Control the quality of your labeled data with the Python tools you already know.
Not a backward-compatible change:
Improvements:
GoldMajorityVote
true_labels
argument now supports multiple ground truth values for a single task.tol
optimization parameter as a tolerance stopping criteria for iterative methods with a variable number of steps.entropy_threshold
methodfit
on_missing_skill
and default_skill
params for models that use skillscrowd-kit==0.0.6 release