Abstract: With the development of railway informatization, the importance of using algorithms to automatically parse a large number of railway dispatching commands has become increasingly prominent. This paper proposed a railway dispatching commands parsing algorithm based on a generative summarization model and a knowledge distillation algorithm. The algorithm used a generative summarization model to analyze railway dispatching commands end-to-end, which had high accuracy and strong robustness, and was suitable for railway dispatching commands of various writing styles. The paper used multiple lightweight strategies such as knowledge distillation algorithm to design new loss functions and multiple model initialization strategies to reduce model size and improve algorithm speed. The algorithm achieved a Rouge-2 score of 21.634 2 on the railway dispatching commands dataset, with an inference time of 103 ms. It provides a reference for the deployment of railway dispatching commands parsing technology in railway scenarios.