ABSTRACT
This chapter explores the issues involved in implementing web-based language tests using automated short-answer scoring, specifically automated scoring using keyword or regular expression matching. The chapter begins by contrasting this approach with other approaches to the automated scoring of limited production tasks. It then provides an overview of how to implement this type of automated scoring, touching upon how to develop scoring keys, how a scoring engine performs regular expression matching, the suitability of this approach for various task formats and constructs, challenges to regular expression matching, test delivery options, and requirements for exporting data from the delivery system. It concludes by discussing reliability and validity issues, starting first with some implications of this approach for construct definitions, and then discussing steps in the scoring key authoring process that can maximize the reliability of scores.
