Processing of Machine Translation Errors by Korean Learners of English: An Eye-tracking Study

The present study investigates second language (L2) processing of machine translation (MT) errors in Korean-to-English translations using an eye-tracker. Twenty-five Korean learners of English participated in an Acceptability Judgment Task in which they had to rate the accuracy of Korean-to-English translated sentences while their reading times, fixations, and revisits were recorded. The effect of proficiency level (high vs. low), six error types, and language (English vs. Korean) on the participants’ responses and reading measures were examined and analyzed. The findings indicate that high-proficiency learners demonstrated a clearer differentiation between accurate and inaccurate translations and were better at recognizing errors related to proverbs and subjects, while low-proficiency learners exhibited a higher number of revisits and fixations especially when processing subject-related errors. These findings highlight the importance of proficiency level in shaping how learners process MT errors and have important pedagogical implications for MT-assisted language instruction.