Detecting and categorizing indices in lecture video using supervised machine learning

Abstract

This work reports on the evaluation of detecting scene transitions in lecture video through supervised machine learning. It expands on previous work by gathering training data from multiple human raters. We include a robust evaluation that compares predictions against the entire set of expert classifications in disagreement. Finally, we explore some of the issues around constructing training data from multiple human experts, specifically emphasizing that evaluation strategies should be carefully considered when using aggregated training data. © 2013 Springer-Verlag.