Extract Page Content from olmOCR Results — olmocr_extract

Transforms olmOCR output into the Tensorlake-compatible page format used by the other providers. olmOCR returns natural-reading-order markdown per page with a YAML front matter header (primary_language, is_rotation_valid, rotation_correction, is_table, is_diagram). This function strips the front matter into a metadata field and returns the body as text.

Usage

olmocr_extract_pages(result, pages = NULL)

Arguments

result: List. The parsed response from olmocr_ocr().
pages: Integer vector. Page numbers to extract. If NULL (default), extracts all pages.

Value

List with one element per page, each containing:

page_number: Integer page number
page_header: Empty character vector (olmOCR removes headers/footers)
section_header: Empty character vector
text: Character string with page content in markdown
tables: Empty list (tables are inline in text)
other: Empty list
metadata: Named list parsed from the olmOCR YAML front matter

Details

Tables and equations are emitted inline in the markdown text by olmOCR, so the tables and other fields are returned empty for compatibility.

Author

Nathan C. Layman

Examples

if (FALSE) { # \dontrun{
result <- olmocr_ocr("document.pdf")
pages <- olmocr_extract_pages(result)
pages[[1]]$text
} # }