Morphological Processing of Low-Resource Languages: Where We Are and What’s Next

Findings of the Association for Computational Linguistics: ACL 2022, pages 988 - 1007 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Morphological Processing of Low-Resource Languages: Where We Are and What’s Next Adam Wiemerslage♯∗ Miikka Silfverberg♭∗ Changbing Yang♭ Arya D. McCarthy♮ Garrett Nicolai♭ Eliana Colunga♯ Katharina Kann♯ ♯University of Colorado Boulder ♭University of British Columbia ♮Johns Hopkins University Abstract Automatic morphological processing can aid downstream natural language processing appli- cations, especially for low-resource languages, and assist language documentation efforts for endangered languages. Having long been multilingual, the field of computational mor- phology is increasingly moving towards ap- proaches suitable for languages with minimal or no annotated resources. First, we survey re- cent developments in computational morphol- ogy with a focus on low-resource languages. Second, we argue that the field is ready to tackle the logical next challenge: understand- ing a language’s morphology from raw text alone. We perform an empirical study on a truly unsupervised version of the paradigm completion task and show that, while existing state-of-the-art models bridged by two newly proposed models we devise perform reason- ably, there is still much room for improvement. The stakes are high: solving this task will in- crease the language coverage of morphologi- cal resources by a number of magnitudes. 1 Introduction Automatic morphological processing tools have the potential to drastically speed up language docu- mentation (Moeller et al., 2020) and thereby help combat the language endangerment crisis (Austin and Sallabank, 2011). Explicit morphological in- formation also benefits myriad NLP tasks, such as parsing (Hohensee and Bender, 2012; Seeker and C¸ etino˘glu, 2015), language modeling (Blevins and Zettlemoyer, 2019; Park et al., 2021; Hofmann et al., 2021), and machine translation (Dyer et al., 2008; Tamchyna et al., 2017). For low-resource languages, valuable morpho- logical resources are typically small or non-existent. Of late, the field of computational morphology has increased its efforts to extend the coverage of multi- lingual morphological resources (Kirov et al., 2016, ∗*The first two authors contributed equally. 2018; McCarthy et al., 2020a; Metheniti and Neu- mann, 2020). Simultaneously, there has been a revival of minimally supervised and unsupervised models for morphological tasks, such as segmenta- tion (Eskander et al., 2019), inflection (Kann et al., 2017b), and lemmatization (Bergmanis and Gold- water, 2019). Given the speed of recent develop- ments, it is important to reflect on where we are as a field and what future challenges lie ahead. To this end, we survey recent computational mor- phology: we review existing multilingual resources (§2) and tasks and systems (§3), with a focus on low-resource languages. Given recent develop- ments in unsupervised segmentation, low-resource morphological inflection, and unsupervised mor- phological paradigm completion (Jin et al., 2020; Erdmann et al., 2020)—which we argue is not fully unsupervised—we believe the community is poised for the next logical step: inferring a language’s mor- phology purely from raw text. In §4, we formalize a new task: truly un- supervised morphological paradigm completion (tUMPC). We then introduce a pipeline with two novel components (§4.3): one model for …

Content truncated to 500 words