Elamur Research⌕
☰
home / papers / 2026 / Learning reduplicative templates as hidden structures: the case of reduplication-phonology interactions

Learning reduplicative templates as hidden structures: the case of reduplication-phonology interactions

Summary

Computational models of morphophonology have mostly dealt with concatenative morphology. This paper addresses reduplication, where a learner must work out three things at once: the prosodic template that controls copying, the underlying representations (URs) of morphemes, and the phonological grammar. Yang Wang extends the EM-MaxEnt framework (O'Hara; Wang and Hayes) by treating the reduplicative template as a latent UR of the reduplicative morpheme, so the same EM procedure that selects segmental URs also selects templates. The hypothesis space adds three things:

  • a set of candidate templates;
  • one template-binding constraint, RED=X, alongside Base-Reduplicant Correspondence Theory (BRCT) constraints;
  • a parametric GEN that produces reduplication candidates.

The learner was tested on three toy languages showing normal application, overapplication and underapplication of final devoicing. In all three it found the correct stem UR and the heavy-syllable template. It learned normal application and overapplication as categorical patterns. For underapplication it settled on free variation. This mirrors the classic asymmetry noted by McCarthy and Prince, here arising from the structure of the hypothesis space. The study points toward a single learning route for both segmental and prosodic hidden structure.

Key points

  • The candidate templates for the reduplicative morpheme are {σµ, σµµ, Ft, PrWd}. The constraint set has 12 BRCT and OT constraints plus the new RED=X, for 13 in total.
  • Each toy language has only six observed surface forms. GEN still produces 1,992 hidden-structure/surface-form pairs, and this number scales with morpheme length and the alternation set, not with lexicon size.
  • Learning used EM with L-BFGS-B in the M-step, a regularization setting of 2σ² = 10⁵, non-negative weights, and a stopping threshold of ΔL < 10⁻³.
  • In all three languages the learner chose the UR /bed/ and the template σµµ with posterior probability above .999. IDENT-BR[VOICE] weights were 0.05 (normal), 21.8 (overapplication) and 15.92 (underapplication).
  • In the underapplication language, *FINALVOICEDOBS and IDENT-IO both converged to a weight of 8.56. This produces a 0.50/0.50 split between variants, which the paper compares to free variation in Tagalog.

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 10.18653/v1/2026.scil-main.43

@inproceedings{wang2026learning,
  title = {{Learning reduplicative templates as hidden structures: the case of reduplication-phonology interactions}},
  author = {Wang, Yang},
  booktitle = {Proceedings of the Society for Computation in Linguistics 2026},
  year = {2026},
  doi = {10.18653/v1/2026.scil-main.43},
  url = {https://doi.org/10.18653/v1/2026.scil-main.43}
}

References

18 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. 2022. Synthesizing theories of human language with bayesian program induction. Nature Communications, 13(1):5024. find
  2. John J. McCarthy and Alan S. Prince. 1986. Prosodic morphology. Ms., University of Massachusetts, Amherst, and Brandeis University, Waltham, Mass. find
  3. John J. McCarthy and Alan S. Prince. 1995. Faithfulness and reduplicative identity. In Papers in Optimality Theory, Amherst, MA. Graduate Linguistic Student Association, Dept. of Linguistics, University of Massachusetts. find
  4. Max Nelson, Hossep Dolatian, Jonathan Rawski, and Brandon Prickett. 2020. Probing RNN encoder-decoder generalization of subregular functions using reduplication. Society for Computation in Linguistics, 3(1). find
  5. Charlie O’Hara. 2017. How abstract is more abstract? Learning abstract underlying representations. Phonology, 34(2):325–345. find
  6. Jian Leat Siah, Sam Zukoff, and F. F. Hsieh. under review. Resolving reduplicative opacity in Malay nasal spreading: Argument for Base-Reduplicant correspondence theory. Language. find
  7. Simon Todd, Annie Huang, Jeremy Needle, Jennifer Hay, and Jeanette King. 2022. Unsupervised morphological segmentation in a language with reduplication. In Proceedings of the 19th SIGMORPHON Workshop, pages 12–22, Seattle, Washington. Association for Computational Linguistics. find
  8. Clara Vania and Adam Lopez. 2017. From characters to words to in between: Do we capture morphology? In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2016–2027, Vancouver, Canada. Association for Computational Linguistics. find
  9. Yang Wang. 2024. Studies in morphophonological copying: Analysis, experimentation and modeling. Ph.D. thesis, University of California, Los Angeles. find
  10. Gašper Beguš. 2021. Identity-based patterns in deep convolutional networks: Generative adversarial phonology and reduplication. TACL, 9:1180–1196. find
  11. Ricardo Bermúdez-Otero. 2012. The architecture of grammar and the division of labour in exponence. In Jochen Trommer, editor, The Morphology and Phonology of Exponence, pages 8–83. Oxford University Press, Oxford. find
  12. Arthur P Dempster, Nan M Laird, and Donald B Rubin. 1977. Maximum likelihood from incomplete data via the em algorithm. Journal of the royal statistical society: series B (methodological), 39(1):1–22. find
  13. Hossep Dolatian and Jeffrey Heinz. 2018. Learning reduplication with 2-way finite-state transducers. In Proceedings of the 14th International Conference on Grammatical Inference, volume 93 of Proceedings of Machine Learning Research, pages 67–80. PMLR. find
  14. Kevin Ellis, Adam Albright, Armando Solar-Lezama, Joshua B Tenenbaum, and Timothy J O’Donnell. Yang Wang and Bruce Hayes. to appear. Learning phonological underlying representations: the role of abstractness. Linguistic Inquiry. find
  15. Ronnie Bring Wilbur. 1973. The phonology of reduplication. University of Illinois at Urbana-Champaign. find
  16. Colin Wilson. 2018. Modeling morphological affixation with interpretable recurrent networks: sequential rebinding controlled by hierarchical attention. In CogSci. find
  17. Hongzhi Xu, Jordan Kodner, Mitch Marcus, and Charles Yang. 2020. Modeling morphological typology for unsupervised learning of language morphology. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6672– 6681. find
  18. Kie Zuraw. 2002. Aggressive reduplication. Phonology, 19(3):395–439. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-learning-reduplicative-templates-as-hidden/