Using Machine learning Techniques to Improve the Accuracy and Fluency of Japanese Translation

This study focuses on improving the performance of NMT models and explores three main optimization strategies through experimental methods: introducing attention mechanisms, applying data augmentation techniques, and performing hyperparameter tuning. Firstly, we established a benchmark NMT model based on the Transformer architecture and trained and tested it on publicly available WMT datasets. Subsequently, we integrated a self attention mechanism to enhance the model’s ability to handle long-range dependencies. In terms of data augmentation, we improved the robustness and generalization ability of the model by implementing random perturbations and noise addition on the training data. The experimental results show that each optimization strategy improves the BLEU (Bilingual Evaluation Understudy) score of the model to varying degrees, with hyperparameter tuning having the most significant impact on the performance improvement of the model. This study confirms the effectiveness of attention mechanisms and data augmentation techniques in improving the translation quality of NMT models, and emphasizes the importance of hyperparameter tuning in the training process of neural networks.