<![CDATA[The rapid development of Neural Machine Translation (NMT) has significantly improved automatic translation quality for globally dominant languages. However, low-resource languages continue to face substantial challenges in achieving accurate and culturally sensitive translation. This article investigates the limitations of NMT systems in translating culturally marked lexical units and expressions from the Karakalpak language, a low-resource Turkic language spoken mainly in Uzbekistan. Special attention is given to the phenomenon of cultural untranslatability, which occurs when linguistic units contain cultural meanings, traditions, or social concepts that lack direct equivalents in target languages. The research examines how NMT systems process idioms, folklore expressions, kinship terminology, and culturally specific vocabulary in Karakalpak-English translation. The study combines linguistic analysis with examples generated through machine translation platforms to evaluate semantic loss, contextual distortion, and cultural simplification. The findings reveal that NMT systems often prioritize lexical equivalence over cultural interpretation, resulting in mistranslations or incomplete representation of cultural meaning. The article also discusses the influence of limited corpora, insufficient parallel datasets, and structural differences between Turkic and Indo-European languages on translation quality.]]>
Copyrights © 2026
