Knowledge Fusion Of Multi-source Question Bank And Knowledge Graph Construction For Primary And Secondary Education

Knowledge graph represents the relationship between entities in the real world in a structured way,and expresses the digital information in a form closer to human cognition.The larger the scale of the knowledge graph,the more knowledge can be stored,and the greater the amount of stored knowledge,the greater the role of the knowledge graph.However,because the knowledge graph can be constructed by any individual or institution according to their own wishes,the data sources contained in it are extremely wide and the quality may be uneven,resulting in heterogeneity and diversity among the knowledge graphs.In this context,how to align the entities in multiple knowledge graphs and integrate the knowledge graphs is extremely important,because it will help to establish a unified large-scale knowledge graph.At present,the research results of entity alignment are still not ideal,and there are still difficulties in real application scenarios.Therefore,this paper attempts to explore based on the general data set and self-built application data set,study the model that can make more full use of the associated node information in the knowledge graph for entity alignment,and apply the model to the real scene of "primary and secondary education question bank",realize the integration of multiple question bank knowledge graphs,and complete the construction of the question bank system based on the final knowledge graph.The main research work of this paper includes the following aspects:(1)Two Chinese question bank knowledge atlas data sets are constructed.In order to study the effect of entity alignment model on Chinese item bank data set and build a complete item bank application system supported by large-scale knowledge base,the construction of domain knowledge graphs of two Chinese primary and secondary school item banks was completed.The main implementation process is to crawl multi-source question bank data through crawlers,and then construct two heterogeneous Chinese question bank knowledge graphs through knowledge extraction technology.(2)An entity alignment model NEI＿KBA.based on multi hop neighbor node embedding learning is proposed The existing entity alignment models based on embedded representation usually only consider the attribute embedding and structure embedding of the entity itself,but do not make full use of the information of other associated entities in the atlas.Taking this as the starting point,based on the methods of structure embedding and attribute embedding fusion,at the same time,the information of surrounding entities is used to enrich entity relationship triples and attribute triples,and the corresponding distance weight is increased.Then,the similarity between entities is predicted by embedding learning entity vector representation,entity alignment is carried out,and finally the knowledge graph fusion is completed.Comparing the NEI＿KBA model and the baseline method on two general datasets and two Chinese question bank datasets constructed in this paper.The experimental results show that NEI＿KBA model can make full use of the associated node information of the knowledge graph,and can well represent the relationship information and attribute information between entities.It has achieved better results than the baseline method in the task of entity similarity calculation and entity alignment.(3)A primary and secondary education question bank system based on large-scale knowledge base is constructed.Based on the knowledge atlas of two Chinese question banks constructed and fused,the design and development of primary and secondary school question bank system are completed by using the commonly used relational database and Middleware in the industry.The system adopts B/S architecture,the back end adopts springboot as the overall framework,mybatis as the data access layer framework,and the front end adopts Vue.js + Ant V framework.