Knowledge Graph

A knowledge graph stores information as a network of entities (nodes) and their relationships (edges). For example: [Anthropic] –[developed]→ [Claude] –[is a]→ [Large Language Model]. This structured representation lets AI systems reason about how concepts relate, not just retrieve documents about them.

Knowledge graphs have experienced a resurgence in the LLM era through GraphRAG — a technique pioneered by Microsoft that combines traditional RAG with graph traversal. Instead of only retrieving the most similar document chunks, GraphRAG can navigate entity relationships to surface insights that span multiple documents.

In enterprise AI, knowledge graphs are used to represent product catalogs, organizational hierarchies, medical ontologies, and regulatory frameworks. Graph databases like Neo4j and Amazon Neptune are commonly used to store and query knowledge graphs, often in conjunction with vector search for hybrid retrieval.