Xu, Q., & Sunaoka, K. (2025). A Data-Driven Approach to Optimizing Beginner Chinese Classrooms: Integrating Language-Use Proportion Visualization with Generative AI. Journal of Technology and Chinese Language Teaching, 16(2), 1-22.
[徐勤, & 砂冈和子. (2025). 数据驱动的初级汉语课堂优化研究：语言使用比例可视化与生成式AI的结合. 科技与中文教学 (Journal of Technology and Chinese Language Teaching), 16(2), 1-22.]

Full paper

Abstract/摘要：

This study investigates the optimization of beginner-level Chinese language classrooms in Japan through a data-driven approach that integrates speech visualization and generative AI. Classroom recordings were analyzed using a self-developed voice-to-text app, which automatically transcribes classroom recordings of mixed Chinese-Japanese classroom discourse and then visualizes the proportion of L2 (Chinese) use. The app demonstrates high transcription accuracy and operational convenience, offering significant advantages in efficiency, usability, and teacher autonomy compared with traditional classroom language analysis frameworks. Based on the app-generated data and two types of instructional prompts, ChatGPT was used to generate feedback and suggestions for classroom improvement, in order to explore how AI could be applied in Chinese language class design and identifying its limitations. The findings reveal that while AI can effectively propose revisions at the linguistic and surface-level reasoning stages, it lacks deeper cognitive and creative capacities necessary for generating pedagogically insightful designs.

本研究以日本的大学初级汉语课堂的两段教学录音为样本，利用我们自主开发的语音可视化应用Voice-to-Text App进行分析。结果显示，该应用在处理中日语码混合的课堂录音时，具有较高的转写准确率与操作便利性，能迅速识别课堂中的L2（汉语）使用比例。与传统的课堂语言行为评估框架相比，该系统在效率、可操作性与教师自主分析能力方面均表现出显著优势。在此基础上，研究将APP生成的数据结合两种类型的提示词（prompt），输入ChatGPT，由生成式AI提供课堂改进建议，以探讨AI在汉语课堂设计中的潜在优势与应用局限。结果发现，AI能够在语言形式重组与表层推理层面提出较为合理的课堂优化方案，但由于缺乏深层认知能力与创新语法概念的生成能力，难以提出具有启发性的教学设计。