KGRAG-Ex: Explainable Retrieval-Augmented Generation with Knowledge Graph-based Perturbations

2025-07-11

AI Summary

To address the weak interpretability and difficulty in verifying factual grounding of Retrieval-Augmented Generation (RAG) systems when retrieving from unstructured text, this paper proposes KG-RAG: a knowledge graph–enhanced RAG framework. First, it constructs a domain-specific knowledge graph (KG) via prompt-driven information extraction; then, semantic paths within the KG are converted into pseudo-paragraphs to improve retrieval relevance. Furthermore, it introduces a KG structure–based attribution method that quantifies the influence of node types and topological features—such as centrality and path length—on answer generation, thereby rendering reasoning paths transparent. Experiments demonstrate significant improvements in factual accuracy and explanation consistency of generated answers. Notably, this work is the first to empirically reveal statistically significant correlations between KG structural metrics and explanation importance, establishing a structured, quantifiable methodology for interpretable RAG.

Technology Category

• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Knowledge Representation and Reasoning: Knowledge Acquisition
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Retrieval-Augmented Generation (RAG) enhances language models by grounding responses in external information, yet explainability remains a critical challenge, particularly when retrieval relies on unstructured text. Knowledge graphs (KGs) offer a solution by introducing structured, semantically rich representations of entities and their relationships, enabling transparent retrieval paths and interpretable reasoning. In this work, we present KGRAG-Ex, a RAG system that improves both factual grounding and explainability by leveraging a domain-specific KG constructed via prompt-based information extraction. Given a user query, KGRAG-Ex identifies relevant entities and semantic paths in the graph, which are then transformed into pseudo-paragraphs: natural language representations of graph substructures that guide corpus retrieval. To improve interpretability and support reasoning transparency, we incorporate perturbation-based explanation methods that assess the influence of specific KG-derived components on the generated answers. We conduct a series of experiments to analyze the sensitivity of the system to different perturbation methods, the relationship between graph component importance and their structural positions, the influence of semantic node types, and how graph metrics correspond to the influence of components within the explanations process.

Problem

• Enhancing explainability in Retrieval-Augmented Generation using knowledge graphs
• Improving factual grounding via structured KG-based retrieval paths
• Assessing KG component influence through perturbation-based explanation methods

Innovation

• Leverages domain-specific knowledge graph for RAG
• Transforms graph paths into pseudo-paragraphs
• Uses perturbation methods for explainability