Variation in the Adaptation of English <x> into Korean: A ChatGPT-based Study

Abstract

This paper examines how the English letter <x> in words, such as xylophone and xylitol, is adapted into Korean: as [s] in the former and as [c] in the latter. It is particularly notable that some lexical items show variation in the adaptation of <x> in Korean, such as xanthan: [s], [c], and [kʰɨ.s]. By using ChatGPT (version: ChatGPT-4; OpenAI 2023), one of the most widely used large language models, the present study investigates whether and how this generative artificial intelligence chatbot predicts variation in Korean pronunciations of English words with <x> that were selected from the Corpus of Contemporary American English (Davies 2008-). Results show that ChatGPT did not correctly predict reported variants in most cases: it either overpredicted or underpredicted the reported variants. Based on the results, this paper suggests two important factors that could help improve ChatGPT’s predictions: different origins of words (e.g., Greek) and different groups of words (e.g., chemical terminology).