Computer Assisted Assessment (CAA) of Free-Text: Literature Review and the Specification of an Alternative CAA System

Abstract

So far there are a number of computer assisted assessment approaches that are based on a variety of features. However, those approaches exploit the whole set of training documents in order to assess a provided free-text answer against a given question. Recent text classification approaches are orientated to mine average class documents and consequently they provide cheap classification methods that compute similarities between new unclassified documents and the average class documents. The same idea can be applied for scoring free-text answers. In this case the methodology should be able to mine one average free-text (reference) answer for each given question.