Skip to content

Entity Resolution

Deciding which surface forms denote the same real thing, and linking them to one canonical identity. "Sam", "Samira", "Sammy", "my partner" and a bare "She" are five ways of referring to one person; resolution is what makes them one node.

Why it matters in a memory layer

Without it, evidence about a person is split across records that never meet, so nothing accumulates and no contradiction between them is detectable — the system holds four half-pictures and cannot tell they are the same picture. It is invisible in testing, because every individual record looks correct.

Distinct from deduplication: dedupe asks whether two memories say the same thing, resolution asks whether two mentions point at the same thing. Two memories about one person are usually both worth keeping.

The property that makes it tractable is that resolution links rather than rewrites — the canonical id goes in a separate field and the content stays exactly as spoken, so provenance survives and the operation can be re-run or undone.