§ 1Annotation & provenance

The annotate() helper tags any value with @confidence, @source, @method, @extractedAt, and @humanVerified. The result is an expanded JSON-LD value object that downstream consumers can inspect or filter on.

Input

value

confidence0.92

source (IRI)

method

extractedAt (ISO 8601)

humanVerified

Output

{
  "@value": "San Francisco",
  "@confidence": 0.92,
  "@source": "https://model.example.org/geo-v3",
  "@extractedAt": "2026-05-18T12:00:00Z",
  "@method": "geocoding",
  "@humanVerified": false
}

Filter a graph by confidence

Once values carry confidence, you can drop low-confidence assertions before they reach a downstream model. Drag the slider:

min confidence0.70

[
  {
    "@type": "Person",
    "name": {
      "@value": "Alice",
      "@confidence": 0.95,
      "@source": "https://ner.example/v2",
      "@method": "NER"
    }
  },
  {
    "@type": "Person",
    "name": {
      "@value": "Bob",
      "@confidence": 0.72,
      "@source": "https://ner.example/v2",
      "@method": "NER"
    }
  }
]