What a Creole Wants, What a Creole Needs

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 6439–6449 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 6439 What a Creole Wants, What a Creole Needs Heather Lent1, Kelechi Ogueji2, Miryam de Lhoneux1,3,4, Orevaoghene Ahia5, Anders Søgaard1 1 University of Copenhagen, Denmark 2 University of Waterloo, Canada 3 Uppsala University, Sweden 4 KU Lueven, Belgium 5 University of Washington, United States {hcl, ml, soegaard}@di.ku.dk kelechi.ogueji@uwaterloo.ca, oahia@cs.washington.edu Abstract In recent years, the natural language processing (NLP) community has given increased attention to the disparity of efforts directed towards high-resource languages over low-resource ones. Efforts to remedy this delta often begin with translations of existing English datasets into other languages. However, this approach ignores that different language communities have different needs. We consider a group of low-resource languages, Creole languages. Creoles are both largely absent from the NLP literature, and also often ignored by society at large due to stigma, despite these languages having sizable and vibrant communities. We demonstrate, through conversations with Creole experts and surveys of Creole-speaking communities, how the things needed from language technology can change dramatically from one language to another, even when the languages are considered to be very similar to each other, as with Creoles. We discuss the prominent themes arising from these conversations, and ultimately demonstrate that useful language technology cannot be built without involving the relevant community. Keywords: natural language processing, low-resource languages, Creole 1. Introduction The field of natural language processing (NLP) has be- come aware that most of the world’s languages are un- fortunately under-represented, or entirely absent, from the field’s body of work (Joshi et al., 2020). In re- cent years, there has been a push in efforts to ame- liorate this discrepancy (Nekoto et al., 2020; Mirza- khalov et al., 2021; Ogueji et al., 2021). Among these low-resourced languages1 are Creole languages, which are particularly under-resourced due to barri- ers like societal stigma (Siegel, 1999), despite the fact that these languages are spoken by many people glob- ally. One line of work has focused on creating datasets for low-resource languages via the translation of ex- isting high-resource language datasets (Conneau et al., 2018; Artetxe et al., 2020b; Budur et al., 2020). De- spite the popularity of this method, it poses several is- sues, which can negatively affect the communities of these low-resource languages. One such issue lies in translation artifacts, which have been shown to have notable impacts on the performance of models trained with such datasets (Artetxe et al., 2020a). Furthermore, translated datasets are often simplified and unnatural, a phenomenon referred to as translationese (Volansky et al., 2013). This has also been shown to adversely af- fect the evaluation of machine translation models (Gra- ham et al., 2020). Creoles, too, are not immune to the shortcomings of this approach. Moreover, many trans- 1This term is often largely ambiguous, and all “low- resource” languages should not be conflated together into one large group, but rather considered independently, in the con- …

Content truncated to 500 words