What Is a “Grounding”?

When parses a document, it breaks the content into chunks, which are discrete elements extracted from a document, like blocks of text or tables. Each chunk includes a grounding, which represents the location of the chunk in the document. The grounding includes:
  • the page number that the chunk is on
  • the relative coordinates of the bounding box of the chunk
For example, below is the JSON output for a text chunk. The grounding object indicates that the text is on the first page, and the box object indicates the bounding box coordinates.
{
      "text": "## INSURANCE COMPANY",
      "grounding": [
        {
          "box": {
            "l": 0.35,
            "t": 0.22619999999999998,
            "r": 0.565,
            "b": 0.24033749999999998
          },
          "page": 0
        }
      ],
      "chunk_type": "text",
      "chunk_id": "9475461e-0686-4b16-b503-ccec7d7f115c"
    }