Currently submitted to: JMIR Formative Research
Date Submitted: Sep 3, 2026
Open Peer Review Period: Sep 7, 2026 - Nov 2, 2026
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Modeling Speech Fluency Improvements Following Personalized Stutter Mitigation Training with Self-Supervised ML Models: A Preliminary Study
ABSTRACT
Background:
Stuttering is a neurodevelopmental speech disorder, with several non-pharmacological treatments such as operant conditioning and speech restructuring. However, traditional approaches consistently note high individual variation in treatment response, highlighting the need for personalized therapies targeting each speaker’s disfluent phonemes.
Objective:
This formative research study aims to evaluate the preliminary effectiveness of APT (Adaptive Phonetic Training), a personalized AI-assisted approach for stutter mitigation prototyped and deployed as the OratorSync web application. The objective is to determine the approach's efficacy in decreasing the frequency of stuttering events for specific disfluent phonemes and improving overall speech fluency across the alphabet.
Methods:
Adult participants (N=14) were recruited through the National Stuttering Association. Participants used the OratorSync app for four weeks. The app prototype functions as a digital speech coach, using the W2V-BERT machine learning architecture to detect stuttering and increase the frequency of phonetic speech exercises on the user’s high-disfluency sounds. Pre-test and post-test fluency scores, indicating the inverse frequency of stuttering, were captured and compared using paired sample t-tests.
Results:
At the letter level (N=26), average fluency scores significantly increased by 13.5% (P < .001). At the participant level, overall fluency improved from a mean of 72.0% to 84.2% (P < .001), indicating a 12.2% fluency improvement and a significant reduction in stutter frequency. The Mean Absolute Deviation (MAD) of participant improvement scores was 4.95%, demonstrating that the participants across the cohort consistently improved fluency as opposed to facing the individual variation observed in traditional treatments.
Conclusions:
The Adaptive Phonetic Training (APT) approach presents a feasible and accessible approach to personalized stutter mitigation, with prototype web deployment and significant preliminary success. Though rigorous future trials must be conducted to validate these results and mitigate potential selection bias, this study serves as a strong preliminary signal to justify and inform those future investigations.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.