Automatic Classification of Discourse in Chinese Classroom Based on Multi-feature Fusion

In the analysis of classroom teaching behaviors, the current systems of quantitative analysis mainly segment, identify, classify and code classroom teaching behaviors manually, including language behaviors, without using automatic methods. Therefore, this paper applied text classification method to classroom discourse analysis to optimize and improve the classification and coding process. In this paper, a feature extraction method fusing TF-IDF feature with word2vec word vector feature was proposed based on the chi-square test for feature selection of the three types of teacher discourse, namely lecture, instruction and question in class. SVM was used as a classifier to realize the automatic classification of teacher discourse in class. The classification accuracy was up to 86.07%.