Showing results for 
Search instead for 
Did you mean: 

Decipher requires "refresh region" to accurately capture characters

Level 3


I am training Decipher on a document that has a 1x3 column of numbers. Characters should all be numerical with the exception of the "-" character occasionally appearing (which I can filter out). 

For whatever reason, Decipher is struggling to read two of the three numbers in this column (the middle and bottom). It consistently transposes digits with letters (ex. "6" read as an "e"). 

I understand that PDF quality can be a factor, but I do find it strange that it does read the first number with consistent accurately. Additionally, Decipher does correct itself and accurately extract the numbers correctly if I refresh the relevant region, so it can read the numbers, it just consistently needs a second or third try via a refresh. 

Is there a misc parameter or formula I can employ that would force the refresh until it meets certain criteria without a HIL? Ex. refresh until all characters are read as numerical or something like that.