Knowledge Graph RAG: Agentic Crawling and Graph Construction in Enterprise Documents

TL;DR

This paper introduces Agentic Knowledge Graphs with Recursive Crawling to improve semantic search accuracy in complex enterprise documents, outperforming standard RAG systems in regulatory query tasks.

Contribution

It presents a novel knowledge graph construction method that captures hierarchical and interconnected information, enhancing retrieval accuracy in enterprise ecosystems.

Findings

Achieves 70% accuracy improvement over standard vector-based RAG systems.

Demonstrates effectiveness on the Code of Federal Regulations (CFR).

Provides exhaustive and precise answers for complex regulatory queries.

Abstract

This research paper addresses the limitations of semantic search in complex enterprise document ecosystems. Traditional RAG pipelines often fail to capture hierarchical and interconnected information, leading to retrieval inaccuracies. We propose Agentic Knowledge Graphs featuring Recursive Crawling as a robust solution for navigating superseding logic and multi-hop references. Our benchmark evaluation using the Code of Federal Regulations (CFR) demonstrates that this Knowledge Graph-enhanced approach achieves a 70% accuracy improvement over standard vector-based RAG systems, providing exhaustive and precise answers for complex regulatory queries.