【作 者】：
林梓柔， 闫寒冰
【关 键 词 】：
信息化教学； 微能力； 听评课； 深度学习； 预训练
【栏 目】：
学科建设与教师发展
【中文摘要】：
教师信息化教学能力是体现教育信息化深度融合成效的关键。为探索一种证据易获得、分析自动化、反馈可持续的测评手段，文章提出了基于预训练的教师信息化教学微能力自动识别方法。具体为，基于包含学情分析、教学设计、学法指导和学业评价4个维度的13项微能力，利用基于预训练的深度学习模型对16,371条信息化课堂评课反馈文本进行实验。递进式实验结果表明，相比传统深度学习模型，预训练模型普遍具有更高准确率，最高能提升16%。领域通用语料的预训练有效解决了教育领域自建数据匮乏的现实问题；在预训练模型中，ERNIE以86.43%的准确率取得了最优性能，混淆误差小。该方法能够基于评课文本自动化识别并可视化表征教师信息化教学微能力，提供常态化反馈支持。最后，文章阐述了其在教师教研、能力画像和发展决策中的应用场景，并提出后续研究方向。
【英文摘要】：
Teachers' informatization teaching ability is the key to reflect the effectiveness of deep integration of education informatization. In order to explore an assessment method with easy access to evidence, automatic analysis and sustainable feedback, this paper proposes an automatic identification method of teachers' micro-competence in informatization teaching based on pre-training. Specifically, based on 13 micro-competence including four dimensions of learning situation analysis, instructional design, learning method guidance and academic evaluation, a deep learning model based on pre-training is used to conduct an experiment on 16,371 feedback texts of informatization classroom. The progressive experimental results show that compared to traditional deep learning models, the pre-training model generally has higher accuracy, with a maximum improvement of 16%. The pre-training of domain-general corpus effectively solves the problem of lack of self-constructed data in education. In the pre-training model, ERNIE achieves the optimal performance with 86.43% accuracy and small confusion deviation. This method can automatically identify and visually represent teachers' micro-competence of informatization teaching based on the evaluation text, and provide regular feedback support. Finally, this paper expounds its application scenarios in teacher teaching and research, competency portrait and development decision-making, and puts forward future research directions.