Elamur Research⌕
☰
home / papers / 2026 / Digital Preservation of Aromanian Through Knowledge Management and Automatic Speech Recognition Evaluation

Digital Preservation of Aromanian Through Knowledge Management and Automatic Speech Recognition Evaluation

Paper · 2026 · Marija Pendevska and Hristina Nastevska

Summary

Aromanian is an endangered Eastern Romance language with only a few thousand mostly adult speakers in fragmented Balkan communities. It has incomplete standardization and is absent from speech technology. The authors treat its preservation as a knowledge management (KM) problem. They propose a three-module framework built on the Knowledge Management Cycle model: localizing existing resources, distributing them through digital platforms, and creating new content with technology. A survey of 176 respondents on knowledge-sharing habits supports the framework's design.

Within this framework they run what they describe as the first ASR benchmark for Aromanian. They test zero-shot OpenAI Whisper models (medium, large-v2, large-v3) under several language settings on two North Macedonian varieties, Gramosteanj (southern group) and Crushova (northern group). Every configuration fails at the word level, although character-level scores show the models partly capture the sound patterns. Romanizing output in non-Latin scripts lowers errors noticeably, which points to script mismatch as a key error source. The work gives a baseline for Aromanian speech technology and shows that endangered varieties need dedicated data and model adaptation.

Key points

  • Survey results (176 respondents): Information quality (28.3%) and personal utility (24.1%) were the most valued criteria for sharing knowledge. Digital information seeking (27.1%) was the most common behaviour.
  • Whisper error rates: WER stayed above 88% in all configurations. CER fell as low as about 34% in the best setup.
  • Best configuration: Whisper large-v3 with the Latin language setting performed best throughout.
  • Romanization: Converting non-Latin script output to Latin, with tools such as Uroman, substantially reduced CER.
  • Resources and context: Other language resources listed include Praat, CTC Forced Aligner and Mozilla Common Voice. The orthographic context is the 35-character Bitola standard from 1997. The paper frames Aromanian as an extreme out-of-distribution case compared with Greek dialect ASR benchmarks.

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

🎓Ss. Cyril and Methodius University in SkopjeSkopje · North Macedonia6 works · 147 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 2Research subjects and languages the work is about.Evaluation & benchmarks · 1,341 worksSpeech & spoken language · 519 works

Cite

Cite as Marija Pendevska and Hristina Nastevska. 2026. Digital Preservation of Aromanian Through Knowledge Management and Automatic Speech Recognition Evaluation. Proceedings of the First Workshop on Dialects in NLP –- A Resource Perspective.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/34e4awcc69g4

@inproceedings{pendevska2026digital,
  title = {{Digital Preservation of Aromanian Through Knowledge Management and Automatic Speech Recognition Evaluation}},
  author = {Marija Pendevska and Hristina Nastevska},
  booktitle = {Proceedings of the First Workshop on Dialects in NLP –- A Resource Perspective},
  year = {2026},
  doi = {10.63317/34e4awcc69g4},
  url = {https://doi.org/10.63317/34e4awcc69g4}
}

References

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

  1. Tanel Alumäe, Jiaming Kong, and Daniil Robnikov. 2023. Dialect Adaptation and Data Augmentation for Low-Resource ASR: TalTech Systems for the MADASR 2023 Challenge. find
  2. Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, 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.
  3. Stamatis Beis and Fanis Dasoulas. 2017. Proposal for the writing system of the Vlach language. Glossologia, 25:51–69. find
  4. Laurent Besacier, Etienne Barnard, Alexey Karpov, and Tanja Schultz. 2014. Automatic speech recognition for under-resourced languages: A survey. Speech Communication, 56:85–100. find
  5. Seham Nasr, Rehab Duwairi, and Muhannad Quwaider. 2023. End-to-end speech recognition for arabic dialects. Arabian Journal for Science and Engineering, 48. find
  6. Steven Bird and Gary Simons. 2002. Seven dimensions of portability for language documentation and description. find
  7. Božidar Nastev. 1988. Aromanski studii, prilozi kon balkanistikata. MANU, Skopje. find
  8. Verena Blaschke, Miriam Winkler, Constantin Förster, Gabriele Wenger-Glemser, and Barbara Plank. 2025. A Multi-Dialectal Dataset for German Dialect ASR and Dialect-to-Standard Speech Translation. In Interspeech 2025, pages 913–917. ISCA.
  9. Theodor Capidan. 1932. Aromânii: Dialectul aromân. Imprimeria Naţională, Bucharest. find
  10. David Crystal. 2000. Language Death. Cambridge University Press. find
  11. Tiberius Cunia. 2000. On the standardization of the Aromanian system of writing. The Farsarotul. find
  12. M. Evans, K. Dalkir, and C. Bidian. 2014. A holistic view of the knowledge life cycle: The knowledge management cycle (KMC) model. Electronic Journal of Knowledge Management, 12(2):85– 97. find
  13. Victor A. Friedman. 2012. Aromanian in the Balkan linguistic league: A comparative approach. volume 14, pages 51–66. find
  14. Pratik Joshi, Sebastin Santy, Amar Budhiraja, Kalika Bali, and Monojit Choudhury. 2020. The state and fate of linguistic diversity and inclusion in the NLP world. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6282–6293, Online. Association for Computational Linguistics.
  15. Omnilingual ASR Team, Gil Keren, Artyom Kozhevnikov, Yen Meng, Christophe Ropers, Matthew Setzler, Skyler Wang, Ife Adebara, Michael Auli, Can Balioglu, Kevin Chan, Chierh Cheng, Joe Chuang, Caley Droof, Mark Duppenthaler, Paul-Ambroise Duquenne, Alexander Erben, Cynthia Gao, Gabriel Mejia Gonzalez, Kehan Lyu, Sagar Miglani, Vineel Pratap, Kaushik Ram Sadagopan, Safiyyah Saleem, Arina Turkatenko, Albert Ventayol-Boada, Zheng-Xin Yong, Yu-An Chung, Jean Maillard, Rashel Moritz, Alexandre Mourachko, Mary Williamson, and Shireen Yates. 2025. Omnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages. find
  16. Marija Pendevska. 2019. The Impact of Knowledge Management on Innovation in Enterprises in the Republic of Macedonia. Ph.D. thesis, Faculty of Economics, Ss. Cyril and Methodius University in Skopje, Skopje. find
  17. Marija Pendevska, Branislav Gerazov, and Branko Prlja. 2024. Digital preservation of Aromanian through a knowledge management framework: The first digital riddle generator. In Proceedings of the 2nd International Conference Dedicated to 100 Years Prof. Dr. Božidar Nastev. find
  18. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. Robust Speech Recognition via Large-Scale Weak Supervision. ↗
  19. Thede Kahl. 2008. Does the Aromanian have a chance of survival? Some thoughts about the loss of language and language preservation. The Romance Balkans, pages 123–140. find
  20. Georg Rehm and Hans Uszkoreit. 2012. Language technology 2012: Current state and opportunities. In The META-NET Strategic Research Agenda for Multilingual Europe 2020, pages 27– 31. find
  21. Marjan Marković. 2007. Govorot na Aromancite Frašeroti vo Ohridsko-Struškiot region (vo Balkanski kontekst). MANU, Skopje. Doctoral dissertation defended in 2000. find
  22. Clement Sicard, Kajetan Pyszkowski, and Victor Gillioz. 2023. Spaiche: Extending State-of-the-Art ASR Models to Swiss German Dialects. ↗
  23. Angelika Mittelmann, Gabriele Vollmar, et al. 2022. Wissensmanagement-kompetenzkatalog. Version 2.0. Gesellschaft für Wissensmanagement e.V. (GfWM). find
  24. Naoki Takahashi, Shogo Miwa, Yuta Kamiya, Takumi Toyama, Raufun Nahar, and Atsuhiko Kai. 2024. Comparison of Large Pre-trained Models and Adaptation Methods for Japanese Dialects ASR. In 2024 IEEE 13th Global Conference on Consumer Electronics (GCCE), pages 811–814. find
  25. Christopher Moseley. 2010. Atlas of the World’s Languages in Danger, 3rd edition. UNESCO Publishing, Paris. find
  26. Chara Tsoukala, Stavros Bompolas, Antigoni Margariti, Konstantina Panagiotou, Maria Elisavet Plaiti, Nefeli Tzanakaki, Petros Karatsareas, Angela Ralli, Antonios Anastasopoulos, and Stella Markantonatou. 2026. Extending ASR evaluation resources for Modern Greek dialects. In Proceedings of the 13th Workshop on NLP for Similar Languages, Varieties and Dialects, pages 210–222, Rabat, Morocco. Association for Computational Linguistics.
  27. Socrates Vakirtzian, Vivian Stamou, Yannis Kazos, and Stella Markantonatou. 2025. Dialectal treebanks and their relation with the standard variety: The case of East Cretan and Standard Modern Greek. In Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025), pages 776–784, Tallinn, Estonia. University of Tartu Library. find
  28. Gustav Weigand. 1895. Die Aromunen: Ethnographisch-Philologisch-Historische Untersuchung über das Volk der sogenannten Makedo-Romanen oder Zinzaren. J.A. Barth, Leipzig. 11. find
  29. Language Resource References Boersma, Paul and Weenink, David. 2024. Praat: Doing Phonetics by Computer. University of Amsterdam. PID https://www.praat.org/. ↗
  30. CTC Forced Aligner Team. 2024. CTC Forced Aligner. Open Source. PID https://github.com/huggingface/ctc-forced-aligner. ↗
  31. Hermjakob, Ulf and May, Jonathan and Knight, Kevin. 2018. Out-of-the-Box Universal Romanization Tool Uroman. USC Information Sciences Institute. PID https://github.com/isi-nlp/uroman. ↗
  32. Mozilla Foundation. 2025. Mozilla Common Voice. Mozilla Foundation. PID https://commonvoice.mozilla.org/. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pendevska-2026-digital-preservation-aromanian-through-knowledge/