Enriching Grammatical Error Correction Resources for Modern Greek

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 4984–4991 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 4984 Enriching Grammatical Error Correction Resources for Modern Greek Katerina Korre†, John Pavlopoulos‡ †Universit`a di Bologna, Forl`ı, Italy ‡Athens University of Economics and Business, Athens, Greece †aikaterini.korre2@unibo.it, ‡annis@aueb.gr Abstract Grammatical Error Correction (GEC), a task of Natural Language Processing (NLP), is challenging for under-represented languages. This issue is most prominent in languages other than English. This paper addresses the issue of data and system sparsity for GEC purposes in the modern Greek Language. Following the most popular current approaches in GEC, we develop and test an MT5 multilingual text-to-text transformer for Greek. To our knowledge this the first attempt to create a fully-fledged GEC model for Greek. Our evaluation shows that our system reaches up to 52.63% F0.5 score on part of the Greek Native Corpus (GNC), which is 16% below the winning system of the BEA-19 shared task on English GEC. In addition, we provide an extended version of the Greek Learner Corpus (GLC), on which our model reaches up to 22.76% F0.5. Previous versions did not include corrections with the annotations which hindered the potential development of efficient GEC systems. For that reason we provide a new set of corrections. This new dataset facilitates an exploration of the generalisation abilities and robustness of our system, given that the assessment is conducted on learner data while the training on native data. Keywords: GEC, MT5, Encoder-Decoder 1. Introduction This article addresses the issue of low-resourced lan- guages in Grammatical Error Correction (GEC) by pro- viding the prerequisites for the expansion of GEC in the Modern Greek language. In the past decade, espe- cially since the Helping Our Own (HOO) shared task (Dale and Kilgarriff, 2011), the performance of GEC systems has increased greatly, while a variety of new datasets has emerged. The advances in GEC, how- ever, are mostly focused on the English language with progress in other languages being quite limited. There have been some recent attempts to enrich the resources for GEC purposes in the following lan- guages: Spanish (Davidson et al., 2020); German (Boyd, 2018); Russian (Rozovskaya and Roth, 2019); Czech (N´aplava and Straka, 2019a); Chinese (Rao et al., 2018); Arabic (Solyman et al., 2019). Our work fo- cuses on the Modern Greek language, a highly inflec- tional language that consists of many declinable parts of speech that produce a vast set of morphological word forms. As Gakis et al. (2016) explain, from a single verb or adjective lemma more than 100 word forms can be produced (including both active and passive word forms, and the comparative and superlative forms). In the same study, it becomes clear that the development of Natural Language Processing systems for Modern Greek is a challenging task as Greek is also a “free- word-order” language, which gives its speakers a free- dom of use that can often lead to multiple errors. We address the issue …

Content truncated to 500 words