Research On Construction Of Education Knowledge Graph Based On Crowdsourcing

The internet and education,is known as the two "wheel" which can promote social progress in the 21 st century.Online learning as a classic application of the "Internet+Education",is profoundly affecting our education theory and method.Learning based on network resources is drawing more and more peoples' attention.A large number of digital learning resources are released online.But these large amount of resources are multi-source and heterogeneous,and loss their organization,which brings great challenge to knowledge learning and knowledge sharing under the background of big data.Knowledge graph is a structured semantic knowledge base,which presents knowledge resources and their carriers with the method of graphical description,and shows the overall structure?development process and associations of knowledge by visual graph.Crowdsourcing is a new internet-based social collaboration mechanism,which directly releases questions online and makes full use of the wide web users to produce results that beyond individual intelligence.Crowdsourcing model can fully mining and use of individual and swarm intelligence,and has been widely used in many fields.Faced with the current problems,we apply knowledge graph as carrier,and crowdsourcing as the method of acquisition and integration of the knowledge subgraph,to build the framework of the education knowledge graph,which can implement "individual knowledge" construct and "intelligence of crowds" fusion.In this thesis,our main work is as follows:(1)We define the formal representation of the knowledge graph,and give the construct framework of knowledge graph based on crowdsourcing technology.The education knowledge graph is abstracted as a weighted undirected graph,which is formed with knowledge point as node,knowledge relationship as edge,knowledge point's importance as node's weight and knowledge relationship's importance as edge's weight.The framework contains 3 modules: knowledge subgraph acquisition,knowledge subgraph fusion,and knowledge graph visualization.First,the task publisher set crowdsourcing tasks-keywords,then learners execute these tasks by associating knowledge to build knowledge subgraph which contains "individual intelligence",afterwards we update keywords though the new words in the subgraph then learners can continue the association.Second,we pretreat all the data,calculate the weight of knowledge point and knowledge association,and combine the individual knowledge subgraph to produce the knowledge graph that contains the "wisdom of crowds".Finally,the graph database Neo4 j is used to store and visualize the knowledge graph.(2)A new strategy to fusion knowledge graphs based on crowdsourcing is given.This strategy converts the knowledge subgraph into the adjacency matrix,then we delete duplicate,redundant,and low-frequency words,calculate the knowledge point weight by its difficulty degree and the subjective scores of learners,calculate the knowledge association weight by the quality of the knowledge subgraph contain it.At last,the fusion of knowledge subgraphs is realized through matrix weighted arithmetic.(3)The comparison and evaluation method of the knowledge graph is given.Learners make subjective assessments of the knowledge points and the subgraph through crowdsourcing.Then we evaluate the quality of knowledge subgraphs and the ability of learners according to the accuracy rate,recall rate and subjective scores of learners through the comparison of the knowledge subgraph with the finall knowledge graph.(4)We build a knowledge graph of software development area in the field of IT industry,and analysis the effectiveness of our approach by compare it with the "HowNet" method and the human-annotated data sets "Words-240".Experiments show that the method we proposed can give full play to the "intelligence of crowds",and has a good experimental effect,which provides a new way to construct education knowledge graph.