Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TagRAG: Tag-guided Hierarchical Knowledge Graph Retrieval-Augmented Generation

Wenbiao Tao, Yunshi Lan, Weining Qian

TL;DR

TagRAG introduces a tag-guided hierarchical knowledge graph for retrieval-augmented generation to address the global reasoning and incremental update limitations of prior graph-based RAG methods. It constructs a Tag Knowledge Graph with object tags and domain tag chains linked into a DAG, and fuses domain-centric knowledge to support efficient retrieval. During inference, domain-centric tags and their chains are retrieved and fused to generate grounded responses, enabling effective operation with smaller language models and scalable knowledge updates. Empirical results on UltraDomain datasets show TagRAG achieves an average win rate of 95.41% over baselines and offers significant construction and retrieval efficiency gains, demonstrating strong practical impact for cross-domain knowledge-intensive QA.

Abstract

Retrieval-Augmented Generation enhances language models by retrieving external knowledge to support informed and grounded responses. However, traditional RAG methods rely on fragment-level retrieval, limiting their ability to address query-focused summarization queries. GraphRAG introduces a graph-based paradigm for global knowledge reasoning, yet suffers from inefficiencies in information extraction, costly resource consumption, and poor adaptability to incremental updates. To overcome these limitations, we propose TagRAG, a tag-guided hierarchical knowledge graph RAG framework designed for efficient global reasoning and scalable graph maintenance. TagRAG introduces two key components: (1) Tag Knowledge Graph Construction, which extracts object tags and their relationships from documents and organizes them into hierarchical domain tag chains for structured knowledge representation, and (2) Tag-Guided Retrieval-Augmented Generation, which retrieves domain-centric tag chains to localize and synthesize relevant knowledge during inference. This design significantly adapts to smaller language models, improves retrieval granularity, and supports efficient knowledge increment. Extensive experiments on UltraDomain datasets spanning Agriculture, Computer Science, Law, and cross-domain settings demonstrate that TagRAG achieves an average win rate of 95.41\% against baselines while maintaining about 14.6x construction and 1.9x retrieval efficiency compared with GraphRAG.

TagRAG: Tag-guided Hierarchical Knowledge Graph Retrieval-Augmented Generation

TL;DR

TagRAG introduces a tag-guided hierarchical knowledge graph for retrieval-augmented generation to address the global reasoning and incremental update limitations of prior graph-based RAG methods. It constructs a Tag Knowledge Graph with object tags and domain tag chains linked into a DAG, and fuses domain-centric knowledge to support efficient retrieval. During inference, domain-centric tags and their chains are retrieved and fused to generate grounded responses, enabling effective operation with smaller language models and scalable knowledge updates. Empirical results on UltraDomain datasets show TagRAG achieves an average win rate of 95.41% over baselines and offers significant construction and retrieval efficiency gains, demonstrating strong practical impact for cross-domain knowledge-intensive QA.

Abstract

Retrieval-Augmented Generation enhances language models by retrieving external knowledge to support informed and grounded responses. However, traditional RAG methods rely on fragment-level retrieval, limiting their ability to address query-focused summarization queries. GraphRAG introduces a graph-based paradigm for global knowledge reasoning, yet suffers from inefficiencies in information extraction, costly resource consumption, and poor adaptability to incremental updates. To overcome these limitations, we propose TagRAG, a tag-guided hierarchical knowledge graph RAG framework designed for efficient global reasoning and scalable graph maintenance. TagRAG introduces two key components: (1) Tag Knowledge Graph Construction, which extracts object tags and their relationships from documents and organizes them into hierarchical domain tag chains for structured knowledge representation, and (2) Tag-Guided Retrieval-Augmented Generation, which retrieves domain-centric tag chains to localize and synthesize relevant knowledge during inference. This design significantly adapts to smaller language models, improves retrieval granularity, and supports efficient knowledge increment. Extensive experiments on UltraDomain datasets spanning Agriculture, Computer Science, Law, and cross-domain settings demonstrate that TagRAG achieves an average win rate of 95.41\% against baselines while maintaining about 14.6x construction and 1.9x retrieval efficiency compared with GraphRAG.
Paper Structure (34 sections, 9 equations, 5 figures, 10 tables, 1 algorithm)

This paper contains 34 sections, 9 equations, 5 figures, 10 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Task Definition
  4. TagRAG
  5. Tag Knowledge Graph Construction
  6. Object Tag Keyword Extraction
  7. Domain Tag Chain Organization
  8. Domain-centric Knowledge Fusion
  9. Knowledge Incremental Insertion
  10. Tag-guided Retrieval-Augmented Generation
  11. Domain-centric Knowledge Retrieval
  12. Hierarchical Tag Chain Integration
  13. Tag Knowledge-Fused Generation
  14. Experimental Setup
  15. Datasets
  16. ...and 19 more sections

Figures (5)

  • Figure 1: Inefficient graph construction and reasoning.
  • Figure 2: The proposed TagRAG framework.
  • Figure 3: Performance-efficiency analysis: Comparative win rates, graph construction time and inference time results of TagRAG and baselines across four datasets. The larger the bubble and the closer to the lower left corner, the better the method.
  • Figure 4: Visualization of knowledge graph construction with Qwen3-4B on UltraDomain Mix.
  • Figure 5: Demonstration of tag chains on UltraDomain Mix.