Spotlights

[Research News] Towards a Fully Automated Approach for Assessing English Proficiency

Michael McGuire, at the Department of English, Faculty of Letters, and the colleague indicated that an automated speech recognition system shows promise for making language testing more accessible and scalable.
While elicited imitation (EI) tests can effectively measure language proficiency by having learners repeat complex sentences, they require time-consuming human scoring. McGuire and his colleague have developed a fully automated system that combines EI for speech recognition with a computational scoring metric. Their study showed that the automated system produced nearly identical overall scores to human raters, suggesting that this approach could enable efficient, large-scale language assessment and potentially transform educational practices.

Reference
Michael McGuire, Jenifer Larson-Hall, Assessing Whisper automatic speech recognition and WER scoring for elicited imitation: Steps toward automation, Research Methods in Applied Linguistics, Volume 4, Issue 1, 2025,100197, ISSN 2772-7661,
https://doi.org/10.1016/j.rmal.2025.100197

This achievement has also been featured in the “EurekAlert!.”
https://www.eurekalert.org/news-releases/1080114

Title: Comparison between the proposed computer-automated assessment method and the established manual scoring method
Caption: The good linearity depicted in this plot indicates excellent agreement between manual scoring and fully automated scoring, showcasing the potential of the proposed system.
Credit: Michael McGuire from Doshisha University, Japan
Image license: CC BY 4.0
Usage restrictions: You are free to share and adapt. Attribution is required.