Machine learning extraction trainer

Hello,

I’m trying to figure out how I can use the ML Extraction Trainer activity together with the Data manager in the UI Center.

Also, is there a solution to do it automatically, through Uipath Studio? For example, the exported data from the ML Extraction Trainer should be automatically uploaded to the Data Manager.

Many thanks,

Hey, @DragosPadurariu,

Next week we will release a new preview version of the IntelligentOCR activities pack and it will contain an improvement to the ML Extractor Trainer which will allow the trained data to automatically be uploaded in a dataset from AI Center.
The steps to train a ML skill in AI Center using the ML Extractor trainer are:

  • You create a workflow with the DU framework and using the Validation Station you set the corresponding values to the fields you wish to extract
  • The output of the Validation Station is fed in the Train Extractors Scope (which should contain the ML Extractor Trainer). After this step is run, you should see the data uploaded in the dataset
  • Then, in AI Center, you import the dataset in Data Manager and then export it - a new dataset is created, this time with the format needed by the Train pipeline
  • Finally, you go to the Pipelines page, create a training pipeline for your model using the above dataset

Hope this helps,
Adi

2 Likes

Finally…! :star_struck: :partying_face:

@AdiPopa
Thank you Adrian for your answer !

Do you have any news regarding this new release ?

Thank you !

@DragosPadurariu,
The preview release is out since last week! :partying_face:
Read all about the new features here: Document Understanding - 21.6 Community Preview

1 Like

This new " Exporting ML Extractor Training Data into a Dataset" is described in more detail here (The Auto-retraining Loop (Preview)), correct?

Is this only available in the Community Edition or also Enterprise Edition (if not, when) or is there no difference?

I have the following packages installed (UiPath.DocumentUnderstanding.ML.Activities 1.7.0, UiPath.IntelligentOCR.Activities 4.13.2) but the new option as shown in videos from UiPath (DevCon) or in the documentation aboive are not showing up for me?!

Here’s how the activity looks for me:
grafik

But this is what i see in the docu:

Any suggestions what i need to change to have the new option available?

One more question: In this DevCon presentation (UiPath DevCon 2021: AI Enabled Document Understanding - Latest Updates and Data Manager Demo - YouTube) the scheduled export files option in AI Center data labelling session shows dates (calender):
grafik

For me, dates are not showing up but only a start time. Is this updated not launched yet?

grafik

In order to enable the upload into an AI Center dataset you need to load the UiPath.DocumentUnderstanding.ML.Activities 1.8.0-preview into your automation project.

1 Like