An Analysis of Lexical Accuracy in Meta AI Automatic Translation of Buttonscarves Beauty’s Instagram Captions

Abstract

This research examines the lexical accuracy of Meta AI’s machine translation in translating Instagram’s captions in English into Indonesian on the @buttonscarvesbeauty account. Using a descriptive qualitative analysis based on Molina and Albir’s (2002) taxonomy, the research evaluates lexical fidelity, translation strategies, and the accuracy of meaning transfer through the aspects of accuracy, appropriateness, and acceptability. The results show varied translation performance, with significant challenges in handling beauty related technical terms and marketing language. Simple terms such as “hydration” (hidrasi) demonstrate high accuracy, while more complex terms like “fine,” “airbrushed,” and idiomatic expressions such as “Make every wink count” are translated literally, losing their persuasive function. The system tends to rely on literal translation with limited use of modulation, transposition, and adaptation strategies. The findings suggest that machine translation is still inadequate for marketing content, which requires cultural adaptation and persuasive language. Beauty brands still need a human touch to keep their brand identity and message consistent. This study looks at how effective machine translation is in social media marketing, and gives practical suggestions to improve the quality of automatic translations in commercial settings.