Analyzing teacher talk using topics inferred by unsupervised modeling from textbooks
We propose a method that automatically describes teacher talk. The method allows us to describe and compare classroom lessons, as well as visualizing changes in teacher discourse throughout the course of a lesson. The proposed method uses a machine learning model to infer topics from school textbooks. Certain topics are related to different contents (e.g. kinematics, solar system, electricity), while others are related to different teaching functions (e.g. explanations, questions, numerical exercises). To describe teacher talk, the machine learning method measures the appearance of the inferred topics throughout each lesson. We apply the proposed method to a collection of transcripts from physics lessons, as well as discussing the potentialities of integrating the proposed method with other kinds of automatic and manual classroom lesson descriptions.