Scholarlink

<p>Classroom assessment is a decisive task that provides the insights of teaching-learning process. In general, the traditional way of classroom assessments is tedious and time-consuming, particularly when analyzing the verbal component. The emerging applications of deep learning can be applied to optimize the process with minimal intervention. In this paper, we propose to automate the assessment of classroom learning environment through a comprehensive analysis of verbal dynamics. For verbal dynamics based analysis and evaluation, a new dataset is prepared from available YouTube videos. The dataset is annotated and grounded in the Flanders interaction model. These verbal dynamics are extracted from classroom interaction using deep learning techniques such as LSTM, BiLSTM, and TCN employed for the speaker diarization task. The performance of classroom assessment obtained through automation is compared with the scores of human evaluators, thereby showing a comparable relationship between the two. The results emphasizes on the relationship between teacher and student engagement with the help of correlation analysis. The correlation of teacher and student participation presents a consistent interaction. In addition, the model performance and interpretability is also validated through NetTrustScore and feature importance analysis. For speaker change detection, the highest accuracy achieved was 97.94 and 95.17 on two benchmark datasets AMI and ICSI respectively. The result suggests that both BiLSTM and TCN are more effective in detecting the speaker change points as they capture the contextual and temporal information respectively. This study suggests the crucial aspects of verbal dynamics for the evaluation teacher’s speech. It therefore offers assistance to human evaluators in assessing trainee teachers’. In addition, it also provides an assessment of teacher-student participation, and highlights various verbal elements such as strength, pitch, energy, and modulation which are an important part of effective teaching.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated Assessment of Classroom Interaction Based on Verbal Dynamics: A Deep Learning Approach

  • 期刊论文

  • 发表日期：

    2025年2月20日

查看全文
  • Sadhana Singh,
  • Lotika Singh,
  • Nandita Satsangee

摘要

Classroom assessment is a decisive task that provides the insights of teaching-learning process. In general, the traditional way of classroom assessments is tedious and time-consuming, particularly when analyzing the verbal component. The emerging applications of deep learning can be applied to optimize the process with minimal intervention. In this paper, we propose to automate the assessment of classroom learning environment through a comprehensive analysis of verbal dynamics. For verbal dynamics based analysis and evaluation, a new dataset is prepared from available YouTube videos. The dataset is annotated and grounded in the Flanders interaction model. These verbal dynamics are extracted from classroom interaction using deep learning techniques such as LSTM, BiLSTM, and TCN employed for the speaker diarization task. The performance of classroom assessment obtained through automation is compared with the scores of human evaluators, thereby showing a comparable relationship between the two. The results emphasizes on the relationship between teacher and student engagement with the help of correlation analysis. The correlation of teacher and student participation presents a consistent interaction. In addition, the model performance and interpretability is also validated through NetTrustScore and feature importance analysis. For speaker change detection, the highest accuracy achieved was 97.94 and 95.17 on two benchmark datasets AMI and ICSI respectively. The result suggests that both BiLSTM and TCN are more effective in detecting the speaker change points as they capture the contextual and temporal information respectively. This study suggests the crucial aspects of verbal dynamics for the evaluation teacher’s speech. It therefore offers assistance to human evaluators in assessing trainee teachers’. In addition, it also provides an assessment of teacher-student participation, and highlights various verbal elements such as strength, pitch, energy, and modulation which are an important part of effective teaching.