Effectiveness of ChatGPT in Korean Grammatical Error Correction

Junghwan Maeng, Jinghang Gu, Sun-A Kim

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

As the first study in applying ChatGPT to L2 Korean grammar correction and learning, this study investigates the effectiveness of ChatGPT models as tools for Korean Grammatical Error Correction (GEC). The model comparison revealed that ChatGPT 4 outperforms ChatGPT 3.5
and the baseline model in all measures related to the precision of error detection and correction. Furthermore, a human evaluation shows that ChatGPT 4 outperforms its previous version in handling multiple grammatical errors and correcting semantic-level mistakes. The
findings of the present study suggest that ChatGPT 4 as a self-learning tool would be more suitable for advanced to near-native level L2 learners of Korean to improve their semantic fluency of sentences with minimal errors than for less proficient L2 learners because the latest version of
ChatGPT 4 still demonstrates a relatively lower accuracy rate in Korean GEC tasks.
Original languageEnglish
Title of host publicationProceedings of the 37th Pacific Asia Conference on Language, Information and Computation
EditorsChu-Ren Huang, Yasunari Harada, Jong-Bok Kim, Si Chen, Yu-Yin Hsu, Emmanuele Chersoni, Pranav A, Winnie Huiheng Zeng, Bo Peng, Yuxi Li, Junlin Li
PublisherAssociation for Computational Linguistics (ACL)
Pages464-472
Number of pages9
Publication statusPublished - Dec 2023
EventPacific Asia Conference on Language, Information and Computation (PACLIC 37) -
Duration: 2 Dec 20235 Dec 2023
https://paclic2023.github.io/

Conference

ConferencePacific Asia Conference on Language, Information and Computation (PACLIC 37)
Period2/12/235/12/23
Internet address

Fingerprint

Dive into the research topics of 'Effectiveness of ChatGPT in Korean Grammatical Error Correction'. Together they form a unique fingerprint.

Cite this