Accurately identifying and intelligently analyzing teachers’ verbal behavior is a pivotal means of intelligent classroom teaching evaluation and provides an objective basis for constructing digital teacher profiles. However, existing studies have largely overlooked inappropriate behaviors such as teachers’ pet phrases, non-standard language, and even verbal violence. To address this gap, the paper incorporates these inappropriate behaviors into the analysis of teachers’ verbal behavior and, in view of their small sample size and strong concealment, proposes a novel intelligent recognition method based on a dual-stream graph convolutional network with adaptive fusion. First, the method extends the Flanders Interaction Analysis System to include teachers’ inappropriate speech, thereby establishing a coding scheme for classroom verbal behavior. Next, it builds a dual-stream graph convolutional network with adaptive fusion to recognize such behavior: a graph convolutional network and a graph attention network are employed to extract global and local features of teachers’ speech, respectively; an adaptive feature-fusion strategy is introduced to achieve deep integration of these features; and a text self-training approach is adopted to reduce semantic errors. Experiments conducted on a purpose-built dataset of teachers’ verbal behavior demonstrate the proposed method’s effectiveness and accuracy in detecting inappropriate speech. The approach can be applied to the supervision of classroom teaching processes and to improving teachers’ professional competence.
该类方法通常先以固定时长分解师生言语,然后通过人工编码标注教师行为类别,最后使用NVivo等质性分析软件进行统计分析。代表性工作有:1963年,Flanders[1]首次提出的弗兰德斯互动分析系统(Flanders Interaction Analysis System,FIAS),将课堂行为编码为10类,其中教师言语行为包括“表达情感、鼓励表扬、采纳意见、提问”4类间接影响和“讲授、指令、批评”3类直接影响。FIAS中教师的“批评”言语涵盖了Criticizing or Justifying Authority相关的4类情况,其中“Bawling Someone Out”即大声呵斥某人,具有不尊重学生和人身攻击成分;“Extreme Self-Reference”即教师过度以自我为中心,采用“我说什么就是什么”“不听我的就滚出去”等话语威胁学生,具有过度使用权威的特征。这两类言语行为具有明显的不当言语行为特征,属不当言语行为。
在文献[1]的基础上,文献[8]提出包含18个编码的基于信息技术的互动分析编码系统,将“提问”扩展为“提问开放性问题”和“提问封闭性问题”,其教师行为包含8类言语行为与1类教师操纵设备或软件的非言语行为;文献[9]以移动学习终端为代表构建1∶1数字化环境,提出包含24个编码的课堂教学互动行为分析编码体系,它沿用文献[8]中的8类教师言语行为,将教师操纵设备或软件细化为“资源演示”等4类;2021年,Zhao等[10]新增“Other”标识教师管理课堂话语,将FIAS的“批评”替换为“Comments on Students’ Answer”,没有考虑教师不当言语;2024年,李红美等[11]从传统书本或板书和信息技术两个角度将“讲授”扩展为“讲授、演示或示范、展示交流”等5类,把FIAS的教师言语从7类拓展到11类。
FLANDERSN A. Intent, action and feedback: A preparation for teaching[J]. Journal of Teacher Education, 1963, 14(3): 251-260. DOI:10.1177/002248716301400305 .
LIUC Y. Reflections on teachers’ language violence in the process of building a country ruled by law[J]. Educational Practice and Research, 2022(12): 23-27. DOI:10.14160/j.cnki.13-1259/g4-c.2022.12.012(Ch ).
LIUQ T, HEH Y, WUL J, et al. Classroom teaching behavior analysis method based on artificial intelligence and its application[J]. China Educational Technology, 2019(9): 13-21. DOI:10.3969/j.issn.1006-9860.2019.09.003(Ch ).
LUOZ Y, ZHAOQ Q, DUANF Q. Automatic analysis of classroom teaching process based on teachers’ near-field speech[J]. Modern Educational Technology, 2021, 31(8): 76-84. DOI:10.3969/j.issn.1009-8097.2021.08.009(Ch ).
[8]
JENSENE, DALEM, DONNELLYP J, et al. Toward automated feedback on teacher discourse to enhance teacher learning[DB/OL]. [2025-01-02]. DOI: 10.1145/3313831.3376418 .
OUZ G, LIUY P, LIR L, et al. Research on teacher speech emotion recognition in international Chinese language classrooms[J]. Modern Educational Technology, 2023, 33(8): 87-95. DOI:10.3969/j.issn.1009-8097.2023.08.010 (Ch ).
GUX Q, WANGW. A new exploration of classroom analysis technology to support teachers’ professional development[J]. China Educational Technology, 2004(7): 18-21. DOI:10.3969/j.issn.1006-9860.2004.07.004 (Ch ).
HANH, WANGD Q, CAOC. An analysis of classroom interaction behaviors in a 1∶1 digital learning environment[J]. e⁃Education Research, 2015, 36(5): 89-95. DOI:10.13811/j.cnki.eer.2015.05.013(Ch ).
[17]
ZHAOK L, BOONYAPRAKOBK. A case study of Chinese students learning Thai as a foreign language: Flanders interaction analysis category system[J]. International Journal of TESOL & Education, 2021, 2(1): 145-169. DOI:10.54855/ijte.222110 .
LIH M, BIANP, XUM Q. The development and pattern mining of double coding analysis system for interactive teaching of mathematics classroom under the smart classroom environment[J]. Modern Educational Technology, 2024, 34(3): 105-115. DOI:10.3969/j.issn.1009-8097.2024.03.011(Ch ).
[20]
XUL L, HEX L, ZHANGJ, et al. Automatic classification of discourse in Chinese classroom based on multi-feature fusion[C]//2019 International Conference on Computer, Information and Telecommunication Systems (CITS). New York: IEEE Press, 2019: 1-5. DOI:10.1109/CITS.2019.8862101 .
[21]
ZHANGX L, WANGJ Q, WANZ H, et al. Classification of classroom teachers’ speech intention based on deep learning[C]//2021 International Conference on Computer Engineering and Application (ICCEA). New York: IEEE Press, 2021: 230-233. DOI:10.1109/ICCEA53728.2021.00053 .
[22]
KIMY. Convolutional neural networks for sentence classification[C]//Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). Stroudsburg: Asscoication for Computational Linguistics, 2014: 1746-1751. DOI:10.3115/v1/d14-1181 .
LIAOS B, QIF. Research on machine analysis of classroom teacher-student interaction behavior[J]. Journal of Central China Normal University (Natural Sciences), 2024, 58(2): 279-285. DOI:10.19603/j.cnki.1000-1190.2024.02.014(Ch ).
[25]
GAZDAG M, BALZERF J, CHILDERSW C, et al. Human Relations Development: A Manual for Educators[M]. 6th Edition. Boston: Allyn and Bacon, 1999.
YINW. A teaching analysis system based on knowledge graph and text analysis[D]. Shanghai:Shanghai Jiao Tong University, 2020. DOI:10.27307/d.cnki.gsjtu.2020.003208(Ch ).
[28]
腾讯云. 腾讯AI开放平台[EB/OL]. [2018-09-18].
[29]
TENCENTCLOUD. Tencent AI Open Platform[EB/OL]. [2018-09-18].
[30]
CUIH Y, WANGG K, LIY X, et al. Self-training method based on GCN for semi-supervised short text classification[J]. Information Sciences, 2022, 611: 18-29. DOI:10.1016/j.ins.2022.07.186 .
LIUP F, QIUX P, HUANGX J. Recurrent neural network for text classification with multi-task learning[DB/OL].[2025-01-03]. DOI: 10.24963/ijcai.2017/473 .
[33]
ZHOUP, SHIW, TIANJ, et al. Attention-based bidirectional long short-term memory networks for relation classification[C]// Proceedings of the Annual Meeting of the Association for Computational Linguistics. Stroudsburg: Association for Computational Linguistics, 2016: 207-212. DOI: 10.18653/v1/p16-2034 .
[34]
JOHNSONR, ZHANGT. Deep pyramid convolutional neural networks for text categorization[DB/OL].[2025-01-02]. DOI: 10.18653/v1/p17-1052 .
[35]
MURFIH, GOWANDIT, ARDANESWARIG, et al. BERT-based combination of convolutional and recurrent neural network for Indonesian sentiment analysis[J]. Applied Soft Computing, 2024, 151: 111112. DOI:10.1016/j.asoc.2023.111112 .