Parameters for errors
Recall or Sensitivity =
	Proportion of answers found
Precision =
	Proportion of "true" answers in those found
Specificity
	Proportion of" "true" non-answers in those not found
F-score or F-measure =
	Single measure capturing both precision and recall
	Recall is  times more important than precision
	When  =1, it is the harmonic mean of precision and recall
Accuracy =
	Proportion of objects correctly classified as answers and non-answers
ROC (Receiver Operating Characteristics) Curve: Sensitivity (y-axis) vs. 1 - Specificity (x-axis)
Confusion matrix: Found out by algorithm (predictions) on rows vs "true" answers (actuals) on columns