Optimizing Large Language Models for Arabic Healthcare Communication: A Focus on Patient-Centered NLP Applications

Abstract

Recent studies have highlighted the growing integration of Natural Language Processing (NLP) techniques and Large Language Models (LLMs) in healthcare. These technologies have shown promising outcomes across various healthcare tasks, especially in widely studied languages like English and Chinese. While NLP methods have been extensively researched, LLM applications in healthcare represent a developing area with significant potential. However, the successful implementation of LLMs in healthcare requires careful review and guidance from human experts to ensure accuracy and reliability. Despite their emerging value, research on NLP and LLM applications for Arabic remains limited particularly when compared to other languages. This gap is largely due to challenges like the lack of suitable training datasets, the diversity of Arabic dialects, and the language’s structural complexity. In this study, a panel of medical experts evaluated responses generated by LLMs, including ChatGPT, for Arabic healthcare inquiries, rating their accuracy between 85% and 90%. After fine tuning ChatGPT with data from the Altibbi platform, accuracy improved to a range of 87% to 92%. This study demonstrates the potential of LLMs in addressing Arabic healthcare queries especially in interpreting questions across dialects. It highlights the value of LLMs in enhancing healthcare communication within the Arabic-speaking world and points to a promising area for further research. This work establishes a foundation for optimizing NLP and LLM technologies to achieve greater linguistic and cultural adaptability in global healthcare settings.