Precision
Measurement
The fraction of positive predictions that are correct against a reference.
Precision equals true positives divided by true positives plus false positives. If positive means a detected annotation error, it asks how many flagged items actually contain errors. Read it alongside recall, which asks how many real errors were found. Define what happens when there are no positive predictions.
In practice
If 80 of 100 flagged annotations contain an error, error detection precision is 80%.