Research Of Relation Extraction Method Of Aviation Emergency Text

Ontology learning refers to applying machine learning methods to automatically acquire concepts and relations between concepts.With the rapid development of civil aviation industry and Internet information technology,management of aviation safety puts forward more important requests to automatic learning of domain ontology.This paper applies deep learning methods into the domain of civil aviation emergency,and uses unstructured civil aviation emergency texts as data source to research on the relation extraction between concepts in the domain ontology.This paper makes a detailed study of relation extraction models based on deep learning methods,and focuses on the application of attention mechanism in natural language processing,then proposes a classification model combining attention mechanism and BiGRU(Att-BiGRU).Firstly,the text words are mapped into vectors;Secondly,BiGRU is constructed to obtain context semantic information of word sequence;Thirdly,word-level and sentence-level attention mechanism are used to allocate more weights to words and sentences that are more important to semantic representation,and weights of noisy data are reduced;Finally,the model is trained and optimized,and experiments are conducted on the public data set for measurement and evaluation,and the results prove the validation of the model.To address the problem that the current accuracy of relation extraction in the domain of civil aviation emergency is low,the Att-BiGRU model is applied to the relation extraction of civil aviation emergency domain ontology.Firstly the preprocessing of civil aviation emergency texts is described.Then the necessary parameters,functions,optimization and regularization methods are analyzed and chosen,and the detailed steps of relation extraction are explained.The analyses of experimental results and comparisons with traditional methods show that this model has further improved the accuracy of relation extraction,which provides better method support for automatic learning of relation extraction of civil aviation emergency domain ontology.