Research On Classification Of Teachers’ Classroom Questioning Based On Deep Integrated Learning

In recent years,with the development of the Internet,more and more online education platforms and MOOC courses have sprung up,resulting in a large number of teacher classroom questions requiring classification and analysis.Classroom questioning is an indispensable part of teaching.Teachers can test students’ mastery of knowledge and inspire students’ thinking by asking questions in class.Under the traditional teaching mode,teachers’ questions are often too frequent and monotonous,leading to a large number of repetitive and meaningless questions,which will reduce students’ enthusiasm and participation,thus affecting their learning effect.Classifying classroom questions can help researchers better understand the level of teachers’ questioning ability,adjust teaching strategies and contents,and improve teaching effect.How to classify classroom questions accurately is of great significance for researchers to study teachers’ classroom questions.Currently,there are few data sets of teacher classroom questioning available for research,which makes it difficult for researchers to carry out relevant work.The deep neural network model is effective in text classification.As a special kind of text,the traditional text classification methods based on machine learning often need to design features manually,which cannot adapt to the processing requirements of large-scale text data.Deep learning text classification method has powerful adaptive learning ability,which can automatically learn rich feature representation from large-scale data and improve classification performance and efficiency.In this paper,deep learning techniques will be used to classify the questions asked by teachers in class.At present,most researches on text classification focus on one aspect.For example,text convolutional neural network(CNN)can only extract text features within a specific range through sliding Windows,so it has a good classification effect for short texts.For long text,cyclic neural network(RNN)is usually used to extract sequence information,but there are problems of gradient vanishing and gradient explosion.However,in practical applications,the length of classroom questions varies greatly,and using a single model cannot achieve the best results.Therefore,in order to further improve the classification performance of classroom questions and reduce the burden of manual design features,this paper puts forward two kinds of classification methods suitable for classroom questions and conducts in-depth research.The main research contents are as follows:(1)In view of the lack of teacher’s classroom questioning data set in the current text classification field,this paper constructs a teacher’s classroom questioning data set containing28800 questions by collecting teacher’s questioning data on various online course platforms,and divides it into 6 categories according to Bloom’s teaching objectives.(2)In order to address the significant variation in text length of classroom questions and enhance the performance of existing algorithms in class question classification tasks,this paper proposes a Hybrid Neural Network(HNN)classification model specifically designed for teacher classroom questioning.The model integrates the advantages of CNN,BILSTM,and attention mechanisms in the field of text classification,enabling effective extraction of local features,contextual information,and long-term dependencies within the text.Additionally,it utilizes a multi-head attention mechanism to weight and fuse different features,facilitating comprehensive feature extraction and further improving the model’s performance.Experimental results demonstrate that the Hybrid Neural Network model exhibits superior performance in class question classification tasks.(3)Considering the unique advantages of each classification model,this paper employs the concept of ensemble learning to combine multiple neural network models(weak learners)and construct a more powerful deep ensemble model(strong learner)for the task of classroom question classification.By adopting different ensemble strategies,two deep ensemble models suitable for classroom questioning are proposed: the Voting ensemble model and the Stacking ensemble model.The deep ensemble models can effectively leverage the strengths of individual models and compensate for the limitations and deficiencies of a single model,thereby enhancing the accuracy and generalization capability of the model.Through multiple sets of comparative experiments,both algorithms have achieved promising experimental results,demonstrating the high accuracy and robustness of the proposed algorithmic models on the classroom question dataset.