A Cogno-Prosodic Approach to Translating Arabic Poetry into English: Human vs. Machine

Abstract

This study offers a schema-theoretic framework for comprehending the translation process of classical Arabic poetry into English, highlighting the significance of schematic ideation in the target text and schematic comprehension in the source text. Four categories of schemata are distinguished in this model: culture-free, culture-bound, culture-sensitive, and language-bound. These four schematic categories are used to evaluate human, ChatGPT, and Gemini translations of a poetic corpus consisting of six individual verses and six stanzas. The findings show that the three translators perform very well in rendering thematic elements that feature culture-free and culture-bound schemata. However, ChatGPT and Gemini lag seriously behind human translation when handling culture-bound and culture-sensitive schemata, especially when they feature allusions to proper nouns, which causes them to offer incomprehensible translations. In terms of prosody, humans and ChatGPT lead the way, with Gemini falling behind by a small margin, especially when aligning thematic elements with prosodic features in culture-free and culture-bound schemata. In fact, the two AI systems present themselves as strong competitors to human translation in this regard. The conclusion emphasises the necessity of human expertise in capturing cultural nuances in poetic texts, as well as stressing the clear and valuable advantages of using AI-based systems in translating poetry while referring to some weaknesses.

Keywords: Poetry Translation; Schema Theory; Cogno-Prosodic Approach; Human Translation; AI-Powered Translation