icris-ocr is a Python automation tool for extracting information from documents purchased from the Hong Kong government's Integrated Companies Registry Information System. It provides high-level abstractions in the form of classes to make extraction of data from multiple documents convenient.

It provides utilities for each part of the OCR process--from PDF conversion and document classification to edge detection and data cleaning. Classes representing documents have been defined in the document_layouts module. The subpackage document_processing is composed of modules aimed at preparing documents, processing them, and cleaning the data extracted; these utilities are provided by the doc_utils, ocr_tools, and string_processing modules respectively.

Currently, it provides utilities for annual return documents. Support for other documents can be added with the optimized functions defined in document_layouts.ocr_tools.

The package relies on Google's Tesseract engine and its Chinese character recognition data set. See installation instructions for the Tesseract engine here. After installation of the Tesseract engine on your machine, download the Chinese character recognition data set here and add it to your tessdata directory, which can be found in the directory for the base Tesseract engine.

The package itself, along with its dependencies, can be installed by following the instructions listed here.

Documents can be converted and categorized with the convert_and_categorize function in the document_processing.document_preparation module. The following commands convert unique PDF files in the target directory to a directory of images of each page of a file grouped in a single directory. Then, it categorizes those directories based on the title of the document on the first page. This results in another directory being created at the level of the target directory and contains categorical directories for different document types.

>>> from icris_ocr.document_processing import document_preparation
>>> document_directory = 'path/to/directory'
>>> document_preparation.convert_and_categorize(document_directory)

Once categorized, data can be collectively extracted from categorical directories using the process_dir function. The function returns a Pandas DataFrame object containing information about every document contained in the categorical directory and can be used for further analysis.

>>> from icris_ocr import *
>>> directory = 'path/to/directory'
>>> df = process_dir(directory, doc_type='Annual Return', parallel=True)
>>> df.to_excel('OCR Results.xlsx') # Write to Excel file

The same result can be achieved from the command line.

$ python -m icris_ocr path/to/directory -t Annual\ Return -p

Despite the autor's optimization efforts, the computational nature of the project makes the process very time consuming. The project can be implemented in a lower language such as C in the future for improved performance.

This project makes liberal usage of the algorithms shared by Doctor Adrian Rosebrock on his website.

This project has been distributed under the MIT license.