When you make a second data labeling session, you have to re-create the fields that you want to label. If I make a new data labeling session, and add the same labels I have from another data labeling session that I used to train a model, can I use the new data labeling session to train the document understanding model more that I had trained on the preexisting session?
I want to do this because I noticed that recently, UiPath updated document understanding so that new values on the document are extractable.
Yes, you can use the new data labeling session to further train your existing Document Understanding model in UiPath. First, create a new data labeling session and re-create the fields from your previous session. Then, label new documents with the same fields. After that, train a new version of your model using both the old and new labeled datasets. Since UiPath has updated Document Understanding, retraining your model will help improve its accuracy and ensure better extraction of the newly available values.
If you found this helpful, please mark it as a solution.
Also as per recommendation as well it is always good to retrain the full dataset rather than incremental trainings…so you can go with new session with all labels and train the base model again
In a perfect world I would retrain from the base model every time. Sadly, AI units are so expensive that anything that cuts down on their use is helpful.