Research on Emergency Management of Coal Mine Accidents Based on Information Extraction and Knowledge Graphs

Currently, the experience and knowledge of coal mine accident description and disposal are mainly in electronic documents and other unstructured forms, which are complicated, seriously fragmented, and poorly shared and cannot meet the demand for digitization of case knowledge for emergency response. The emergency of natural language processing and Knowledge Graph technology reorganize texture knowledge, improve data utilization efficiency, and explore information utilization value. With the goal extracted from automatic accident reporting entity, relationship and event information, a rule-based approach is used to match long-tail entities, a unified text-to-structure generation system is constructed using the Universal Information Extraction model, and the target structure of information extraction is generated adaptively to achieve unified extraction of entities, relationships and events. According to the experiment, the ontology model proposed in this paper can better digitally represent coal mine accidents. By pre-training and parameter adjustment of the UIE model, the extraction from entity relationship achieves 60.7% accuracy, 62.3%recall, and 61.5% f1 value in the case of the small-scale sample labeling. The construction of the knowledge graph about the coal mine accident provides a basis for intelligent search and Q&A, knowledge inference and accurate analysis, and intelligent recommendation and decision-making, which is of great significance to improve further the digitalization and intelligence level of emergency management in a coal mine field.