The measures used to evaluate the checker are the measures of precision and recall. These measures indicate:

- how often is grammatical incorrectness rejected (precision of a grammar checker)
- how often is grammatical correctness accepted (recall of a grammar checker)

Precision measures the percentage of grammatical errors correctly detected by the grammar checker. Recall measures the percentage of the grammatically correct items accepted by the grammar checker. At the same time, the recall percentage may be used to calculate the percentage of false flaggings, i.e. grammatically correct items rejected by the grammar checker: The percentage of false flaggings is obtained by subtracting the recall percentage from 100.

The two measures complement each other and are both necessary in the evaluation of a grammar checker. To know whether the checker detects the errors that we test for without knowing whether it in fact finds errors in all the input sentences is not useful. Similarly, to know whether it accepts all the correct input sentences without knowing whether it accepts all the input sentences, incorrect as well as correct ones, is useless.

ceditor@tnos.ilc.pi.cnr.it