wjosielct/aymara-spanish-parallel-corpus

Aymara-Spanish Parallel Corpus

Dataset Summary

This dataset contains aligned sentences in Aymara (ay) and Spanish (es), designed for training and evaluating Neural Machine Translation (NMT) models for low-resource Andean languages.

The data has been collected from religious texts and educational resources, shuffled to disrupt the original narrative flow, and split into training, validation, and test sets.

⚠️ Usage Warning: This dataset is distributed under a Non-Commercial license (CC BY-NC 4.0). It is intended strictly for academic research and educational purposes.

Dataset Structure

The dataset is partitioned as follows:

• Train: Used for model training.
• Validation: Used for hyperparameter tuning and evaluation during training.
• Test: Reserved for final model evaluation.

File Format

• Files are plain text (.txt), with one sentence per line.
• Parallel files (e.g., train.aymara and train.spanish) have the same number of lines and are aligned line-by-line.

Citation (Important)

If you use this dataset in your research, please cite the canonical version hosted on Zenodo using the following DOI:

BibTeX:

@dataset{corbera_terrones_2026_18193320,
  author       = {Corbera Terrones, Josiel},
  title        = {Aymara-Spanish Parallel Corpus for Neural Machine Translation},
  month        = jan,
  year         = 2026,
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.18193320},
  url          = {https://doi.org/10.5281/zenodo.18193320}
}

APA:

Corbera Terrones, J. (2026). Aymara-Spanish Parallel Corpus for Neural Machine Translation [Data set]. Zenodo. https://doi.org/10.5281/zenodo.18193320

Disclaimer

This is a derived dataset. The original copyrights of the source texts belong to their respective owners. The shuffling and processing were performed to support Fair Use in computational linguistics research.