Elamur Research⌕
☰
home / papers / 2026 / Simple Additions, Substantial Gains: Expanding Scripts, Languages, and Lineage Coverage in URIEL+

Simple Additions, Substantial Gains: Expanding Scripts, Languages, and Lineage Coverage in URIEL+

Summary

URIEL+ is a linguistic knowledge base that describes languages with geographic, genetic and typological vectors. It is widely used to pick source languages for cross-lingual transfer. However, it suffers from three kinds of data sparsity: it has no information on writing systems, many catalogued languages are missing, and lineage imputation covers only six parent languages.

The authors fix each gap separately:

  • They add a new script vector type built from ScriptSource, with properties such as script type, case and ligatures turned into binary features via one-hot encoding.
  • They import all remaining Glottolog languages, so the full Glottolog family trees are represented.
  • They generalise lineage imputation so that child languages inherit typological and script values from their ancestors across the whole genealogy.

These changes greatly increase coverage and improve imputation quality. In transfer-language selection benchmarks focused on low-resource languages, results are mixed but sometimes better than the original URIEL+. This matters because poorly documented languages often have little data apart from script and genealogy.

Key points

  • Script vectors now cover 7,488 languages. Examples include features like SC_ALPHABET, SC_ABJAD, SC_CASE and SC_LIGATURES. The authors report a 14% sparsity reduction tied to this vector type.
  • Glottolog integration adds 18,710 languages. Total language coverage rises by up to 19,015 languages, about 1,007%.
  • Expanded lineage imputation now reaches 26,449 languages, up from dialects of only six parents (Spanish, French, English, German, Malay, Arabic). Imputation quality metrics improve by up to 35%.
  • Script distance and typological distance capture different relationships. For example, Russian–Kazakh and Spanish–Finnish are close in script but distant typologically, both before and after SoftImpute imputation.
  • In cross-lingual transfer benchmarks, performance sometimes diverges from the original URIEL+. The best setups gain up to 6%.

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

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/5e6qapi74t97

@inproceedings{shipton2026simple,
  title = {{Simple Additions, Substantial Gains: Expanding Scripts, Languages, and Lineage Coverage in URIEL+}},
  author = {Shipton, Mason and Ng, York Hay and Khan, Aditya and Hoang, Phuong H. and Lu, Xiang and Doğruöz, A. Seza and Lee, Annie En-Shiun},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/5e6qapi74t97},
  url = {https://doi.org/10.63317/5e6qapi74t97}
}

References

9 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Dryer, Matthew S. and Haspelmath, Martin. 2013. WALS Online (v2020.3). Zenodo. find
  2. Harald Hammarström and Robert Forkel. 2022. Glottocodes: Identifiers Linking Families, Languages and Dialects to Comprehensive Reference Information. find
  3. Holloway, Steph. 2013. ScriptSource - Writing systems, computers and people. SIL International. find
  4. Khan, Aditya and Shipton, Mason and Anugraha, David and Duan, Kaiyao and Hoang, Phuong H. and Khiu, Eric and Doğruöz, A. Seza and Lee, En-Shiun Annie. 2025. URIEL+: Enhancing Linguistic Inclusion and Usability in a Typological and Multilingual Knowledge Base. Association for Computational Linguisti… find
  5. Koopman, Hilda. 2009. Syntactic Structures of the World’s Languages (SSWL). find
  6. Kortmann, Bernd and Lunkenheimer, Kerstin and Ehret, Katharina. 2020. eWAVE. Zenodo. find
  7. Littell, Patrick and Mortensen, David R. and Lin, Ke and Kairis, Katherine and Turner, Carlisle and Levin, Lori. 2017. URIEL and lang2vec: Representing languages as typological, geographical, and phylogenetic vectors. Association for Computational Linguistics. find our card
  8. Marsico, Egidio and Flavier, Sebastien and Verkerk, Annemarie and Moran, Steven. 2018. BDPROTO: A Database of Phonological Inventories from Ancient and Reconstructed Languages. European Language Resources Association (ELRA). find
  9. Skirgård, Hedvig and Haynie, Hannah J and Blasi, Damián E and Hammarström, Harald and Collins, Jeremy and Latarche, Jay J and Lesage, Jakob and Weber, Tobias and Witzlack-Makarevich, Alena and Passmore, Sam and others. 2023. Grambank Reveals the Importance of Genealogical Constraints on Linguistic D… find

This page in Markdown → https://md.research.elamur.ai/free/en/papers/shipton-2026-simple-additions-substantial-gains-expanding/