Knowledge Graph

A knowledge graph represents information as nodes (entities) and edges (relationships), rather than as documents or as rows in a table. The structure matters because it lets questions be asked across relationships — what connects these two things, what else shares this property — which neither prose nor a rigid hierarchy supports.

Formal knowledge graphs use typed relationships and identifiers so that separate statements about the same entity compose: the model behind Schema.org markup, where a page emits a reference by identifier and it resolves against a description published elsewhere. See Structured Data.

Informal knowledge graphs use untyped links between notes and get much of the benefit for almost none of the cost. What a personal wiki gains from being a graph rather than a hierarchy is that a concept can sit in many contexts at once, and that connections can be discovered rather than planned. See Zettelkasten and Backlink.

The failure mode of the formal kind is over-modelling — an ontology so precise that adding anything requires a decision nobody wants to make. The failure mode of the informal kind is under-modelling — links so indiscriminate that the graph carries no signal. The useful middle is a small number of relationship kinds used consistently.