Enhancing Retrieval-Augmented Generation Systems by Text-Representing Centroid
Yanakorn Ruamsuk, Anirach Mingkhwan, Herwig Unger · 2024
This paper introduces a novel approach to enhance Retrieval-Augmented Generation (RAG) systems by integrating Text-Representing Centroid (TRC) methodology.Addressing the limitations of traditional vector databases, this method preserves structural relationships and adapts to content complexity, improving information retrieval efficiency and accuracy.Key contributions include advanced graph construction, relevance scoring algorithms, and extensive validation, with discussions on potential applications and future research.Empirical evidence demonstrates that TRC methods achieve a 75 percent success rate on 100 questions, outperforming traditional vector methods.