Reflections on Translating Poetry through ChatGPT: Insights from Arabic, English, and French

Abstract

Each language has its unique poetic traditions and stylistic nuances. ChatGPT excels in some aspects of these and falls short in others. This study explored these strengths and weaknesses across three languages: Arabic, English, and French. Through a comparative analysis of three short poems, the study evaluated ChatGPT-enabled translations alongside human translations. Findings show that ChatGPT-based translations exhibit specific limitations requiring human intervention with poetic expertise in both the source and target languages. Such limitations illustrate how AI tools fail to preserve poetic devices accurately, convey emotional connotations, and capture cultural context. Elements such as metaphors, similes, and alliteration rely heavily on subtle linguistic nuances, which are challenging for AI to translate accurately.  AI-generated translations hardly capture the artistic and historical contexts that imbue poems with deep emotional connotations. With these findings, the study adds to the ongoing debate about the reliability of AI-enabled translations of poetic texts. It suggests a balanced approach that leverages AI tools while acknowledging the irreplaceable value of human creativity and cultural insight in translating poetry.