In response to the insufficient information obtained from traditional text sentiment analysis, as well as the underutilization of emotional resources and other textual information, which leads to inadequate text information for determining the sentiment polarity of comments, an ERNIE-BiGRU-Attention model was proposed for sentiment classification of civil aviation passenger evaluations. Firstly, Easy Data Augmentation (EDA) was applied to enhance the dataset. Subsequently, based on Baidu's proposed pre-trained language model, Enhanced Representation through Knowledge Integration (ERNIE), emotional knowledge was extracted from the text. In terms of passenger evaluation feature extraction, Bi-directional Gate Recurrent Unit (BiGRU) was utilized and an attention mechanism was incorporated to extract features from the input. Finally, The results indicate that the proposed model performs excellently in classification, achieving a binary classification accuracy of 97.21%, a comprehensive F1 average of 0.975 9, and a 0.73% improvement in accuracy compared to the baseline model.