metadata

license: mit
language:
  - ru
  - az
  - lez
size_categories:
  - 10K<n<100K
configs:
  - config_name: default
    data_files:
      - split: train
        path:
          - parallel/train.csv
      - split: test
        path:
          - parallel/test.csv
      - split: mono
        path: mono/lezgi_gazet_05102025.csv

Neural machine translation system for Lezgian, Russian and Azerbaijani languages

We release the first neural machine translation system for translation between Russian, Azerbaijani and the endangered Lezgian languages, as well as monolingual and parallel datasets collected and aligned for training and evaluating the system.

Parallel corpus sources:

Bible: parsed from https://bible.com, aligned by verse numbers (merged if necessary). "Oriental translation" (CARS) version is used for the Russian language as it is more coherent with Lezgian translation
Quran: parsed from https://quranacademy.org. Russian translation by Abu Adel is used as it is more coherent with Lezgian translation
Qusar encyclopedia: manually parsed and aligned from the book "Qusar, qusarılar. КцӀар, кцӀарвияр" by Sedaqet Kerimova. Contains only Lezgian-Azerbaijani pairs

Mono corpus sources:

Lezgi gazet: the largest newspaper in Lezgian language. Parsed all available archives as of 05/Oct/2024 from the website (https://lezgigazet.ru/)

Citation

If you use that results in your research, please cite our paper:

@misc{asvarov2024neuralmachinetranslationlezgian,
      title={Neural machine translation system for Lezgian, Russian and Azerbaijani languages}, 
      author={Alidar Asvarov and Andrey Grabovoy},
      year={2024},
      eprint={2410.05472},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2410.05472}, 
}