Getting different confidence for the same field and same document in Document understanding extraction result

Hello friends…

i have two workflow, when i run the first process and check for the confidence of a field it gives me 98%.

when i run the second process for the same document using same ml skill the confidence level is low.

Any one know what may be the proble for this.

below are the extraction results.
ResultsValue { Components=ResultsDataPoint[0] { }, Confidence=0.9894537, DerivedFields=ResultsDerivedField[0] { }, OcrConfidence=1, OperatorConfirmed=false, Reference=ResultsContentReference { TextLength=7, TextStartIndex=1782, Tokens=ResultsValueTokens[1] { ResultsValueTokens { Boxes=Box[1] { T:284.601,L:426.7682,W:25.0726,H:6.4015 }, Page=0, PageHeight=841, PageWidth=595, TextLength=7, TextStartIndex=1782 } } }, TextType=Unknown, UnformattedValue=“15814-1”, Value=“15814-1” }

ResultsValue { Components=ResultsDataPoint[0] { }, Confidence=0.82645494, DerivedFields=ResultsDerivedField[0] { }, OcrConfidence=1, OperatorConfirmed=false, Reference=ResultsContentReference { TextLength=7, TextStartIndex=781, Tokens=ResultsValueTokens[1] { ResultsValueTokens { Boxes=Box[1] { T:283.2674,L:426.7682,W:25.6061,H:9.869 }, Page=0, PageHeight=841, PageWidth=595, TextLength=7, TextStartIndex=781 } } }, TextType=Unknown, UnformattedValue=“15814-1”, Value=“15814-1” }


The difference in confidence levels between the two processes may be due to variations in the input document’s quality or clarity, potentially affecting the ML model’s ability to accurately extract information.


Dear @sai_gupta im using the same document to test

Hi @aslam_ali1 ,

Could you let us know if there is any change in the Workflow mainly the Document Understanding part ? We see that the TextStartIndex property differs in the two extraction results.

Also check if different types were used for the same field in the Taxonomy of the two workflows.