Possible to improve data extracting from document understanding activities?

Train / Retrain the model with more number of sample documents to increase the efficiency if you are using document understanding to increase the accuracy