Developing NLP Tools with a New Corpus of Learner Spanish

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 7238–7243 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 7238 Developing NLP Tools with a New Corpus of Learner Spanish Sam Davidson∗, Aaron Yamada†, Paloma Fern´andez-Mira∗, Agustina Carando∗, Claudia H. S´anchez Guti´errez∗, Kenji Sagae∗ ∗University of California Davis {ssdavidson, pfernandezmira, acarando, chsanchez, sagae}@ucdavis.edu †Creighton University aaronyamada@creighton.edu Abstract The development of effective NLP tools for the L2 classroom depends largely on the availability of large annotated corpora of language learner text. While annotated learner corpora of English are widely available, large learner corpora of Spanish are less common. Those Spanish corpora that are available do not contain the annotations needed to facilitate the development of tools beneficial to language learners, such as grammatical error correction. As a result, the field has seen little research in NLP tools designed to benefit Spanish language learners and teachers. We introduce COWS-L2H, a freely available corpus of Spanish learner data which includes error annotations and parallel corrected text to help researchers better understand L2 development, to examine teaching practices empirically, and to develop NLP tools to better serve the Spanish teaching community. We demonstrate the utility of this corpus by developing a neural-network based grammatical error correction system for Spanish learner writing. Keywords: L2 Spanish, error correction, learner corpus 1. Introduction Studies in second language acquisition benefit from large quantities of corpus data from L2 learners, which facilitate the analysis of developmental patterns with respect to their emerging grammar (Hawkins and Buttery, 2009). In addi- tion, the development of effective NLP systems for second language instruction, such as automated grammatical error correction and automated student assessment, depends on the availability of large annotated learner corpora (Kasewa et al., 2018). While annotated learner corpora of English are widely available, large learner corpora of Spanish are less common, and as a result, the field has seen little data- driven research on the developmental processes that under- lie Spanish language learning, or on the development of NLP tools to assist teachers and students of Spanish. This may come as unexpected, considering the fact that there ex- ists a relatively high demand for learning Spanish; in 2013, fifty-one percent of students enrolled in university language courses in the United States studied Spanish (AAAS, 2016) and there are over 21 million learners of L2 Spanish across the globe (Instituto Cervantes, 2019). Given the large demand for Spanish courses, a need exists for NLP tools to help instructors better assess incoming stu- dents’ abilities and to assist with grading and correction of student writing. However, without adequate datasets, the development of such tools is challenging. We describe a project that seeks to address the limited availability of an- notated Spanish learner data through the development of COWS-L2H (the Corpus Of Written Spanish – L2 and Her- itage speakers). This growing corpus aims to be the largest and most comprehensive corpus of learner Spanish freely available to the research community. In addition to raw …

Content truncated to 500 words