Elamur Research⌕
☰
home / papers / 2026 / Cross-lingual and Word-Independent Methods for Quantifying Degree of Grammaticalization

Cross-lingual and Word-Independent Methods for Quantifying Degree of Grammaticalization

Paper · 2026 · Ryo Nagata, Daichi Mochihashi, Misato Ido et al.

Summary

The paper addresses the automatic measurement of how far a word has moved from content-word status toward function-word status, a process called grammaticalization. Earlier automated approaches were few. The strongest of them works only for English deverbal prepositions, and earlier studies disagreed about which features actually help. The authors compare three approaches. The first is a recurrence-time baseline, reformulated as the shape parameter β of a Weibull distribution. The other two are classifiers built on static word embeddings. Because no graded training labels exist, those classifiers are trained with either PU-learning (Elkan and Noto, 2008) or a cross-validation-style scheme the authors call CV-learning. They also build new evaluation datasets covering English and Japanese.

The CV-learning variant agrees best with human judgments, reaching moderate to strong correlations on English deverbal prepositions and on Japanese nouns undergoing grammaticalization. The analysis shows that recurrence-time distribution and contextual diversity are necessary but not sufficient signals. Contextual similarity proves more informative. Applied to CCOHA, the best method suggests candidate grammaticalizing words and possible counterexamples to the unidirectionality hypothesis. This offers linguists a scalable tool that does not depend on a particular language or word.

Key points

  • Methods compared: (1) the Weibull shape parameter β of recurrence time, (2) a PU-learning classifier on static word embeddings, and (3) a CV-learning classifier on the same embeddings. CV-learning performed best.
  • Evaluation data: English human judgments come from Hayashi (2015), who rated 37 deverbal prepositions on a 0–10 scale. The authors added new datasets covering three types of grammaticalizing items across two languages, each containing only dozens of words. The datasets and code are released at github.com/nagata-github/grammaticalization_tools.
  • Feature finding: recurrence-time distribution and contextual diversity, whose usefulness earlier work disputed, act only as necessary conditions. Contextual similarity is the more effective cue.
  • Diachronic pilot study: the authors compared CCOHA 1800s and 2000s sub-corpora for words occurring more than 5,000 times in each.
    • such showed a possible drift toward adjective-like use.
    • upon appeared degrammaticalized, but only because the tokenizer split hyphenated compounds such as agreed-upon.
  • Stated limitations:
    • The scores serve for finding candidates and supporting existing hypotheses, not as definitive measures.
    • The methods assume a word keeps a fixed form, so they cannot directly detect grammaticalizing phrases.
    • Contextualized embeddings are left for future work.

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

 Institutions

🎓The University of TokyoTokyo · Japan90 works · 9,156 citations🔬National Institute of Advanced Industrial Science and TechnologyTsukuba · Japan17 works · 2,760 citations

Also on the byline: Tokyo University of Foreign Studies · The Institute of Statistical Mathematics.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 828 worksHistorical linguistics & language change · 249 works

Cite

Cite as Ryo Nagata, Daichi Mochihashi, Misato Ido, Yusuke Kubota, Naoki Otani, Yoshifumi Kawasaki and Hiroya Takamura. 2026. Cross-lingual and Word-Independent Methods for Quantifying Degree of Grammaticalization. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-long.221

@inproceedings{nagata2026cross,
  title = {{Cross-lingual and Word-Independent Methods for Quantifying Degree of Grammaticalization}},
  author = {Ryo Nagata, Daichi Mochihashi, Misato Ido, Yusuke Kubota, Naoki Otani, Yoshifumi Kawasaki and Hiroya Takamura},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.221},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.221}
}

References

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

  1. Reem Alatrash, Dominik Schlechtweg, Jonas Kuhn, and Sabine Schulte im Walde. 2020. CCOHA: Clean corpus of historical American English. In Proc. of the 12th Language Resources and Evaluation Conference, pages 6958–6966.
  2. Eduardo G. Altmann, Janet B. Pierrehumbert, and Adilson E. Motter. 2009. Beyond word frequency: Bursts, lulls, and scaling in the temporal distributions of words. PLOS ONE, 4:1–7. find
  3. Joan Bybee. 2003. Cognitive Processes in Grammaticalization, pages 145–167. Lawrence ErLbaum Associates. find
  4. Joan Bybee. 2006. From usage to grammar: The mind’s response to repetition. Language, 82:711–733. find
  5. Lyle Campbell. 2000. What’s wrong with grammaticalization? Language Sciences, 23:113–161. find
  6. Mark Davies. 2012. Expanding horizons in historical linguistics with the 400-million word corpus of historical American English. Corpora, 7(2):121–157. find
  7. Robbert De Troij and Freek Van de Velde. 2020. Beyond mere text frequency: Assessing subtle grammaticalization by different quantitative measures. A case study on the dutch Soort construction. Languages, 5(4). find
  8. Charles Elkan and Keith Noto. 2008. Learning classifiers from only positive and unlabeled data. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 213–220. find
  9. Teruhiko Fukaya. 1997. The Emergence of -ing Prepositions in English: A Corpus-Based Study, pages 285–300. Taishukan, Tokyo. find
  10. Stefan T. Gries. 2010. Dispersions and adjusted frequencies in corpora: further explorations, pages 197– 212. Brill, Leiden. find
  11. Martin Haspelmath. 1999. Why is grammaticalization irreversible? Linguistics, 37:1043–1068. find
  12. Hideo Teramura. 1992. Syntax and semantics of noun modification – part 4, pages 1–34. Kuroshio, Tokyo. find
  13. Tomoaki Hayashi. 2015. Prepositionalities of deverbal prepositions: Differences in degree of grammaticalization. Papers in Linguistic Science, 21:129–151. find
  14. A Paul Hopper. 1991. On some Principles of Grammaticalization, pages 17–35. John Benjamins Publishing Company. find
  15. Paul J. Hopper and Elizabeth Closs Traugott. 2003. Grammaticalization, second edition. Cambridge University Press, New York. find
  16. Brian Joseph. 2000. Is there such a thing as “grammaticalization?”. Language Sciences - LANG SCI, 23:163–186. find
  17. Slava M. Katz. 1996. Distribution of content words and phrases in text and language modelling. Natural Language Engeneering, 2(1):15–59. find
  18. Bernd Kortmann and Ekkehard König. 1992. Categorial reanalysis: The case of deverbal prepositions. Linguistics, 30:671–698. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/nagata-2026-cross-lingual-word-independent-methods/