Improved Correction Detection in Revised ESL Sentences

Abstract

This work explores methods of automatically detecting corrections of individual mistakes in sentence revisions for ESL students. We have trained a classifier that specializes in determining whether consecutive basic-edits (word insertions, deletions, substitutions) address the same mistake. Experimental result shows that the proposed system achieves an F1-score of 81% on correction detection and 66% for the overall system, out-performing the baseline by a large margin.