Supervised learning: How to persist data definition with activity UiPath.IntelligentOCR.Activities.ValidationStation.PresentValidationStation

Hi All,
I’m working on the following workflow:

Firstly, I’ve defined the taxonomies using the Taxonomy Manager successfully.

I run the workflow, I define the fields with the information to extract, and the information is extracted successfully. But everyTime I run the workflow, I have to define again the fields to extract the key information (invoice number, total amount, …). How can I export the ExtractionResult data (obtained on activity PresentValidationStation) to be reused when I start the workflow again?

Best regards,

Alberto G. Rivas

1 Like

Check the following post

Cheers

Thanks for your reply J0ska.
We’ve seen that post and that doesn’t seems complete (need activities to get all the functionality). We want to know what are the extractor activities that can learn (to train them, trainable) and using in local to get the behaivior that we need. Or maybe there is another way to get some supervised learning in the UiPath platform?

Hi @alberto.gonzalez
I understand there is no trainable extractor available from UiParh at this moment.
In case you want a trainable solution you will probably need to point your attention to Abbyy Flexi Capture. But I have no experience with this.

Cheers

Hello @alberto.gonzalez,
It seems all you do in the workflow is

  • digitize document
  • present validation station for human intervention.

If you want to perform automatic classification, try the Classify Document Scope with the Keyword Based Classifier or the FlexiCapture Classifier (your choice). If you want to train the Keyword Based Classifier, use the trainer in the Train Classifeirs Scope activity.

If you want to perform data extraction, try the Extract Data Scope with the available extractors (FlexiCapture, Regex Based for now, more to come). Currently no out-of-the-box extractors have training capabilities, but we are working on providing training capabilities for the machine learning extractor.

We will keep you posted when we will have more updates on OOB learning capabilities :slight_smile:

1 Like

Thanks for your answer.

@Ioana_Gligan, Do you know when is UiPath going to publish extractors with training capabilities?

Thank you

1 Like

I can say we are working on it :slight_smile:

1 Like

Hi @Ioana_Gligan,

I have a question related to Keyword Based Classifier Trainers. I posted an update in another thread and therefore won’t repeat it here. Can you please help me figure out why the trainer isn’t building the JSON learning file using the results from the Present Validation station?

Thanks!
Andy

Hello @AndyMenon,

Can you please check you used the configure classifiers wizard and activated the learning?

Ioana

How do I activate learning? This is how I have configured the classifier. The keyword.json file contains keywords as shown below.

image

image

I have multiple taxonomies and keywords working the way they should and the documents are being processed as expected.

thanks.

Hi @Ioana_Gligan,

This might sound crazy - I went back to take a look at my RPA process once more to see if I had missed something. Instead it looks like the Trainer has worked!

But I did not do anything except run the process multiple times and process at least three different documents! :confused:

Here is my trainer classifier and I see that it has indeed added a few more keywords by itself!

image

This is a timestamp trace of the last tests I ran. All of the files including the learning.json have been modified between 3:51 PM and 4:01 PM which is the duration of my last test.

Is there a certain number or variety of documents I must give to the process before it triggers the learning ?

image

Thanks again!