Design Of Knowledge Query System For Transmission Regulations Based On Knowledge Graph

Overhead transmission line regulations texts are mainly used to guide power grid staff in the safe maintenance of power transmission lines.The storage and query of regulations documents is of great significance to the safe construction and standard operation of power transmission lines.In this paper,based on knowledge graph technology,combined with natural language processing and graph database technology,a knowledge query system is constructed,which realizes the accurate query of regulations.The following is the main work of this article:(1)According to the characteristics of the power transmission regulations text,a bottom-up knowledge map construction method is adopted.First,the semantic relationship of the knowledge graph was designed,and then the training data was manually annotated,which made the preliminary preparations for the construction of the knowledge graph.In the process of constructing the knowledge graph,the relational extraction model is used to extract knowledge,and the extracted knowledge is manually reviewed and stored in the graph database.Finally,the knowledge map of transmission regulations is stored in the graph database in the form of triples,which realizes the storage of knowledge of transmission regulations.(2)The relation extraction model is used for knowledge extraction in the process of knowledge graph construction,and the named entity recognition model is used for the recognition of electric power terminology of query sentences in the process of knowledge query.On the one hand,by building a half-pointer and half-labeling model based on Seq2 seq to solve the problem of fuzzy entity boundaries in relation extraction,the experimental results show that the F1 value of this model in the transmission procedure text triple test set is 0.8376;on the other hand,it will be based on The BERTBi LSTM-CRF model is used for named entity recognition.Experimental results show that the F1 value of this model in the electric power terminology test set is 0.8145.(3)This article innovatively proposes a triple query method for knowledge query,mainly adopts a query matching coefficient algorithm in the screening and sorting of triple data sets.On the one hand,through comparison with algorithms such as Levenshtein Distance algorithm,it is proved that the query matching coefficient algorithm is more suitable for the screening and sorting of triples;on the other hand,in order to verify the superiority of the triple query method,the accuracy and retrieval The rate is the evaluation index,and the experiment shows that the triple query method is better than the node query method.(4)Based on the completion of the knowledge graph and knowledge query system,the software development of the power transmission regulation knowledge query system was completed through the Python language and related function libraries.The system realizes the functions of knowledge extraction,graph data import,knowledge query and knowledge visualization.