Abstract :
A natural language based system has been used to author and mark shortanswer
free-text assessment tasks. Students attempt the questions online and
are given tailored and relatively detailed feedback on incorrect and incomplete
responses, and have the opportunity to repeat the task immediately so as to
learn from the feedback provided. The answer matching has been developed in
the light of student responses to the questions. A small number of the questions
are nowin low-stakes summative use, alongside other e-assessment tasks
and tutor-marked assignments, to give students instantaneous feedback on
constructed response items, to help them to monitor their progress and to
encourage dialogue with their tutor. The answer matching has been demonstrated
to be of similar or greater accuracy than specialist human markers.
Students have been observed attempting the questions and have been seen to
respond in differing ways to both the questions themselves and the feedback
provided.We discuss features of appropriate items for assessment of this type.
