To POS Tag or Not to POS Tag: The Impact of POS Tags on Morphological Learning in Low-Resource Settings

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 966–978 August 1–6, 2021. ©2021 Association for Computational Linguistics 966 To POS Tag or Not to POS Tag: The Impact of POS Tags on Morphological Learning in Low-Resource Settings Sarah Moeller Ling Liu Mans Hulden University of Colorado Boulder first.last@colorado.edu Abstract Part-of-Speech (POS) tags routinely appear as features in morphological tasks. POS taggers are often one of the first NLP tools devel- oped for low-resource languages. However, as NLP expands to new languages it cannot as- sume that POS tags will be available to train a POS tagger. This paper empirically examines the impact of POS tags on two morphological tasks with the Transformer architecture. Each task is run twice, once with and once with- out POS tags, on otherwise identical data from ten well-described languages and five under- documented languages. We find that the pres- ence or absence of POS tags does not have a significant bearing on the performance of either task. In joint segmentation and gloss- ing, the largest average difference is an .09 im- provement in F1-scores by removing POS tags. In reinflection, the greatest average difference is 1.2% in accuracy for published data and 5% for unpublished data. These results are indi- cators that NLP and documentary linguistics may benefit each other even when a POS tag set does not yet exist for a language. 1 Introduction Parts of speech (POS), also known as word classes or lexical categories, communicate information about a word, its morphological structure and in- flectional paradigm, and its potential grammatical role in a clause. POS tagging is a well-studied problem in NLP. It is one of the first tasks under- taken for a new data set and a POS tagger is of- ten one of the first NLP resources built for low- resource languages (Yarowsky and Ngai, 2001; Cox, 2010; De Pauw, 2012; Baldridge and Gar- rette, 2013; Duong, 2017; Anastasopoulos, 2019; Millour and Fort, 2019; Eskander et al., 2020b). Although this priority on early POS tagging may be simply due to the relative ease of building a POS tagger, it seems to reflect an assumption that POS Figure 1: Average F1-scores on joint segmentation and glossing on interlinear glossed texts from fieldwork in five languages found that POS-tags have little and irreg- ular impact. tags simplify or improve other NLP tasks (Krauwer, 2003). As far as we are aware, this assumption has not been methodically tested. This paper examines the impact of POS tags on morphological learning, an important area for low-resource languages, many of which are more morphologically complex than English, Mandarin, or other large-resource languages. Morphologi- cal learning can help reduce the out-of-vocabulary problem in morphologically complex languages, especially in low-resource settings. Morphological learning also holds high priority in documentary and descriptive linguistics as a necessary founda- tion for further descriptive work. We focus on two related tasks that involve morphological learning: joint morpheme …

Content truncated to 500 words