TOWARDS GENERATING EXERCISE QUESTIONS WITH LOD FOR WEB-BASED INVESTIGATIVE LEARNING

Web allows learners to investigate any question to learn with a large number of Web resources. In such investigative learning, leaners are expected to investigate the question by navigating Web resources/pages to construct their knowledge and decomposing the question into sub-questions. In acquiring skills in such investigative learning, learners need to practice with exercise questions. However, it is hard to define correct knowledge to be constructed for the questions since Web-based investigative learning could result in diverse knowledge as correct one for the same question. Towards this issue, this paper proposes the method with Linked Open Data (LOD) for generating exercise questions, which includes an initial question and sub-questions to be decomposed from an initial question. These sub-questions are extracted and selected as keywords with LOD and Word2vec. This paper also reports a case study with the generation method. The results suggest that it is effective as scaffolding particularly for novice learners.