Article: Multimodal generative adversarial networks for dynamic mitigation of foreign language anxiety Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.45 pp.80 - 107 Abstract: Foreign language anxiety is a key psychological barrier to language acquisition. In this study, we propose a dynamic mitigation framework based on multi-modal conditional generative adversarial networks, which uses cross-modal transformers to fuse speech, text, and facial features in real time to identify anxiety states, and conditionalise them to generate text rewriting, speech adjustment, and visual guidance feedback. Experiments show that the anxiety recognition accuracy of the system reaches 85.3%, and the naturalness score of the generated feedback is significantly better than the baseline (mean opinion score 4.2). Longitudinal studies found that participants who used the system had an average 30% decrease in state anxiety scores and a 25% increase in spoken fluency. This study provides an effective paradigm for developing personalised and adaptive emotion regulation systems. Inderscience Publishers - linking academia, business and industry through research

Title: Multimodal generative adversarial networks for dynamic mitigation of foreign language anxiety

Authors: Ming Li

Addresses: College of Foreign Languages, Ningbo University of Finance and Economics, Ningbo, 315175, China

Abstract: Foreign language anxiety is a key psychological barrier to language acquisition. In this study, we propose a dynamic mitigation framework based on multi-modal conditional generative adversarial networks, which uses cross-modal transformers to fuse speech, text, and facial features in real time to identify anxiety states, and conditionalise them to generate text rewriting, speech adjustment, and visual guidance feedback. Experiments show that the anxiety recognition accuracy of the system reaches 85.3%, and the naturalness score of the generated feedback is significantly better than the baseline (mean opinion score 4.2). Longitudinal studies found that participants who used the system had an average 30% decrease in state anxiety scores and a 25% increase in spoken fluency. This study provides an effective paradigm for developing personalised and adaptive emotion regulation systems.

Keywords: adversarial networks; dynamic mitigation; affective computing; conditional generation.

DOI: 10.1504/IJICT.2026.153513

International Journal of Information and Communication Technology, 2026 Vol.27 No.45, pp.80 - 107

Received: 17 Dec 2025
Accepted: 19 Jan 2026
Published online: 12 May 2026 *

Title: Multimodal generative adversarial networks for dynamic mitigation of foreign language anxiety

Keep up-to-date