Train a use case
Train the document classification use case with user input from completed document tasks to improve Document Intelligence recommendations over time.
Avant de commencer
There must be at least one reviewed document task associated with the use case to train it. Across all reviewed tasks, there must be at least two attachments for each type you defined in your use case.
Begin by creating a document task and completing it using the Document Intelligence workspace.
Remarque :You won’t be able to process the task until the use case is trained, but you can complete it by completing all the fields and submitting it.- Role required: sn_docintel.manager
Pourquoi et quand exécuter cette tâche
Document classification use cases don’t begin with pre-trained AI models, so it’s important to train the models with user input from completed document tasks.
Procédure
Résultats
The train use case job begins. This job may take several hours to complete.
Test a use case
The following example shows how to train a document classification use case to process identification documents.
-
You start by creating a document classification use case called Identification. You create a field for each type of category you want to apply to your identification documents.
In this case, there are fields for a Drivers license, Passport, Military ID, and ID card.
-
Then you create document tasks containing attached examples of the categories you defined in the fields.
Conseil :You should submit at least 50 documents to adequately train the AI model. -
You complete the tasks by opening the documents in Document Intelligence workspace, completing all the fields, and submitting them.
- Select each document task to open it.
-
Select Show in DocIntel.
- Fill in the category.
-
Select Submit.
The document tasks will be in a Done state and so become available to train the model.
-
You go back to the Identification use case screen and select Train Use Case.