subordinate relationship of the course concept,we use the implicit relationship in course outline and design an algorithm to obtain the basic skeleton of subordinate relationship.Then we add the course concepts into the subordinate relationship to get compete subordinate relationship of the course concept.For the prerequisites relationship of the course concept,we extract the three basic characteristics of the prerequisites.How to extract the prerequisites relationship of the course concept converts into a binary classification problem.we use three different classifiers SVM?NB and RF to experimrnt it.Finally,we select the RF classifier to extract the prerequisites of the course concept.