companies_house_ocr_runner

Runs Tesseract OCR over PDF paper filings from Companies House.

This project is tightly coupled to a specific deployment scenario. It also makes assumptions about what data is available in the environment.

For running in another environment it would probably be simplest to review how this works and then create a similar application.

To learn how to call Tesseract from the command line in an efficient way see: the run_ocr function in tesseract_wrapper.

NOTE: This project was created as a prototype

Running locally

To run locally:

• Install Vagrant/VirtualBox
• Then:

  vagrant up
  vagrant ssh
  vagrant@ubuntu-bionic:~$ ch_ocr_runner

Environment variables

Batch allocation is done with an environment variable:

CH_OCR_MACHINE_ID="UNIQUE_NAME"

To maximise CPU usage for batch processing we start multiple Tesseract processes, but restrict each one to use a single processor by setting:

OMP_THREAD_LIMIT = 1

Stop/start running

Each batch gets a directory in the working area. When a batch is finished a batch lock file is created. As long as that file is present future runs won't repeat that batch.