Improving retrieval augmented generation via denoising and reconstruction of knowledge graphs
Abstract
Retrieval-Augmented Generation (RAG) has greatly enhanced the performance of large language models (LLMs) in terms of knowledge intense area by retrieving relevant knowledge. Recent researches that combine RAG and knowledge graph (KG) is a new approach to utilize the advantages of both RAG and KG, providing structured facts with node descriptions and relationships, allowing LLM to generate more precise, various answers. However, during the extraction, KG may contain excessive nodes and edges, which will mislead the LLM to produce irrelevant answers. This structural noise and redundancy can inadvertently mislead the LLM, potentially degrading the relevance and accuracy of the generated response. To overcome these limitations, we introduce a new method called DenoiseKG, which denoises knowledge graph by post-processing similar nodes, together with relations between them, by refining retrieved KG subgraphs before they are utilized, DenoiseKG focuses specifically on reducing noise by identifying and deduplicating semantically similar nodes and consolidating their associated relations. This denoising process optimizes the overall performance of KG-augmented RAG systems. To address these limitations, we propose DenoiseKG, a post-processing method that denoises knowledge graphs by identifying and merging semantically similar nodes and consolidating their associated relations before use. By refining retrieved KG subgraphs, DenoiseKG effectively reduces noise in KG-augmented RAG…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.