Abstract:

In order to solve the knowledge redundancy caused by data accumulation during coal mine construction, some researches were conducted about domain knowledge management based on KG in this paper. The systematic analysis of the field of standard specification text was conducted using the safety management system structure and security risk management. 12 types of entities and 10 types of relations in the field of coal mine construction safety management were defined, and the knowledge structure model was also improved. 43 standard specifications were selected as data sources for text entity and relationship recognition by rules, dictionaries and deep learning methods. For different entity types, the framework of domain knowledge integrated approaches was proposed, and two models of bidirectional long-short-term memory(BiLSTM) and conditional random field(CRF) and bidirectional encoder representations from transformers(BERT)-BiLSTM-CRF were also compared. The accuracy, recall rate and F₁ value of the BERT-BiLSTM-CRF model are more than 7% higher than that of the BiLSTM-CRF model, which verifies the superiority of the selected model. Through knowledge extraction, knowledge storage and visualization, the entities contained in different types of entities in the field of coal mine construction safety and the relationship between different entities were investigated.

Key words: knowledge graph(KG), coal mine construction, safety field, safety management, knowledge structure, entity types