摘要：We propose a system that allows university instructors to check teaching behaviors in their lecture videos for lecture improvement. The system offers two functions: time-series graphing, which visualizes real-time changes in student evaluations during a lecture, and teaching behavior prediction, which shows instructors information on their own teaching behavior as predicted from student evaluations. We develop the system and conduct experiments to evaluate its functions. Subjective instructor evaluations of each function indicate that (1) the graphing function was useful for identifying portions of lecture videos containing teaching behaviors needing improvement, and (2) that the prediction function was useful for determining teaching behavior tendencies during lectures.
