Hi
I have setup a document understanding framework which works really well.
I am a developer, not a data scientiest , so not overly knowledgable with Machine learning , but get the concept .
I have created a MS SKill using a document understanding out of the box package.
The data I used was created using Data manager in AI Fabric.
I have incorporated this MSSKill into my DU as a Maching Learning Extractor and it works ok .
My question is about retraining the MSSKill .
I have watched the AI FAbric training in Academy , as well as other instructional videos I have found in this forumn.
Can you please advise what is the best approach to take when retraining a machine learning model.
-
Upload file method
Which allows you to upload a csv file to a dataset folder and then re-run your piperline, which will update the MSSKill .
In this event does the skill in studio need to be refreshed and redeployed? -
Machine LEarning Extractor Trainer
Which creates two folders, document and metadata, I believe you zip this and import it into the Datamanager and then re-label manual (i assume) and recreate your data that way . Of course this can only be done if you have 10 or more documents,
Also
Can you setup a model to have structured forms and unstructured forms .
ie Structured forms have 3 pages of information where as unstrucutred forms , have only one page,
destination is the same , but the larger forms just has more options.
Thanks