GRASP: the group learning assessment platform

Abstract

We demonstrate a prototype assessment technology designed to enable unobtrusive, real time assessment of group dynamics from speech. As part of that work, we describe a test bed for experimentation with alternative approaches for automatic processing of speech for this purpose. Furthermore, we present a specific successful technique for predicting activity levels and amount of overlapping speech in recordings of actual student group meetings recorded over a semester of a graduate engineering design project course.