run GOCR and clean up afterwards

ocerrors.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
Replace common (G)OCR errors (in German)

Usage: python ocerrors.py inputfile [outputfile]
"""

import locale
import os, sys 
import shutil
import re

if len(sys.argv) < 2:
	print __doc__
	sys.exit(1)
	
infilename = sys.argv[1]
if not os.path.isfile(infilename):
	print u"%s is not a file!" % infilename
	sys.exit(2)
	
infile = open(infilename, 'rU')
text = u''.join(infile.readlines())
infile.close()


if len(sys.argv) > 2:
	print sys.argv
	outfilename = sys.argv[2]
else:
	shutil.move(infilename, infilename.replace('.txt','')+'.bak')
	outfilename = infilename


textre = (
#	('\r',	'\n'), # not necessary with filemode rU
	('-\n', ''),
)

rere = (
	# combine lines
	(u'([^\.][\.\!\?])\n',	r'\1\n###\n'), # CR -> #
	(u'\n\n+',	r'\n###\n'), # CR -> #
	(u'\n+',	u' '), # CR -> _
	(u'###\s*', r'\n\n'),
	# typical (G)OCR errors
	(u'_', ''), # _ are always errors (?)
	(u'(\w)(N)([a-zäöüß])',	u"\\1v\\3"), # N-v
	(u'(\w)(F)([a-zäöüß])',	u"\\1f\\3"), # F-f
	(u'(\w)(5|S)([a-zäöüß])',	u"\\1s\\3"), # 5/S-s
	(u'(\w)(2)([a-zäöüß])',	u"\\1z\\3"), # 2-z
	(u'(\w)(0)([a-zäöüß])',	u"\\1o\\3"), # 0-o
	(u'(\w)(1|I)([a-zäöüß ])',	u"\\1l\\3"), # 1/I-l
	
	(u'([A-ZÄÖÜ])(l)([A-ZÖÄÜ])',	u"\\1I\\3"), # l-I
	(u'([A-ZÄÖÜ])(0)([A-ZÖÄÜ])',	u"\\1O\\3"), # 0-O

	(u'lc([^h])', u'k\\1'), # lc-k
	(u'[1Il]C', u'K'), # 1/I/lC-K
	
	(u'(\d+)(o|O)',	u'\\g<1>0'), # o-0
	(u'(\d+)(o|O)',	u'\\g<1>0'), # again
	
	(u'[ \t]+',	u' '),
	(r"(\w+)u?\s?'\s?'\s?(\w+)",	u'\\1ü\\2'), # u''-ü
	(u'(ie|ei|eu|au|äu|ai|rö)ss',	u'\\1ß'), # manuscript ß errors
	(u'(<<|_<|<_|«)',	u'„'), # opening quotes
	(u'\s?(>>|_>|>_|»)',	u'“'), # closing quotes
	
	(u'\n\. ?\n',	''), # whitespace
)

text = text.replace('-\n','') # remove hyphenation
for (k, v) in rere:
	try:
		pat = re.compile(k, re.L|re.U)
		(text, n) = re.subn(pat, v, text)
		#print u'replace "%s"\twith "%s":\t%d replacements' % (k, v, n)
		#print text
		#raw_input('?')
	except Exception, ex:
		print ex
		print k
		sys.exit(3)

outfile = open(outfilename, 'wb')
outfile.write(text.encode('utf-8'))
outfile.close()

ocr.sh

#!/bin/bash
prefix=$1
prefix=${prefix:-noname}
inputfilter=$prefix

if [ "$prefix" = "noname" ]
then
	inputfilter=
fi
outputfile=$prefix.txt

for scan in pbm/$inputfilter*.pbm
do
	echo processing $scan to $outputfile
	gocr -i $scan -f UTF8 -l 0 -d -1 -m 130 -p ~/.db/ >> $outputfile
done

python ocerrors.py $outputfile

aspell check $outputfile -lde

edit $outputfile

tif2pbm.sh

#!/bin/bash
# gocr understands only PBM pictures
mkdir tif
mkdir pbm
for tif in *.tif
do
	pbm=${tif%tif}pbm
	echo converting $tif to $pbm ...
	convert $tif pbm/$pbm
done
mv *.tif tif/