Text Data vs Text Data revised

[D Alemán]

Hello,

I am working with a project that use Tesseract for invoices reading. I have text data and data revised. 

My goal is that less and less data  need to be revised. How I can model the data output of Tesseract to be corrected automatically? (using a model that previously learn certain patterns)