Construction and Application of Knowledge Graph in Nuclear Safety Field
Nuclear safety knowledge has the characteristic of fragmentation,which lacks structured knowledge organization and intuitive display,making it difficult to be rapidly popularized.In order to solve this problem,a method for building a nuclear safety knowledge graph based on Neo4j has been proposed.Firstly,entities and relations is defined by combining with the nuclear safety registered engineer examination syllabus,which takes the nuclear safety comprehensive knowledge and nuclear safety laws and regulations as research objects.Then,a knowledge graph based on Neo4j graph database is constructed,combining with manual extraction of entities and relations in 41 nuclear safety related laws to construct a corpus,using BiLSTM-CRF model to extract entities in nuclear comprehensive knowledge.The constructed knowledge graph contains 6743 nodes and 8052 relations,and comparing with the traditional BiLSTM model,the resultant F1 value of the BiLSTM-CRF model is improved to 89.06%,which realizes the visual display of nuclear safety knowledge and Q&A function,and the Q&A system is verified to have an average accuracy rate of 86.00%.The nuclear safety knowledge graph facilitates the structured and objective presentation of nuclear safety information,advancing the dissemination of nuclear safety knowledge,and serving as a reference for the application of knowledge graphs in the nuclear safety field.
万方数据
