Medical Chatbot System Using Natural Language Processing

Abstract

Access to timely and reliable healthcare is one of the primary issues in Nepal because of the challenging topography, lack of medical facilities, and the absence of healthcare professionals, especially in rural and remote areas. In order to overcome these issues, this research project is going to present a Medical Chatbot System, which is based on NLP and which is aimed at the provision of the initial disease diagnosis and situation-specific medical advice based on the symptoms described by the user. The system combines a fine-tuned BERT model to predict multiple labels of a disease with a Large Language Model (LLM) to be able to engage in natural dialogue and ask follow-ups. The system was trained and tested on a secondary medical dataset, which included descriptions of symptoms and disease names. The results of the experiment show that the proposed chatbot achieved an accuracy of 91.53%, as well as high precision and recall, which is evidence of reliable performance in terms of disease prediction. Safety mechanisms such as red-flag detection and fallback responses were incorporated to encourage responsible use and timely medical consultation. The system is launched as a web-based interface so that it is easy to use by people with different degrees of digital literacy, and it is supposed to be used during early health evaluation and not to substitute professional medical diagnosis. In general, the results imply the potential of smart conversational agents to enhance the accessibility of healthcare, lessen unneeded hospital admissions, and facilitate early medical care in resource-limited countries, like Nepal.