【作 者】：
杨彦军， 徐 刚， 童 慧
【关 键 词 】：
多模态数据； 学习场景； 智慧课堂； 深度学习
【栏 目】：
学习环境与资源
【中文摘要】：
随着信息技术与教育教学深度融合的加速，基于智能学习环境的线上线下混合式智慧教学将成为学校教学的新常态。智慧教学过程中产生的教育多模态数据为深度学习监测评价与反馈干预提供了新机遇。文章根据三元交互决定论与具身认知理论构建了多模态数据的分类框架，将多模态数据分为个体身体、交互行为、智能环境三类。从教师与学生、个体与群体、传统信息技术与现代信息技术三个维度对教学模式进行分类，并归纳出十二种典型学习场景。根据多模态数据分类与学习场景分类总结归纳出典型学习场景下的监测模态，进而构建基于多模态数据的深度学习监测模型，主要包括智能学习环境下的学习活动、场景学习与识别、多模态数据采集、多模态数据处理与分析、深度学习状态监测五个环节。
【英文摘要】：
With the acceleration of the deep integration of information technology and teaching, online and offline hybrid intelligent teaching based on intelligent learning environment will become the new normal of school teaching. The educational multimodal data generated in the process of smart teaching provides new opportunities for deep learning monitoring, evaluation and feedback intervention. Based on the three-way interaction determinism and embodied cognitive theory, this paper constructs a classification framework for multimodal data, and divides the multimodal data into three types of individual body, interactive behavior, and intelligent environment. This study classifies teaching models from three dimensions of teachers and students, individuals and groups, traditional information technology and modern information technology, and summarizes twelve typical learning scenarios. Based on the classification of multimodal data and learning scenarios, the monitoring models in typical learning scenarios are summarized. Then the deep learning monitoring model based on the multimodal data is constructed, which mainly includes five links: learning activities in intelligent learning environment, learning and recognition of learning scenarios, multimodal data acquisition, multimodal data processing and analysis, and deep learning status monitoring.