Title: Knowledge structures and their applications in CALL
Authors: Ivo Düntsch , Dept of Computer Science , Brock University , St Catherines, Ontario, L2S 3A1, Canada
Günther Gediga , Institut für Evaluation und Marktanalysen; Brinkstr. 19; D-49143 Jeggen; Germany
(Equal authorship implied)
Status: In: Language Teaching and Language Technology , Sake Jager, John Nerbonne and Arthur van Essen (eds.) (1998), 177 - 186
Abstract: The paper describes two approaches towards knowledge assessment which can be applied within any CALL application. The first approach uses test items to define a universe of empirical knowledge states. We show that the relations between items (`are about equal', `is harder than') and subjects (`are about equal', `is better than') can be easily computed within a CALL system, and that the empirical results give insight into the structure of problems and their solutions. Given a standardized test procedure, a CALL system can use this method to describe the knowledge state of a tested subject either using a reference population or a knowledge structure generated by experts (see below), or by computing a knowledge structure using the given sample.

The second approach is based on querying four experts about the skills minimally necessary to solve problems of the item set of the first approach. The results of these queries can be transformed into theoretical knowledge spaces, and the empirical data can be used to test the prediction of the experts. The results show that %

  • The experts disagree remarkably,
  • Most of the experts did not express even simple relationships which are observable by the empirical data analysis,
  • Situational factors such as the appearance of learning tasks in the same lecture are observable in the empirical data as well as in the opinions of the experts.

View technical report version