Research on Semantic Verification Method of Control Instruction Based on Ensemble Learning

For flight safety, the semantic verification of radiotelephony communication is of great significance, which can avoid aviation accidents caused by communication errors between pilots and air traffic controllers. To achieve this goal, we used an ensemble learning strategy in machine learning and applied it to deep learning to build an ensemble learning model for semantic verification. Firstly, we used three different pre-trained language models based on the Transformer architecture for fine-tuning on the control instruction semantic verification dataset. Then, we used the Bagging ensemble classification method to integrate the trained base models, and finally used a voting method based on accuracy weights to output the ensemble model. The experimental results show that our ensemble model can significantly improve the discriminative performance of the model without increasing its complexity. Compared with the best-performing base model, the discriminative accuracy has increased by 1.4%.