Machine translation is a hard problem in Natural Language Processing (NLP). Today, based on deep learning, the quality of machine translation systems is increasing strongly in rich-resource language pairs, but it is limited in low-resource language pairs, such as the Vietnamese-Laos language pair. In this work, we propose a method that depends on augmenting data and using a bilingual dictionary to improve the quality of machine translation for low-resource language pairs. Applying it to the Vietnamese-Laos language pair, it improved 14.0 BLEU scores for translating from Laos to Vietnamese and 3.5 BLEU scores for translating from Vietnamese to Laos. In addition, we built a good quality Vietnamese-Laos bilingual corpus that includes 182.208 sentence pairs to share with the community.