---
title: "Automatic Speech Recognition for Documenting — Taguchi 2026"
description: "Ikema, a Ryukyuan language of the Miyako Islands in Okinawa, has about 1,300 speakers, most of them over 60."
authors: ["Chihiro Taguchi", "Yukinori Takubo", "David Chiang"]
year: 2026
venue: "Proceedings of the Fifteenth Language Resources and Evaluation Conference"
source: "https://aclanthology.org/2026.lrec-1.284/"
doi: "10.63317/4im4f6vuxk42"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/taguchi-2026-automatic-speech-recognition-documenting-endangered/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/taguchi-2026-automatic-speech-recognition-documenting-endangered/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/2/) / Automatic Speech Recognition for Documenting Endangered Languages: Case Study of Ikema Miyakoan

# Automatic Speech Recognition for Documenting Endangered Languages: Case Study of Ikema Miyakoan

**Authors:** Chihiro Taguchi, Yukinori Takubo, [David Chiang](https://md.research.elamur.ai/free/en/authors/david-chiang/) · **Year:** 2026 · **Venue:** Proceedings of the Fifteenth Language Resources and Evaluation Conference · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.lrec-1.284/) · [PDF](https://aclanthology.org/2026.lrec-1.284.pdf) · [DOI 10.63317/4im4f6vuxk42](https://doi.org/10.63317/4im4f6vuxk42) · [OpenAlex](https://openalex.org/W7160057842)

**Citations:** 0 citations (openalex-snapshot · 2026-10-14)

## Summary
Ikema, a Ryukyuan language of the Miyako Islands in Okinawa, has about 1,300 speakers, most of them over 60. It is projected to become critically endangered within roughly three decades. Documentation is slowed by the "transcription bottleneck", meaning that turning recordings into text takes a lot of time and expert skill.

The authors built the first Ikema speech corpus, about 6.33 hours drawn from:
- nearly twenty years of fieldwork video,
- spoken dictionary entries,
- audiobooks.

Speech was transcribed exactly as pronounced, in a phonemic kana orthography, and converted to romaji by deterministic rules. Tags mark Japanese code-switching, disfluencies, songs, personal names and unclear passages.

They fine-tuned multilingual Wav2Vec2-family models with a CTC output layer, separately for kana and romaji. The best system reaches a character error rate of about 14.8%. Embedded in the ELAN annotation tool, ASR drafts noticeably shortened transcription and reduced annotators' mental effort.

This provides rare quantitative evidence that ASR-assisted post-editing can speed up endangered-language documentation. The corpus, models and code are public.

## Key points
- **Corpus composition:**
  - Field: 11,126 segments, about 16,058 s.
  - Dictionary: 5,680 segments, about 5,379 s.
  - Audiobooks: 285 segments, about 1,348 s.
  - Most material comes from one male speaker; the audiobooks are read by a female speaker.
- **Models compared:** wav2vec2-xls-r-300m, wav2vec2-xls-r-1b and mms-1b, using an 80/10/10 split. The smallest model, xls-r-300m, performed best, with a CER of 14.80%.
- **CER vs. WER:** WER stayed high, at roughly 62–77%. The authors consider it a weak metric for Ikema because word boundaries are poorly defined, even in the reference transcriptions.
- **Why CTC instead of Whisper:** CTC suits small data and keeps per-frame output faithful to the speech, including fillers and disfluencies.
- **Availability:** Dataset, models and code are released at github.com/ctaguchi/ikema_asr. The ASR is integrated into ELAN.

*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](https://md.research.elamur.ai/free/en/shelves/linguistics/)

## Cite

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{taguchi2026automatic,
  title = {{Automatic Speech Recognition for Documenting Endangered Languages: Case Study of Ikema Miyakoan}},
  author = {Taguchi, Chihiro and Takubo, Yukinori and Chiang, David},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/4im4f6vuxk42},
  url = {https://doi.org/10.63317/4im4f6vuxk42}
}
```

- [← previous Automatic Prediction of Prominence and Boundary Strength from Text](https://md.research.elamur.ai/free/en/papers/mas-2026-automatic-prediction-prominence-boundary-strength/)
- [next → Automatic Suggestions Help Extending Eventive Ontology: A Case Study on SynSemClass](https://md.research.elamur.ai/free/en/papers/strakova-2026-automatic-suggestions-help-extending-eventive/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/2/)

## References

26 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, and Michael Auli. 2021. XLS-R: Self-supervised cross-lingual speech representation learning at scale. [find](https://scholar.google.com/scholar?q=Kritika+Singh%2C+Patrick+von+Platen%2C+Yatharth+Saraf%2C+Juan+Pino%2C+Alexei+Baevski%2C+Alexis+Conneau%2C+and+Michael+Auli%2E+2021%2E+XLS%2DR%3A+Self%2Dsupervised+cross%2Dlingual+speech+representation+learning+at+scale%2E) [our card](https://md.research.elamur.ai/free/en/papers/babu-2021-xls-r-self-supervised-cross/)
2. Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, and Michael Auli. 2020. wav2vec 2.0: A framework for self-supervised learning of speech representations. [find](https://scholar.google.com/scholar?q=Alexei+Baevski%2C+Henry+Zhou%2C+Abdelrahman+Mohamed%2C+and+Michael+Auli%2E+2020%2E+wav2vec+2%2E0%3A+A+framework+for+self%2Dsupervised+learning+of+speech+representations%2E) [our card](https://md.research.elamur.ai/free/en/papers/baevski-2020-wav2vec-2-0-framework-self/)
3. Blaine Billings and Bradley McDonnell. 2025. Connecting automated speech recognition to transcription practices. In Proceedings of the Eighth Workshop on the Use of Computational Methods in the Study of Endangered Languages (ComputEL), pages 128–132. [find](https://scholar.google.com/scholar?q=Blaine+Billings+and+Bradley+McDonnell%2E+2025%2E+Connecting+automated+speech+recognition+to+transcription+practices%2E+In+Proceedings+of+the+Eighth+Workshop+on+the+Use+of+Computational+Methods+in+the+Study+)
4. Ilaria Chizzoni and Alessandro Vietti. 2024. Towards an ASR system for documenting endangered languages: A preliminary study on Sardinian. In Proceedings of the 10th Italian Conference on Computational Linguistics (CLiC-it 2024), pages 214–220. [find](https://scholar.google.com/scholar?q=Ilaria+Chizzoni+and+Alessandro+Vietti%2E+2024%2E+Towards+an+ASR+system+for+documenting+endangered+languages%3A+A+preliminary+study+on+Sardinian%2E+In+Proceedings+of+the+10th+Italian+Conference+on+Computationa)
5. Alex Graves, Santiago Fernández, Faustino Gomez, and Jürgen Schmidhuber. 2006. Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks. In Proceedings of the 23rd International Conference on Machine Learning, pages 369–376. [find](https://scholar.google.com/scholar?q=Alex+Graves%2C+Santiago+Fern%C3%A1ndez%2C+Faustino+Gomez%2C+and+J%C3%BCrgen+Schmidhuber%2E+2006%2E+Connectionist+temporal+classification%3A+labelling+unsegmented+sequence+data+with+recurrent+neural+networks%2E+In+Proceedings)
6. Séverine Guillaume, Guillaume Wisniewski, Cécile Macaire, Guillaume Jacques, Alexis Michaud, Benjamin Galliot, Maximin Coavoux, Solange Rossato, Minh-Châu Nguyên, and Maxime Fily. 2022. Fine-tuning pre-trained models for automatic speech recognition, experiments on a fieldwork corpus of Japhug (Tran… [find](https://scholar.google.com/scholar?q=S%C3%A9verine+Guillaume%2C+Guillaume+Wisniewski%2C+C%C3%A9cile+Macaire%2C+Guillaume+Jacques%2C+Alexis+Michaud%2C+Benjamin+Galliot%2C+Maximin+Coavoux%2C+Solange+Rossato%2C+Minh%2DCh%C3%A2u+Nguy%C3%AAn%2C+and+Maxime+Fily%2E+2022%2E+Fine%2Dtuning+pr)
7. Yuka Hayashi. 2013. Minami ryuukyuu miyakogo ikema hougen no bunpou. Ph.D. thesis, Kyoto University. [find](https://scholar.google.com/scholar?q=Yuka+Hayashi%2E+2013%2E+Minami+ryuukyuu+miyakogo+ikema+hougen+no+bunpou%2E+Ph%2ED%2E+thesis%2C+Kyoto+University%2E)
8. Robbie Jimerson and Emily Prud’hommeaux. 2018. ASR for documenting acutely under-resourced indigenous languages. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018). [find](https://scholar.google.com/scholar?q=Robbie+Jimerson+and+Emily+Prud%E2%80%99hommeaux%2E+2018%2E+ASR+for+documenting+acutely+under%2Dresourced+indigenous+languages%2E+In+Proceedings+of+the+Eleventh+International+Conference+on+Language+Resources+and+Evalu)
9. Zhaolin Li, Monika Rind-Pawlowski, and Jan Niehues. 2024. Speech recognition corpus of the khinalug language for documenting endangered languages. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 15171… [find](https://scholar.google.com/scholar?q=Zhaolin+Li%2C+Monika+Rind%2DPawlowski%2C+and+Jan+Niehues%2E+2024%2E+Speech+recognition+corpus+of+the+khinalug+language+for+documenting+endangered+languages%2E+In+Proceedings+of+the+2024+Joint+International+Confer)
10. Siyu Liang and Gina-Anne Levow. 2025. Breaking the transcription bottleneck: Fine-tuning ASR models for extremely low-resource fieldwork languages. In Proceedings of the Fourth Workshop on NLP Applications to Field Linguistics, pages 26–37. [find](https://scholar.google.com/scholar?q=Siyu+Liang+and+Gina%2DAnne+Levow%2E+2025%2E+Breaking+the+transcription+bottleneck%3A+Fine%2Dtuning+ASR+models+for+extremely+low%2Dresource+fieldwork+languages%2E+In+Proceedings+of+the+Fourth+Workshop+on+NLP+Applica)
11. Julia Mainzinger and Gina-Anne Levow. 2024. Fine-tuning ASR models for very low-resource languages: A study on Mvskoke. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop), pages 76–82. [find](https://scholar.google.com/scholar?q=Julia+Mainzinger+and+Gina%2DAnne+Levow%2E+2024%2E+Fine%2Dtuning+ASR+models+for+very+low%2Dresource+languages%3A+A+study+on+Mvskoke%2E+In+Proceedings+of+the+62nd+Annual+Meeting+of+the+Association+for+Computational+L)
12. Alexis Michaud, Oliver Adams, Trevor Cohn, Graham Neubig, and Séverine Guillaume. 2018. Integrating automatic transcription into the language documentation workflow: Experiments with Na data and the Persephone toolkit. Language Documentation & Conservation, 12:393–429. [find](https://scholar.google.com/scholar?q=Alexis+Michaud%2C+Oliver+Adams%2C+Trevor+Cohn%2C+Graham+Neubig%2C+and+S%C3%A9verine+Guillaume%2E+2018%2E+Integrating+automatic+transcription+into+the+language+documentation+workflow%3A+Experiments+with+Na+data+and+the+P)
13. Hiroyuki Nakama, Yukinori Takubo, Shoichi Iwasaki, Yosuke Igarashi, Daniel Wymark, and Natsuko Nakagawa. 2025. Dictionary of Ikema spoken in Nishihara village, a variety of Miyako Ryukyuan, 3rd edition. National Institute for Japanese Language and Linguistics. [find](https://scholar.google.com/scholar?q=Hiroyuki+Nakama%2C+Yukinori+Takubo%2C+Shoichi+Iwasaki%2C+Yosuke+Igarashi%2C+Daniel+Wymark%2C+and+Natsuko+Nakagawa%2E+2025%2E+Dictionary+of+Ikema+spoken+in+Nishihara+village%2C+a+variety+of+Miyako+Ryukyuan%2C+3rd+editio)
14. Alexander O’Neill, Marieke Meelen, Rolando Coto-Solano, Sonam Phuntsog, and Charles Ramble. 2023. Language preservation through ASR. In Cambridge Language Sciences Annual Symposium. [find](https://scholar.google.com/scholar?q=Alexander+O%E2%80%99Neill%2C+Marieke+Meelen%2C+Rolando+Coto%2DSolano%2C+Sonam+Phuntsog%2C+and+Charles+Ramble%2E+2023%2E+Language+preservation+through+ASR%2E+In+Cambridge+Language+Sciences+Annual+Symposium%2E)
15. Thomas Pellard. 2015. The linguistic archeology of the Ryukyu Islands. In Patrick Heinrich, Shinsho Miyara, and Michinori Shimoji, editors, Handbook of the Ryukyuan languages: History, structure, and use, pages 13–37. De Gruyter Mouton, Berlin; Boston. [find](https://scholar.google.com/scholar?q=Thomas+Pellard%2E+2015%2E+The+linguistic+archeology+of+the+Ryukyu+Islands%2E+In+Patrick+Heinrich%2C+Shinsho+Miyara%2C+and+Michinori+Shimoji%2C+editors%2C+Handbook+of+the+Ryukyuan+languages%3A+History%2C+structure%2C+and+)
16. Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi, Xiaohui Zhang, Wei-Ning Hsu, Alexis Conneau, and Michael Auli. 2023. Scaling speech technology to 1,000+ languages. [find](https://scholar.google.com/scholar?q=Vineel+Pratap%2C+Andros+Tjandra%2C+Bowen+Shi%2C+Paden+Tomasello%2C+Arun+Babu%2C+Sayani+Kundu%2C+Ali+Elkahky%2C+Zhaoheng+Ni%2C+Apoorv+Vyas%2C+Maryam+Fazel%2DZarandi%2C+Alexei+Baevski%2C+Yossi+Adi%2C+Xiaohui+Zhang%2C+Wei%2DNing+Hsu%2C) [our card](https://md.research.elamur.ai/free/en/papers/pratap-2023-scaling-speech-technology-1-000/)
17. Emily Prud’hommeaux, Robbie Jimerson, Richard Hatcher, and Karin Michelson. 2021. Automatic speech recognition for supporting endangered language documentation. Language Documentation & Conservation, 15:491–513. [find](https://scholar.google.com/scholar?q=Emily+Prud%E2%80%99hommeaux%2C+Robbie+Jimerson%2C+Richard+Hatcher%2C+and+Karin+Michelson%2E+2021%2E+Automatic+speech+recognition+for+supporting+endangered+language+documentation%2E+Language+Documentation+%26+Conservation%2C+)
18. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. Robust speech recognition via large-scale weak supervision. [find](https://scholar.google.com/scholar?q=Alec+Radford%2C+Jong+Wook+Kim%2C+Tao+Xu%2C+Greg+Brockman%2C+Christine+McLeavey%2C+and+Ilya+Sutskever%2E+2022%2E+Robust+speech+recognition+via+large%2Dscale+weak+supervision%2E) [our card](https://md.research.elamur.ai/free/en/papers/radford-2022-robust-speech-recognition-large-scale/)
19. Frank Seifart, Nicholas Evans, Harald Hammarström, and Stephen C. Levinson. 2018. Language documentation twenty-five years on. Language, 94(4):e324–e345. [find](https://scholar.google.com/scholar?q=Frank+Seifart%2C+Nicholas+Evans%2C+Harald+Hammarstr%C3%B6m%2C+and+Stephen+C%2E+Levinson%2E+2018%2E+Language+documentation+twenty%2Dfive+years+on%2E+Language%2C+94%284%29%3Ae324%E2%80%93e345%2E)
20. Michinori Shimoji. 2008. A Grammar of Irabu, a Southern Ryukyuan Language. Ph.D. thesis, Australian National University. [find](https://scholar.google.com/scholar?q=Michinori+Shimoji%2E+2008%2E+A+Grammar+of+Irabu%2C+a+Southern+Ryukyuan+Language%2E+Ph%2ED%2E+thesis%2C+Australian+National+University%2E)
21. Chihiro Taguchi and David Chiang. 2024. Language complexity and speech recognition accuracy: Orthographic complexity hurts, phonological complexity doesn’t. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15493–15503. [find](https://scholar.google.com/scholar?q=Chihiro+Taguchi+and+David+Chiang%2E+2024%2E+Language+complexity+and+speech+recognition+accuracy%3A+Orthographic+complexity+hurts%2C+phonological+complexity+doesn%E2%80%99t%2E+In+Proceedings+of+the+62nd+Annual+Meeting+o)
22. Chihiro Taguchi, Jefferson Saransig, Dayana Velásquez, and David Chiang. 2024. Killkan: The automatic speech recognition dataset for kichwa with morphosyntactic information. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LRE… [find](https://scholar.google.com/scholar?q=Chihiro+Taguchi%2C+Jefferson+Saransig%2C+Dayana+Vel%C3%A1squez%2C+and+David+Chiang%2E+2024%2E+Killkan%3A+The+automatic+speech+recognition+dataset+for+kichwa+with+morphosyntactic+information%2E+In+Proceedings+of+the+2024)
23. Yukinori Takubo. 2021a. Hougen wo kana de kaku: Ryuukyuu miyakogo ikemahougen wo rei ni. Kotoba to Mozi. [find](https://scholar.google.com/scholar?q=Yukinori+Takubo%2E+2021a%2E+Hougen+wo+kana+de+kaku%3A+Ryuukyuu+miyakogo+ikemahougen+wo+rei+ni%2E+Kotoba+to+Mozi%2E)
24. Yukinori Takubo. 2021b. Morphophonemics of Ikema Miyakoan. In Studies in Asian Historical Linguistics, Philology and Beyond, chapter 5. Brill, Leiden, The Netherlands. [find](https://scholar.google.com/scholar?q=Yukinori+Takubo%2E+2021b%2E+Morphophonemics+of+Ikema+Miyakoan%2E+In+Studies+in+Asian+Historical+Linguistics%2C+Philology+and+Beyond%2C+chapter+5%2E+Brill%2C+Leiden%2C+The+Netherlands%2E)
25. Peter Wittenburg, Hennie Brugman, Albert Russel, Alex Klassmann, and Han Sloetjes. 2006. ELAN: a professional framework for multimodality research. In Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC). [find](https://scholar.google.com/scholar?q=Peter+Wittenburg%2C+Hennie+Brugman%2C+Albert+Russel%2C+Alex+Klassmann%2C+and+Han+Sloetjes%2E+2006%2E+ELAN%3A+a+professional+framework+for+multimodality+research%2E+In+Proceedings+of+the+Fifth+International+Conference) [our card](https://md.research.elamur.ai/free/en/papers/wittenburg-2006-elan-professional-framework-multimodality-research/)
26. Masahiro Yamada, Yukinori Takubo, Shoichi Iwasaki, Celik Kenan, Soichiro Harada, Nobuko Kibe, Tyler Lau, Natsuko Nakagawa, Yuto Ni-inaga, Tomoyo Otsuki, Manami Sato, Rihito Shi-rata, Gijs van der Lubbe, and Akiko Yokoyama. 2020. Experimental study of inter-language and inter-generational intelligibi… [find](https://scholar.google.com/scholar?q=Masahiro+Yamada%2C+Yukinori+Takubo%2C+Shoichi+Iwasaki%2C+Celik+Kenan%2C+Soichiro+Harada%2C+Nobuko+Kibe%2C+Tyler+Lau%2C+Natsuko+Nakagawa%2C+Yuto+Ni%2Dinaga%2C+Tomoyo+Otsuki%2C+Manami+Sato%2C+Rihito+Shi%2Drata%2C+Gijs+van+der+Lubb)