Research And System Implementation Of Educational Knowledge Graph Completion Based On Crowdsourcing

Intelligent education based on artificial intelligence,big data and other technologies is the development direction of today's education.The core of intelligent education is personalized education,that is,taking students as the center and carrying out teaching work around students' learning needs.However,personalized education currently has technical difficulties in terms of guidance and recommendation,such as: predicting learners' knowledge blind spots;helping learners to plan learning paths,and a perfect curriculum knowledge graph is the cornerstone of solving these problems.as it saves time and effort,large-scale curriculum knowledge graphs are often constructed using automated methods such as machine learning,but the effect is difficult to be satisfactory,so the completion of curriculum knowledge graphs is important,but the current mainstream knowledge graph completion studies are often ignored Group wisdom,the completion effect is not very good.In response to the above problems,this thesis uses learners and the degree of understanding of the course to improve the course knowledge graph,and designs and implements a set of crowd-sourced course knowledge graph completion error correction system.This thesis first defines and constructs the curriculum knowledge graph according to the needs of wisdom education,analyzes the completeness and correctness of the curriculum knowledge graph,and clarifies the key points of the course knowledge graph error correction and completion.In the process of using crowdsourcing error correction to complete the knowledge graph of the course,the accuracy of the learner's answer was estimated,and the crowdsourcing task allocation algorithm was improved.In this thesis,the graph estimation theory of text similarity of crowdsourcing tasks is used.Based on the text relevance of crowdsourcing tasks and the historical answer quality of learners,we can estimate the accuracy of learners' answers to unanswered crowdsourcing tasks.The improved crowdsourcing task allocation algorithm considers the following four factors:(1)In order to understand the learner to the greatest extent,that is,every time the learner answers the crowdsourcing question,the algorithm judges the quality of the learner's answer,and then Update the accuracy of learners' answers to other crowdsourcing tasks according to learner's answer quality;(2)the purpose of crowdsourcing tasks that can be completed fastest;(3)the purpose of crowdsourcing tasks assigned to learners with the highest answer accuracy;(4)the thesis considers the situation where learners are unwilling to answer the assigned crowdsourcing tasks.After experimental verification,when the crowdsourcing rejection rate is below 10%,the accuracy of crowdsourcing only slightly decreases.Compared with the allocation algorithm that does not consider dynamic allocation and the random allocation algorithm,the accuracy of crowdsourcing has been greatly improved.Finally,this paper develops a course knowledge graph error correction and completion system based on crowdsourcing.Its core functions are course knowledge graph visualization,course knowledge graph completion,course knowledge graph error correction,and crowdsourcing task details statistics.Both the learner and the teacher can complete the knowledge graph,but the processing strategy is different.The learner's completion will join the crowdsourcing process,and the teacher's completion will directly update the knowledge graph.The course knowledge graph error correction function distributes crowdsourcing tasks to learners through a series of algorithms constructed above,and gives them answers.Finally,the answers are aggregated to form a consensus answer that is fed back to the course knowledge graph.