Research And Application Of Recommendation Algorithm Based On Knowledge Graph In Question Answering System

Education has always been the cornerstone of national development,and mathemat-ics is the foundation of all disciplines.The state,teachers,and parents have been con-tinuously strengthening the teaching of students' mathematics.With the development of science and technology and the popularization of intelligent terminals,online teaching models have begun to appear in people's vision,and a large number of students have started to learn independently on the Internet.At the same time,in the context of the rapid development of artificial intelligence in recent years,a variety of personalized intelligent teaching software has gradually become popular.These teaching systems can develop personalized learning programs for students,and automatically handle differences in stu-dent learning processes,allowing students to learn independently without the guidance of teachers and parents.The research in this article is an intelligent teaching system in the field of elemen-tary mathematics.This teaching system is based on knowledge points.By strengthen-ing students' understanding of mathematical knowledge points,students can master the definitions and relationships of various mathematical concepts,and master the problems of mathematical problem solving.Keys and skills to discover the laws of mathematical problem solving,thus improving students' mathematical performance.In the text,a com-prehensive and clear multi-dimensional mathematical knowledge map is constructed by referring to the authoritative materials from multiple parties,starting from the three dimen-sions of knowledge point classification,knowledge point attributes,and the relationship between knowledge points.The knowledge map contains 721 mathematical knowledge points and 786 mathematical relationships.Based on the established knowledge map,this article proposes four question-and-answer forms related to knowledge points,each of which can enable students to understand and practice knowledge points from different perspectives.The four question-and-answer forms are recommending related knowledge points based on the topic,recommending similar topics based on the topic,recommend-ing related topics based on the knowledge point,and recommending related knowledge points based on the knowledge point.Among these four types of recommendation,topic-to-knowledge recommendation is very important.The recommendation algorithm used in this article is a text space similarity calculation method based on vector space.First,more than 10,000 mathematical questions are used to train the BERT word vector model.,And then used the cosine similarity to calculate the degree of matching between the topic and the knowledge points.With the help of the system studied in this article,students can input unclear knowledge points or unsolvable math problems into the system at any time,and the system can immediately return the local knowledge map of the knowledge points or related problems.Through this self-directed learning method,students can continuously promote self-improvement,achieve inferences from each other,and integrate.In the text-structured system,this article tested the system with 500 mathematical questions.The average accuracy rate of a single knowledge point recommended by the question reached above 85%,and the accuracy rate of the related questions recommended by the knowledge point reached 88%.From the testing process and results,it can be seen that the system constructed in this paper can run stably and has certain value for practical application.