The NTNU System at the S&I Challenge 2025 ASR Open Track

Jhih Rong Guo; Kuan Tang Huang; An Ci Peng; Berlin Chen

doi:10.21437/SLaTE.2025-10

The NTNU System at the S&I Challenge 2025 ASR Open Track

Jhih Rong Guo
, Kuan Tang Huang
, An Ci Peng
, Berlin Chen

Department of Computer Science and Information Engineering

Research output: Contribution to conference › Paper › peer-review

Abstract

This paper explores an effective approach to advancing nonnative (L2) automatic speech recognition (ASR). To this end, we design and implement a pragmatic architecture for L2 ASR and use an effective fine-tuning mechanism tailored for L2 ASR. Experiments conducted on the Speak & Improve 2025 corpus demonstrate that our approach can offer a significant improvement of 53.7% in terms of SpWER compared to the baseline. Building on top of the proposed approach, we came out in first place on the open track of the ASR task in Speak & Improve Challenge 2025 (S&I Challenge).

Original language	English
Pages	46-50
Number of pages	5
DOIs	https://doi.org/10.21437/SLaTE.2025-10
Publication status	Published - 2025
Event	10th ISCA Workshop on Speech and Language Technology in Education, SLaTE 2025 - Nijmegen, Netherlands Duration: 2025 Aug 22 → 2025 Aug 24

Conference

Conference	10th ISCA Workshop on Speech and Language Technology in Education, SLaTE 2025
Country/Territory	Netherlands
City	Nijmegen
Period	2025/08/22 → 2025/08/24

Keywords

automatic speech recognition
non-native speech

ASJC Scopus subject areas

Computer Science (miscellaneous)
Computational Mathematics
Education

Access to Document

10.21437/SLaTE.2025-10

Cite this

@conference{acf4a56e0a2c41bfaa9554c3b567fe0e,

title = "The NTNU System at the S\&I Challenge 2025 ASR Open Track",

abstract = "This paper explores an effective approach to advancing nonnative (L2) automatic speech recognition (ASR). To this end, we design and implement a pragmatic architecture for L2 ASR and use an effective fine-tuning mechanism tailored for L2 ASR. Experiments conducted on the Speak \& Improve 2025 corpus demonstrate that our approach can offer a significant improvement of 53.7\% in terms of SpWER compared to the baseline. Building on top of the proposed approach, we came out in first place on the open track of the ASR task in Speak \& Improve Challenge 2025 (S\&I Challenge).",

keywords = "automatic speech recognition, non-native speech",

author = "Guo, \{Jhih Rong\} and Huang, \{Kuan Tang\} and Peng, \{An Ci\} and Berlin Chen",

note = "Publisher Copyright: {\textcopyright} SLaTE 2025. All right reserved.; 10th ISCA Workshop on Speech and Language Technology in Education, SLaTE 2025 ; Conference date: 22-08-2025 Through 24-08-2025",

year = "2025",

doi = "10.21437/SLaTE.2025-10",

language = "English",

pages = "46--50",

}

TY - CONF

T1 - The NTNU System at the S&I Challenge 2025 ASR Open Track

AU - Guo, Jhih Rong

AU - Huang, Kuan Tang

AU - Peng, An Ci

AU - Chen, Berlin

PY - 2025

Y1 - 2025

N2 - This paper explores an effective approach to advancing nonnative (L2) automatic speech recognition (ASR). To this end, we design and implement a pragmatic architecture for L2 ASR and use an effective fine-tuning mechanism tailored for L2 ASR. Experiments conducted on the Speak & Improve 2025 corpus demonstrate that our approach can offer a significant improvement of 53.7% in terms of SpWER compared to the baseline. Building on top of the proposed approach, we came out in first place on the open track of the ASR task in Speak & Improve Challenge 2025 (S&I Challenge).

AB - This paper explores an effective approach to advancing nonnative (L2) automatic speech recognition (ASR). To this end, we design and implement a pragmatic architecture for L2 ASR and use an effective fine-tuning mechanism tailored for L2 ASR. Experiments conducted on the Speak & Improve 2025 corpus demonstrate that our approach can offer a significant improvement of 53.7% in terms of SpWER compared to the baseline. Building on top of the proposed approach, we came out in first place on the open track of the ASR task in Speak & Improve Challenge 2025 (S&I Challenge).

KW - automatic speech recognition

KW - non-native speech

UR - https://www.scopus.com/pages/publications/105045841704

UR - https://www.scopus.com/pages/publications/105045841704#tab=citedBy

U2 - 10.21437/SLaTE.2025-10

DO - 10.21437/SLaTE.2025-10

M3 - Paper

AN - SCOPUS:105045841704

SP - 46

EP - 50

T2 - 10th ISCA Workshop on Speech and Language Technology in Education, SLaTE 2025

Y2 - 22 August 2025 through 24 August 2025

ER -

The NTNU System at the S&I Challenge 2025 ASR Open Track

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this