Elamur Research⌕
☰
home / papers / 2026 / Amulwe Kimün: A Community-Grounded Demo, Resource, and ASR Baseline for Mapuzugun

Amulwe Kimün: A Community-Grounded Demo, Resource, and ASR Baseline for Mapuzugun

Summary

Mapuzugun, the agglutinative and polysynthetic language of the Mapuche people of Chile and Argentina, keeps losing fluent speakers. It has no single standard orthography, and few of its digital tools stay publicly available for long. The authors present Amulwe Kimün, a proof-of-concept web platform co-designed with a Mapuzugun-speaking developer, native instructors and learners. It was built within a FACSO (University of Chile)–CONADI immersion teaching program (PEM). The interface defaults to Mapuzugun, with an option to switch to Spanish. It offers course-based quizzes (multiple-choice, ordering and open-answer items) plus forums and chat to support a learner community. In a pilot, participants scored high on recognition-type items but very low on free-text answers. The authors attribute this mainly to exact-match grading, which cannot handle the language's spelling variation. Separately, they fine-tune Whisper-small with LoRA as a first ASR baseline and compare it with zero-shot output. This model is groundwork for future speaking-practice feedback and was not used in the pilot. The work shows how participatory design and lightweight evaluation can produce a practical tool for an endangered language.

Key points

  • Participants: The pilot involved 32 young Mapuche learners living in Chile's Metropolitan Region: 23 women and 9 men, aged 22–37, with a mean age of 27.
  • Pilot data: Learners gave 562 answers to 43 questions.
    • Multiple-choice accuracy was 92.3%.
    • Ordering accuracy was 55.2%.
    • Free-text accuracy was 7.1%, which points to a need for scoring that tolerates orthographic variants.
  • Program evaluation: The underlying PEM program lasted four months. Its qualitative evaluation, conducted in April 2024, used a grounded-theory approach (Teorización Anclada). It drew on 32 semi-structured interviews of about 45 minutes each, 10 naturalistic observations and audiovisual records.
  • ASR prototype: The model is Whisper-small adapted with LoRA, reported as a fine-tuned baseline against zero-shot performance. It is planned as an optional in-app feature for oral practice.
  • Related ecosystem: The paper situates the work among earlier resources, including the AVENUE data and Mapudungun Corpus (Duan et al., 2020), a Mapudüngun FST morphological analyser, MAPU, Mapudungun mew, and CENIA's now-restricted translator. The authors describe this landscape as fragmented.

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

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/386vegsicf5x

@inproceedings{oliva2026amulwe,
  title = {{Amulwe Kimün: A Community-Grounded Demo, Resource, and ASR Baseline for Mapuzugun}},
  author = {Oliva, Cristian Eduardo Ahumada and Sadat, Fatiha},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/386vegsicf5x},
  url = {https://doi.org/10.63317/386vegsicf5x}
}

References

37 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. C. Ahumada, C. Gutierrez, and A. Anastasopoulos. 2022. Educational tools for mapuzugun. In Proceedings of the 17th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2022), pages 183–196, Seattle, Washington. Association for Computational Linguistics. find
  2. M. Alvarado. 2012. Sistema para el aprendizaje del mapudungun, incluyendo características de reconocimiento de voz y bot conversacional. Master’s thesis, Pontificia Universidad Católica de Valparaíso. find
  3. A. Alvarez, R. Aranovic, R. Brown, J. Carbonell, C. Fasola, A. Lavie, L. Levin, A. Font-Llitjos, C. Monson, E. Peterson, K. Probst, and R.M. Vega. 2005. Informe final proyecto avenue/mapudungún: Desarrollo de herramientas informáticas para el mapudungún que se habla en chile. Technical report, Insti… find
  4. Antti Arppe, Jordan Lachler, Trond Trosterud, Lene Antonsen, and Sjur N. Moshagen. 2016. Basic language resource kits for endangered languages: A case study of plains cree. In CCURL 2016—Collaboration and Computing for Under-Resourced Languages: Towards an Alliance for Digital Language Diversity (LR… find
  5. Luis Camacho and Rodrigo Zevallos. 2019. Siminchikkunarayku. lingüística computacional para la revitalización y el poliglotismo: Hoja de ruta. Technical report, Fundación Siminchikkunarayku and Pontificia Universidad Católica del Perú, Lima, Perú. Roadmap document. find
  6. S. Castillo, S. Mayo, and J. Soto. 2022. Production of educational materials for mapuzugun teaching: An educators’ perspective and experience approach. In Addressing linguistic inequality in Latin America. Unknown. find
  7. CENIA. 2023. Chile a la vanguardia de la ia: Los avances que nos deja el centro nacional de inteligencia artificial este 2023. Accessed: 2025-05-08. find
  8. A. Chandía. 2022. A mapudüngun fst morphological analyser and its web interface. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 6540–6547, Marseille, France. European Language Resources Association. find
  9. Pascual Coña. 2019. Kuyfi mapuche chumgechi ñi azmogekeel egün. Genlol. find
  10. Mohamed Daouad, Fadoua Ataa Allah, and El Wardani Dadi. 2024. Optimizing whisper models for amazigh asr: a comparative analysis. International Journal of Speech Technology, 28(1):27– 37. find
  11. Tim Dettmers, Artidoro Pagnoni, Ari Holtzman, and Luke Zettlemoyer. 2023. Qlora: Efficient fine-tuning of quantized llms. Relevant for memory-/compute-efficient parameter-efficient finetuning. find
  12. M. Duan, C. Fasola, S. Rallabandi, R. Vega, A. Anastasopoulos, L. Levin, and A. Black. 2020. A resource for computational experiments on mapudungun. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 2872–2877, Marseille, France. European Language Resources Association… find
  13. M. Duan, C. Fasola, S.K. Rallabandi, R. M. Vega, A. Anastasopoulos, L. Levin, and A.W. Black. 2019. A resource for computational experiments on mapudungun. find
  14. Mingjun Duan and contributors. 2019. Mapudungun speech/text corpus (github repository). https://github.com/mingjund/mapudung un-corpus. Accessed 2025-10-15. ↗
  15. M. Inostroza, E. Quero, and A. Berrios. 2021. Profile of participants in mapuche language immersion programs organised by mapuzuguletuaiñ. Nueva Revista del Pacífico, (74). find
  16. Eduardo Iturrieta, Cristian Lagos, and Elizabeth Ávila. 2024. Los grafemarios de la lengua mapuche como herramienta de revitalización lingüística: una revisión bibliográfica. find
  17. Yuxuan Liu, Haipeng Liu, and Ting Long. 2024. Hierllm: Hierarchical large language model for question recommendation. find
  18. E. Loncon, B. Villena, and S. Fernández-Silva. 2023. Chumleafel chi anültuzugu chile mapu mew: The role of mapuzugun in the chilean constituent process (2019–2022). Revista Signos, 56(113):582–609. find
  19. Manuel Mager, Ximena Gutiérrez-Vásquez, Gerardo Sierra, and Iván Meza. 2018. Challenges of language technologies for the indigenous languages of the americas. In Proceedings of the 27th International Conference on Computational Linguistics (COLING 2018), Santa Fe, New Mexico, USA. Association for Co… find
  20. F. Maldonado, A. Arjona, and D. Johnson. 2024. Revitalización lingüística: Diseño, implementación y evaluación de un programa de intervención para jóvenes mapuche. Intercien-cia: Revista de ciencia y tecnología de América, 49(8):471–480. find
  21. A. Mariano. 2021. Mapuche language revitalization through the mapuzuguletuaiñ wallmapu mew project. Documentos Lingüísticos y Literarios, (40). find
  22. Ministry of Social Development and Family. 2024. Informe final. Estado de la lengua mapuzugun. Accessed: 2024-04. find
  23. Benjamin Molineaux. 2023. The corpus of historical mapudungun: morpho-phonological parsing and the history of a native american language. Corpora, 18(2). find
  24. Jean de Dieu Ntirenganya, Didier Habineza, Yvan Niyigena, et al. 2024. Kinyawhisper: Enhancing asr for kinyarwanda using whisper models. arXiv preprint arXiv:2405.08072. Demonstrates fine-tuning of Whisper models for Kinyarwanda ASR as a low-resource case. — [arXiv:2405.08072](https://arxiv.org/abs/2405.08072) find
  25. M. Pell. 2023. La región que traza el proceso de revitalización del mapuzugun. In C. Hammerschmidt, L. Anapios, C. Tomadoni, F. Oliveira de Souza, and S. Espul, editors, América Latina en discusión: una apuesta por las metodologías horizontales, pages 136–146. Editorial Universidad de Guadalajara, M… find
  26. N. Quiero. 2025. Estudio confirma que la continuidad de la lengua mapuche está en riesgo. Accessed: 2025-01-02. find
  27. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. Robust speech recognition via large-scale weak supervision. find our card
  28. M. Rivas and M. Del Pino. 2023. Enseñanza del mapuzugun en chile: ecosonoridad como apresto a la adquisición de la lengua indígena. Educación, 32(62):185. find
  29. Scott Sadowsky, Héctor Painequeo, Gastón Salamanca, and Heriberto Avelino. 2013. Mapudungun. Journal of the International Phonetic Association, 43(1):87–96. find
  30. V. Sarmiento. 2023. Traducción español-mapuzugun en la universidad de playa ancha: un proyecto de innovación educativa con impacto social, cultural y político. Revista digital de políticas lingüísticas, (19). find
  31. I. Smeets. 1989. A mapuche grammar. Leiden: Rijksuniversiteit te Leiden. find
  32. Zheshu Song, Jianheng Zhuo, Yifan Yang, Ziyang Ma, Shixiong Zhang, and Xie Chen. 2024. Lora-whisper: Parameter-efficient and extensible multilingual asr. find
  33. CEDETI UC. Mapudungun mew. http://www.cedeti.cl/tecnologias-inclusivas/ software-educativo/mapudungun-mew. Accessed: 2025-05-08. ↗
  34. UNESCO. 2010. Atlas of the World’s Languages in Danger, 3rd edition. UNESCO, Paris. find
  35. F. Zúñiga. 2006. Mapudungun. El habla mapuche. Centro de Estudios Públicos, Santiago de Chile. 9. find
  36. Language Resource References Mapudungun Corpus. 2020. find
  37. Mapudungun Corpus. LREC 2020. Freely available speech/written corpus. PID https://lremap.elra.info/search_by_map _responsive.php?query=Mapudungun+Corpus. ↗

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/oliva-2026-amulwe-kimun-community-grounded-demo/