Fact Edges: How Link Relationships Become Machine Knowledge
In the SiteNexis graph model, an internal link between two pages is not just a connection — it is a fact edge: a typed, verifiable relationship between two entities.
The SiteNexis graph architecture distinguishes between fact edges and perception edges. A fact edge is extracted directly from the DOM: if page A has an anchor tag pointing to page B with anchor text "our machine trust methodology," that link is recorded as a fact edge with confidence 1.0 — it is an observable, verifiable relationship in the page structure. A perception edge is AI-inferred: a semantic relationship that the AI Perception Graph derives from the content of the two pages, not from an explicit link.
What Fact Edges Capture
Every fact edge in the SiteNexis graph captures: source page, target page, anchor text (the declared relationship), link position in DOM (navigation vs body vs footer — body links carry stronger topical signal), and surrounding sentence context (the text immediately before and after the link, which describes the relationship between the pages). These five fields make a fact edge a rich, typed relationship declaration, not just a pointer from A to B.
Fact Edges vs Navigation Links
Not all links are equally meaningful as fact edges. Navigation links (in header, footer, or sidebar navigation) create structural links but weak topical signals — they appear on every page and therefore carry no page-specific relationship information. Body links, appearing in the flow of content, are the most signal-rich fact edges because they appear in context and the surrounding text explains why this page links to that page.
●The fact_edges table in SiteNexis stores only body links and contextual links — not navigation links. Navigation links are stored separately in a navigation_structure table. This separation keeps the fact graph clean and prevents navigation-driven authority concentration from distorting the topical authority analysis.
How Fact Edges Feed the Perception Graph
The AI Perception Graph is derived from, but never merged with, the Fact Graph. Fact edges are one of the primary inputs to Perception Graph construction: when two pages share a fact edge and both contain strong entity signals, the Perception Graph infers a relationship between their primary entities. But the inferred perception edge is a separate record, with a separate confidence score, that cannot overwrite or replace the fact edge. The derivation is one-directional.
Link Graph Intelligence — Part 5 of 10