This commit is contained in:
Patrick Jentsch
2019-04-14 14:33:40 +02:00
parent 9f3c71a118
commit 0d3efe167e
2 changed files with 87 additions and 74 deletions

View File

@ -4,6 +4,7 @@
import xml.etree.ElementTree as ET
from xml.sax.saxutils import escape
import os
import re
import sys
input_files = filter(lambda x: x.endswith(".hocr"), sorted(os.listdir(sys.argv[1])))
@ -25,7 +26,8 @@ output_file.write('<?xml version="1.0" encoding="UTF-8"?>\n' +
for input_file in input_files:
tree = ET.parse(os.path.join(sys.argv[1], input_file))
output_file.write(' <pb n="%s"/>\n' % (input_file.split(".")[0]))
page_number = int(re.search(r'\d+', input_file.split(".")[0]).group(0))
output_file.write(' <pb n="%i"/>\n' % (page_number))
for para in tree.findall(".//*[@class='ocr_par']"):
output_file.write(' <p>\n')
for line in para.findall(".//*[@class='ocr_line']"):
@ -41,4 +43,4 @@ for input_file in input_files:
output_file.write(' </body>\n' +
' </text>\n' +
'</TEI>')
output_file.close()
output_file.close()