run_ocr.sh

#!/bin/bash

# requires ghostscript (http://www.ghostscript.com/)
# requires ImageMagick
# requires tesseract (https://code.google.com/p/tesseract-ocr/)
# requires GNU parallel (https://www.gnu.org/software/parallel/)
# all of these are typically available through yum/apt/brew/etc.

# number of cores over which the process will be parallelized
num_cores=$1

# converts each of the PDFs into TIFF images so that tesseract can interact with them
ind . -name '*.pdf' | parallel --gnu -j $NUMCORES convert -depth 8 -density 200 {}[0-19] {}.tif

# runs OCR on the found TIFF files and converts them to text. Assumes English, but you can supply
# extra arguments to tesseract
find . -name '*.tif' | parallel -j $NUMCORES tesseract -l eng {} {}