Elamur Research⌕
☰
home / papers / 2022 / Morphologically annotated corpora of Pomak

Morphologically annotated corpora of Pomak

Summary

Pomak is a Slavic language from the eastern part of the South Slavic branch. It is spoken mostly in the Rhodopes, including Greek Thrace. The language is endangered, still exists mainly in oral form and has no generally accepted writing system. Within the Philotis platform, the authors wanted to show that researchers of low-resource languages can build high-quality annotated corpora on their own, using open tools. In collaboration with the Pomak community, the team collected scattered texts: folklore, journalism, translations and proverbs. They developed a Latin-based orthography (K&K) for the texts. Texts written in various alphabets were automatically converted into K&K and then corrected manually. Morphological annotation was obtained using the electronic dictionary Rodopsky. On the gold corpus, spaCy, Stanza, UDify and UDPipe were compared on lemmatization, PoS tagging and morphological feature tagging. Stanza achieved the highest accuracy in lemmatization and UPOS, and spaCy with XLM-RoBERTa in feature tagging. The authors also describe practical difficulties with these tools, such as outdated instructions. Another problem is that morphology is hard to evaluate separately from syntax. The work provides an open resource for an endangered language and also a ready-made workflow that can be applied to other oral languages.

Key points

  • The corpus contains about 130 thousand words across different genres. The largest part is folk tales, about 44 thousand words. The gold part was split 80:10:10: 5000 / 671 / 679 sentences, i.e. 67 345 / 9 736 / 9 701 tokens.
  • The Rodopsky dictionary contains about 61.5 thousand lemmas and about 3.5 million word forms with lemma, PoS and morphological features. It was automatically converted into the K&K alphabet and proofread manually.
  • Lemmatization accuracy: Stanza — 97.82, spaCy (XLM-RoBERTa-large) — 93.85, UDPipe v1.2 — 92.04, UDify — 90.27. In UPOS, Stanza also has the best result, about 98.7. For FEATS, the extracted table is corrupted, so exact figures are not given here. The paper text states only that spaCy performs best on this task.
  • On the UNESCO scale, the language has about 35 thousand speakers, about 3.2% of the population. There is no education in Pomak in any country where Pomaks live.
  • The work took about 8 months. The corpora are planned to be released on the Philotis website, and the gold annotation in a UD treebank repository. At the time of publication, UD syntactic annotation was still in progress.

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

Topics: Treebanks

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2022.computel-1.22

@inproceedings{karahoa2022morphologically,
  title = {{Morphologically annotated corpora of Pomak}},
  author = {Karahóǧa, Ritván Jusúf and Krimpas, Panagiotis G. and Stamou, Vivian and Arampatzakis, Vasileios and Karamatskos, Dimitrios and Sevetlidis, Vasileios and Constantinides, Nikolaos and Kokkas, Nikolaos and Pavlidis, George and Markantonatou, Stella},
  booktitle = {Proceedings of the Fifth Workshop on the Use of Computational Methods in the Study of Endangered Languages},
  year = {2022},
  doi = {10.18653/v1/2022.computel-1.22},
  url = {https://doi.org/10.18653/v1/2022.computel-1.22}
}

References

18 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Evangelia Adamou and Davide Fanciullo. 2018. Why Pomak will not be the next Slavic literary language. In D. Stern, M. Nomachi, and B. Belić, editors, Linguistic regionalism in Eastern Europe and beyond: minority, regional and literary microlanguages, pages 40–65. Peter Lang. find
  2. Antonios Anastasopoulos, Marika Lekakou, Josep Quer, Eleni Zimianiti, Justin DeBenedetto, and David Chiang. 2018. Part-of-speech tagging on an endangered language: a parallel Griko-Italian resource. In Proceedings of the 27th International Conference on Computational Linguistics, pages 2529–2539, Sa… find
  3. Alexandre Arkhipov and Nick Thieberger. 2018. Reflections on software and technology for language documentation. In Andrea L. Berez-Kroeker Bradley Mc-Donnell and Gary Holton, editors, Reflections on Language Documentation. 20 Years after Himmelmann 1998 Language Documentation & Conservation Special… find
  4. Tim Hutchinson. 2020. Natural language processing and machine learning as practical toolsets for archival processing. Records Management Journal, 30:155– 174. find
  5. Dan Kondratyuk and Milan Straka. 2019. 75 languages, 1 model: Parsing Universal Dependencies universally. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2779–279… find our card
  6. Panagiotis Krimpas, Nikolaos Constantinides, Rit-van Karahoǧa, Stella Markantonatou, and George Pavlidis. 2021. «Pomak: An idiosyncratic South East Slavic language?». In VII Scientific Conference on "The Traditional Culture of Greece"», Lomonosov State University in Moscow. find
  7. Spyros Armostis, Christodoulou Kyriaci, Katsoyannou Marianna, and Charalambos Themistocleous. 2014. Addressing writing system issues in dialectal lexicography: the case of Cypriot Greek, page 23–38. find
  8. Panagiotis G. Krimpas. 2020. Language and origin of Pomaks in the light of the Balkan Sprachbund. In A. Bartsiokas & N. Macha-Bizoumi M. Varvounis, editor, The Pomaks of Thrace: Multidisciplinary and interdisciplinary approaches, pages 167–204. Thessaloniki: K&M Stamoulis. find
  9. Steven Bird. 2020. Decolonising speech and language technology. In Proceedings of the 28th International Conference on Computational Linguistics, COLING 2020, Barcelona, Spain (Online), December 8-13, 2020, pages 3504–3519. International Committee on Computational Linguistics. find
  10. William Lane, Mat Bettinson, and Steven Bird. 2021. A computational model for interactive transcription. In Proceedings of the Second Workshop on Data Science with Human in the Loop: Language Advances, pages 105–111, Online. Association for Computational Linguistics. find
  11. Bradley McDonnell. 2018. Reflections on linguistic analysis in documentary linguistics. In Andrea L. Berez-Kroeker & Gary Holton Bradley McDonnell, editor, Reflections on Language Documentation. 20 Years after Himmelmann 1998. Language Documentation Conservation Special Publication, volume 20, pages… find
  12. Pronunciation Character Minh Van Nguyen, Viet Dac Lai, Amir Pouran Ben Veyseh, and Thien Huu Nguyen. 2021. Trankit: A light-weight transformer-based toolkit for multilingual natural language processing. In EACL (System Demonstrations), pages 80–90. find our card
  13. Panayotis G. Papadimitriou. 2013. Dialects of the Pomaks of the Greek Rhodope. Regional Analytical Slavic and Muslim speakers in Southeastern Europe. Thessaloniki: Balkan Peninsula Research Institute. [In Greek]. find
  14. Peng Qi, Yuhao Zhang, Yuhui Zhang, Jason Bolton, and Christopher D. Manning. 2020. Stanza: A python natural language processing toolkit for many human languages. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pages 101–108, Online.… find our card
  15. Sally Rice. 2018. Reflections on documentary corpora. In Andrea L. Berez-Kroeker & Gary Holton Bradley McDonnell, editor, Reflections on Language Documentation. 20 Years after Himmelmann 1998. Language Documentation Conservation Special Publication, 15, pages 157–172. University of Hawai’i Press. find
  16. Georg Renatus Solta. 1980. Einführung in die Balkan-linguistik mit besonderer Berücksichtigung des Sub-strats und des Balkanlateinischen. Darmstadt: Wissenschaftliche Buchgesellschaft. find
  17. A, a [E5], [æ] [b] Æ, æ B, b [t“ s] [Ù] C, c Č,č [d] [e] D, d E, e [f] [g] [g j ] [d“ z] [d“Z ] F, f G, g Ǵ, ǵ Ǧ, ǧ [x] [i] [j] [k], [k j ] H, h I, i J, j K, k [ë], [l j ] L, l [L] L , , l , [m] M, m [n], [n j ] N,n [ñ] N , , n , [o], [u], [a], O, o [5] [ø] Ø, ø Milan Straka, Jan Hajič, and J… find
  18. Anthony C Woodbury. 2003. Defining documentary linguistics. In Peter K. Austin, editor, Language Documentation and Description, volume 1, pages 35–51. London: SOAS. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/a-2022-morphologically-annotated-corpora-pomak/