A Comparative Analysis of English to Thai Translation by Human translators VS. ChatGPT

Abstract

Currently, Artificial Intelligence (AI) program is becoming increasingly popular in various fields including language studies and translation. There are two objectives of this study. It is aimed to analyze errors in English-to-Thai translations. There is also a comparison between the translation made by a group of students and those made by AI ChatGPT. Another objective is to divide the errors found in this study into groups.  The content chosen for translation in this study primarily focuses on the topic of "Machine Translation." This case study involved 20 students who enrolled in the course “14319 Translation Skill” in semester 2/2563 at an Open University in Thailand. The group of students was selected through simple random sampling method. This study used the qualitative method using an 11-item English-to-Thai translation test as the instrument for data collection and the analysis method followed the translation framework proposed by Laisattrueklai (2011).

The results of this study revealed that there were errors from translation produced by both the group of students and the AI.  30% of translation errors were made by the students, while the remaining 70% of errors were made by the AI (ChatGPT). Grammatical errors identified in the translations, were produced by both the students and the AI, They are included by issues with passive voice and the use of dummy subjects. Semantic errors, on the other hand, appeared to stem from participants' misconceptions and misunderstandings of homonyms. Examples of errors from both the students and AI translations are provided and categorized according to the framework of this study. They are composed of errors in the word, phrase and sentence levels.  It is recommended from this study that AI should not be relied upon alone and that human translation should be used along with it.