Article: Dynamic diagnosis of students' state in English classes based on feature decoupling and improved active learning algorithm Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.71 pp.80 - 103 Abstract: Traditional subjective observation for diagnosing students' state in English classes suffers from bias, low efficiency, and inability to capture multi-dimensional information. This study builds a dynamic diagnosis system integrating text, audio, and video data. It employs an improved active learning algorithm with diversity constraints to optimise annotation, a semantic-state decoupling framework using triplet loss to reduce interference, feature alignment and cross-modal attention fusion to improve feature quality, and parallel deployment to accelerate response. Experimental results show multi-modal fusion achieves 90.5% accuracy, 8% higher than the best single modality. The active learning strategy yields 83.1% sample utilisation and 91.9% model accuracy within 180 minutes. The decoupling mechanism lowers diagnostic error to 9.0% in high-semantic complexity scenarios, and parallel deployment cuts response delay to 80ms. The system significantly enhances diagnostic accuracy and efficiency, providing reliable technical support for personalised and real-time teaching intervention in English classes. Inderscience Publishers - linking academia, business and industry through research

Title: Dynamic diagnosis of students' state in English classes based on feature decoupling and improved active learning algorithm

Authors: Dongfeng Liu; Xiaolong Ren

Addresses: School of Foreign Languages, North China Institute of Aerospace Engineering, Langfang, 065000, China ' School of Foreign Languages, North China Institute of Aerospace Engineering, Langfang, 065000, China

Abstract: Traditional subjective observation for diagnosing students' state in English classes suffers from bias, low efficiency, and inability to capture multi-dimensional information. This study builds a dynamic diagnosis system integrating text, audio, and video data. It employs an improved active learning algorithm with diversity constraints to optimise annotation, a semantic-state decoupling framework using triplet loss to reduce interference, feature alignment and cross-modal attention fusion to improve feature quality, and parallel deployment to accelerate response. Experimental results show multi-modal fusion achieves 90.5% accuracy, 8% higher than the best single modality. The active learning strategy yields 83.1% sample utilisation and 91.9% model accuracy within 180 minutes. The decoupling mechanism lowers diagnostic error to 9.0% in high-semantic complexity scenarios, and parallel deployment cuts response delay to 80ms. The system significantly enhances diagnostic accuracy and efficiency, providing reliable technical support for personalised and real-time teaching intervention in English classes.

Keywords: students' status in English class; dynamic diagnosis; multi-modal data fusion; feature decoupling; improved active learning.

DOI: 10.1504/IJICT.2026.154470

International Journal of Information and Communication Technology, 2026 Vol.27 No.71, pp.80 - 103

Received: 16 Jan 2026
Accepted: 19 Mar 2026
Published online: 29 Jun 2026 *

Title: Dynamic diagnosis of students' state in English classes based on feature decoupling and improved active learning algorithm

Keep up-to-date