Research On The Construction And Intelligent Question Answering Of Knowledge Graph Of Coalmine Occupation

Coalmine safety production has always been a major issue related to the overall situation in the coal mining industry.As coalmine cover a wide range of work types,it is of basic and important practical significance to effectively improve the professional knowledge and ability of coalmine operators to reduce coal mine accidents and promote the level of safety production.Through theoretical and practical research on knowledge graph and intelligent question answering,this thesis constructs a knowledge graph of coalmine occupations based on the collection of a large number of professional knowledge,and studies intelligent question answering system based on it,aiming to improve the lack of structured knowledge of coalmine work types and to provide a beneficial way for coalmine occupations to acquire relevant professional knowledge quickly and accurately.The main work of this thesis is as follows:Research on the construction of knowledge graph of coalmine occupations.The Schema of the graph is preliminarily constructed based on the domain data,and based on this Schema the relevant professional knowledge is collected and sorted out through the crawler framework,and subsequently the coalmine professional theory is used to classify the types of work.The BIO annotation strategy is used to annotate the data,and the BERT-Bi LSTM-CRF model is constructed to extract the knowledge of work types to form the work types entity and attribute information.Then the relationships of the knowledge graph are defined,and then the Schema of the knowledge graph of occupations is refined based on entities,relationships and attribute information,and the Neo4 j graph database is used to save the triple information to realize the construction of the knowledge graph of coalmine occupations.Research on the intelligent question and answer method of knowledge graph.Based on the characteristics of coalmine worker data and the constructed knowledge graph,the intelligent question answering process is designed and the method of BERT intention recognition and slot extraction is studied on this basis.The single-sentence classification task and the slot extraction task of BERT are studied respectively,based on which a joint BERT-wwm-ext model of intent recognition and slot extraction is constructed and compared with the metrics of the individual model tasks.Based on the characteristics of the knowledge graph data of coalmine occupations,the SentenceBERT model is used to associate the question mention words with the candidate target entities of the knowledge graph,and the final results of entity linking is obtained by combining the similarity matching and overlapping word comparison.Based on the identified intentions and extracted slot results,combined with the target entities of the entity linking,Cypher query statements are constructed,and the corresponding answer is obtained from the knowledge graph of coalmine occupations.Research and implementation of intelligent question answering system based on knowledge graph of coalmine occupations.The system is constructed by using languages such as Python and the Django framework,and the human-computer interaction of knowledge graph display,intent recognition and slot extraction,entity linking and question answering module of work types is completed.