Negative language transfer in learner English: A new dataset

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3129–3142 June 6–11, 2021. ©2021 Association for Computational Linguistics 3129 Negative language transfer in learner English: A new dataset Leticia Farias Wanderley Nicole Zhao EdTeKLA Research Group Department of Computing Science University of Alberta {fariaswa,zzhao1,cdemmansepp}@ualberta.ca Carrie Demmans Epp Abstract Automatic personalized corrective feedback can help language learners from different back- grounds better acquire a new language. This paper introduces a learner English dataset in which learner errors are accompanied by in- formation about possible error sources. This dataset contains manually annotated error causes for learner writing errors. These causes tie learner mistakes to structures from their first languages, when the rules in English and in the first language diverge. This new dataset will enable second language acquisition re- searchers to computationally analyze a large quantity of learner errors that are related to language transfer from the learners’ first lan- guage. The dataset can also be applied in per- sonalizing grammatical error correction sys- tems according to the learners’ first language and in providing feedback that is informed by the cause of an error. 1 Introduction English has become an international language. It is the lingua-franca that unites native speakers of other languages around the world (Lysandrou and Lysandrou, 2003). For that reason, it is not hard to believe that the teaching of English as a Sec- ond Language1 has caught a lot of attention from the research community (Caine, 2008). Over the years, computational linguistics researchers have collected corpora containing text written by lan- guage learners. These corpora have made possible several advances in language teaching, such as au- tomatic writing assessment (Rahimi et al., 2017; Yannakoudakis et al., 2011) and automatic error detection and correction (Chollampatt et al., 2016; Nadejde and Tetreault, 2019; Omelianchuk et al., 2020). 1Throughout this manuscript, we use the term second lan- guage to refer to any additional language beyond the mother tongue, whether the speaker is in a second or foreign language learning context. Although learner corpora are used to model grammatical error correction systems, they are not as often employed in the enhancement of learner feedback. Language learners benefit from direct corrective feedback (Sheen, 2007). Moreover, feed- back that makes them reflect upon their errors and distinguish a cause for their mistakes correlates to increased performance (Demmans Epp and Mc- Calla, 2011; Sheen, 2007; Shintani and Ellis, 2013; Karim and Nassaji, 2020). In this paper, we intro- duce a learner English dataset enhanced with error cause information and concrete examples of learner errors that relate to the learners’ first language. It has the potential to help create computational mod- els that provide personalized feedback to English language learners based on the learners’ native lan- guages. This new dataset can be accessed by fol- lowing the instructions described in our research group’s repository2. The dataset presented in this paper contains sup- plementary explanations for errors made by Chi- nese native speakers when writing in …

Content truncated to 500 words