Analytical comparison of sbert, claude and gemini large language models on Turkish health chat bots via performance metrics

In this study, the applicability of Large Language Models (LLM) in Turkish healthcare chatbots was investigated. The performance of SBERT, Claude and Gemini models for chatbot applications in the healthcare field was evaluated comparatively. Considering the unique morphological difficulties of the Turkish language, the ability of these models to produce context-adaptive responses was tested. In the study, performance metrics such as accuracy, precision, sensitivity and F1 score of the models were compared and inferences were made based on these metrics. The findings showed that large language models offer a strong potential in terms of user satisfaction and information accuracy in Turkish healthcare chatbots. However, due to the structural features of the language, optimization and customization processes were seen to be of great importance. In this direction, methods that will increase performance were proposed for future studies. The thesis aims to contribute to the development of Turkish healthcare chatbots and to determine the limitations of large language models in this area.