Determining-Threshold-in-an-Active-Learning-setting-for-Disambiguation
-
In an active learning setting, the oracle or domain expert is presented with candidate data points for disambiguation.
-
The quality of label deteriorates when the annotator's efficiency reduces.
-
To combat this problem, we determine the threshold to select the candidate points, and update the threshold in each iteration.
-
We propose efficient method for determining the threshold in each iteration for selecting the best candidates.