Research On Entity Recognition And Relation Extraction Method Of Civil Aviation Emergency

The core of emergency management of civil aviation emergency is to make fast and effective decision-making on civil aviation emergency,so that decision makers can timely and creatively resolve the crisis.With the rapid development of the civil aviation industry and Internet information technology,higher standards and requirements for emergency management capabilities of civil aviation under the new normal have been proposed.Knowledge graph is a new type of knowledge representation tool that can effectively realize the expression,retrieval and reasoning of domain knowledge.Research on the key issues of civil aviation emergency Knowledge graph construction is of great significance to further improve the emergency management capabilities of civil aviation emergency.The core content of civil aviation emergency Knowledge graph construction includes entity recognition and relation extraction.Among them,civil aviation emergency entity recognition and relation extraction are the key tasks for constructing civil aviation emergency Knowledge graph.Therefore,based on textual information related to civil aviation emergency investigation and tracking reports,Micro Blog,WeChat,and TikTok,this paper deeply studies the issues of civil aviation emergency entity recognition and relation extraction.To address the problem that traditional methods cannot automatically obtain entities of civil aviation emergency,this paper proposes a method combining Bi LSTM with CRF.Firstly,entities are automatically marked by domain lexicons.Then labeled data is converted to distributed character vectors.Secondly,contextual features of the sequence text are got via BiLSTM.Finally,sequence label results are gained by CRF.The experimental results show that,compared with the traditional CRF method,the f-value of this method is increased by about 6.6%,which effectively solves the problem that the traditional method is inefficient in identifying the composite entity or mixed entity in the text information of civil aviation emergencies and uses a large number of manually defined feature templates.Aiming at the problem of low accuracy in the extraction of civil aviation emergency relations,combined with the data characteristics of civil aviation emergency text information,this paper proposes a model of civil aviation emergency relations extraction,Datt-Bi LSTMs,based on the combination of multi-level attention mechanism and double-level BiLSTM.Firstly,each word of text information is vectorized and fused with feature vectors such as part of speech,entity position,etc.,and then all feature vectors are spliced and input into Bi LSTMs model to capture higher-level context representation;secondly,attention mechanism is introduced at word level and sentence level to obtain the importance of each word,and at the same time,semantic information is effectively used to reduce the impact of noise;finally The output relation is extracted by softmax classification.In the data set of civil aviation emergency relation extraction,the F value of this method is 79.3%,which is further improved than the traditional neural network model method,and provides a better method support for the automatic construction of civil aviation emergency Knowledge graph.