Better Evaluation for Grammatical Error Correction

We present a novel method for evaluating grammatical error correction. The core of our method, which we call MaxMatch (M), is an algorithm for efficiently computing the sequence of phrase-level edits between a source sentence and a system hypothesis that achieves the highest overlap with the goldstandard annotation. This optimal edit sequence is subsequently scored using F1 measure. We test our M scorer on the Helping Our Own (HOO) shared task data and show that our method results in more accurate evaluation for grammatical error correction.