An Enhanced Automated Test Item Creation Based on Learners Preferred Concept Space

Recently, research has become increasingly interested in developing tools that are able to automatically create test items out of text-based learning contents. Such tools might not only support instructors in creating tests or exams but also learners in self-assessing their learning progress. This paper presents an enhanced automatic question-creation tool (EAQC) that has been recently developed. EAQC extracts the most important key phrases (concepts) out of a textual learning content and automatically creates test items based on these concepts. Moreover, this paper discusses two studies for the evaluation of EAQC application in real learning settings. The first study showed that concepts extracted by the EAQC often but not always reflect the concepts extracted by learners. Learners typically extracted fewer concepts than the EAQC and there was a great inter-individual variation between learners with regard to which concepts they experienced as relevant. Accordingly, the second study investigated whether the functionality of the EAQC can be improved in a way that valid test items are created if the tool was fed with concepts provided by learners. The results showed that the quality of semi-automated creation of test items were satisfactory. Moreover, this depicts the EAQC flexibility in adapting its workflow to the individual needs of the learners. Keywords—Automated Assessment; Automatic Test-Item Creation; Self-Regulated Learning; Evaluation of CAL systems; Pedagogical issues; Natural-Language Processing