【作 者】：
胡婉青， 李 新， 黄睿妍，李艳燕
【关 键 词 】：
协作学习投入； 协作学习； 深度学习； 自动分析； 应用案例
【栏 目】：
课程与教学
【中文摘要】：
协作学习投入的自动分析是动态追踪协作学习过程、提升协作学习效果的重要手段。现有研究大多使用浅层机器学习分析个体学习投入，较少关注协作学习投入，并且存在准确度受限且泛化能力较差等问题。为了解决相应问题，首先，研究提出以BERT-BiLSTM深度学习模型为核心的协作学习投入自动分析方法，包括收集数据、建立数据库、建立训练集、构建深度学习模型、检验模型效果、批量识别数据以及统计分析等流程。其次，研究应用该方法分析了某高校数学建模活动中20个小组的协作学习投入，验证了方法的有效性，并进一步探索了高低成就小组协作学习投入的差异以及动态变化特点。研究突破了传统学习投入自动检测方法在准确性和泛化能力上的局限性，将学习投入的自动分析对象从个体拓展到了小组，并揭示了协作学习投入的时序变化特征及其与协作成效的复杂联系，为未来实时监测及干预协作学习提供了重要支持，进一步推动了该研究领域的理论与实践发展。
【英文摘要】：
The automatic analysis of collaborative learning engagement is an important means to dynamically track the process of collaborative learning and enhance the effectiveness of collaborative learning. Most of the existing studies employ shallow machine learning to analyze individual learning engagement, but pay less attention to collaborative learning engagement. And there are problems such as limited accuracy and poor generalizability ability. In order to address these issues, this study firstly proposes an automatic analysis method of collaborative learning engagement based on BERT-BiLSTM deep learning model, which includes collecting data, establishing a database, setting up a training set, constructing a deep learning model, examining the effect of the model, identifying the data in batches, and statistically analyzing the data. Secondly, this study applies the method to analyze the collaborative learning engagement of 20 groups in a university's mathematical modeling activity, verifies the effectiveness of the method, and further explores the differences in the collaborative learning engagement of high- and low-achievement groups as well as the characteristics of dynamic changes. This study breaks through the limitations of traditional automatic detection methods in terms of accuracy and generalization ability of learning engagement, expands the object of automatic analysis of learning engagement from individuals to groups, and reveals the temporal variation characteristics of collaborative learning engagement and its complex relationship with collaborative effectiveness, which provides an important support for real-time monitoring and intervention of collaborative learning in the future, and further promotes the theoretical and practical development of this research field.