Toward the Automated Measurement of Teacher Talk: Introducing a Comprehensive Framework for Teacher Discourse

• In Event: 34.094 - Fairmont Royal York Roundtable Session Six
  In Roundtable Session: 34.094-1 - Online and Computer-Aided Assessment

Abstract

This study analyzes a new coding framework, based on automatic speech recognition transcripts, for assessing the quality of teacher discourse. Two samples of teacher observation data were used to test our framework: archival data originally analyzed in Author et al. (2018) and a newly collected sample of 127 classroom observations from 16 teachers in western Pennsylvania. To test the viability of the new system, we assessed the central tendency and variability of discourse features, the inter-rater reliability of expert coders, and covariance structure of teacher discourse. We confirmed the viability of identifying dialogic discourse features. In addition, several new features that hold promise for observational study, largely orthogonal to dialogic discourse, were identified.

Authors

• Sean P. Kelly, University of Pittsburgh

• Amanda J. Godley, University of Pittsburgh

• Sarah Capello, Judson University

• Meghan Dale, University of Pittsburgh

• Patrick Donnelly, California State University - Chico

• Sidney K. D'Mello, University of Notre Dame