AI-Integrated Computational Framework for Pronunciation Error Analysis Through Native–Learner Speech Comparison
Abstract
This study proposes an AI-integrated computational framework for pronunciation error analysis through native–learner speech comparison. The framework integrates Whisper-based Automatic Speech Recognition (ASR), acoustic feature extraction, statistical analysis, and machine learning to analyze learner pronunciation. Unlike previous studies that primarily rely on individual transcription-based or acoustic measures, the proposed framework integrates multiple pronunciation indicators into a composite Pronunciation Accuracy Score (PAS) formulated in this study. The framework was evaluated using a dataset comprising 45 non-native English learners and one native English speaker. PAS combines Pronunciation Error Rate (PER), Word Error Rate (WER), Pitch Similarity, Rhythm Similarity, and Mel-Frequency Cepstral Coefficient (MFCC) Similarity. Pearson correlation analysis showed that Rhythm Similarity (r = 0.544) and Pitch Similarity (r = 0.510) exhibited the strongest positive correlations with PAS, whereas Word Error Rate showed the strongest negative correlation (r = −0.495). K-Means clustering identified three learner groups with distinct pronunciation characteristics within the tested dataset. Although evaluated on a relatively small dataset without validation against expert human ratings, the proposed framework provides a promising foundation for Computer-Assisted Pronunciation Training (CAPT) and future multilingual pronunciation assessment.
Related
- A computational comparison of handling distance constraints in MINLP
- A Computational Framework for the Moduli of Supersingular Abelian Varieties
- Global Regularity for the 3D Incompressible Navier–Stokes Equations via the Frohmanian Symplectic Tether: Lean 4 Formalization and Conditional Cubic Comparison Framework (v2.1)
- Risk Reduction Through Reverse Engineering of a Comparison Functional
- Computational framework NetBiF predicts and classifies steady-state bifurcations in biochemical systems via reaction network structure
BibTeX
@article{sunardi2026aiintegrated,
title = {{AI-Integrated Computational Framework for Pronunciation Error Analysis Through Native–Learner Speech Comparison}},
author = {Hastha Sunardi and Rendra Gustriansyah and Fery Antony},
journal = {Journal of Information Systems and Informatics},
year = {2026},
month = oct,
doi = {10.63158/journalisi.v8i5.1812},
url = {https://doi.org/10.63158/journalisi.v8i5.1812},
note = {CC BY},
}
RIS
TY - JOUR
TI - AI-Integrated Computational Framework for Pronunciation Error Analysis Through Native–Learner Speech Comparison
AU - Hastha Sunardi
AU - Rendra Gustriansyah
AU - Fery Antony
PY - 2026
DA - 2026/10/24
JO - Journal of Information Systems and Informatics
DO - 10.63158/journalisi.v8i5.1812
UR - https://doi.org/10.63158/journalisi.v8i5.1812
AB - This study proposes an AI-integrated computational framework for pronunciation error analysis through native–learner speech comparison. The framework integrates Whisper-based Automatic Speech Recognition (ASR), acoustic feature extraction, statistical analysis, and machine learning to analyze learner pronunciation. Unlike previous studies that primarily rely on individual transcription-based or acoustic measures, the proposed framework integrates multiple pronunciation indicators into a composite Pronunciation Accuracy Score (PAS) formulated in this study. The framework was evaluated using a dataset comprising 45 non-native English learners and one native English speaker. PAS combines Pronunciation Error Rate (PER), Word Error Rate (WER), Pitch Similarity, Rhythm Similarity, and Mel-Frequency Cepstral Coefficient (MFCC) Similarity. Pearson correlation analysis showed that Rhythm Similarity (r = 0.544) and Pitch Similarity (r = 0.510) exhibited the strongest positive correlations with PAS, whereas Word Error Rate showed the strongest negative correlation (r = −0.495). K-Means clustering identified three learner groups with distinct pronunciation characteristics within the tested dataset. Although evaluated on a relatively small dataset without validation against expert human ratings, the proposed framework provides a promising foundation for Computer-Assisted Pronunciation Training (CAPT) and future multilingual pronunciation assessment.
SN - 2656-5935
SN - 2656-4882
ER -