Enhancing Medical Knowledge Access: Development and Assessment of an Open-Domain Medical Question Answering System with Deep Learning

Arabic open question answering system (OQAS) in health informatics (HI) can promote health education and support individuals in making informed decisions. It may contribute to health education and awareness initiatives by offering precise and reliable information. It enables individuals to comprehend health-related topics, preventative actions, and treatment alternatives in their preferred language. In addition, it facilitates the accessibility of pertinent health information to a broader audience, consequently addressing language barriers. There is a limited number of Arabic OQAS in HI. Thus, this study built an OQAS using graph convolutional network (GCN) and Transformers. The author conducted data preparation and extracted named entity-relationship (NER) from raw data. A graph generator was developed to represent the NER relationship for improving the efficiency of GCN in generating a required outcome. Similarly, Generative Pre-trained Transformer (GPT 2.0) was fine-tuned using the Hyperband optimization. Finally, the author integrated GCN and GPT 2.0 to improve the performance of the proposed model. The performance evaluation was conducted using the MAQA dataset. The experimental outcome highlighted the importance of the proposed OQAS in healthcare informatics. The extended fine-tuning of the proposed model could improve its interpretability.