Generation of Fill-in-the-Blank Questions from Concept Map and Preliminary Comparison between Multiple-Choice Task and Kit-Build Task

Abstract

In previous research, we proposed a framework for a kit-build concept map (KB map), in which a learner is provided a set of components with which to build a concept map. In that framework, the learner's constructed map can be diagnosed automatically. The task of building a concept map out of provided components is a promising exercise for strengthening and assessing a learner's comprehension of a learned topic. In general, the multiple-choice task of fill-in-the-blank (FIB) questions can also be used to strengthen and assess a learner's comprehension, and the answers can be automatically evaluated. In this paper, we discuss our process for designing a set of FIB questions that can be generated from a concept map. Then, we compare the KB map task with the FIB multiple choice task. Both tasks can be generated from the same concept map: that is, from the same content. We compare the two tasks, using three science lessons for fifth graders in two classes. One class uses the KB map and the other uses the FIB questions. In this experiment, the KB task evaluated the learners' comprehension more adequately than did the FIB question task, and the students in the KB class performed at a higher level than did the students in the FIB class.