| As the time of earthquake occurrence changes incrementally,and the types of information become more and more,which makes it difficult to identify the key information of earthquake emergency.It is possible to quickly acquire and process massive earthquake emergency information after the earthquake,but these information often contains a large number of "dirty data",that is,useless information that is not related to the earthquake.How to quickly deal with the quality problems in the emergency data,Extract valuable key information of earthquake emergency,such as the three elements of earthquake(time,place,person),casualties and economic losses;The intelligent analysis,processing and display of these disaster data is one of the key issues that need to be solved in the current earthquake emergency work.Based on natural language processing technology,this paper designs and improves the named entity recognition model based on BERT-BILSTM-CRF to extract key seismic information,and verifies the model performance through self-built data sets,and designs and develops a visualization system for earthquake emergency rescue on this basis.The main research work is as follows:Firstly,based on the three elements of earthquake information,this paper explores the request data acquisition framework for earthquake emergency events,and combines the web page cleaning algorithm based on node text density and coincidence density to realize the real-time acquisition and cleaning of post-earthquake earthquake news data,and on this basis,establishes the seismic named entity recognition data set.In addition,an intelligent identification algorithm of earthquake emergency information based on BERT-BILSTM-CRF is proposed in this paper.Based on the seismic named entity recognition data set,the sequence annotation method BIO is used to classify the seismic entities,and the BERT pre-training model is constructed to represent the sentence-level feature vector of the seismic emergency text;The Bi LSTM algorithm is used to obtain the context information of the two-way earthquake emergency text.In order to enhance the recognition effect of the key earthquake emergency information in the sentence,the conditional random field algorithm is used to extract the dependency between adjacent vectors,and the Self-Attention mechanism is used to extract the key earthquake emergency information from the earthquake emergency text,so as to realize the intelligent recognition of the earthquake emergency information.This model has been tested and validated in three earthquake emergency events in 2022,including the 6.1 magnitude earthquake in Lushan County,Ya’an,Sichuan,on June 1,the 6.0 magnitude earthquake in Malkang City,Aba Prefecture,Sichuan,on June 10,and the 6.8 magnitude earthquake in Luding County,Ganzi Prefecture,Sichuan,on September 5.The results show that the calculation efficiency and accuracy of this model are within one minute compared to traditional Bert model methods,and the recognition accuracy is improved from 86.99% to 92.49%.Experience shows that this model can effectively and accurately extract the earthquake emergency information from the network media.Finally,on the basis of the above research work,a visualization platform for earthquake emergency rescue is designed and implemented based on the model to improve the efficiency and quality of the processing,analysis and application of earthquake emergency data and provide technical support for scientific researchers in the earthquake industry. |
