A Comparative Analysis of AI and Machine Translation versus Human Translation of English Literary Works

الملخص

The study at hand investigates the quality of translation produced by two machine translation systems, namely; Google Translate and ChatGPT-4, in comparison with human translation (HT) of a literary work from English into Arabic. The corpus of the study consists of the Arabic translation of The Forty Rules of Love by Elif Shafak. Five excerpts were purposefully selected from the novel to serve as the main dataset for analysis. These excerpts were chosen because they contain figurative expressions, culturally embedded idioms, metaphorical language, and emotional nuances that typically pose challenges for machine translation. To evaluate the translations, the study adopts the Multidimensional Quality Metrics (MQM) framework, which provides a systematic assessment of translation quality through dimensions such as accuracy, fluency, style, and cultural adequacy. The findings reveal clear distinctions between the three types of translations. Human translation demonstrates the highest level of accuracy and fluency, maintaining not only the semantic content of the source text (ST) but also its literary tone, stylistic richness, and cultural depth. ChatGPT-4, while producing coherent and fluent Arabic output, struggles with cultural sensitivity and with rendering figurative language in a contextually appropriate manner. It occasionally simplifies metaphors or reinterprets them in ways that weaken the stylistic effect intended by the author. Google Translate, on the other hand, performs the weakest among the three. It frequently misinterprets idiomatic expressions, distorts metaphorical meanings, and fails to convey the emotional and contextual layers of the ST. Overall, the study highlights the limitations of current machine translation tools when dealing with literary texts and emphasizes the irreplaceable role of human translators in preserving artistic and cultural integrity.