The use of the Bidirectional Long Short-Term Memory (Bi-LSTM) network can solve problems such as the imbalance of conversion efficiency between English-Korean and Korean-English, and improve translation quality by applying attention mechanisms, data enhancement and other methods. In order to improve the generalization performance of the model, data and other strategies were adopted to achieve the diversity of training samples through methods such as enhancement and synthesis, and different improvement methods were evaluated through interactive verification to strictly evaluate the model performance and ensure its effectiveness. To ensure that it is suitable for the actual use, application parameters were adjusted and how the model performs under various scenarios in various scenarios was monitored. Through comparative analysis, it was found that the average translation quality of the two sentences was 0.1 higher. Although there is not much difference between long sentences and long sentences, there is a significant difference in their translation quality. It can be seen that this method has certain difficulties in processing large sentences, which reduces the translation quality and efficiency.