Enrichment

Document AI uses Enterprise Knowledge Graph to normalize and enrich entity extraction results (for supported fields). For example, the addresses 123 Main St Apt 1 and 123 Main street # 1 could be normalized to the same standardized address.

For each supported field, Document AI also returns a normalizedValue in addition to the raw extracted field, normalizing the literal text. This contains the data in a standardized format to reduce post-processing.

Most data belongs to one of the following categories:

• Money
• Date
• Timestamp
• Address
• Boolean
• Integer
• Float

Sample response

The enriched values can be found in the entities.normalizedValue field as shown in the following truncated sample:

{
  "entities": [
    {
      "textAnchor": {
        "textSegments": [ ... ],
        "content": "Google Singapore"
      },
      "type": "employer_name",
      "mentionText": "Google Singapore",
      "confidence": 0.69933707,
      "pageAnchor": {
        "pageRefs": [
          {
            "boundingPoly": {
              "normalizedVertices": [ ... ]
            }
          }
        ]
      },
      "id": "9",
      "normalizedValue": {
        "text": "Google Asia Pacific, Singapore"
      }
    }
  ]
}

In the sample, the original employer_name "Google Singapore" has been normalized to "Google Asia Pacific, Singapore".

In the Google Cloud console, the enriched and normalized fields are annotated with G. For example:

Supported processors

Here are the processors and fields that support entity enrichment.