Using Grammar Checkers in an ESL Context: An Investigation of Automatic Corrective Feedback

Our study examines written corrective feedback generated by two online grammar  checkers (GCs), Grammarly and Virtual Writing Tutor, and by the grammar  checking function of Microsoft Word. We tested the technology on a wide  range of grammatical error types from two sources: a set of authentic ESL compositions  and a series of simple sentences we generated ourselves. The GCs were  evaluated in terms of (1) coverage (number of errors flagged), (2) appropriacy of  proposed replacement forms, and (3) rates of “false alarms” (forms mistakenly  flagged as incorrect). Although Grammarly and Virtual Writing Tutor outperformed  Microsoft Word, neither of the online GCs had high rates of overall coverage  (<50%). Consequently, they cannot be relied on to supply comprehensive  feedback on student compositions. The finding of higher identification rates for  errors from simple rather than authentic sentences reinforces this conclusion.  Nonetheless, since few inaccurate replacement forms and false alarms were  observed, only rarely is the feedback actively misleading. In addition, the GCs  were better at handling some error types than others. Ultimately, we suggest  that teachers use GCs with specially designed classroom activities that target  selected error types before learners apply the technology to their own writing.