Assessing the Accuracy and Functional Adequacy of AI-Generated Urdu Translations of HR Policy Documents

Abstract

This study examines the accuracy and quality of AI-generated Urdu translations of two HR policy documents, the General Code of Conduct and the Harassment Policy, using ChatGPT and Gemini. The research aimed to evaluate how well each tool preserves meaning, handles terminology, maintains structure, and reflects the professional tone required in HR communication. Using Skopos Theory as the guiding framework, the study analyzed lexical, semantic, structural, and pragmatic aspects of both translations. Findings show that ChatGPT produced clearer, more fluent, and more contextually appropriate Urdu translations, making them more suitable for practical HR use. Gemini’s translations, although generally accurate, were more literal and contained formatting artifacts that reduced readability. The study concludes that while AI tools can support HR translation tasks, human post-editing remains essential to ensure clarity and accuracy, especially for sensitive policies. The research highlights the value of AI-assisted translation while emphasizing the need for careful human oversight.