Elamur Research⌕
☰
home / papers / 2026 / From Corpus to Concept Scheme: Developing a SKOS Vocabulary for Armenian Epigraphic Heritage

From Corpus to Concept Scheme: Developing a SKOS Vocabulary for Armenian Epigraphic Heritage

Summary

Armenian epigraphy is one of the oldest inscriptional traditions, yet it is barely represented in digital humanities infrastructure. It has no machine-readable corpus and no controlled vocabulary for inscription types. Existing ontologies such as EAGLE and FAIR Epigraphy are also built around Greco-Roman categories. The authors take a first step toward an Armenian SKOS vocabulary, centred on the EAGLE "Type of Inscription" domain.

They OCR'd and cleaned authoritative printed sources, chiefly the Divan Hay Vimagrut'yan (DHV) series, into a sentence-level corpus. Specialists, with GPT-5 as an assistive tool, compiled an expert-reviewed list of inscription-type terms. This list was used to build an annotated NER dataset, on which spaCy, HyeBERT, mBERT and XLM-RoBERTa were trained. mBERT performed best.

On books held out from all training, the model reliably found known terms in new contexts. However, it almost never surfaced genuinely new inscription types. The authors conclude that expert curation remains essential. The corpus also shows Armenian-specific categories missing from international vocabularies. Data and code are openly released.

Key points

  • The corpus draws on 29 printed publications, including DHV volumes 1–10 and regional, museum and thematic studies. An early baseline used only DHV–10 (about 1,200 sentences), while the full dataset has over 56,000 sentences. The train split has 51,223 sentences, of which only 530 contain a term.
  • The curated expert list contains 41 inscription-type terms, covering funerary, donor, legal, architectural-placement, tax-exemption, water-management and cryptographic types. The generalization analysis, however, cites a 54-term reference vocabulary, and the paper does not explain the difference.
  • On the full dataset, mBERT reached P 0.82, R 0.50 and F1 0.63. XLM-RoBERTa scored F1 0.53, spaCy 0.11 and HyeBERT 0.10.
  • On unseen books the model inferred 98 terms. Of these, 40 overlap with the 136 benchmark terms and 5 are shared by all three sets. The 53 model-only terms were mostly morphological variants or ambiguous expressions.
  • Planned work covers synonymy and variant detection, hierarchy building, choosing preferred and alternative labels, and alignment with EAGLE and FAIR Epigraphy. Resources are hosted at github.com/dhlab-epfl/dhlab-epigraphy-studies.

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 · History & humanities

Institutions

🎓École Polytechnique Fédérale de LausanneLausanne · Switzerland19 works · 6 324 citations

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.latechclfl-1.1

@inproceedings{tamrazyan2026corpus,
  title = {{From Corpus to Concept Scheme: Developing a SKOS Vocabulary for Armenian Epigraphic Heritage}},
  author = {Tamrazyan, Hamest and Nour, Kamal and Boros, Emanuela},
  booktitle = {Proceedings of the 10th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature 2026},
  year = {2026},
  doi = {10.18653/v1/2026.latechclfl-1.1},
  url = {https://doi.org/10.18653/v1/2026.latechclfl-1.1}
}

References

36 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. S. Avagyan and H. Janpoladyan. 1977a. Corpus Inscriptionum Armeniacarum, volume 6. find
  2. Arsen Harutyunyan. 2017. Armaviri marzi vimagrakan zharangutyan usumnasirutyune. In Metsamoryan Yntertsumner I. Patmamshakutayin Argelots-Thangaranneri Tsarayutyun, Yerevan. find
  3. Suren Avagyan and Hovhannes Janpoladyan. 1977b. Divan Hay Vimagrutyan, Prak VI. Academy of Sciences of the Armenian SSR, Yerevan. find
  4. Arsen Harutyunyan. 2019a. Antip vimagrer yerevani katoghike yekeghetsuts. In G. G. Sargsyan and A. E. Harutyunyan, editors, Sedrak Barkhudaryan – 120. Gitakan Hodvatsneri Zhoghovatsu. HAI, Yerevan. find
  5. Ashot Ayvazyan. 2011. Dadivanki Verakangnumy (1997–2011 tt.). HChU, Yerevan. find
  6. M. Barkhudaryan. 1999. Aghvants Yerkir yev Dratsik: Artsakh (The Land of Aghvank and Its Neighbors: Artsakh). Gandzasar, Yerevan. find
  7. Arsen Harutyunyan. 2019b. Sisiani n. adontsi anvan patmutyan tangarani ardzanagir ararkanere. Hnagitutyun: Vem, 11(2(66)). find
  8. S. Barkhudaryan. 1967a. Corpus Inscriptionum Armeniacarum, volume 3. find
  9. Arsen Harutyunyan. 2022. Tarerayin aghetnern u yerknaayin yerevuytnere vimagrerum. mas arayin: Yerkracharzh. Hnagitutyun: Vem, 14(1(77)). find
  10. S. Barkhudaryan. 1973a. Corpus Inscriptionum Armeniacarum, volume 4. find
  11. Arsen Harutyunyan. 2023a. Tarerayin aghetnern u yerknaayin yerevuytnere vimagrerum. mas yerrord: Yerash, morekh. Hnagitutyun: Vem, 15(1(81)). find
  12. S. Barkhudaryan. 1982a. Corpus Inscriptionum Armeniacarum, volume 5. find
  13. Arsen Harutyunyan. 2023b. Tatevi mets anapati vimagrakan zharangutyune. find
  14. S. G. Barkhudaryan, K. Ghafadaryan, and S. T. Saghumyan. 2012a. Corpus Inscriptionum Armeniacarum, volume 9. find
  15. Arsen Harutyunyan and Armine Melkonyan. 2021. S. ejmiatsni tapanagreri: Ghevond vrd. pirghalemiani norahayt zhoghovatsu. Ejmiatsin. find
  16. S. G. Barkhudaryan, K. Ghafadaryan, and S. T. Saghumyan. 2012b. Divan Hay Vimagrutyan, Prak IX. Institute of Archaeology and Ethnography, NAS RA, Yerevan. find
  17. S. Karapetyan. 2021a. The armenian lapidary inscriptions of boon aghvank. find
  18. Sedrak Barkhudaryan. 1967b. Divan Hay Vimagrutyan, Prak III. Academy of Sciences of the Armenian SSR, Yerevan. find
  19. Samvel Karapetyan. 1998. Vratsi Petakan Kaghakakanutyune yev Hay Mshakuyti Hushardzannere (1988–1998), Book II. Gitutyun, NAS RA, Yerevan. find
  20. Sedrak Barkhudaryan. 1973b. Divan Hay Vimagrutyan, Prak IV. Academy of Sciences of the Armenian SSR, Yerevan. find
  21. Samvel Karapetyan. 2004. Kovkasyan Tangarani Haykakan Havaqatsun. HChU, Yerevan. find
  22. Sedrak Barkhudaryan. 1982b. Divan Hay Vimagrutyan, Prak V. Academy of Sciences of the Armenian SSR, Yerevan. find
  23. Samvel Karapetyan. 2014. Akunk yev Katnaghbyur Gyugheri Patmakan Hushardzannere. HChU, Yerevan. find
  24. Francesco Bianchini. 2023. Looking beyond the text: Opportunities and challenges in the digitisation of sanskrit inscriptions. Can’t Touch This. find
  25. Samvel Karapetyan. 2021b. Bun aghvanki hay vima-grutyunnere. find
  26. David Espinosa Espinosa and Isabel Velázquez Soriano. 2021. Epigraphy in the digital age: opportunities and challenges in the recording, analysis and dissemination of inscriptions. find
  27. Samvel Karapetyan. 2021c. Hyusisayin artsakh. https://raa-am.org/northern-artsakh/. ↗
  28. Samvel Karapetyan. 2024. Khojivank. HChU, Yerevan. find
  29. Artak Ghulyan. 2001. Artsakhi yev Syuniki Melikakan Aparanknere. HChU, Yerevan. find
  30. Raffi Kortoshyan. 2013a. Halepi Vimagrere, Prak 16. RAA Publishing, Yerevan. find
  31. Tim Greenwood. 2004. A corpus of early medieval armenian inscriptions. Dumbarton Oaks Papers, 58:27–? find
  32. Raffi Kortoshyan. 2013b. The Inscriptions of Aleppo, volume 16. RAA Publishing, Yerevan. find
  33. Raffi Kortoshyan. 2024. Hndkastani Hayahots Bnakavayreri Vimagrere. HChU, Yerevan. find
  34. Yingwei Yan, Kenneth Dean, Chen-Chieh Feng, Guan Thye Hue, Khee-heong Koh, Lily Kong, Chang Woei Ong, Arthur Tay, Yi-chen Wang, and Yiran Xue. 2020. Chinese temple networks in southeast asia: a webgis digital humanities platform for the collaborative study of the chinese diaspora in southeast asia.… find
  35. Fangchao Liu, John Hindmarch, and Mona Hess. 2023. A review of the cultural heritage linked open data ontologies and models. find
  36. Husik Melkonyan. 2024. Tsaghats Kar Vanke (Patmah-naagitakan Usumnasirutyun). HAI Publishing, Yerevan. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tamrazyan-2026-corpus-concept-scheme-developing-skos/