Scholarlink

<p>The increase of global communication has highlighted the importance of successful cross-cultural English education strategies that blend language and cultural understanding. Current monomodal assessments (text-based tests, surveys) do not represent the complexity of multimodal classroom interactions and learning engagement. This study introduces CulTeach-MMA (Cultural Teaching Multimodal Assessment), a multimodal assessment approach that was trained and tested on a dataset of 4200 labeled samples (1400 text, 1400 audio, 1400 images) to capture and assess linguistic, cultural, and behavioural features for cross-cultural English teaching evaluation. The model uses various data sources to offer a comprehensive assessment and inform teaching practices based on evidence. The proposed deep learning-based multimodal model pre-processes and embeds text (student feedback, transcripts), audio (pronunciation and tone), and video (gestures, engagement levels) using LSTMs, CNNs/3D-CNNs, and attention-based layers. Multimodal embeddings predict student's success in education. The model was assessed using accuracy, F1-score, precision and recall, with the main result being that the proposed model is able to capture nuanced cross-cultural differences and multimodal learning insights, improving on text-only assessment. The comparative experiments yield an average increase in accuracy of 12–15% over baseline models, with fivefold cross-validation. The results show the system effectively links cultural settings, learning participation and teaching quality, helping teachers modify their teaching strategies. The model’s stability and scalability ensure its cross-cultural and cross-linguistic compatibility. This study ends with a deep learning model for cross-cultural English teaching efficacy assessment via multimodal analysis.</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

Construction of a multimodal assessment model of cross-cultural English teaching effectiveness based on deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月30日

查看全文
  • Xueyu Sun,
  • Weiwei Dou,
  • Yuan Yang,
  • Zhonglin Jiang

摘要

The increase of global communication has highlighted the importance of successful cross-cultural English education strategies that blend language and cultural understanding. Current monomodal assessments (text-based tests, surveys) do not represent the complexity of multimodal classroom interactions and learning engagement. This study introduces CulTeach-MMA (Cultural Teaching Multimodal Assessment), a multimodal assessment approach that was trained and tested on a dataset of 4200 labeled samples (1400 text, 1400 audio, 1400 images) to capture and assess linguistic, cultural, and behavioural features for cross-cultural English teaching evaluation. The model uses various data sources to offer a comprehensive assessment and inform teaching practices based on evidence. The proposed deep learning-based multimodal model pre-processes and embeds text (student feedback, transcripts), audio (pronunciation and tone), and video (gestures, engagement levels) using LSTMs, CNNs/3D-CNNs, and attention-based layers. Multimodal embeddings predict student's success in education. The model was assessed using accuracy, F1-score, precision and recall, with the main result being that the proposed model is able to capture nuanced cross-cultural differences and multimodal learning insights, improving on text-only assessment. The comparative experiments yield an average increase in accuracy of 12–15% over baseline models, with fivefold cross-validation. The results show the system effectively links cultural settings, learning participation and teaching quality, helping teachers modify their teaching strategies. The model’s stability and scalability ensure its cross-cultural and cross-linguistic compatibility. This study ends with a deep learning model for cross-cultural English teaching efficacy assessment via multimodal analysis.