Construction And Applications Of Knowledge Graph Of Railway Travel

With the continuous development of railway infrastructure construction in China and the sweeping wave of the Internet,a large amount of railway travel data has been generated.As a way of the structured organization of domain knowledge,knowledge graph ha s emerged in various fields.Based on this background,the thesis builds the railway travel knowledge graph and utilizes the knowledge graphs of Chinese tourist attractions and Chinese major cities’ information as the third-party knowledge bases to expand it,to provide users with visual display services and link prediction services of railway travel knowledge.The main work of this thesis is as follows:(1)By exploring the knowledge of fields related to regions,tourism,and railway,the knowledge graph of railway travel is constructed based on mixing bottom-up and top-down knowledge modeling methods.First,the tool of the requests is applied to obtain the railway station information,passing and stopping information,corresponding fare,and other data from a website.Second,the visual ontology modeling tool Protégé is used to construct the ontology of railway travel knowledge graph.Then,the mapping relationship between the railway travel data and the ontology of the railway travel knowledge graph is established through instantiation.Then,through the fusion of the ontology layer and the instance layer,the knowledge graph of railway travel is completed and expanded.Finally,the data in the railway travel knowledge graph is supplemented by reasoning rules and stored in the Neo4 j diagram database.(2)A knowledge representation learning model named TConv2 Conv based on CNN is proposed.First,the feature input module is to generate the input in the model that is fed into the residual convolution module to extract the interaction features within the triple,and then the eigenvectors are obtained by combining the different levels of interactive features in the feature fusion module.Finally,the probability of a positive triple is predicted in the feature mapping module.The effectiveness of the model is verified by comparison experiments with the baseline models in the task of link prediction.(3)A knowledge representation learning model named Desc HAKE based on entities’ text information is proposed.In the model,the entities are mapped into the polar coordinate system,and the relation patterns such as symmetry,antisymmetry,inversion and interchangeable composition are modeled through the two coordinate information of the polar coordinate system,and the text description information of the entities is introduced.The distance vector of the modulus module is projected into the entity semantic hyperplane through the hyperplane projection technology to model the interaction between the triple and the entities’ text description information and further improve the performance of knowledge representation.The effectiveness of the model is verified by comparison experiments with the baseline models in the task of link prediction.(4)Based on the railway travel knowledge graph and the KRL model,a visualization system of railway travel knowledge retrieval is built using Spring Boot framework.It provides users with retrieval service and link prediction service of railway travel knowled ge,and achieves visualization of atlas data in the front end through D3.js.