Scholarlink

<p>Based on neural network technology, this study investigates the evaluation and optimization of English classroom teaching effectiveness. In real teaching scenarios, existing evaluation methods still face several open challenges: they rely heavily on subjective scoring, overlook students’ emotional dynamics and participation patterns, and rarely transform model outputs into actionable teaching feedback. To address these issues, we design a multimodal neural-network-based evaluation framework that jointly models three dimensions of student behavior: facial expressions, vocal emotions, and textual performance. The proposed model integrates CNN-based visual feature extraction, LSTM-based audio and text sequence modeling, and a fusion network that outputs both holistic effectiveness scores and interpretable indicators related to engagement, affect, and language ability. Experiments on multimodal classroom data from multiple institutions, including five deeply analyzed case students, show that the model achieves high predictive ability (accuracy above 92%) and provides fine-grained diagnostic feedback. Teachers can accurately identify individual strengths and weaknesses and then generate targeted, personalized teaching strategies. This method offers an effective supplement to traditional evaluation approaches and provides a scientific basis for data-driven, personalized optimization of English classroom 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

Evaluation and optimization of English classroom teaching effect based on neural network

  • 期刊论文

  • 发表日期：

    2026年2月5日

查看全文
  • Cui Wang

摘要

Based on neural network technology, this study investigates the evaluation and optimization of English classroom teaching effectiveness. In real teaching scenarios, existing evaluation methods still face several open challenges: they rely heavily on subjective scoring, overlook students’ emotional dynamics and participation patterns, and rarely transform model outputs into actionable teaching feedback. To address these issues, we design a multimodal neural-network-based evaluation framework that jointly models three dimensions of student behavior: facial expressions, vocal emotions, and textual performance. The proposed model integrates CNN-based visual feature extraction, LSTM-based audio and text sequence modeling, and a fusion network that outputs both holistic effectiveness scores and interpretable indicators related to engagement, affect, and language ability. Experiments on multimodal classroom data from multiple institutions, including five deeply analyzed case students, show that the model achieves high predictive ability (accuracy above 92%) and provides fine-grained diagnostic feedback. Teachers can accurately identify individual strengths and weaknesses and then generate targeted, personalized teaching strategies. This method offers an effective supplement to traditional evaluation approaches and provides a scientific basis for data-driven, personalized optimization of English classroom teaching.