PARSE_DOCUMENT (SNOWFLAKE.CORTEX)¶

Syntax¶

SNOWFLAKE.CORTEX.PARSE_DOCUMENT( '@<stage>', '<path>', [ <options> ] )

Arguments¶

Required:

stage

Name of the Snowflake stage.

path

Relative path to the document on the Snowflake stage.

Optional:

options

An OBJECT value that contains options for parsing documents. The supported keys are shown below. All are optional.

• 'mode': Specifies the parsing mode. The supported modes are:

  • 'OCR': The function extracts text only. This is the default mode.

  • 'LAYOUT': The function extracts layout as well as text, including structural content such as tables.

• 'page_split': If set to TRUE, the function splits the document into pages and processes each page separately. Only PDF, PowerPoint (.pptx), and Word (.docx) documents are supported. Documents in other formats return an error. The default is FALSE.

  Tip

  To process long documents that exceed the token limit of PARSE_DOCUMENT, set this option to TRUE.

Returns¶

A JSON object (as a string) that contains the extracted data and associated metadata. The options argument determines the structure of the returned object.

If the 'page_split' option is set, the output has the following structure:

• "pages": An array of JSON objects, each containing text extracted from the document. If the document has only one page, the output still contains a "pages" array (which contains a single object). Each page has the following fields:

  • "content": Plain text (in OCR mode) or Markdown-formatted text (in LAYOUT mode).

  • "index": The page index in the file, starting at 0. Page numbers and formats specified in the document are ignored.

• "errorInformation": Contains error information if document can’t be parsed.

• "metadata": Contains metadata about the document, such as page count.

Note

The "pages" and "metadata" fields are present in the output when parsing succeeds. "errorInformation" is present only if parsing fails.

If 'page_split' is FALSE or is not present, the output has the following structure:

• "content": Plain text (in OCR mode) or Markdown-formatted text (in LAYOUT mode).

• "errorInformation": Contains error information if the document can’t be parsed.

• "metadata": Contains metadata about the document, such as page count.

Note

The "content" and "metadata" fields are present in the output when parsing succeeds. "errorInformation" is present only if parsing fails.

Examples¶

SELECT TO_VARCHAR(
    SNOWFLAKE.CORTEX.PARSE_DOCUMENT(
        '@PARSE_DOCUMENT.DEMO.documents',
        'document_1.pdf',
        {'mode': 'OCR'})
    ) AS OCR;

{
    "content": "content of the document"
}

LAYOUT mode¶

This example parses a document with a table shown in the following screenshot:

SELECT
  TO_VARCHAR (
    SNOWFLAKE.CORTEX.PARSE_DOCUMENT (
        '@PARSE_DOCUMENT.DEMO.documents',
        'document_1.pdf',
        {'mode': 'LAYOUT'} ) ) AS LAYOUT;

Copy

Output:

{
  "content": "# This is PARSE DOCUMENT example
     Example table:
     |Header|Second header|Third Header|
     |:---:|:---:|:---:|
     |First row header|Data in first row|Data in first row|
     |Second row header|Data in second row|Data in second row|

     Some more text."
 }

SELECT
  TO_VARCHAR (
    SNOWFLAKE.CORTEX.PARSE_DOCUMENT (
        '@PARSE_DOCUMENT.DEMO.documents',
        'document_1.pdf',
        {'mode': 'OCR', 'page_split': TRUE} ) ) AS MULTIPAGE;

{
  "pages": [
    {
      "content": "content of the first page",
      "index": 0
    },
    {
      "content": "content of the second page",
      "index": 1
    },
    {
      "content": "content of the third page",
      "index": 2
    }
  ],
  "metadata": {
    "pageCount": 3
  }
}