EVALUATING ARTIFICIAL INTELLIGENCE VERSUS HUMAN EXPERTISE IN LITERARY TRANSLATION

Abstract

This study explores the comparative capabilities of artificial intelligence systems (such as Large Language Models like ChatGPT and Gemini, along with tools like DeepL) and professional human linguists in the domain of literary translation. The text outlines the historical progression of machine translation from early rule-based systems (RBMT) and statistical models (SMT) to neural machine translation (NMT) and contemporary generative LLMs. It highlights the distinct nature of literary translation, which requires aesthetic equivalence, dynamic adaptation (domestication vs. foreignization), global narrative consistency, and ethical responsibility toward the source author's voice. Furthermore, the study addresses the methodological complexities of translation quality assessment, contrasting traditional frameworks like Multidimensional Quality Metrics (MQM) and n-gram metrics like BLEU with hybrid approaches and LLM-driven evaluations. Ultimately, while modern AI systems demonstrate strong localized fluency and syntactic handling, expert human translators remain vital for navigating subtle stylistic nuances, cultural realia, and overall aesthetic artistic integrity.