SEMANTIC TRANSFORMATION MODELS OF CULTURE-SPECIFIC REALIA IN GENERATIVE AI TRANSLATION BETWEEN ENGLISH AND UZBEK

Abstract

This article examines semantic transformation models of culture-specific realia in generative AI translation between English and Uzbek. Particular attention is paid to the preservation, modification, generalization, explicitation, substitution, and reduction of culturally significant semantic components. The research applies comparative, contextual, componential, semantic, and pragmatic methods of analysis. The findings demonstrate that generative AI translation does not simply reproduce culture-specific meaning in another linguistic code; rather, it reconstructs such meaning through a variety of semantic operations. Six principal transformation models are identified: direct cultural transfer, semantic approximation, semantic generalization, explicitation, cultural substitution, and semantic-pragmatic reduction. The analysis indicates that linguistically fluent AI-generated translations may still demonstrate partial cultural loss, especially when complex Uzbek concepts such as mahalla, hashar, dasturxon, to‘r, choyxona, and kelin salom are reduced to more general English equivalents. The article argues that the evaluation of generative AI translation should therefore incorporate cultural-semantic adequacy alongside grammatical accuracy, lexical correspondence, and textual fluency.