Modeling Word-Internal Structures: Morphological Segmentation Across 58 Languages

Vojtěch John, Zdeněk Žabokrtský, Benjamin Reeves


Abstract
We present the largest multilingual experiment to date on word-to-morph segmentation, covering 58 typologically diverse languages. We describe a newly compiled collection of linguistically annotated resources for the task, providing broad coverage and enabling systematic cross-lingual evaluation. Second, we train two neural models on surface morphological segmentation, achieving 81% average word accuracy on the original datasets, slightly outperforming previous methods. Experiments on custom test sets reveal substantial variation in performance, highlighting the need for further harmonization and more robust multilingual approaches.
Anthology ID:
2026.slide-1.16
Volume:
Proceedings of the Workshop on Structured Linguistic Data and Evaluation (SLiDE)
Month:
May
Year:
2026
Address:
Palma de Mallorca, Spain
Editors:
Erhard Hinrichs, Joakim Nivre, Petya Osenova, James Pustejovsky, Claus Zinn
Venues:
SLiDE | WS
SIG:
Publisher:
ELRA Language Resources Association (ELRA)
Note:
Pages:
180–190
Language:
External URL:
https://lrec.elra.info/lrec2026-ws-slide-16
DOI:
10.63317/4qgccaokf8mz
Bibkey:
Cite (ACL):
Vojtěch John, Zdeněk Žabokrtský, and Benjamin Reeves. 2026. Modeling Word-Internal Structures: Morphological Segmentation Across 58 Languages. In Proceedings of the Workshop on Structured Linguistic Data and Evaluation (SLiDE), pages 180–190, Palma de Mallorca, Spain. ELRA Language Resources Association (ELRA).
Cite (Informal):
Modeling Word-Internal Structures: Morphological Segmentation Across 58 Languages (John et al., SLiDE 2026)
Copy Citation: