Analyzing Syntactic Complexity in ESL and EFL Learners’ Language Production Using an AI Classification Model

• Published in: 언어학 no. 100; pp. 283 - 306
• Main Author: Wonbin Kim
• Format: Journal Article
• Language: English
• Published: 사단법인 한국언어학회 01.12.2024
• Subjects: 언어학; Korean learners of English; XGBoost classifier; ESL learners; EFL learners; syntactic complexity; communication modes
• ISSN: 1225-7494; 2508-4429
• DOI: 10.17290/jlsk.2024..100.283

This study aims to (i) identify the key syntactic complexity-related characteristics that distinguish between learners studying English as a second language (ESL) and learners studying English as a foreign language (EFL) across three communication modes, (ii) investigate whether Korean learners are closer to ESL or EFL in terms of the identified syntactic characteristics in each mode, and (iii) test whether a machine learning-based classification model can effectively perform in addressing these two objectives. For the first objective, this study utilized the feature importance metric within the XGBoost classifier to assess the importance of fourteen syntactic complexity measures in differentiating between ESL and EFL learners in essays, dialogues, and monologues. For the second objective, this study trained the XGBoost classifier to sort new input data into ESL and EFL based on the key measures obtained from the first objective. For the third objective, evaluation metrics to assess the XGBoost classifier’s performance were employed. The results demonstrated that the XGBoost classifier can successfully identify the main syntactic characteristics that differentiate between ESL and EFL learners, Korean learners are closer to EFL learners in every mode, and the XGBoost classifier has the potential to serve as a new approach to reveal these two findings. KCI Citation Count: 0