Improving Khmer-Vietnamese Machine Translation with Data Augmentation methods
Hanh Pham Van, Lê Thanh Hương · 2022
Machine translation has achieved significant improvements with the development of neural models. However, such approaches require large-scale parallel data, which are hard to collect for low-resource language pairs. This paper solves this problem by applying a pretrained multilingual model and fine-tuning it with a low-resource bilingual dataset. In addition, we propose two data-augmentation strategies to receive new training data, including: (i) back-translating with the dataset from the source language; (ii) translating sentences from the source language to the target one through a pivot language. The proposed approach is applied to the Khmer-Vietnamese machine translation. Experimental results show that our proposed approach gains 4.426% BLEU score higher than the Google translator model using a test set of 2000 Khmer-Vietnamese sentence pairs.