The performance of neural machine translation on low-resource language pairs remains sub-optimal, primarily due to the lack of large parallel corpora or sufficient linguistic knowledge. To address this challenge, we propose a machine translation approach for Chinese-Turkish, leveraging data augmentation and contrastive learning. By integrating a bilingual dictionary as an external resource, we enhance word alignment and bring the vector representations of aligned words closer, effectively expanding the training dataset and improving the model’s generalization capabilities. Additionally, we introduce contrastive learning to further refine translation performance. Experimental results demonstrate that our proposed method significantly outperforms baseline models in the translation tasks of Chinese-Turkish and other low-resource languages. Ablation studies further confirm the effectiveness of combining data augmentation with contrastive learning in enhancing translation model performance.