Generative AI Potentials and Limits in Translating Iraqi Dialect in Literature: An Evaluation of Accuracy in Azher Jirjees’s At Rest in the Cherry Orchard

Abstract

Recent academic scrutiny has considered the usage of artificial intelligence (AI) tools in transforming literary texts and studies have largely shown that AI translation tools can generate accurate outputs for common vocabulary. However, their performance may vary when dealing with difficult expressions related to dialects, cultural distinctions, and context-dependent meanings. In literary translation from Arabic, maintaining high level of accuracy in translating cultural expressions is challenging for AI systems. The main aim of this paper is to evaluate the accuracy level of four AI translations (ChatGPT, Microsoft Copilot, DeepSeek, and Gemini) and one human translation of Iraqi dialect in a novel by the Iraqi writer Azher Jirjees’ At Rest in the Cherry Orchard which has been translated by Jonathan Wright into English. In the current study, the researcher will not only assess the level of accuracy of translation but also evaluate the similarities and differences between the human translation and AI translations of the Iraqi dialect to see which is more reliable in understanding it. The researcher adopted Nababan’s translation accuracy theory to analyse the data in terms of accurate, less accurate, and inaccurate renditions. The analysis revealed that the level of accuracy was equal in the human translation, DeepSeek, and Gemini at 67% and 16% inaccuracy, while Copilot and ChatGPT recorded the lowest level of accuracy respectively. Overall, all the AI tools recorded 60% accuracy, 25% less accuracy, and 15% inaccuracy.