Improving Indigenous Language Machine Translation with Synthetic Data and Language-Specific Preprocessing
This work addresses the data scarcity problem for low-resource indigenous language translation, but the gains are incremental and limited to specific language pairs.
The authors augment parallel corpora for indigenous languages (Guarani-Spanish, Quechua-Spanish) with synthetic data from a multilingual model and apply language-specific preprocessing, achieving consistent chrF++ improvements over baseline mBART models.
Low-resource indigenous languages often lack the parallel corpora required for effective neural machine translation (NMT). Synthetic data generation offers a practical strategy for mitigating this limitation in data-scarce settings. In this work, we augment curated parallel datasets for indigenous languages of the Americas with synthetic sentence pairs generated using a high-capacity multilingual translation model. We fine-tune a multilingual mBART model on curated-only and synthetically augmented data and evaluate translation quality using chrF++, the primary metric used in recent AmericasNLP shared tasks for agglutinative languages. We further apply language-specific preprocessing, including orthographic normalization and noise-aware filtering, to reduce corpus artifacts. Experiments on Guarani-Spanish and Quechua-Spanish translation show consistent chrF++ improvements from synthetic data augmentation, while diagnostic experiments on Aymara highlight the limitations of generic preprocessing for highly agglutinative languages.