Title: Modelling and optimisation of intelligent speech feedback mechanisms for French pronunciation correction
Authors: Ge Song; Wenyong Guo
Addresses: College of Foreign Languages, Hebei North University, Zhangjiakou, 075000, China ' College of Foreign Languages, Hebei North University, Zhangjiakou, 075000, China
Abstract: In order to improve the accuracy of French pronunciation correction, this study develops a multimodal feedback system, which adopts the improved wav2vec2 model to integrate the physiological features of articulation, and combines time-frequency analysis to extract the acoustic parameters. The developed system generates the targeted training materials through the dynamic knowledge graph and integrates the articulatory organ visualisation module. The selective spectrum enhancement strategy is designed to assist in the listening discrimination training. Experiments show that the feedback delay of the system is ≤ 155 ms, and the VOT recognition error is reduced by 9.2%; after ten weeks of training, the confusion rate of articulatory parts is reduced by 5.1%, and the accuracy rate of question rhymes reaches 79.2%. The results confirm that moderate multimodal feedback has a progressive optimisation effect on French pronunciation.
Keywords: multimodal feedback systems; French language; acoustics; dynamic knowledge mapping.
DOI: 10.1504/IJICT.2025.150399
International Journal of Information and Communication Technology, 2025 Vol.26 No.43, pp.20 - 35
Received: 26 Jul 2025
Accepted: 18 Sep 2025
Published online: 12 Dec 2025 *


