A COMPARATIVE STUDY OF TRANSLATION QUALITY BETWEEN HUMAN TRANSLATION AND CHATGPT OF SAFETY STANDARD TRAINING QUALITY CONTROL (SST QC) MATERIALS AT TOYOTA MOTOR MANUFACTURING INDONESIA (TMMIN)

This study aims to analyze the comparative translation quality between generative artificial intelligence, ChatGPT, and human translators on the Safety Standard Training Quality Control (SST QC) materials at PT Toyota Motor Manufacturing Indonesia (TMMIN). The focus of this research is to identify the types of imperative sentences used in the SST QC materials and to compare the translation quality of both methods based on accuracy, acceptability, and readability. This study employs a qualitative descriptive approach with a comparative research design. Data were collected through document analysis and expert assessment. Imperative sentence types were classified using Quirk et al.’s (1985) framework, while translation quality was evaluated using Nababan’s (2012) assessment model. The results indicate that among the 65 imperative sentence data analyzed, instruction is the dominant type with 71% (46 data), followed by prohibition at 23% (15 data), while warning and advice each account for 3% (2 data). The translation quality assessment reveals that human translation achieved a slightly higher overall average score of 2.61 compared to ChatGPT’s 2.56. Human translators performed better in the accuracy aspect, scoring 2.7 compared to ChatGPT's 2.4, as they were more capable of preserving technical terminology and workplace safety context. In contrast, ChatGPT demonstrated superior performance in acceptability (2.7) and readability (2.8), producing translations that were more fluent and easier to understand. This study is expected to provide practical recommendations for the manufacturing industry regarding the implementation of Post-Editing Machine Translation (PEMT) to optimize the use of artificial intelligence in translating technical documents while maintaining translation quality and safety standards.