Telegram Bot for Tracking User's Medical Tests by Photo/File

The main objectives of this research are threefold: firstly, to perform a comprehensive comparative analysis of the performance of three different approaches—optical character recognition (PaddleOCR), large language models (LLM), and a hybrid algorithm combining EasyOCR and Table Transformer—in handling and analyzing medical diagnostic documents. Secondly, it aims at developing a methodology for efficient structured output and table data recognition. Thirdly, the research focuses on implementing a Retrieval-Augmented Generation (RAG) framework to enhance the interpretation and contextual understanding of the extracted data, ensuring accurate medical insights and reliable diagnostic information extraction. This integrated approach combines the strengths of OCR technologies, advanced language models, and context-aware data processing to create a robust solution for medical document analysis, providing not only accurate data extraction but also meaningful interpretation of the results within the medical domain.