| With the rapid growth of power demand,a large number of data,including a large number of unstructured text,has been accumulated in the field of power operation and inspection,which makes it inconvenient for full-time personnel to read and organize information.To carry out the automation and intelligent work in the field of power operation and inspection is the inevitable demand of modern transmission-line system of electric power.The knowledge graph expresses the way humans perceive everything in the world in the form of graphs.The knowledge graph technology can be used to discover knowledge,effectively organize knowledge and efficiently retrieve knowledge,thereby helping the electric power transportation inspection field to realize business intelligence.At present,there are two problems in knowledge extraction of electric power operation and inspection domain:one is the lack of knowledge extraction algorithm model for electric power transportation inspection domain;the other is the lack of publicly available annotation corpus.Therefore,this paper studies and refers to the research materials and specific application status of the knowledge extraction algorithm in other fields,and finally combines the actual situation of this paper to design and implement the knowledge extraction algorithm for electric power transportation inspection.Specifically,the main work of this paper is as follows.(1)Aiming at the lack of publicly available annotation corpus in the field of power operation and inspection,a pseudo annotation dataset"BUPT-Power" for power operation inspection entity extraction is constructed.This paper proposes two entity extraction methods,BERTHuffinan based on Huffinan encoding and BERT-Balanced based on balanced encoding,which reduce the class imbalance problem in label construction.(2)In order to solve the problem of relation extraction caused by the complex text structure of power operation and inspection,this paper proposes a method of power operation and inspection relation extraction based on dependency syntax analysis.This method first extracts entities and then performs the syntactic analysis.According to the sentence components and the relationship between sentences,the triples are initially determined,and then the available triples are obtained through the rule conversion and filtering of the power operation inspection.This method can effectively discover the relationship of the power operation inspection and reduce the difficulty of extracting the power operation inspection relationship.(3)Based on the research results of the above-mentioned data and algorithms,this paper designs and implements the power operation inspection knowledge graph management system.Specifically,this paper uses the power transportation inspection triad to construct the power transportation inspection knowledge graph and save it in the graph database.The system visually displays the query results through the frontend interface,helping full-time personnel to quickly and intuitively learn and understand the knowledge of power operation inspection.The system test results show that the system has perfect functions and can run well online. |
