Methodological principles
The methodology separates what was observed from what is inferred, and what is published from the evidence that supports it.
A source observation is captured with its origin and time. It becomes a claim only after normalization and governance.
A claim can refer to an entity without proving identity. Resolution is a separate step with its own evidence and conflict handling.
Evidence is the supporting material. Assurance expresses how strongly a governed claim is supported; they are not interchangeable.
State is temporal. Freshness, supersession and correction are explicit parts of the model.
ENTIA Home, API, MCP, grounding and deltas are delivery surfaces over a canonical graph, not independent truths.
1. Source and rights layer
ENTIA starts with observations from public, official and otherwise lawfully usable sources. Each source is treated with its own provenance, authority, temporal context and rights conditions. Source material is not flattened into an anonymous pool before governance.
2. Evidence and claims
Normalized observations are represented as governed claims linked back to supporting evidence. Where evidence conflicts, the conflict is preserved rather than silently collapsed. Derived or inferred values are identified as such and are not presented as direct source facts.
3. Entity resolution
Resolution determines whether observations belong to the same economic entity. Matching can use names, identifiers, addresses, domains and other corroborating signals. Ambiguous cases remain unresolved or conflict-marked rather than being forced into a match.
4. Canonical graph
The primary asset is the canonical economic-identity graph: governed claims, evidence, provenance, temporal state, assurance, corrections and relations between entities. Public pages and machine interfaces are projections of that graph.
Canonical Entity
Stable identity and governed state are maintained once, then projected consistently across delivery surfaces.
5. Temporal state and assurance
ENTIA distinguishes when something was observed, when it was valid, when it changed and when a newer claim superseded it. Assurance is attached to claims and reflects support quality; it is not a substitute for the underlying evidence.
6. Machine delivery
Canonical identity can be projected through ENTIA Home, JSON-LD, API, MCP, grounding surfaces and delta feeds. Each projection should preserve stable identifiers and enough provenance to trace machine-readable output back to governed claims.
7. Measurement boundaries
ENTIA keeps machine events separate. Access does not prove retrieval; retrieval does not prove selection; selection does not prove citation; citation does not prove entity absorption; and none of these events by itself proves model training.
ACCESS → RETRIEVAL → SELECTION → CITATION → ENTITY ABSORPTION → ACTION
Each transition requires its own evidence.