INTEGRATED STUDY OF AUTOMATED TRANSLATION QUALITY AND PHRASEOLOGICAL EQUIVALENCE IN ENGLISH-UZBEK TRANSLATION

Abstract

Machine translation (MT) of highly idiomatic text remains a formidable challenge, especially for low-resource, agglutinative languages like Uzbek. This paper presents a comprehensive study integrating (1) MT quality evaluation methods and (2) translation of English phraseological units (idioms, proverbs, phrasal verbs) into Uzbek. We review structural, cultural, and semantic issues in translating English idiomatic expressions into Uzbek, and evaluate how current MT systems (e.g. Google Translate, ChatGPT) handle them. Methodologically, we detail automatic metrics (BLEU, METEOR), human evaluation (fluency/adequacy ratings, post-editing effort), and comparative translation strategies (literal, idiomatic, calque, paraphrase, etc.). Our analysis includes tables contrasting machine outputs with human translations for representative idioms. We discuss the impact of Uzbek’s agglutinative morphology and cultural specificity on translation accuracy. Finally, we offer recommendations for improving Uzbek MT - including richer phraseological corpora, better morphological processing, and integration of cultural knowledge - and suggest how phraseological insights can be embedded in MT models.