Article: A multimodal generative AI dialogue system for real-time intervention in English learning anxiety Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.69 pp.92 - 121 Abstract: English learning anxiety significantly impairs learners' cognitive performance and language acquisition, yet existing interventions lack real-time responsiveness and personalisation. This paper introduces multimodal generative artificial intelligence for anxiety intervention in conversation, a multimodal generative artificial intelligence dialogue system that continuously perceives a learner's anxiety level through audio, video, and text, and generates adaptive supportive responses to alleviate anxiety in real time. The system integrates a cross-modal transformer with bidirectional long short-term memory for anxiety perception, a conditional variational autoencoder for generating empathetic responses, and deep reinforcement learning to optimise when to intervene. A new English learning anxiety corpus comprising 120 real learners is constructed for training and evaluation. Experiments demonstrate that MAGIC significantly reduces self-reported anxiety (Δ<i>a</i> = 0.31, <i>p</i> < 0.01) compared to baseline methods, confirming its effectiveness in providing timely and personalised emotional support. Inderscience Publishers - linking academia, business and industry through research

Title: A multimodal generative AI dialogue system for real-time intervention in English learning anxiety

Authors: Qingbin Huang

Addresses: College of Foreign Languages, Guangdong Technology College, Zhaoqing, 526100, China

Abstract: English learning anxiety significantly impairs learners' cognitive performance and language acquisition, yet existing interventions lack real-time responsiveness and personalisation. This paper introduces multimodal generative artificial intelligence for anxiety intervention in conversation, a multimodal generative artificial intelligence dialogue system that continuously perceives a learner's anxiety level through audio, video, and text, and generates adaptive supportive responses to alleviate anxiety in real time. The system integrates a cross-modal transformer with bidirectional long short-term memory for anxiety perception, a conditional variational autoencoder for generating empathetic responses, and deep reinforcement learning to optimise when to intervene. A new English learning anxiety corpus comprising 120 real learners is constructed for training and evaluation. Experiments demonstrate that MAGIC significantly reduces self-reported anxiety (Δa = 0.31, p < 0.01) compared to baseline methods, confirming its effectiveness in providing timely and personalised emotional support.

Keywords: multimodal perception; generative dialogue system; English learning anxiety; ELA; real-time intervention; cognitive load theory; CLT.

DOI: 10.1504/IJICT.2026.154382

International Journal of Information and Communication Technology, 2026 Vol.27 No.69, pp.92 - 121

Received: 06 Mar 2026
Accepted: 09 Apr 2026
Published online: 25 Jun 2026 *

Title: A multimodal generative AI dialogue system for real-time intervention in English learning anxiety

Keep up-to-date