@InProceedings{ilman:2026:latell,
  author    = {Ilman, Tetiana},
  title     = {Worldview Annotation for Low-Resource Languages},
  booktitle      = {Proceedings of the First International Conference on Language Technologies for Low-resource Languages (LaTeLL 2026)},
  month          = {September},
  year           = {2026},
  address        = {Fes, Morocco},
  publisher      = {Association for Computational Linguistics},
  pages     = {429--437},
  abstract  = {Low-resource language technology has advanced significantly in preserving linguistic form, yet a critical gap remains: the conceptual systems that languages encode - how communities structure personhood, agency, time, evidence, relationality, and moral responsibility - are systematically lost in current documentation pipelines. We argue that the problem is not only the annotation scale but annotation level. Annotation pipelines are not neutral: they systematically remove meaning-encoding distinctions that are not directly representable in dominant-language categories, even when data originates from native speakers. We introduce the Worldview Annotation Schema (WAS), a three-layer framework that starts at community member level in the field and ends with machine-readable structured data, catching culturally grounded meaning before downstream AI analysis can homogenize it. We track where worldview meaning dissolves in existing pipelines, propose the WAS architecture with a worked example from Maa (Eastern Africa), and outline a governance model that maintains interpretive authority with speaker communities while producing graph-compatible structured output. In the age of AI, language preservation without worldview preservation risks producing languages that survive archivally while disappearing cognitively.},
  url       = {https://aclanthology.org/2026.latell-1.44}
}