Knowledge Graphsconcept

A knowledge graph represents entities as nodes and their relationships as typed edges, enabling multi-hop reasoning and community detection that flat vector search cannot.

Organizing documents into a graph lets a system answer "how do these concepts connect?" not just "which chunk is most similar?". It is the relational backbone of compiled knowledge.

Where it appears

Descend

Historical evolution

From semantic networks and RDF to Microsoft GraphRAG (2024) demonstrating graph retrieval beats naive vector search on sensemaking queries.

Implementation details

SKCE builds several graph views over one ConceptNode set — concept, prerequisite, decision, ontology, misconception (see KNOWLEDGE_GRAPH_MODEL).

Sources

View in graph →