hte 0.0.25

Extracting content from spesific address books

historical-text-extraction (hte)

Package to extract text from historical documents. The package is written for personal use.

Installation

The current release from the PyPI repository:

pip install hte

The development version from GitHub with:

pip install git+ssh://[e-mail]/eirikberger/hte.git

Note that it is nessecary with a SSH key for this approach to work.

Using it

Import the package.

from hte import digitize

The basic setup is the following:

# Define class
book = digitize.Book("data/finnmark_1968.pdf", "books")

# Run methods on the class
book.CreateFolderStructure()
book.PdfImport(page_info=False, from_page=21, to_page=263)
book.Split(multiple_columns=True)
book.RunOCR(type="splits", export_image=False)
book.CombineCleanGroup(ocr_grouping=True, group_type='norway')
book.RegexStructure("norway")

Make sure to install the correct language package for Tesseract.

# Check languages already installed: 
tesseract --list-langs

# Languages available for installation
apt-cache search tesseract-ocr

# Install the Norwegian language pack
sudo apt-get install tesseract-ocr-nor

Extracting headers

Start by converting XML files using the Pascal VOC data format to JSON. These files can be created by using the free software labelImg.

import os 
import hte.headers as hteheaders
os.chdir('/home/eirikb/Desktop')

header = Headers('train', 2022)
header.runbbxConverting()

Then convert the json file to csv.

hteheaders.convertFromJson('/home/eirikb/Desktop/training_xml/json-bbox', 'xml')

Finally, read the concent of the boxes.

hteheaders.ReadBoxes('json-bbox/xml.csv', 'hordaland', 'train', print_images=True)

Downloade more content

These functions create a list of relevant content from the API of the Norwegian National Library and then downloads a high resolution pdf version.

import os
import hte.NB as NB
os.chdir('/home/eirikb/Desktop/')

ListOfBooks = NB.ListNB(1920, 1930, 'skatteligning', 'digitidsskrift')
ListOfBooks = NB.ListOfBooks[0:1]