Classification of Classroom Teachers’ Speech Intention Based on Deep Learning

Teachers use language to guide classroom teaching activities. The automatic classification of teacher speech according to intention is helpful for the quantitative analysis and evaluation of classroom teaching process. Teachers’ speech in real classroom teaching of middle school Chinese and mathematics is used to construct a corpus, and deep convolutional neural network (CNN) is trained to classify teachers’ speech and identify three kinds of teacher-led teaching activities, including teaching, questioning and classroom management. The experimental data show that:(1) compared with the classical shallow network classification algorithm SVM, the classification accuracy of CNN is increased by 10% to 95.5%, which can meet the requirements for accuracy of automatic analysis of classroom teaching process; (2) Classifying and statistical analysis of classroom teaching behaviors by using CNN classification algorithm can provide useful ideas for classroom analysis and research.