Research on Name Entity Recognition Method in Civil Aviation Text

Named entity recognition technology has been applied to various fields in recent years, obtaining business entities in civil aviation passenger reviews can quickly locate pain points in the service process. The benchmark BiLSTM model performs well for general text, but there are a large number of composite entities in civil aviation public opinion, which needs to improve the model's ability to recognize local text. The paper propose an ODCNN-BiLSTM-CRF model integrated with convolutional neural network. We obtain the standard data set through manual annotation and semisupervised learning firstly. Then learn the word embedding vector from the dataset. We use one-dimensional convolutional neural network (ODCNN) to obtain local context features of text data, and use long and short-term (LSTM) neural network to obtain context features that depend on long text, the two features were fused to obtain the predictive labels through conditional random field (CRF). Experimental results show that our three proposed models improved the recognition performance compared with the previous benchmark model, with the F value reaching 74.01%, 73.72% and 68.06%.