Knowledge-Based Characterization of Test Questions

Abstract

The recent advances in knowledge engineering entail us to represent knowledge associated with a course in an expressive yet computable format as a hierarchical prerequisite relation-based weighted ontology. A schema called the course concept dependency schema written in Web ontology language (OWL) is designed to represent the prerequisite concept dependency. The knowledge associated with educational resources, like the knowledge required for answering a particular test question correctly, can be mapped to subgraphs in the course ontology. A novel approach for selectively extracting these subgraphs is given and some interesting inferences are made by observing the clustering of knowledge associated with test questions. We argue that the difficulty of a question is not only dependent on the knowledge it tests but also the structure of the knowledge it tests. Some assessment parameters are defined to quantify these properties of the knowledge associated with a test question. It is observed that the parameters are very good indicators of question difficulty.