Elamur Research⌕
☰
home / papers / 2026 / Giving Voice to the Constitution: Low-Resource Text-to-Speech for Quechua and Spanish Using a Bilingual Legal Corpus

Giving Voice to the Constitution: Low-Resource Text-to-Speech for Quechua and Spanish Using a Bilingual Legal Corpus

Paper · 2026 · John E. Ortega, Rodolfo Joel Zevallos and Fabrício Carraro

Summary

Quechua-speaking communities in Peru often cannot easily access legal and political information, which is mostly published in Spanish. This short paper aims to make the Peruvian Constitution available as audio in both Quechua and Spanish, and to create reusable speech data for Quechua research. The authors trained three open text-to-speech systems: XTTS v2, F5-TTS and DiFlow-TTS. For Quechua they used the Siminchik and Lurin corpora, which shrank to about 40 hours after quality filtering and morphological normalization. For Spanish they used roughly 218 hours of multi-dialect read speech. Cross-lingual transfer from Spanish is meant to offset the scarcity of Quechua data. The constitution text was split by article into aligned JSON and normalized for numbers, abbreviations and Roman numerals. The models were evaluated only on synthesized Quechua constitutional text, using automatic proxies for naturalness, speaker consistency, intelligibility and prosody. The smallest model, DiFlow-TTS, scored best on most metrics. The authors release checkpoints, inference code and per-article audio as a resource for accessibility and for future ASR and speech-translation work.

Key points

  • Training data: Siminchik (~97.5 h) and Lurin (~83.3 h) gave about 180 h of Southern Quechua speech. Removing clips under one second and noisy material left about 40 h. Spanish training data came from Guevara-Rukoz et al. (2020), totaling about 218 h.
  • Models compared: XTTS-v2 (470M parameters), F5-TTS (336M) and DiFlow-TTS (164M), all trained on NVIDIA H100 GPUs following each system's official recipe.
  • Evaluation metrics:
    • UTMOS as a proxy for perceived naturalness
    • SIM-O for speaker consistency
    • WER from the QUESPA Quechua ASR model for intelligibility
    • RMSE of F0 and energy for prosody
  • Results: DiFlow-TTS led on UTMOS (3.31), WER (0.16), RMSE F0 (10.24) and RMSE E (0.011). F5-TTS had the highest SIM-O (0.60).
  • Released resources: the model checkpoints produced by training, code for running inference, and generated speech covering every article of the Peruvian Constitution, drawn from the official 2020 Spanish–Quechua edition.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Universitat Pompeu FabraBarcelona · Spain47 works · 11,723 citations🎓Northeastern UniversityBoston · United States57 works · 6,533 citations🔬Barcelona Supercomputing CenterBarcelona · Spain14 works · 53 citations

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 worksSignal Processing · 452 works

🧭 Topics 3Research subjects and languages the work is about.Speech & spoken language · 480 worksLow-resource languages · 267 worksLegal AI · 134 works

Cite

Cite as John E. Ortega, Rodolfo Joel Zevallos and Fabrício Carraro. 2026. Giving Voice to the Constitution: Low-Resource Text-to-Speech for Quechua and Spanish Using a Bilingual Legal Corpus. Proceedings of the 3rd Workshop on Natural Language Processing for Political Sciences (PoliticalNLP 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/39457jk69sij

@inproceedings{ortega2026giving,
  title = {{Giving Voice to the Constitution: Low-Resource Text-to-Speech for Quechua and Spanish Using a Bilingual Legal Corpus}},
  author = {John E. Ortega, Rodolfo Joel Zevallos and Fabrício Carraro},
  booktitle = {Proceedings of the 3rd Workshop on Natural Language Processing for Political Sciences (PoliticalNLP 2026)},
  year = {2026},
  doi = {10.63317/39457jk69sij},
  url = {https://doi.org/10.63317/39457jk69sij}
}

References

29 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Milind Agarwal, Sweta Agrawal, Antonios Anastasopoulos, Luisa Bentivogli, Ondřej Bojar, Claudia Borg, Marine Carpuat, Roldano Cattoni, Mauro Cettolo, Mingda Chen, et al. 2023. Findings of the iwslt 2023 evaluation campaign. In Proceedings of the 20th International Conference on Spoken Language Translation (IWSLT 2023), pages 1–61.
  2. Victor Agostinelli, Tanel Alumäe, Antonios Anastasopoulos, Luisa Bentivogli, Ondřej Bojar, Claudia Borg, Fethi Bougares, Roldano Cattoni, Mauro Cettolo, Lizhong Chen, et al. 2025. Findings of the iwslt 2025 evaluation campaign. In Proceedings of the 22nd International Conference on Spoken Language Translation (IWSLT 2025), pages 412–481. find
  3. Ibrahim Sa’id Ahmad, Antonios Anastasopoulos, Ondřej Bojar, Claudia Borg, Marine Carpuat, Roldano Cattoni, Mauro Cettolo, William Chen, Qianqian Dong, Marcello Federico, et al. 2024. Findings of the iwslt 2024 evaluation campaign. In Proceedings of the 21st International Conference on Spoken Language Translation (IWSLT 2024), pages 1–11. ↗
  4. Nouman Ahmed, Natalia Flechas Manrique, and Antonije Petrović. 2023. Enhancing spanish–quechua machine translation with pre-trained models and diverse data sources. In AmericasNLP 2023, pages 156–162. find
  5. Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, et al. 2022. Speecht5: Unified-modal encoder-decoder pre-training for spoken language processing. In Proceedings of the 60th annual meeting of the association for computational linguistics (volume 1: Long papers), pages 5723–5738. ↗
  6. Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick Von Platen, Yatharth Saraf, Juan Pino, et al. 2021. Xls-r: Self-supervised cross-lingual speech representation learning at scale. arXiv preprint arXiv:2111.09296. ↗
  7. Loïc Barrault, Yu-An Chung, Mariano Cora Meglioli, David Dale, Ning Dong, Paul-Ambroise Duquenne, Hady Elsahar, Hongyu Gong, Kevin Heffernan, John Hoffman, et al. 2023. Seamlessm4t: massively multilingual & multimodal machine translation. arXiv preprint arXiv:2308.11596. ↗
  8. Wyndham A Bewes. 1920. The new constitution of peru (january 18, 1920). Journal of Comparative Legislation and International Law, 2(3):266–269. find
  9. Marcely Zanon Boito, Vivek Iyer, Nikolaos Lagos, Laurent Besacier, and Ioan Calapodescu. 2024. mhubert-147: A compact multilingual hubert model. arXiv preprint arXiv:2406.06371. ↗
  10. Ronald Cardenas, Rodolfo Zevallos, Reynaldo Ba-querizo, and Luis Camacho. 2018. Siminchik: A speech corpus for preservation of southern quechua. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Paris, France. European Language Resources Association (ELRA). find
  11. Edresson Casanova, Kelly Davis, Eren Gölge, Görkem Göknar, Iulian Gulea, Logan Hart, Aya Aljafari, Joshua Meyer, Reuben Morais, Samuel Olayemi, and Julian Weber. 2024. XTTS: a Massively Multilingual Zero-Shot Text-to-Speech Model. In Interspeech 2024, pages 4978–4982. find
  12. Chih-Chen Chen, William Chen, Rodolfo Zevallos, and John E. Ortega. 2024. Evaluating self-supervised speech representations for indigenous american languages. In Proc. LREC-COLING 2024. find
  13. Yushen Chen, Zhikang Niu, Ziyang Ma, Keqi Deng, Chunhui Wang, JianZhao JianZhao, Kai Yu, and Xie Chen. 2025. F5-TTS: A fairytaler that fakes fluent and faithful speech with flow matching. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6255–6271, Vienna, Austria. Association for Computational Linguistics.
  14. Abteen Ebrahimi, Manuel Mager, Shruti Rijhwani, Enora Rice, Arturo Oncevay, Claudia Garcia Baltazar, María Elena Méndez Cortés, Cynthia Montaño, John E. Ortega, Rolando Coto-Solano, Hilaria Cruz, Alexis Palmer, and Katharina Kann. 2023. Findings of the americasnlp 2023 shared task on machine translation into indigenous languages. In AmericasNLP 2023, pages 206–219. find
  15. Edward Gow-Smith, Alexandre Berard, Marcely Zanon Boito, and Ioan Calapodescu. 2023. Naver labs europe’s multilingual speech translation systems for the iwslt 2023 low-resource track. find
  16. Adriana Guevara-Rukoz, Isin Demirsahin, Fei He, Shan-Hui Cathy Chu, Supheakmungkol Sarin, Knot Pipatsrisawat, Alexander Gutkin, Alena Butryna, and Oddur Kjartansson. 2020. Crowdsourcing Latin American Spanish for Low-Resource Text-to-Speech. In Proceedings of The 12th Language Resources and Evaluation Conference (LREC), pages 6504–6513, Marseille, France. European Language Resources Association (ELRA). find
  17. Ministerio de Justicia y Derechos Humanos. 2020. Constitución Política del Perú en Castellano y Quechua, cuarta edición oficial edition. Ministerio de Justicia y Derechos Humanos, Lima, Perú. Sistema Peruano de Información Jurídica (SPIJ). find
  18. Ngoc-Son Nguyen, Thanh VT Tran, Hieu-Nghia Huynh-Nguyen, Truong-Son Hy, and Van Nguyen. 2025. Diflow-tts: Compact and low-latency zero-shot text-to-speech with factorized discrete flow matching. arXiv preprint arXiv:2509.09631. ↗
  19. John E Ortega, Richard Castro Mamani, and Kyunghyun Cho. 2020. Neural machine translation with a polysynthetic low resource language. Machine Translation, 34(4):325–346. ↗
  20. John E. Ortega, Rodolfo Joel Zevallos, William Chen, and Idris Abdulmumin. 2025. QUESPA submission for the IWSLT 2025 dialectal and low-resource speech translation task. In Proceedings of the 22nd International Conference on Spoken Language Translation (IWSLT 2025), pages 260–268, Vienna, Austria (in-person and online). Association for Computational Linguistics. find
  21. John E. Ortega, Rodolfo Zevallos, and William Chen. 2023. Quespa submission for the iwslt 2023 dialect and low-resource speech translation tasks. In Proc. IWSLT 2023, pages 261–268. find
  22. John E. Ortega, Rodolfo Zevallos, William Chen, and Ibrahim Said Ahmad. 2024. Quespa submission for the iwslt 2024 dialectal and low-resource speech translation task. In Proc. IWSLT 2024, pages 173–181. find
  23. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318. ↗
  24. Vineel Pratap, Awni Hannun, Qiantong Xu, Jeff Cai, Jacob Kahn, Gabriel Synnaeve, Vitaliy Liptchinsky, and Ronan Collobert. 2019. Wav2letter++: A fast open-source speech recognition system. In ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 6460–6464. IEEE. find
  25. Mónica Romero, Sandra Gómez, and Iván G. Torre. 2024a. Asr advancements for indigenous languages: Quechua, guarani, bribri, kotiria, and wa’ikhana. arXiv. find
  26. Mónica Romero, Sandra Gómez-Canaval, and Iván G. Torre. 2024b. Automatic speech recognition advancements for indigenous languages of the americas. Applied Sciences, 14(15):6497. ↗
  27. Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon, and Saining Xie. 2023. Convnext v2: Co-designing and scaling convnets with masked autoencoders. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16133–16142. find
  28. Rodolfo Zevallos, Núria Bel, Guillermo Cámbara, Mireia Farrús, and Jordi Luque. 2022a. Data Augmentation for Low-Resource Quechua ASR Improvement. In Interspeech 2022, pages 3518– 3522. ↗
  29. Rodolfo Zevallos, Luis Camacho, and Nelsi Melgarejo. 2022b. Huqariq: A multilingual speech corpus of native languages of peru forspeech recognition. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 5029–5034. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ortega-2026-giving-voice-constitution-low-resource/