Active learning

A technology used in machine learning training whereby the training documents that the interim statistical model is least confident in its predictions, are given highest priority to annotators to annotate, thus presenting the model with a more varied set of training documents up front.