#!/bin/bash
limit=30;
function fromPDF(){
echo `date +"%Y-%m-%d +%H"` > $1.txt;
pdftotext $1 - | sed 's/[^a-zA-Z0-9, ]*//g' | sed 's/Page/@@MARKER@@ Page/g' | sed 's/CHAPTER/@@MARKER@@ CHAPTER/g' | sed 's/SUMMARY/@@MARKER@@ SUMMARY/g' | sed 's/REFERENCES/@@MARKER@@ REFERENCES/g' | sed 's/OVERVIEW/@@MARKER@@ OVERVIEW/g' | sed 's/Chapter/@@MARKER@@ Chapter/g' | sed 's/Index/@@MARKER@@ Index/g' >> $1.txt;
mv $1.txt tmp-test;
cat tmp-test/$1.txt | egrep -v "^[0-9]*[A-Z]$|[0-9][0-9]$|[0-9]$|^[*]$" | data-freq --limit $limit
}
function fromImage(){
echo `date +"%Y-%m-%d +%H"` > $1.txt;
tesseract $1 $1.ocr
cat $1.ocr.txt | sed 's/[^a-zA-Z0-9, ]*//g' >> $1.txt;
rm $1.ocr.txt
mv $1.txt tmp-test;
cat tmp-test/$1.txt | egrep -v "^[0-9]*[A-Z]$|[0-9][0-9]$|[0-9]$|^[*]$" | data-freq --limit $limit
}
function find_all(){
for i in `find . -iname "*$1" -print | cut -d '/' -f2`; do
if [ -f tmp-test/$i.txt ] || [ -f $i.txt ] ;then
echo "skip $i";
else
case $1 in
pdf) fromPDF $i ;;
jpeg|jpg|tif|tiff|png) fromImage $i ;;
*) echo "MISSING transform function for $1 in $i!" ;;
esac
fi
done
}
mkdir -p tmp-test
for i in `echo -e 'pdf\npng\njpeg\njpg\ntif\ntiff'`; do
find_all $i;
done