Research On Construction Technology Of Coal Mine Safety Knowledge Graph

Recently,the construction of mine information has been presented the explosive growth.Emerging information technologies represented by the Internet of Things,big data,and artificial intelligence have been extensively developed in coal mine areas.The application of Emerging information technologies,which generally have the characteristics of scattered distribution,complex relationships,lack of effective management and not sufficient exploit,causing coal mine safety knowledge cannot be quickly acquired.The emergence of the knowledge graph provides a new way to solve these problems.This paper focuses on named entity recognition,relationship extraction for coal mine safety data and store the acquired entity relationship triples into the Neo4 j graph database to complete the knowledge base construction.To better practice the research content at the same time,this paper also develops the web visualization system of coal mine safety knowledge graph.For entity recognition in the field of coal mine safety,firstly,this paper obtains text data in two steps.The first step is to collect policies and regulations of coal mine safety,and the second step is to obtain data from the encyclopedia website through crawler technology.The obtained data should be preprocessed.Secondly,in the construction of the named entity recognition model,this paper explores the commonly used Bi LSTM-CRF model,and adds the ALBERT Chinese pre-training model to the basic model,which effectively solves the polysemy phenomenon that the ordinary word vector model cannot solve.Finally,to solve the problem of insufficient coal mine safety data,this paper uses the parameters obtained by the model training of the THUCNews public corpus as the initialization parameters of the corpus training.Through the multi-layer perceptron(MLP),The knowledge obtained from THUCNews corpus is transferred to the field of coal mine safety.Experimental results show that the performance of the ABMC model is better than the original model.After obtaining the result of named entity recognition,for relation extraction in the field of coal mine safety,this paper applies two methods.The first way is to analyze the advantages and disadvantages of the pipeline model and the joint model.The joint extraction model not only can share the coding layer in the entity recognition task and the relational extraction task,but also can use the knowledge acquired by named entity recognition in the relational extraction task,which effectively solves the problem of error propagation in the pipeline model and neglecting the relationship between the two tasks.the ALBERT joint extraction model adapted in this paper to extract entity relationships from Baidu Encyclopedia and literature data.The second method is to extract entity relation triples by designing dependency parsing template and store the extracted triples into the Neo4 j graph database to complete the construction of the coal mine safety knowledge base.Based on the research of entity and relationship extraction,this paper constructs a web visualization system about knowledge graph in the field of coal mine safety.The system realizes functions of knowledge query and knowledge management,which can help users learn coal mine safety knowledge quickly,accurately and intuitively.