Greetings!
I recently had wonderful success with tesseract-ocr on grand jury transcripts but now have a harder problem.
Can tesseract be trained to recognize censoring blocks in text? For example:
Assume this sentence has XXXXXXXXXXXXX a censoring block that obscures all the text it covers. (here represented by the X's, in the text, it is a solid black line)
What I want to do, in addition to recognizing the surrounding text, is to train tesseract to substitute for the black mark, (redaction - N) where N is the length of the redaction.
There aren't that many different sized redactions, well, probably from one character space or a little better up to an entire line so producing examples of all the blackouts would be tedious but not difficult.
Is that pushing tesseract in a direction it is not meant to go?
If so, any suggestions on software that might be better suited to the task?
Thanks!
Patrick