IWSLT2023 - Low-resource Speech Translation Track: Quechua-Spanish Parallel Corpus

Repository for sharing the data in the Quechua language, one of the languages for the low-resource speech translation track at IWSLT 2023.

Parallel data

This corpus is a small fraction of the Siminchik corpus (Cardenas et al., 2018) translated into Spanish. It comprises 1 hour, 40 minutes of clean speech in Quechua, translated into Spanish (que_spa_clean). This version contains a validation and test data set.

The raw text transcriptions are located in que_spa_clean/<split>/txt/<split>.<lang>.

True-cased Spanish target translations are found in que_spa_clean/<split>/txt/<split>.spa.tc.

True-casing was done with a sacremoses Truecaser model trained on the Spanish side of WMT13 EN-ES.

Additional Audio data

In addition to the 1 hour, 40 minutes of Quechua audio data aligned with Spanish translations, we also provided participants with a corpus of 60 hours of fully transcribed Quechua audio. All of this data comes from the Siminchik corpus.

Participants are not required to use any of this data.

Due to their size, this data is made available through this website.

Citation

@article{cardenas2018siminchik,
  title={Siminchik: A speech corpus for preservation of southern quechua},
  author={Cardenas, Ronald and Zevallos, Rodolfo and Baquerizo, Reynaldo and Camacho, Luis},
  journal={ISI-NLP 2},
  pages={21},
  year={2018}
}

Additional Parallel Text data

As part of the constrained task, we allow parallel text from previous work. Participants are also not required to use this data.

The data is found in this repository and called: "additional_mt_text". They are extracted from the JW300 and Hinantin websites and used in the cited work below.

Citation

@article{ortega2020neural,
  title={Neural machine translation with a polysynthetic low resource language},
  author={Ortega, John E and Castro Mamani, Richard and Cho, Kyunghyun},
  journal={Machine Translation},
  volume={34},
  number={4},
  pages={325--346},
  year={2020},
  publisher={Springer}
}

License

All audio recordings are property of Siminchikkunarayku and Llamacha.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.

Acknowledgements

This work has been funded by AmericasNLP-2022 and Llamacha. Special thanks to John Ortega for his support.