Entity Research Methodology.
Entity research begins by deciding what a named thing refers to. Only after identity is resolved can a method assign a type, examine attributes, validate relationships, map evidence and determine whether a corpus represents that entity clearly, consistently and in the right context.
An entity is not merely a repeated word.
The same label may refer to several things; the same thing may be expressed through several labels. Entity research controls that many-to-many problem before counting coverage or constructing relationships.
Entity research methodology is the controlled resolution and examination of distinct identities, their types, aliases, attributes, relations, contexts and supporting evidence across a declared information environment.
An entity is the resolved thing that the mention is judged to represent under a documented identity rule.
An entity is a distinguishable identity that can carry attributes and participate in relations.
A relation requires a defined predicate and evidence supporting that specific connection.
Six objects that must never be collapsed.
Entity analysis becomes unreliable when strings, concepts, identities, documents and structured representations are treated as interchangeable.
Observed expression
A name, alias, abbreviation, pronoun, identifier or markup value found in an eligible source.
Resolved identity
The distinct person, organization, place, product, event, work or other thing represented by one or more mentions.
Declared classification
The analytical class assigned under a versioned type system, including allowed overlaps and unresolved cases.
Property of an entity
A value or state describing the entity, with its evidence, time validity and conflict status preserved.
Directed connection
A subject–predicate–object statement whose direction, meaning, evidence and temporal scope are explicit.
Evidence surface
A page or record can mention many entities. It does not become identical to the primary entity it describes.
Identity is produced through inspectable decisions.
The pipeline preserves the path from raw expression to accepted entity record. An unresolved mention remains unresolved; it is never forced into the nearest candidate to improve completion.
Mention
Record exact expression, source location and surrounding context.
Candidates
Create plausible identities using aliases, identifiers and context.
Signals
Evaluate type, attributes, relations, time and contextual fit.
Resolution
Accept, reject, merge, split or retain ambiguity under declared rules.
Classification
Assign permitted entity class and record competing types.
Evidence
Attach each accepted property and relation to supporting observations.
Record
Freeze canonical identity, aliases, claims, conflicts and provenance.
One entity system. Four research operations.
Each operation uses a different analytical unit and produces a different permitted claim. All default console content is embedded in HTML and remains visible without JavaScript.
Resolve an ambiguous label without forcing certainty
An edge is a claim, not a decorative line.
Every relation must declare direction, predicate, evidence and temporal state. Reversing an edge or changing its predicate creates a different claim.
predicate / ownsRequires evidence of ownership. A link, mention or shared branding alone does not establish legal or operational ownership.
predicate / producesDifferent from distributes, sells, repairs or mentions. Predicate precision protects the identity graph from semantic inflation.
predicate / primary_subjectDifferent from incidental mention. Corpus ownership rules determine which document is responsible for the entity’s central explanation.
Entity claims require typed evidence states.
The record distinguishes what is directly observed from what is derived, inferred, contradicted or unresolved.
| STATE | MEANING | EXAMPLE | ALLOWED USE | PROHIBITED COLLAPSE |
|---|---|---|---|---|
| Observed | Directly present in eligible source | Exact name or identifier appears | MENTION EVIDENCE | Observed mention → confirmed identity |
| Resolved | Identity decision satisfies declared rules | Mention maps to E-042 | ENTITY AGGREGATION | Resolved here → universally identical |
| Derived | Calculated from validated entity records | Attribute coverage rate | TRANSPARENT METRIC | Derived score → external fact |
| Inferred | Reasoned interpretation with alternatives | Likely document owner | QUALIFIED ANALYSIS | Likely → established |
| Contradicted | Eligible evidence supports incompatible states | Two founding dates | CONFLICT REGISTER | Select preferred value silently |
| Unresolved | Evidence cannot distinguish candidates | Shared name, weak context | REVIEW QUEUE | Unknown → absent or false |
Measure the corpus without inventing an authority score.
These TAO measures describe the declared entity model and analyzed corpus. Formulas, denominators and unresolved states remain visible.
Resolution rate
resolved eligible mentions ÷ all eligible mentionsReports how much of the mention population has passed identity resolution.
Entity representation rate
represented required entities ÷ required model entitiesRequires a versioned entity model and explicit threshold for what counts as represented.
Attribute coverage
supported required attributes ÷ required attributesCount supported values separately from mentions, nulls, contradictions and unresolved attributes.
Relation coverage
supported required relations ÷ required relationsPredicate and direction must match. Co-occurrence does not satisfy the denominator.
Conflict rate
entities with unresolved conflicts ÷ reviewed entitiesHigher conflict visibility may reflect better auditing, not worse content.
Document orphan rate
entities without a responsible document ÷ required entitiesUses declared ownership rules to find entities scattered across mentions but explained nowhere.
Concrete decisions with strict inference boundaries.
All examples are synthetic. They demonstrate method logic rather than live entity facts or external system states.
Strong entity systems preserve what they cannot resolve.
Forcing every input into a clean graph creates false certainty. These failure modes remain first-class research states.
Homonym collision
One label refers to several identities and context is insufficient to choose.
Alias fragmentation
Several labels for one entity are counted as distinct identities.
Type drift
The entity class changes across analysis without a versioned rule or valid multi-type state.
Predicate inflation
A weak mention is converted into ownership, production, endorsement or another stronger relation.
Temporal collision
Attributes valid at different times are treated as simultaneous contradictions.
Source circularity
Several pages repeat one unsupported origin and are misread as independent confirmation.
Entity-page collapse
A document identifier is treated as the identity of the entity described by the document.
Authority fiction
An internal coverage score is presented as external proof that a system recognizes or trusts the entity.
Every entity record should be reconstructable.
The manifest preserves identity logic, evidence state, conflicts and corpus responsibility as part of the research output.
Every method node. One controlled research route.
MTH/06 resolves identities and relationships. MTH/07 determines how entities, corpora or systems can be compared without breaking equivalence.