CONTEXT-AWARE RAG OPTIMIZATION USING DYNAMIC QUERY AND DOCUMENT RELEVANCE FEEDBACK

Abstract

Retrieval-Augmented Generation (RAG) has become a fundamental technology for enterprise knowledge management by integrating Large Language Models (LLMs) with external knowledge repositories to generate accurate and evidence-based responses. However, conventional RAG systems typically employ static query generation and fixed retrieval strategies that cannot effectively adapt to changing user intent, document relevance, and contextual information. Consequently, retrieval quality may degrade, resulting in reduced response accuracy and increased retrieval latency. Recent advances in context-aware computing, dynamic query optimization, relevance feedback, Reinforcement Learning, semantic search, and cloud-native architectures enable adaptive retrieval mechanisms capable of continuously improving enterprise knowledge access. This paper proposes a context-aware RetrievalAugmented Generation optimization framework using dynamic query reformulation and document relevance feedback by integrating enterprise knowledge repositories, vector databases, machine learning, semantic retrieval, and intelligent feedback mechanisms. The proposed framework continuously refines retrieval strategies based on contextual information, user interactions, and document relevance to improve response quality and enterprise knowledge utilization. Experimental analysis demonstrates significant improvements in retrieval precision, response relevance, operational efficiency, contextual understanding, and enterprise scalability compared with conventional static RetrievalAugmented Generation systems. Keywords: Retrieval-Augmented Generation, Context-Aware Computing, Dynamic Query Optimization, Relevance Feedback, Large Language Models, Semantic Search, Vector Database, Machine Learning, Enterprise AI, Cloud-Native Computing.