Code for creating an efficient parallel workflow for converting and OCR-ing a large number of text document to text
The original lawsuit scanned pdfs are downloaded by the download
scripts into the
folder
<DATADIR>/CaseFiles. The individual lawsuits are spread in the
folder structure as
<DATADIR>/CaseFiles/State/<county>/<year>/<casenr>/, and all the
original scans for a particular lawsuit are in this folder.
Both the OCR and address extraction scripts assume the files are layd
out like this. It also permits a
special county, Test, to be located among counties. Everything in
the Test county folder is ignored by the ocr script.
The OCR script mirrors the same structure for the ocr-d texts in
folders like
<DATADIR>/txt/State/County/year/casenr/.