Analyzing Speech Data to Detect Work Environment in Group Activities

ABSTRACT

Collaboration is one required skill for the future workforce that requires constant practice and evaluation. However, students often lack formative feedback and support for their collaboration skills during their formal learning. Current technologies for emergent learning due to COVID-19 could make visible digital traces of collaboration to support timely feedback. This work aims to automatically detect the group work environment using speech data captured during group activities. Grounded in literature and students’ perspectives, this work defines and implements three indicators for detecting the work environment namely noise, silence and speech time. Three experts rated two hundred thirty-two video instances lasting 30-secs each to get a group work environment score. We report the results of two machine learning models for detecting the group work environment and briefly reflect on these results. © 2022, Springer Nature Switzerland AG.