Improving large-scale assessment tests by ontology based approach

Knowledge formalized in ontologies can assist Intelligent Tutoring Systems (ITS) in generating question items like multiple choice questions (MCQs), to assess the level of knowledge of a learner. Existing ontology based MCQ generation techniques generate unmanageably large number of questions, but not necessarily all are relevant to the domain. These question items need to be vetted by human experts to choose a suitable subset of questions (as question-set), to conduct an assessment test. Currently, there are no automated methods to achieve this task. We propose three heuristic techniques to choose a desired number of significant MCQs that cover the required knowledge boundaries, from a given ontology. Through experimental results, we show that the question-sets generated based on our approach compare satisfactorily to the ones prepared by domain experts, in terms of precision and recall.