Selective labeling user guide

Selective labeling helps with recommendations on which documents to label. You can create diverse training and test datasets to train representative models.

Getting suggested documents

  1. Create a CDE processor and import documents.

    • At least 100 are required for training (25 for testing).
    • Once sufficient documents are imported and after selective labeling, the information bar should appear.

  2. In case of a CDE processor with 0 suggested documents, import more to have sufficient documents in either split for sampling.

    • This should enable Get suggested documents in Suggested category. You should be able to request suggested documents manually from the UI.
    • There's a new filter on top to filter out suggested documents.

Labeling suggested documents

  1. Go to Suggested category on the left-hand label list panel. Start labeling these documents.

  2. Select Review now on the bar when you have suggested documents in the processor to navigate to. Start labeling.

  3. Select Auto-label on the information bar if the processor is in a warm state. Label suggested documents.

Train after labeling all suggested documents

  1. Move to Train now on the information bar. When the suggested documents are labeled, you should see the following information bar recommending training.

Supported features and limitations

Feature Description Supported
Support for old processors Might not work well with old processors with previously imported dataset

Feedback

To submit feedback, bugs or feature requests, use this form