Entity Disambiguation in the Perception Graph
Disambiguation is the process by which an AI system resolves which specific entity a name refers to. Without strong disambiguation signals, the same name can resolve to the wrong entity — or no entity at all.
Disambiguation is one of the hardest problems in knowledge representation. The word "Apple" could refer to the technology company, the fruit, the record label, a person's name, or a geographic location. An AI system resolving which Apple is meant in a specific context uses disambiguation signals: type declarations, co-occurring attributes, geographic context, industry context, and external validation (does a sameAs link confirm which Apple this is?).
Disambiguation Signals in the Content
Strong disambiguation signals in page content: Organization schema with a specific @type sub-type (e.g., TechCompany, LocalBusiness), explicit category declaration in body text ("a machine trust intelligence platform"), founding or operating context ("founded in 2024"), geographic specificity, industry terms that narrow the entity to one possible interpretation, and sameAs links to external identity sources (LinkedIn, Companies House, Wikidata, Wikipedia).
What Happens Without Disambiguation
Without disambiguation signals, the AI system may: conflate the entity with a more famous entity of the same name, assign the entity the wrong type (treating a software product as a physical product, or a service as a person), create multiple separate entity records for the same real-world entity (one per context in which it appears), or fail to match the entity to any external knowledge base entry — leaving it as an isolated, unvalidated entity with low trust.
◆The most powerful disambiguation combination: (1) Schema @type that is as specific as possible, (2) a sameAs link to a Wikidata entity or LinkedIn company page, (3) an explicit category sentence in the body text within the first paragraph. These three together give an AI system enough context to correctly identify the entity in 95%+ of disambiguation scenarios.
Disambiguation Strength in SiteNexis
The disambiguationStrength field on each perception_node in the Perception Graph reflects how many disambiguation signals are present for that entity. A strength of 1.0 means every available disambiguation signal is present and consistent. A strength of 0.2 means the entity is named but has no type declaration, no sameAs link, no category context, and no distinguishing attributes. Low disambiguation strength is one of the most common causes of low Entity Confidence Scores in the audit.
Perception vs Fact Layer — Part 4 of 10