OCR to pdf description

Finder and Preview.app GetInfo readily display PDF:Subject metadata (not XMP-dc:Subject or XMP-dc:Description).

exiftool -overwrite_original '-Subject<=-' cicero.pdf < cicero.txt 

exiftool -a -G1 -s -Subject cicero.pdf                            
[PDF]           Subject                         : Text from cicero.txt
[XMP-dc]        Subject                         : Text from cicero.txt

And also PDF:Author, Title, and Subject:

exiftool -overwrite_original -Author='Charles Dickens' -Title='A Tale of Two Cities' -Subject='A historical novel set during the French Revolution' cicero.pdf

exiftool -a -G1 -s cicero.pdf         
[PDF]           Author                          : Charles Dickens
[PDF]           Title                           : A Tale of Two Cities
[PDF]           Subject                         : A historical novel set during the French Revolution
[XMP-pdf]       Author                          : Charles Dickens
[XMP-dc]        Subject                         : A historical novel set during the French Revolution
[XMP-dc]        Title                           : A Tale of Two Cities

I have a three paragraph text file named cicero.txt in my home directory. I have a PDF on my Desktop named cicero.pdf containing the same text as the text file. Now, I want to use ExifTool to insert the contents of that text file into the XMP:descriptioin field of the PDF.

exiftool -overwrite_original '-Description<=-' cicero.pdf < ~/cicero.txt

One requires a PDF Editor to view this XMP description data, but I have written a shell script that extracts the XMP data from the PDF to stdout using ExifTool. The pipe to awk removes dozens of spaces at the end of the ExifTool output.

#!/bin/zsh

[[ "${1:e:l}" == "pdf" ]] || {echo "not a pdf."; exit 1}

XMPData="$(which exiftool) -XMP -b "${1}" | awk 'NF' "
[[ ${XMPData} ]] || {echo "no XMP data in PDF"; exit 1}
print ${XMPData}
exit 0

exiftool can pipe tesseract output into file metadata (example below uses english but you can set it to use other OCR languages as well). But tesseract can not read .pdf. And do you want to manually limit what text snippet to use?

tesseract file.jpg - -l eng |exiftool -overwrite_original '-Description<=-' file.jpg

https://exiftool.org/forum/index.php?topic=15872.msg85337#msg85337

Matti,

Fully aware of the standard PDF metadata fields that Finder and Preview can display. Have written tools in Python and Swift to do that as well. I don't see the advantage of sending three paragraphs of text from Cicero.pdf into the Subject metadata field though… 🧐

Yes, those standard metadata fields can be updated by Exiftool, but I think the place for description category content remains in the PDF XMP data container.

Or, without ExifTool, how to extract XMP metadata from a PDF (if it exists), and write to a text file on the Desktop. This assumes one has the command line tools for Xcode installed as it uses the /usr/bin/strings command:

#!/bin/zsh

# force absolute path
INFILE="${1:a:s/\~\//}"
OUTFILE="$HOME/Desktop/${INFILE:r:t}_meta.txt"

hasXMPData=$(egrep -a -c -m1 "x:xmpmeta" "${INFILE}")
(( $hasXMPData )) || {echo "PDF XMP data not found."; exit 1}

awk -v out="${OUTFILE}" '/<x:xmpmeta/,/<\/x:xmpmeta/ {print $0 > out}' <<<"$(strings "${INFILE}")"
exit 0

Usage: pdfxmp_meta.zsh ~/Desktop/cicero.pdf
Output: ~/Desktop/cicero_meta.txt