Research On Teachers’ Classroom Teaching Performance Based On Teaching Behavior

Developing intelligent training and innovating teacher training mode are important ways for AI to promote the construction of teachers in the new era.Precise teaching and research are the starting point and direction of teacher training and intelligent training.Precise teaching research includes an accurate assessment of teachers’ teaching ability and modeling of teachers’ portraits.The performance of teachers’ classroom teaching is the most intuitive external representation of teachers’ teaching ability,and it fits the connotation of teachers’ portraits.It can reflect the process and dynamic characteristics of classroom teaching ecology through teachers’ classroom teaching behavior.Although the current research on teachers’ classroom teaching performance has made some achievements in the automatic identification and intelligent analysis of classroom teaching events or teaching structures,the teaching behavior data referred to by these studies are sparse,ignoring the professional ability and teaching significance reflected by teachers’ behavior,and lacking systematic theoretical support for teachers’ evaluation.Therefore,this research is committed to comprehensively and systematically portraying teachers’ classroom teaching performance based on multimodal and finegrained teachers’ classroom teaching behavior as evidence,to form an effective and meaningful classroom teaching behavior data set and behavior coding system,and finally build an intelligent description model of teachers’ classroom teaching performance.To achieve the above objectives,this study mainly answers three research questions:(1)The construction of the theoretical framework: What are the systematic portrayal elements of teachers’ classroom teaching performance?(2)Construction of behavioral data set: Which behavioral indicators can represent the elements of teachers’ classroom teaching performance?(3)Implementing intelligent description: How to use AI technology to train teachers’ classroom teaching performance intelligent description model? To answer these three research questions,this research is divided into three stages: the construction of theoretical framework,the construction and verification of the behavior data system,and the training and verification of the intelligent characterization model.In the construction of the theoretical framework,this study constructs a theoretical framework to characterize teachers’ classroom teaching performance.Specifically,this study first analyzes the connotation and general evaluation methods of teachers’ classroom teaching performance,and classic classroom evaluation schemes.This theoretical framework could systematically evaluate teachers’ teaching performance from three areas in classroom teaching,namely emotional support,teaching support,and classroom management,based on the Classroom Assessment Scoring System(CLASS).Then,this study explores the teaching performance dimensions or elements that conform to the domain connotation and can be represented by teaching behavior in each classroom teaching field.Therefore,in the framework finally constructed in this study,each classroom teaching practice field is divided into different teaching performance dimensions according to the teaching meaning of the field connotation and behavior.Moreover,each teaching performance dimension can be measured by classroom teaching behavior indicators.In the portrayal framework of teachers’ classroom teaching performance,the field of emotional support includes the dimensions of teaching immediacy and teaching emotional state.The dimension of teaching immediacy includes two sub-dimensions:verbal immediacy and nonverbal immediacy;the emotional state of teaching includes facial emotion and phonetic emotion.Teaching support areas include teachers’ classroom teaching actions and teaching posture.According to the characteristics of teaching behavior,teachers’ classroom teaching actions can be divided into seven categories: illustrative actions,symbolic actions,and so on;the teachers’ classroom teaching posture includes openness,closeness,and fuzziness.The field of classroom management includes two dimensions: teachers’ attention and students’ participation in classroom learning.Teachers’ attention is composed of attention objects(attributes),time,and frequency.Student’s participation in classroom learning is composed of ontask behavior and off-task behavior.The characterization framework of teachers’ classroom teaching performance provides systematic theoretical support for the formulation of behavior coding systems and the training of intelligent characterization models.In the process of constructing and verifying the behavior data system,this study constructs and verifies the behavior coding system of teachers’ classroom teaching performance.Specifically,based on the specific teaching performance dimensions under each field of the portrayal framework of teachers’ classroom teaching performance,this study analyzes the research findings,traditional evaluation methods,and specific evaluation indicators on each teaching performance dimension through a literature review.At the same time,this research synthesizes the connotation of each teaching performance dimension and the actual classroom teaching situation and teachers’ practical behavior,summarizes the behavioral indicators of each teaching performance dimension,and forms the initial behavior coding system that depicts teachers’ classroom teaching performance.This study constructs an initial behavior coding system based on theories and existing measurement indicators.Therefore,the initial behavior coding system may not agree with the classroom teaching practice,but also has the problems of usefulness and reliability,which does not have the reliability and authority to promote its use.Therefore,to further obtain an authoritative and reliable coding system for teachers’ classroom teaching performance behavior,this study uses Delphi and questionnaire surveys to verify,modify and improve the initial behavior coding system in multiple rounds.This research combines the existing research findings,traditional questionnaire items,and actual teaching situations to build the behavior coding systems of teachers’ classroom teaching actions,teaching immediacy,attention,and students’ learning participation.Therefore,the behavior coding system of these four dimensions needs to be revised and improved after verification.However,the existing authoritative and reliable coding system or feature library is used to determine the behavior of the two teaching performance dimensions of teacher posture and emotional state,which are not verified.This research realized the intelligent description and model verification of teachers’ classroom teaching performance.Specifically,this research uses artificial intelligence technologies such as machine learning and deep learning algorithms to achieve feature extraction and automatic recognition of teachers’ classroom teaching performance dimensions,taking teachers’ immediacy,emotional state,teaching action,and teaching posture as examples.Then,this study compares the results of machine recognition with the results of artificial judgment to verify the effectiveness and reliability of the behavior coding system and intelligent description model.Because the behavioral indicators of teachers’ classroom teaching performance are multi-modal,this study not only uses video images to identify explicit behavior but also uses audio signals to identify voice information.Therefore,according to the data type,this research divides the training of intelligent models into two parts: model training based on video image analysis and model training based on speech signal analysis.Finally,this research designs the visualization results of teachers’ classroom teaching performance intelligent characterization model and applies them in actual teaching to provide teachers with the visualization results.The research analyzes the usability and practicability of intelligent models in practice by investigating teachers’ experience in using them.The main conclusions of this study are as follows: first,the characterization framework of teachers’ classroom teaching performance constructed in this study can guide the construction of the behavior coding system of teachers’ classroom teaching performance,and support the systematic characterization of teachers’ classroom teaching performance.Secondly,the coding system of teachers’ classroom teaching performance behavior is reliable,authoritative,and scalable.The system focuses on reflecting the teaching significance of teaching behavior in teachers’ classroom teaching performance.The behavior data set based on the coding system can sufficiently support the training of teachers’ intelligent description model of classroom teaching performance.Third,the intelligent description model of teachers’ classroom teaching performance trained in this study has good performance and interpretability and can realize the automatic analysis of classroom teaching videos.Moreover,teachers have a good sense of experience in using the intelligent characterization model system and have a certain degree of acceptance.This research has unique innovation and significance in theory,method and practice.(1)At the theoretical level,this study highlights the systematic and comprehensive characterization of teachers’ classroom teaching performance,emphasizes the feasibility of teaching behavior as evidence,enriches the characterization theory of teachers’ classroom teaching performance,and clarifies the information about teaching performance and teaching significance conveyed by teachers’ classroom teaching behavior.(2)At the methodological level,first of all,this study ensures that the dimensions of teachers’ classroom teaching performance not only have the connotation of their field but also can be measured by specific teaching behaviors.Secondly,this research combines the verification of the coding system with teaching practice,integrates students’ real feelings,front-line teachers’ practical experience,and experts’ professional knowledge,and revises the coding system of verification behavior from multiple perspectives and multiple subject perceptions.Then,this research trains the model from multi-modal behavior and uses the same learning model to complete different behavior recognition tasks,which improves the efficiency of video analysis.This study uses a framework of a pre-training model combined with fine-tuning to solve the problem of a small sample size.Finally,the validity and reliability of the intelligent description model are analyzed by comparing the results of manual autonomous judgment with the results of classification model judgment,which to some extent makes up for the deficiencies of data-driven methods in model interpretation.(3)At the practical level,this concern is about teachers’ classroom teaching performance reflected by teaching behavior,not limited to the classification of teaching behavior or the judgment of teaching events.The description of teaching performance in this study is based on a wide variety of teaching behavior data with multimodal characteristics,rather than a single category of body behavior data.This research constructs a universal teaching behavior data set that conforms to the cultural background of classroom teaching in China and is used to depict teachers’ classroom teaching performance,systematically summarizing teachers’ classroom teaching performance.Finally,the intelligent description model of teachers’ classroom teaching performance trained in this study has a systematic and intelligent significance.With the service and support of artificial intelligence technology,the intelligent description model system of this study improves the efficiency of classroom teaching evaluation and makes up for the inherent deficiencies brought by the manual observation and evaluation of teachers’ teaching behavior.