Elamur Research⌕
☰
home / papers / 2026 / Amharic DBpedia Chapter: A Knowledge Graph for a Low-Resource Language

Amharic DBpedia Chapter: A Knowledge Graph for a Low-Resource Language

Paper · 2026 · HIzkiel Mitiku Alemayehu, Tilahun Abedissa Taffa, Meti Adane Bayissa et al.

Summary

DBpedia builds open knowledge graphs from Wikipedia through language-specific chapters, but low-resource languages are poorly covered. This paper takes the first steps toward an Amharic chapter. The authors extend the DBpedia Extraction Framework (DEF) to handle what makes Amharic Wikipedia hard: the Ge'ez script, the Ethiopian calendar, Ge'ez numerals, and infobox templates whose definitions often disagree with how pages actually use them. Their additions include a parser for dates written in the Ethiopian calendar, a tool that turns Ethiopian dates into Gregorian ones, a converter between Arabic and Ge'ez numbers, and Amharic template mappings, and they also automate the extraction pipelines.

To reduce manual mapping work, they test two steps:

  • zero-shot NLLB-200 translation of infobox property names, which reaches 45.31 BLEU;
  • ranking DBpedia ontology properties for each mapped property with multilingual encoders, where a fine-tuned Afro-XLM-R reaches 92.1% Top-10 accuracy.

The resulting graph and all tooling are openly published. The work gives Amharic a queryable structured knowledge source and serves as a template for building chapters in other under-served languages.

Key points

  • The DEF is extended with Amharic components: a parser for Ethiopian-calendar dates, a tool converting Ethiopian dates into their Gregorian equivalents, a converter between Arabic and Ge'ez numerals, and Amharic mapping files in the DBpedia mapping framework.
  • Zero-shot NLLB-200 translates Amharic infobox property names at 45.31 BLEU.
  • Ontology alignment covers 58 DBpedia classes and benchmarks three encoders: Afro-XLM-R Base, XLM-R Base and Amharic fine-tuned mBERT. Fine-tuned Afro-XLM-R performs best, with 92.1% Top-10 accuracy and strong MRR.
  • The authors identify two main failure modes:
    • mismatch between infobox templates and actual page usage, such as outdated property names still used on pages;
    • calendar or numeral formats the parsers cannot normalize, which either lose facts or produce invalid values.
  • The graph is released at am.dbpedia.org, on a DBpedia Databus collection, through a Tentris query endpoint and as a Zenodo archive, together with the automation workflows and documentation.

Interesting ideas

  • I1. Values in local formats, such as another calendar or another digit system, are converted into a canonical form before any triple is produced, because unnormalized values either vanish silently or enter the graph as garbage.
  • I2. Ontology alignment is treated as a ranking task scored with Top-k accuracy and MRR, not as a single forced label, so a human only picks from a short candidate list.
  • I3. Mapping is split into two stages: first translate the property label with an off-the-shelf MT model, then match the translation to ontology properties with an encoder.
  • I4. Drift between a declared schema (the template) and its real use on pages is named as a separate source of extraction failures and diagnosed case by case.
  • I5. The extraction design keeps two paths: high-precision triples from curated mappings, and broader but looser raw extraction of labels, links and categories.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (world): Mova's measures and values layer could put dates, spelled-out numbers and unusual numeral notations through explicit converters into one canonical value, and log when a conversion fails rather than dropping the fact.
  • I2 (world): when linking verbs or entities to WordNet/FrameNet/Wikidata, Mova could keep a ranked candidate list with scores, which makes each linking decision explainable and easy to review.
  • I3 (world): a similar translate-then-match chain could help compile Mova's ontology from non-English Wikidata labels, with each stage inspected and scored on its own.
  • I4 (learn): Mova could routinely compare the label inventory of each treebank or silver source against its actual usage, so that outdated or off-spec labels are caught before they reach training data.
  • I5 (world): Mova's knowledge graph could record which path produced each fact, curated mapping or raw harvest, so readers can weigh facts by trust when they explain an answer.

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

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,085 worksArtificial Intelligence · 8,888 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 3Research subjects and languages the work is about.Semitic languages · 1,485 worksKnowledge graphs & ontologies · 103 worksLow-resource languages · 165 works

Cite

Cite as HIzkiel Mitiku Alemayehu, Tilahun Abedissa Taffa, Meti Adane Bayissa, Andargachew Asfaw Zewge, Hamada Zahera, Ricardo Usbeck and Axel-Cyrille Ngonga Ngomo. 2026. Amharic DBpedia Chapter: A Knowledge Graph for a Low-Resource Language. Proceedings of the Fifteenth Language Resources and Evaluation Conference.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/536jn6g8rjnx

@inproceedings{alemayehu2026amharic,
  title = {{Amharic DBpedia Chapter: A Knowledge Graph for a Low-Resource Language}},
  author = {HIzkiel Mitiku Alemayehu, Tilahun Abedissa Taffa, Meti Adane Bayissa, Andargachew Asfaw Zewge, Hamada Zahera, Ricardo Usbeck and Axel-Cyrille Ngonga Ngomo},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/536jn6g8rjnx},
  url = {https://doi.org/10.63317/536jn6g8rjnx}
}

References

29 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Samuael Adnew and Paul Pu Liang. 2024. Semantically corrected amharic automatic speech recognition.
  2. H. Al-Feel. 2015. The roadmap for the arabic chapter of dbpedia. In Proceedings. Available at: https://api.semanticscholar.org/Cor-pusID:207984981. ↗
  3. Haytham Al-Feel. 2013. A step towards the arabic dbpedia. International Journal of Computer Applications, 80(3). ↗
  4. Jesujoba O. Alabi, David Ifeoluwa Adelani, Marius Mosbach, and Dietrich Klakow. 2022. Adapting pre-trained language models to African languages via multilingual adaptive fine-tuning. In Proceedings of the 29th International Conference on Computational Linguistics, pages 4336–4349, Gyeongju, Republic of Korea. International Committee on Computational Linguistics. ↗
  5. Hizkiel Mitiku Alemayehu, Hamada M Zahera, and Axel-Cyrille Ngonga Ngomo. 2024. Error analysis of multilingual language models in machine translation: A case study of English-Amharic translation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 19758–19768, Miami, Florida, USA. Association for Computational Linguistics. find
  6. S. Auer, C. Bizer, G. Kobilarov, J. Lehmann, R. Cyganiak, and Z. Ives. 2007. Dbpedia: A nucleus for a web of open data. In The Semantic Web, volume 4825 of Lecture Notes in Computer Science, pages 722–735, Berlin/Heidelberg. Springer. find
  7. Satanjeev Banerjee and Alon Lavie. 2005. ME-TEOR: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization, pages 65–72, Ann Arbor, Michigan. Association for Computational Linguistics. find
  8. Tadesse Destaw Belay, Atnafu Lambebo Tonja, Olga Kolesnikova, Seid Muhie Yimam, Abinew Ali Ayele, Silesh Bogale Haile, Grigori Sidorov, and Alexander Gelbukh. 2022. The effect of normalization for bi-directional amharic-english neural machine translation. find
  9. Elena Cabrio, Serena Villata, and Fabien Gandon. 2014. Classifying inconsistencies in dbpedia language specific chapters. In LREC, pages 1443–1450. Citeseer. find
  10. Dawit Chali and Peter Parapatics. 2023. Language policy and practices in an ethiopian university towards multilingualism. Languages, 9(6):198. ↗
  11. Julien Cojan, Elena Cabrio, and Fabien Gandon. 2013. Filling the gaps among dbpedia multilingual chapters for question answering. In Proceedings of the 5th Annual ACM Web Science Conference, pages 33–42. find
  12. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Unsupervised cross-lingual representation learning at scale. CoRR, abs/1911.02116. ↗
  13. Peter T. Daniels and William Bright, editors. 1996. The World’s Writing Systems. Oxford University Press, Oxford. find
  14. Encyclopaedia Britannica. 2024. guage. Amharic lan-Joseph L Fleiss. 1971. Measuring nominal scale agreement among many raters. Psychological Bulletin, 76(5):378–382. find
  15. Henry F. Fliegel and Thomas C. Van Flandern. 1968. A machine algorithm for processing calendar dates. Communications of the ACM, 11(10):657. find
  16. Grover Hudson. 2013. Northeast African Semitic: Lexical Comparisons and Analysis, volume 26 of Porta Linguarum Orientalium. Harrassowitz Verlag, Wiesbaden. find
  17. E. K. Kim, M. Weidl, K. S. Choi, and S. Auer. 2010. Towards a korean dbpedia and an approach for complementing the korean wikipedia based on dbpedia. In Proceedings of the 5th Open Knowledge Conference (OKCon 2010), CEUR Workshop Proceedings, pages 12–21. find
  18. D. Kontokostas, C. Bratsas, S. Auer, S. Hellmann, I. Antoniou, and G. Metakides. 2012. Internationalization of linked data: The case of the greek dbpedia edition. Journal of Web Semantics, 15:51–61. ↗
  19. Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas, Pablo N Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick Van Kleef, Sören Auer, et al. 2015. Dbpedia– a large-scale, multilingual knowledge base extracted from wikipedia. Semantic web, 6(2):167–195. ↗
  20. Mohamed Morsey, Jens Lehmann, Sören Auer, Claus Stadler, and Sebastian Hellmann. 2012. Dbpedia and the live extraction of structured data from wikipedia. Program, 46(2):157–181. ↗
  21. Hellina Hailu Nigatu, John Canny, and Sarah Chasins. 2024. Low-resourced languages and online knowledge repositories: A need-finding study. pages 1–21. find
  22. Hellina Hailu Nigatu, Atnafu Lambebo Tonja, Henok Biadglign Ademtew, Hizkel Mitiku Alemayehu, Negasi Haile Abadi, Tadesse Destaw Belay, and Seid Muhie Yimam. 2025. A case against implicit standards: Homophone normalization in machine translation for languages that use the ge’ez script. find
  23. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318. ↗
  24. Maja Popović. 2015. chrF: character n-gram Fscore for automatic MT evaluation. In Proceedings of the Tenth Workshop on Statistical Machine Translation, pages 392–395, Lisbon, Portugal. Association for Computational Linguistics. find
  25. Matthew G Snover, Nitin Madnani, Bonnie Dorr, and Richard Schwartz. 2009. Ter-plus: paraphrase, semantic, and alignment enhancements to translation edit rate. Machine Translation, 23(2):117–127. ↗
  26. T. Tamrat. 2008. Ethiopian calendar & millennia highlights. International Journal of Ethiopian Studies, 3(2):177–188. Available at: http://www.jstor.org/stable/27828897. ↗
  27. NLLB Team, Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Guillaume Wenzek, Al Youngblood, Bapi Akula, Loic Barrault, Gabriel Mejia Gonzalez, Prangthip Hansanti, John Hoffman, Semarley Jarrett, Kaushik Ram Sadagopan, Dirk Rowe, Shannon Spruit, Chau Tran, Pierre Andrews, Necip Fazil Ayan, Shruti Bhosale, Sergey Edunov, Angela Fan, Cynthia Gao, Vedanuj Goswami, Francisco Guzmán, Philipp Koehn, Alexandre Mourachko, Christophe Ropers, Safiyyah Saleem, Holger Schwenk, and Jeff Wang. 2022. No language left behind: Scaling human-centered machine translation. ↗
  28. Gebremichael Tesfagerish et al. 2022. Amharic-english neural machine translation. Journal of Artificial Intelligence Research. 9. find
  29. Language Resource References David Ifeoluwa Adelani. 2021. Davlan/bert-base-multilingual-cased-finetuned-amharic. Accessed: 2025-10-24. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/alemayehu-2026-amharic-dbpedia-chapter-knowledge-graph/