Multi-scenario-oriented Construction And Evolution System Of Smart City Domain Knowledge Graph

Thanks to the continuous development of the smart city and the extensive application of information technologies in recent years,a large amount of data and knowledge have been accumulated in the city.These data and knowledge are distributed in various information systems of the city,without a unified format and semantic description,which makes it difficult to serve the smart city applications.In order to integrate and utilize the diverse and heterogeneous data in the smart city and empower the construction of the smart city,a semantic model of the smart city domain knowledge graph is needed to provide a unified concept and terminology system.However,the existing manual knowledge graph construction methods are inefficient,and the machine-based knowledge graph construction methods cannot meet the requirements of the quality.At the same time,the lack of functions of the existing knowledge management platforms also makes the construction of the knowledge graph challenging.In addition,the existing knowledge graph systems can only provide a single knowledge view,which is difficult to meet the needs of diversified application scenarios in the smart city.Based on this,this thesis mainly studies the construction and evolution,management and application of the smart city domain knowledge graph,and proposes a multi-scenario-oriented knowledge construction and evolution system of smart city domain knowledge graph,covering the construction and evolution,management and application of the smart city domain knowledge graph.The main work of this thesis is as follows:This thesis proposes a human-machine-cooperative based domain knowledge graph construction and evolution method.This method constructs a knowledge graph based on the structured data.In the construction phase,the machine algorithm is used to process the underlying data,transforming the structured data into a temporary knowledge graph.Then,the high-level concepts of the knowledge graph are established on the basis of the machine-built temporary knowledge graph manually to establish the initial knowledge graph.In the evolution phase,the newly-coming structured data is transformed into a temporary knowledge graph by the machine algorithm,and then the temporary knowledge graph is merged with the existing knowledge graph through the human-machine-cooperative based concept matching method.A small number of artificial verifications are used to improve the accuracy of concept matching,and as annotations to optimize the machine algorithm.Finally,the efficient and refined construction and evolution of the domain knowledge graph is realized.This thesis designs a cloud-based knowledge graph management tool,which provides users with browser-based knowledge graph visualization and collaborative editing function,knowledge graph version control function,and knowledge graph evolution traceability function.The visual collaborative editing function supports users to visually edit the knowledge graph in views of both tree structure and graph structure.This function also supports multi-person collaborative editing of the knowledge graph,which can detect and deal with conflicts caused by multi-person collaboration.The map management function provides efficient management of various historical knowledge graph versions in the knowledge graph evolution process.The knowledge graph evolution traceability function can visualize the differences between different versions of the knowledge graph,allowing users to intuitively browse the evolution process of the knowledge graph.This thesis proposes a multi-scenario-oriented trustworthiness evaluation method,which splits the trustworthiness of a knowledge into five knowledge credibility attributes,and specifies the connotation of knowledge trustworthiness.The trustworthiness related data and information in the construction,evolution,and application stages of knowledge are used as trustworthy evidences to comprehensively evaluate the trustworthiness of knowledge.After that,a knowledge trustworthiness grading model is proposed to classify the trustworthiness of knowledge.Finally,a flexible and customizable knowledge view function is proposed to provide different application scenarios with trusted knowledge views that meet their trustworthiness requirements.