Classification is most accurate when the system is trained with blank forms. If blank forms are not available, accurate classification can still be achieved.
Pressing Learn Files in the Batch Class Management administrative interface will cause Ephesoft to OCR the sample documents into a HOCR file. This is an HTML representation of the OCR output.
The HTML file can be edited to remove any content that is not part of the blank form. After the HTML file is updated, click on Learn Files again to update the index files used by Ephesoft. This will not overwrite the changes that have been made to the HTML file; that will only happen if the source TIFF is updated.