Research On Technology Of Automatic Classification Of Civil Aviation Safety Information Based On Text Augment

Safety is the eternal theme of civil aviation,and safety information is an important part of safety management work.With the development of civil aviation industry,safety information is increasing day by day,and it is impossible to meet the needs of information processing and analysis only by manual,so it is necessary to introduce automatic classification technology to process a large number of data.To solve the problem of incomplete safety information type labels,an automatic classification method for civil aviation safety information based on text augment is proposed.The main work is as follows:Firstly,the unsafe event information in the Chinese civil aviation safety information system is divided into three subsets according to the number of single event type event information.The Bert model and Easy Data Augment algorithm are used to construct the Support Vector Machine automatic classifiers,and the model performance of data sets under different orders of magnitude is analyzed after augment.The results show that the F₁value increases by 9.66%on a data set with hundreds of orders of magnitude of events for a single event type.Secondly,for different types of information,classifiers are constructed in combination with the SVM binary classification and unary classification model based on the text augment.According to the actual application scenarios and traditional evaluation indexes,a new index system and a classifier screening scheme are constructed,they are used to obtain the optimal classifier corresponding to each type,then the classifiers are used to re-predict and complement the type labels of event information.Finally,association analysis and verification are carried out on the results before and after complementing labels.The results show that the complement of label is effective,and the use of automatic classification to complement the label is beneficial to the improvement of information quality.