04. Active learning

While you are waiting for your dataset to get labeled and constructed, you can use active learning.

Active learning -> focus on a shorter training with samples from the "hard" instances to create a temporary model or permanent where good enough.