MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling

Rob van der Goot; Anette Jensen; Emil Allerslev Schledermann; Mikkel Wildner Kildeberg; Nicolaj Larsen; Mike Zhang; Elisa Bassignana

MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling

Rob van der Goot, Anette Jensen, Emil Allerslev Schledermann, Mikkel Wildner Kildeberg, Nicolaj Larsen, Mike Zhang, Elisa Bassignana

Aalborg Universitet

Publikation: Konference artikel i Proceeding eller bog/rapport kapitel › Konferencebidrag i proceedings › Forskning › peer review

Abstract

Current language models (LMs) mostly exploit subwords as input units based on statistical co-occurrences of characters. Adjacently, previous work has shown that modeling morphemes can aid performance for Natural Language Processing (NLP) models. However, morphemes are challenging to obtain as there is no annotated data in most languages. In this work, we release a wide-coverage Danish morphological segmentation evaluation set. We evaluate a range of unsupervised token segmenters and evaluate the downstream effect of using morphemes as input units for transformer-based LMs. Our results show that popular subword algorithms perform poorly on this task, scoring at most an F1 of 57.6 compared to 68.0 for an unsupervised morphological segmenter (Morfessor). Furthermore, evaluate a range of segmenters on the task of language modeling.

Originalsprog	Engelsk
Titel	Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025)
Redaktører	Richard Johansson, Sara Stymne
Antal sider	7
Udgivelsessted	Tallinn, Estonia
Forlag	University of Tartu Library
Publikationsdato	1 mar. 2025
Sider	223-229
ISBN (Trykt)	978-9908-53-109-0
Status	Udgivet - 1 mar. 2025

Adgang til dokumentet

Citationsformater

Goot, R. V. D., Jensen, A., Schledermann, E. A., Kildeberg, M. W., Larsen, N., Zhang, M., & Bassignana, E. (2025). MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling. I R. Johansson, & S. Stymne (red.), Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025) (s. 223-229). University of Tartu Library. https://aclanthology.org/2025.nodalida-1.23.pdf

Goot, Rob van der ; Jensen, Anette ; Schledermann, Emil Allerslev et al. / MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling. Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025). red. / Richard Johansson ; Sara Stymne. Tallinn, Estonia : University of Tartu Library, 2025. s. 223-229

@inproceedings{a4555d835f294b04b0c313151f26d4c8,

title = "MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling",

abstract = "Current language models (LMs) mostly exploit subwords as input units based on statistical co-occurrences of characters. Adjacently, previous work has shown that modeling morphemes can aid performance for Natural Language Processing (NLP) models. However, morphemes are challenging to obtain as there is no annotated data in most languages. In this work, we release a wide-coverage Danish morphological segmentation evaluation set. We evaluate a range of unsupervised token segmenters and evaluate the downstream effect of using morphemes as input units for transformer-based LMs. Our results show that popular subword algorithms perform poorly on this task, scoring at most an F1 of 57.6 compared to 68.0 for an unsupervised morphological segmenter (Morfessor). Furthermore, evaluate a range of segmenters on the task of language modeling.",

keywords = "Morphological segmentation, Unsupervised token segmentation, Danish language processing, Transformer-based language models, Morpheme-based input representations",

author = "Goot, \{Rob van der\} and Anette Jensen and Schledermann, \{Emil Allerslev\} and Kildeberg, \{Mikkel Wildner\} and Nicolaj Larsen and Mike Zhang and Elisa Bassignana",

year = "2025",

month = mar,

day = "1",

language = "English",

isbn = "978-9908-53-109-0",

pages = "223--229",

editor = "Richard Johansson and Sara Stymne",

booktitle = "Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025)",

publisher = "University of Tartu Library",

}

Goot, RVD, Jensen, A, Schledermann, EA, Kildeberg, MW, Larsen, N, Zhang, M & Bassignana, E 2025, MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling. i R Johansson & S Stymne (red), Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025). University of Tartu Library, Tallinn, Estonia, s. 223-229. <https://aclanthology.org/2025.nodalida-1.23.pdf>

MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling. / Goot, Rob van der; Jensen, Anette; Schledermann, Emil Allerslev et al.
Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025). red. / Richard Johansson; Sara Stymne. Tallinn, Estonia: University of Tartu Library, 2025. s. 223-229.

Publikation: Konference artikel i Proceeding eller bog/rapport kapitel › Konferencebidrag i proceedings › Forskning › peer review

TY - GEN

T1 - MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling

AU - Goot, Rob van der

AU - Jensen, Anette

AU - Schledermann, Emil Allerslev

AU - Kildeberg, Mikkel Wildner

AU - Larsen, Nicolaj

AU - Zhang, Mike

AU - Bassignana, Elisa

PY - 2025/3/1

Y1 - 2025/3/1

N2 - Current language models (LMs) mostly exploit subwords as input units based on statistical co-occurrences of characters. Adjacently, previous work has shown that modeling morphemes can aid performance for Natural Language Processing (NLP) models. However, morphemes are challenging to obtain as there is no annotated data in most languages. In this work, we release a wide-coverage Danish morphological segmentation evaluation set. We evaluate a range of unsupervised token segmenters and evaluate the downstream effect of using morphemes as input units for transformer-based LMs. Our results show that popular subword algorithms perform poorly on this task, scoring at most an F1 of 57.6 compared to 68.0 for an unsupervised morphological segmenter (Morfessor). Furthermore, evaluate a range of segmenters on the task of language modeling.

AB - Current language models (LMs) mostly exploit subwords as input units based on statistical co-occurrences of characters. Adjacently, previous work has shown that modeling morphemes can aid performance for Natural Language Processing (NLP) models. However, morphemes are challenging to obtain as there is no annotated data in most languages. In this work, we release a wide-coverage Danish morphological segmentation evaluation set. We evaluate a range of unsupervised token segmenters and evaluate the downstream effect of using morphemes as input units for transformer-based LMs. Our results show that popular subword algorithms perform poorly on this task, scoring at most an F1 of 57.6 compared to 68.0 for an unsupervised morphological segmenter (Morfessor). Furthermore, evaluate a range of segmenters on the task of language modeling.

KW - Morphological segmentation

KW - Unsupervised token segmentation

KW - Danish language processing

KW - Transformer-based language models

KW - Morpheme-based input representations

M3 - Article in proceedings

SN - 978-9908-53-109-0

SP - 223

EP - 229

BT - Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025)

A2 - Johansson, Richard

A2 - Stymne, Sara

PB - University of Tartu Library

CY - Tallinn, Estonia

ER -

Goot RVD, Jensen A, Schledermann EA, Kildeberg MW, Larsen N, Zhang M et al. MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling. I Johansson R, Stymne S, red., Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025). Tallinn, Estonia: University of Tartu Library. 2025. s. 223-229

MorSeD: Morphological Segmentation of Danish and its Effect on Language Modeling

Abstract

Adgang til dokumentet

Fingeraftryk

Citationsformater