NLP for Indigenous Languages of the Americas:
Transfer Learning Meets Translation
Katharina Kann1
1
    University of Colorado Boulder, USA


                                         Abstract
                                         Developing human language technology for truly low-resource languages, such as Indigenous languages,
                                         is challenging. Not only do we lack annotated training data, but even unlabeled data are often only
                                         available in small amounts. To make things worse, many truly low- resource languages are not represented
                                         in the pretraining data of multilingual language models. This talk will be centered around how to build
                                         NLP systems for Indigenous languages of the Americas. We will talk about the creation of AmericasNLI,
                                         a natural language inference dataset for Indigenous languages. Then, we will discuss model adaptation
                                         and translation-based approaches for the task. We will end this talk with a discussion of open questions
                                         and challenges for the development of NLP systems for Indigenous languages of the Americas.




The International Conference and Workshop on Agglutinative Language Technologies as a challenge of Natural
Language Processing, ALTNLP’22, June 7-8, Koper, Slovenia
Envelope-Open katharina.kann@colorado.edu (K. Kann)
                                       © 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
    CEUR
    Workshop
    Proceedings
                  http://ceur-ws.org
                  ISSN 1613-0073
                                       CEUR Workshop Proceedings (CEUR-WS.org)