Analisis Kesalahan Terjemahan Otomatis Arab–Indonesia pada Konten TikTok

Abstract

The rapid development of Artificial Intelligence (AI) has encouraged the implementation of machine translation features on various social media platforms, including TikTok. Although this feature facilitates users in understanding foreign-language content, its translation quality, particularly for Arabic–Indonesian language pairs, still requires further evaluation due to significant linguistic and cultural differences. This study aims to analyze translation errors and evaluate the quality of Arabic–Indonesian machine translation on TikTok based on three aspects: accuracy, acceptability, and readability. This research employed a qualitative approach using content analysis. The data consisted of four Arabic-language TikTok contents translated automatically by TikTok's translation feature. Data were analyzed using the translation quality assessment instrument developed by Nababan et al. (2012) and described qualitatively. The findings reveal that the overall quality of TikTok's machine translation remains low. The average accuracy score was 33,33%, indicating that most source-language meanings were not transferred correctly. The acceptability aspect obtained an average score of 46,67, showing that many translations did not conform to Indonesian linguistic norms and sounded unnatural. Meanwhile, the readability aspect achieved an average score of 53,33%, indicating that the translated texts were still difficult for readers to understand. The most dominant errors included lexical errors, contextual errors, failures in translating idiomatic expressions, and deviations from Indonesian grammatical conventions. These findings indicate that TikTok's machine translation feature has not yet been able to accurately convey Arabic messages into Indonesian, particularly for religious and context-dependent expressions, and therefore still requires human post-editing to ensure translation quality.