LEXICAL ERROR PRODUCED BY INSTAGRAM MACHINE TRANSLATION IN TRANSLATING CAPTION OF “NYTIMES” INSTAGRAM ACCOUNT

This research explores Instagram’s machine translation (MT) performance by analyzing caption translations from The New York Times (NYTIMES) Instagram account, focusing on posts from December 2024 related to global political events. Using a qualitative descriptive approach, the study identifies and classifies lexical errors based on Vilar et al.’s (2006) framework and examines the translation techniques applied, guided by Darbelnet’s (1989) model. A total of 90 lexical errors were recorded, falling into three major categories: incorrect word (50%), missing word (28%), and untranslated word (22%). In terms of translation strategies, literal translation was the most commonly used (50%), followed by modulation (27%), transposition (23%), and borrowing (16%). The analysis reveals that while literal translation is prevalent, it often overlooks contextual and cultural nuances, especially in politically sensitive content. These findings underscore the limitations of current MT systems in handling complex lexical and semantic structures. By shedding light on common lexical pitfalls and the role of translation strategies, the study contributes to broader discussions on enhancing MT accuracy and reliability in multilingual digital communication