Machine Learning Extractor Trainer

Designer panel

• Output Folder - The directory where the collected data is stored. Once the data is stored, it can be imported into machine learning training tools.

• Dataset - The dataset where the training data can be uploaded. If the robot is connected to a tenant which has AI Center enabled, you can see all the datasets from AI Center in the dropdown menu and select the folder where to upload the validated documents using the dropdown menu.
• Project - The project where the training data can be uploaded.
  Note: Project and dataset selection are enabled only when connected to Orchestrator. Visit Managing datasets for more information about Public/Private Datasets.

• Dataset ApiKey - The authentication key of the dataset.
• Dataset Endpoint - The endpoint of the dataset where training data can be uploaded. Once a dataset is public, it can be accessed outside UiPath® environment through an endpoint and using API key. Do this if you want to upload datasets to an AI Center instance that you're not connected to (for example in the case of hybrid deployments where the AI Center is on Cloud and the robot is connected to an On premises tenant).

Properties panel

• DisplayName - The display name of the activity.

• Output Folder - The directory where the collected data is stored. Once the data is stored, it can be imported into machine learning training tools.

• Private - If selected, the values of variables and arguments are no longer logged at Verbose level.

• Dataset ApiKey - The authentication key of the dataset.
• Dataset Endpoint - The endpoint of the dataset where training data can be uploaded. Once a dataset is public, it can be accessed outside UiPath® environment through an endpoint and using API key. Do this if you want to upload datasets to an AI Center instance that you're not connected to (for example in the case of hybrid deployments where the AI Center is on Cloud and the robot is connected to an On premises tenant).

• Dataset - The dataset where the training data can be uploaded. If the robot is connected to a tenant which has AI Center enabled, you can see all the datasets from AI Center in the dropdown menu and select the folder where to upload the validated documents using the dropdown menu.

• Project - The project where the training data can be uploaded.

  Note: Project and dataset selection are enabled only when connected to Orchestrator. Visit Managing datasets for more information about Public/Private Datasets.

• RetryOnFailure - Retry on transient failure. This field only supports Boolean values (True, False). The default value is True.
• Timeout (milliseconds) - Specifies the amount of time (in milliseconds) to wait for a response from the server before an error is thrown. The default value is 100000 milliseconds (100 seconds).

1. Use the Taxonomy Manager Wizard to define your document types and fields.
2. Add a Machine Learning Extractor Trainer into a Train Extractors Scope activity.
3. In the Machine Learning Extractor wizard that automatically opens, enter information for the Endpoint field. You can choose one of the public endpoints. Visit Public endpoints for more information about public endpoints.
4. Select the check box for the Update activity arguments if you wish to also use the entered values as input arguments for the activity, more precisely for the Endpoint.
5. Select Get Capabilities.

   The wizard closes after this operation

6. Enter a value for Output Folder.
7. Select the Configure Extractors option in the Train Extractors Scope.

   A wizard is displayed.

   Figure 1. The Configure Extractors wizard
   

   

   
   
8. The Machine Learning Extractor Trainer is now ready for configuration. Expand the document type that you want to apply it for, and start selecting the fields you want to train, by selecting the checkboxes next to the appropriate fields.
9. Fill in the text boxes either manually or by selecting, from the available dropdown list, the correct data you wish to map to each field. The dropdown list contains all fields that the Machine Learning Extractor Trainer, using the endpoint entered in the Machine Learning Extractor wizard, declares as extraction capability.
   Note: If you select the check box but you leave the text box empty, the latter will be automatically filled in with the Document Type ID from the local taxonomy. The changes apply after saving. Should you want to avoid using a long string for the field ID, we would recommend you to manually enter a value in case you do not have access to the internal taxonomy of the extractor.
10. To check if you are using the latest capabilities of the extractor, you can select the Get or refresh extractor capabilities which opens the Machine Learning Extractor wizard.
11. Selecting one of the options from a dropdown list automatically confirms that field.
12. To train an extractor based on its extraction result, you can set the exact alphanumeric value in the Framework Alias field previously used for an extractor.
13. Select Save once all fields are configured properly.
    Important: You cannot choose the same option for two distinct fields.

Document Understanding Integration