Machine Translation of English Idioms into Georgian and Russian: A Comparative Analysis Based on George Orwell’s Animal Farm

Abstract

This study examines the semantic, stylistic, and pragmatic aspects of translating English idioms from George Orwell’s Animal Farm into Georgian and Russian, using Neural Machine Translation (NMT) and Large Language Models (LLMs). It evaluates five contemporary systems - Google Translate, Gemini AI, ChatGPT, PROTRANSLATE.GE, and Glosbe- to assess their strengths, weaknesses, and translation strategies.

Using a comparative, descriptive, and qualitative approach, the study analyses the generated translations for semantic accuracy, idiomatic equivalence, contextual adequacy, and naturalness.

The findings reveal that while AI systems demonstrate high lexical and syntactic accuracy, translating idioms still yields literal renderings, calques, semantic shifts, and stylistic flaws. Limited digital resources and parallel corpora for Georgian negatively affect translation quality. Conversely, established phraseological equivalents in Russian yield more natural, functionally adequate results.

The study’s novelty lies in comparing generative AI models alongside established machine translation tools for literary idioms across two distinct language families. Ultimately, these insights contribute to computational linguistics, idiom-translation methodology, and the enhancement of natural language processing resources for Georgian.