Automatic Classification of Question & Answer Discourse Segments from Teacher's Speech in Classrooms.

Question-answer (Qu0026A) is fundamental for dialogic instruction, an important pedagogical technique based on the free exchange of ideas and open-ended discussion. Automatically detecting Qu0026A is key to providing teachers with feedback on appropriate use of dialogic instructional strategies. In line with this, this paper studies the possibility of automatically detecting segments of Qu0026A in live classrooms based solely on audio recordings of teacher speech. The proposed approach has two steps. First, teacher utterances were automatically detected from the audio stream via an amplitude envelope thresholding-based approach. Second, supervised classifiers were trained on speech-silence patterns derived from the teacher utterances. The best models were able to detect Qu0026A segments in windows of 90 seconds with an AUC (Area Under the Receiver Operating Characteristic Curve) of 0.78 in a manner that generalizes to new classes. Implications of the findings for automatic coding of classroom discourse are discussed.