A knowledge graph represents domain entities and their relationships explicitly, giving retrieval a structured alternative to unstructured documents or isolated vector embeddings. Graph-based retrieval can follow linked relationships rather than returning only separate text fragments. section 3.3
A graph combines three core elements:
- Nodes represent entities; edges connect them through named relationships; properties attach descriptive attributes to nodes or edges. section 3.3.1
- Labeled property graphs store linked facts with attributes. RDF graphs represent knowledge as triples and can use ontologies to define schema and support inference. section 3.3.1
- RDF offers semantic consistency, interoperability, and richer provenance, at the cost of greater modeling and operational complexity. section 3.3.1
Simple relationship
A graph can connect a product to an organization with a “manufactures” edge, and attach attributes such as identifiers or provenance. section 3.3.1














