@inproceedings{mailhot-2026-much,
    title = "How much capacity does {T}urkish inflection require? An empirical study of {GRU} encoder{--}decoder bottlenecks.",
    author = "Mailhot, Fred",
    editor = "Voigt, Rob  and
      Warstadt, Alex  and
      Feldman, Naomi  and
      Linzen, Tal",
    booktitle = "Proceedings of the Society for Computation in Linguistics 2026",
    month = jul,
    year = "2026",
    address = "San Diego, CA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.scil-main.39/",
    pages = "423--426",
    ISBN = "979-8-89176-412-5",
    abstract = "Encoder{--}decoder neural networks with high-dimensional (e.g. $d$=300-{--}500) embeddings and hidden layers can be used to model a variety of morphophonological phenomena as sequence-to-sequence mappings, achieving high accuracy across languages and patterns. We show here that these high-capacity models are overparameterized, at least for the task of morphological inflection, and that simpler and smaller networks can perform near ceiling on the task of inflecting Turkish stems. Moreover these reduced-capacity models encode linguistically relevant information even when they are too small to succeed at the inflectional task."
}