Unified Framework for Context-Aware OCR and Text Classification using LLMs

The inclusion of deep learning methods has greatly changed Optical Character Recognition (OCR) systems. On the other hand, traditional OCR techniques have difficulty with complicated text structures, different orientations, and contextual discrepancies. This work offers a new context-aware OCR system that uses Large Language Models (LLMs) to improve text recognition and classification accuracy. The suggested system adjusts to several real-world situations by including semantic enrichment, data augmentation, and bounding box labeling. A strong text classification tool driven by LLMs guarantees correct reading of extracted text. Moreover, deployment based on Docker guarantees simplicity of implementation and scalability. Key performance indicators-including OCR accuracy, contextual classification accuracy, and user input-are used to assess the framework, which shows notable advances over conventional methods.