Extract layout-aware content from PDF

Supported in: Batch, Faster

Extracts layout-aware content from PDF pages. Returns structured blocks with bounding boxes, text content, and block types. Supports configurable OCR modes and output formats.

Expression categories: Media

Declared arguments

• Languages to detect: Languages to detect in the PDF.
  Set<Enum<Afrikaans, Albanian, Amharic, Arabic, Armenian, Assamese, Azerbaijani, Azerbaijani - Cyrilic, Basque, Belarusian, and more ...>>
• Media reference: The column containing media references to PDF files in a media set.
  Expression<Media reference>
• Scripts to detect: Scripts to detect in the PDF.
  Set<Enum<Arabic, Armenian, Bengali, Canadian Aboriginal, Cherokee, Cyrillic, Devanagari, Ethiopic, Fraktur, Georgian, and more ...>>
• optional End page: The end of the page range (inclusive). If no value is provided, it will default to the last page.
  Expression<Integer>
• optional Error handling: Determines the behavior of the pipeline for inputs that fail to process.
  Enum<FAIL, NULL>
• optional Extract method: Choose how text is extracted. Conditional scan uses best-effort detection to scan pages with little or no digital text. Digital text extracts embedded digital text. Full-page scan scans every page.
  Enum<Conditional Scan, Digital Text, Full-page Scan>
• optional Output format: The format of the extracted text content within blocks.
  Enum<HTML, Markdown, Text, hOCR>
• optional Start page: The start of the page range. If no value is provided, it will default to the first page.
  Expression<Integer>

Output type: Array<Array<Struct<block_index, block_id, page, block_type, content, bounding_box, languages<String>, confidence>>>

Examples

Example 1: Base case

Argument values:

• Languages to detect: {ENG}
• Media reference: mediaReference
• Scripts to detect: {}
• End page: End Page
• Error handling: FAIL_FAST
• Extract method: AUTO
• Output format: TEXT
• Start page: Start Page

---