期刊名称：Journal of Systemics, Cybernetics and Informatics
印刷版ISSN：1690-4532
电子版ISSN：1690-4524
出版年度：2017
卷号：15
期号：7
页码：77-84
出版社：International Institute of Informatics and Cybernetics
摘要：There must be as many concrete indicators as possible ineducation, which will become signposts. People will not beconfident about their learning and will become confused withtenuous instruction. It is necessary to clarify what they can doand what kinds of abilities they can improve. This paperdescribes a case of evidence-based education that acquireseducational data from students’ study activities and not onlyuses the data to enable instructors to check the students’ levelsof understanding but also improve their levels of performance.Our previous research called discussion mining was specificallyused to collect various data on meetings (statements and theirrelationships, presentation materials such as slides, audio andvideo, and participants’ evaluations of statements). This paperfocuses on student presentations and discussions in laboratoryseminars that are closely related to their research activities inwriting their theses. We propose a system that supports tasks tobe achieved in research activities and a machine-learningmethod to make the system sustainable for long-term operationby automatically extracting essential tasks. We conductedparticipant-based experiments that involved students andcomputer-simulation-based experiments to evaluate howefficiently our proposed machine-learning method updated thetask extraction model. We confirmed from the participantbasedexperiments that informing responsible students of tasksthat were automatically extracted on the system we developedimproved their awareness of the tasks. Here, we also explainimprovements in extraction accuracy and reductions in labelingcosts with our method and how we confirmed its effectivenessthrough computer simulations.
