User permissions required: 'View Sources' AND 'Review and label'
Please Note: 'Teach label' is now a training mode solely for labelling unreviewed verbatims and as such the reviewed filter is disabled in this mode. 'Teach label' for reviewed verbatims has been split into the 'Check label' and 'Missed label' training modes (see here).
If you have a label that is struggling to predict accurately, and you're happy with the consistency of the already pinned examples (as discussed in the previous article), then it is likely that you need to provide the model with more varied (and consistent) training examples.
The platform will typically suggest this mode as a recommended action for labels that would benefit from it the most under the Model Rating factors, as well as in the recommended actions for specific labels that you can select in Validation.
The best method for training the platform on the instances where it struggles to predict whether a label applies or not, is using 'Teach' for unreviewed verbatims.
As this mode shows you predictions for a label with confidence scores ranging outwards from 50% (or 66% in the case of a sentiment-enabled dataset), accepting or correcting these predictions sends much more powerful training signals to the model than if you were to accept predictions with confidence scores of 90% or more. In this way, you can quickly improve the performance of a label by providing varied training examples that the platform was previously unsure about.
The actual process of labelling in this mode is discussed in the Explore phase here.
Previous: Training using 'Check label' & 'Missed label' | Next: Understanding and improving coverage