CULTURAL - SPECIFIC ITERMS AND MACHINE TRANSLATION: IS ARTIFICIAL INTELLIGENCE ABLE TO ACCESS CULTURAL CATEGORIES?

Tóm tắt

This study investigates how machine translation (MT) systems render culture-specific items (CSIs)
in Vietnamese–English translation, assessing strategies and cultural accuracy. Drawing on Aixela (1996),
Newmark (1988), and Venuti (1995), 185 CSIs from 50 culturally rich Vietnamese texts were translated
by Google Translate, DeepL, and Microsoft Translator. Qualitative comparative analysis examined the
use and adequacy of strategies such as literal translation, generalization, cultural substitution, and
functional equivalence. All systems predominantly employed literal translation and generalization, with
limited culturally adaptive approaches. DeepL showed slightly better contextual interpretation, yet none
displayed strong cultural competence. Findings highlight persistent MT limitations in conveying cultural
nuance and pragmatic meaning, underscoring the need for human post-editing and culturally informed
AI model design.