Title: Research on automatic annotation of English pronunciation errors based on deep transfer learning

Authors: Fengjuan Zhang

Addresses: School of Foreign Languages, Zhengzhou University of Industrial Technology, Zhengzhou, Henan, China

Abstract: In order to overcome the problems of low accuracy and long-time of traditional English pronunciation error labelling methods, this paper proposes an automatic English pronunciation error labelling method based on deep transfer learning. Firstly, construct an English phonetic corpus to extract English pronunciation features. Then, a hidden Markov model is used to map the pronunciation phoneme sequence. Finally, through deep transfer learning, the pronunciation error annotation function is constructed and the automatic annotation of English pronunciation errors is realised through iterative training, and the final annotation results are output to realise the automatic annotation of English pronunciation errors. The results show that the accuracy of the proposed method for annotation can reach 99.8%, and the automatic annotation time does not exceed 16 minutes, effectively improving the effectiveness of automatic annotation for English pronunciation errors.

Keywords: speech corpus; phoneme sequence; end-to-end model; deep transfer learning; linear prediction coefficient.

DOI: 10.1504/IJCAT.2023.138828

International Journal of Computer Applications in Technology, 2023 Vol.73 No.4, pp.245 - 252

Received: 11 Oct 2023
Accepted: 15 Dec 2023
Published online: 31 May 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article