Abstract:
"Diabetes is a widespread chronic disease posing serious health risks. Patients struggle with self-management due to various reasons such as, inadequate personalized advice, emotional support, and educational resources. This lack of support can lead to complications, decreased quality of life, and increased healthcare costs. This research aims to improve diabetes care with a unique approach.
This research addresses these challenges by proposing and developing an innovative conversational agent (CA) powered framework with the use of generative Artificial Intelligence (AI). The design prioritizes a patient-centric approach, integrating Large Language Models (LLMs) with a specialized diabetes knowledge base to provide personalized, actionable guidance. This approach leverages the strengths of generative AI for natural conversation and combines them with a continuously updatable knowledge base for reliable, domain-specific support.
The proposed framework's applicability was tested using both qualitative and quantitative methods. The evaluation involved using the RAGAS framework's context precision and recall metrics to test the implemented retriever, achieving a context precision of 0.9851 and a context recall of 0.83. Additionally, human evaluations confirmed that the framework's components work together effectively, providing better patient support and enhancing overall diabetes care. However, in order to test this proposed framework effectively this framework should be applied in real-world."
