Automatic classification of activities in classroom discourse

Classroom discourse is the primary medium through which teaching and learning occur. Managed skillfully, it can provide an opportunity for students to develop their understanding and to profit from the ideas of their peers and the teacher. Yet it is difficult for teachers to be mindful of the nature and distribution of classroom discourse at the same time as they juggle other instructional concerns. It is possible to record, transcribe, and analyze classroom discourse, but it is not possible to do this quickly enough to give a teacher timely feedback. We report on the development and validation of an automated system for recording and analyzing aspects of classroom discourse that can result in timely feedback. Based on the LENA system, it aims to identify three common discourse activities: teacher lecturing, whole class discussion and student group work. The system consists of a speech processing module (diarisation performed by the LENA system) and an activity detection module that detects the discourse activities by using classification analysis. Results showed that our automatic detection of discourse activities converged well with those of human coders. The system enables timely and relatively inexpensive generation of a classroom discourse profile, which helps teachers to visualize and potentially improve their classroom discourse management skills.