ABSTRACT

Machine scoring of constructed responses, both text and speech, is an emerging subdomain in the field of computational linguistics that has strong application in the world of assessment. This field was first developed in the area of automated essay scoring (Ajay, Tillett & Page, 1973; Page, 1966), expanded to short-form constructed responses (Burstein, Wolff & Lu, 1999; Kaplan & Bennett, 1994; Leacock & Chodorow, 2003; Martinez & Bennett, 1992) and more recently extended to the evaluation of speech (Bernstein, 1999; Xi, Higgins, Zechner & Williamson, 2008). It also includes the evaluation of constructed responses to a performance assessment (e.g., making a medical diagnosis; Clauser, Kane & Swanson, 2002).