Construction And Application Of Coal Mine Safety Knowledge Graph Based On Historical Accidents

Coal is the pillar industry of China’s energy,coal mine safety occupies an extremely important position,and it is the implementation of coal mining basic requirements.At present,the relevant information of coal mine accidents exists on various websites in the form of accident overview text.These accidents cover the knowledge of accident time,accident place,accident mine,accident casualties and so on,which is of great significance to coal mine safety.How to efficiently manage these heterogeneous and disorganized data has become a hot topic in recent years.The construction of knowledge map of coal mine accidents can realize the effective integration and continuous accumulation of coal mine accidents as well as the rapid retrieval of accident cases.Meanwhile,it can complete the accident analysis by statistical data such as the time,place and cause of accidents,and provide knowledge support for accident prevention.The construction of knowledge map of coal mine accidents is of great significance for improving the efficiency of coal mine safety work.Therefore,based on the above background and from the actual situation,this paper introduces the relevant methods of deep learning to build the coal mine safety knowledge map,and conducts indepth research and analysis on the named entity recognition model and relationship extraction model involved in the construction of coal mine safety knowledge map.The main research contents are as follows:This paper proposes a dual network entity recognition model of coal mine naming based on word features.Then,ALBERT and Word2 vec are used to obtain word and word vectors,and the splined vectors are introduced into the bidirectional LSTM model and the iterative expansive convolutional network respectively.Compared with a single model,the parallel use of IDCNN mainly aims to enrich the feature output of the model.By connecting the two modules in parallel,text features of different granularity can be extracted,and the accuracy of coal mine naming entity recognition can be improved by increasing the output text features.The ALBERT-BiLSTM-ATT model with location features is proposed to effectively solve the feature representation of the same entity in different relations through the special labeling of entity pairs in input sentences.At the same time,dependency parsing is introduced to extract triples of coal mine safety regulations,and the triples of coal mine knowledge are extracted by combining these two methods.Finally,the knowledge map of coal mine accident cases is constructed,and its effectiveness is verified,which confirms the accuracy and feasibility of the construction of coal mine safety knowledge map,and provides data support for the final coal mine safety search system.The coal mine safety knowledge graph search system is developed by Flask framework,which helps users to quickly search knowledge without proficient knowledge of Cypher query statements.Figure [40] Table [15] Reference [98]...