There are two similar values underlined with “Blue” and “Red”.
- Earlier I used datalabelling and trained all the documents for this field on “Blue” field.
- However for one document we found the bot was unable to extract the value from “Blue” field. It is extracting the “35450” part.
- So , we retrained Data Labelling with both the fields and used Post processing “longest value” and scoring “levenshtein”.
- But then the value is coming as “35450 GATX 35450”.
- The catch for all the documents all the values are coming in both “Blue” and “Red” fields but for some the bot is able to extract from “Red” and for some “Blue”. Any fix ?
Try to retrain the Model using UiPath Document Manager by highlighting that Document value and Download that trained document as a data set & using it again create the pipeling and retrain it using the same.
How many variations of the same format has the template been trained to capture the data?
It may be necessary to increase the number of variations to fit the model more closely. The orchestrator allows you to upload up to 20 variations per form. Then you can consume it from the studio.
When to use AI Center and when to use Document Understanding Manager
Extracted from UiPath Document Understanding Overview Course
Are there features overlapping between the Document Understanding service and AI Center?
Yes, some capabilities are available in multiple components, with the UiPath Document Understanding almost always in the picture. For example, you can train a semi-structured AI Extractor in either AI Center or in Document Understanding.
The difference is that the AI Center interface is more appropriate for more technical roles (Data Scientists or Automation Developers), while Document Understanding is meant to be more friendly towards business users. Also, AI Center can be used for projects outside document understanding.
However, whenever in doubt, always start with Document Understanding.