@inproceedings{martins-etal-2026-libras,
    title = "{L}ibras-{UFP}el Corpus: A Parallel Dataset of {B}razilian {S}ign {L}anguage and {P}ortuguese for Multimodal Research and Processing",
    author = "Martins, Antonielle  and
      Santana, Brenda S.  and
      Martins, Francielle  and
      Lebedeff, Tatiana  and
      Nunes, Darley  and
      Bohm, Luisa",
    editor = "Souza, Marlo  and
      de-Dios-Flores, Iria  and
      Santos, Diana  and
      Freitas, Larissa  and
      Souza, Jackson Wilke da Cruz  and
      Ribeiro, Eug{\'e}nio",
    booktitle = "Proceedings of the 17th International Conference on Computational Processing of {P}ortuguese ({PROPOR} 2026) - Vol. 1",
    month = apr,
    year = "2026",
    address = "Salvador, Brazil",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-dnd/2026.propor-1.112/",
    pages = "1068--1073",
    ISBN = "979-8-89176-387-6",
    abstract = "The Libras-UFPel Corpus is a multimodal, multilayer parallel resource designed for the documentation and computational analysis of Brazilian Sign Language (Libras) in systematic alignment with written Portuguese. By integrating controlled recordings with naturalistic data from the Invent{\'a}rio Nacional de Libras-Pelotas, the corpus ensures interoperability through shared methodological standards. The dataset currently comprises 4,800 controlledaudiovisual records (2,400 sentences and 2,400 isolated signs) fully paired with Portuguese translations, supplemented by approximately 10 hours of spontaneous interaction from threenew naturalistic interviews, currently in the editing phase. To date, 1,200 controlled sentences have been lemmatized, gloss-annotatedand translated, providing a structured parallel subset for Libras-to-Portuguese Sign Language Processing tasks such as recognition and machine translation. The annotation model follows a hierarchical structure covering lexical, partially lexical, and non-lexical signs, including independent tiers for non-manual markers. By bridging descriptive linguistics and Natural Language Processing, Libras-UFPel Corpus serves as a reference source for bilingual data-driven modeling, advancing digital inclusion and linguistic accessibility."
}