ChatGPT in Indonesia-Arabic Translation: A Quality Analysis

Abstract

ChatGPT applications are increasingly popular for answering questions and providing information to users on various topics and languages, including text translation. In several previous studies, the use of the ChatGPT was deemed coherent, relatively accurate, informative, and systematic. This research aims to open new avenues for analyzing the quality of ChatGPT translation in Indonesian texts into Arabic, not only to pinpoint translation errors, but also to explore the positive aspects and capabilities of ChatGPT in translation. This study attempts to assemble a layered cascading translation analysis, progressing from individual words to phrases, then to simple sentences, and finally to longer sentences covering popular themes, such as education, health, economics, and tourism. Consequently, this research can serve as a consideration for translators collaborating with AI, aligning with the focus of Industry 5.0 era. This study employs qualitative and quantitative descriptive research models. The qualitative descriptive research model is obtained through textual analysis techniques that focus on translating various models and types of texts, and the quantitative data are provided by the questionnaire results from general readers for the readability of the translation result. The researchers found that the use of ChatGPT in Indonesian Arabic text translation has both drawbacks and advantages. The shortcomings of ChatGPT include difficulties in translating complex words with specific themes, such as education, tourism, and literature. When translating long sentences, it tends to be less appropriate in composition, and in the semantic field or meaning within specific contexts, it may lack precision. The selection of pronouns may not be suitable for the subject, and the use of complex idioms and the arrangement of words in idhafah and wasfi forms may be inaccurate. Evidence shows that ChatGPT's translation has a 66% readability percentage, indicating a relatively good outcome. Nevertheless, human translators can still contribute to translations by collaborating with the advancements of ChatGPT and utilizing their sensitivity to the context and external elements of the text. Further research and analysis, such as the development of a translation framework applicable to learning and translation practice or an analysis of Indonesian-Arabic translation within translation classes, could be implemented.