Research On Automatic Construction Technology Of Chinese Knowledge Graph For Education

With the wide application of knowledge graph in the education,it is more and more urgent to construct knowledge graph automatically based on educational data.Several challenging problems exist in previous Chinese knowledge graph generation technologies for education,such as knowledge points missing,knowledge point relationship error,and human participation.In this thesis,an automatic construction technology of Chinese knowledge graph is proposed to improve the accuracy of automatic generation of Chinese knowledge graph for education.Specifically,the main contributions in this thesis include the following three aspects.(1)An automatic generation method of knowledge graph based on textbooks of computer science(named as KG-T)is studied.KG-T automatically extracts the key knowledge points and the sequence of key knowledge points in teaching textbooks to generate knowledge graph.(2)An automatic generation method of knowledge graph based on Power Point documents for teaching(named as KG-S)is studied,and a simplified version of LSTM model based on KG-S is proposed.KG-S takes advantage of the clear hierarchical structure of teaching Power Point slides,and combines the two key steps of knowledge point extraction and relationship extraction.It is the first time to construct knowledge graph with prerequisite relationship based on Chinese Power Point presentation in the education field.The simplified LSTM model reduces the requirement of computer performance for the automatic generation of knowledge graph based on Power Point presentation,which makes the application of KG-S more universal.In this thesis,real data sets are used to extract the order of knowledge points through the hierarchical relationship of Power Point slides,set the weight between knowledge points,adjust the influence coefficient,and generate the knowledge graph.(3)The adjustment and optimization technology(named as KG-T&S)to generate knowledge graph based on teaching textbooks is studied.KG-T&S adjusts the knowledge graph based on textbooks by using the triplet generated by Power Point slides,and realizes the quality optimization of the existing knowledge graph.The experimental results based on real data sets show that,the proposed KG-T&S can optimize the knowledge graph of education domain generated by KG-T,and the accuracy rate of the optimized knowledge graph is more than 90%.The automatic construction of Chinese knowledge graph for education is a problem worth further study,which has important theoretical and practical significance.In this thesis,the automatic generation methods of knowledge graph based on textbooks and Power Point slides are studied,and the adjustment scheme of knowledge graph are also explored.Experiments based on real education data sets verify the effectiveness of the key technologies proposed in this thesis.