Research on the Construction of Math Teachers' Classroom Speech Portraits in Junior High School Based on Deep Learning
Studies show that teaching behavior based on teachers'speech account for about 80％ of all the behaviors in the classroom, indicating teachers'speech as main carrier of classroom teaching activities. Continuous development of big data and artificial intel-ligence promotes deep learning application in the field of education. This article used deep learning and knowledge graph technology to analyze teachers'classroom speech. Firstly, eight dimensions of junior high school mathematics teachers'classroom speech were constructed. For the first seven dimensions, two models of Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) based on deep learning algorithms were selected to classify more than 10000 teachers' speech texts generated from 70 junior high school mathematics lessons. Results showed that the accuracy of the CNN model and RNN model was 83.09％ and 80.26％. For the last dimension, knowledge graph technology was used to extract sentences with antecedent knowledge points in teacher'classroom speech. Then the above technologies were combined to generate 6 teachers' classroom speech portraits. By comparing the accuracy of CNN model and RNN model, method of deep learning algorithm and knowledge graph technology to classify teachers'classroom speech was shown to be feasible. Meanwhile, the generation of portraits provided a strong reference to standardize teachers'classroom speech.
万方数据
维普
