COMPARATIVE ANALYSIS OF GOOGLE TRANSLATE AND DEEPL IN ENGLISH–UZBEK LITERARY TRANSLATION

: This article presents a comparative analysis of two leading neural machine translation systems — Google Translate and DeepL — applied to English–Uzbek literary translation. A corpus of literary texts (short stories and novel excerpts, ~5,000 words) was translated with both systems, and the outputs were analysed using the five-group error typology. Results show that the systems exhibit different error profiles: Google Translate tends toward generalisation and omission, while DeepL tends toward preservation of source structures and transliteration. For literary translation, neither system reaches publishable quality; both require full post-editing, with DeepL output generally requiring somewhat less effort for morphological correction. The article analyses strengths and weaknesses of each system by error group, proposes a usage strategy (system selection and combination based on text characteristics), and discusses implications for post-editing workflow design. The conclusion argues that system comparison should be based on post-editing effort rather than raw output quality, and that for Uzbek, the choice between systems matters less than the post-editor's competence.