Teen Interns at Stuttering Research Institute
Teen Interns at Stuttering Research Institute
Dear Reader,
I recently enjoyed a two-week internship at Hollins Communications Research Institute (HCRI), during which I learned more than I hoped and felt more inspired than I ever thought. I received an endowment, the William "Bill" Reeves Renaissance Student Award and Endowment, from my high school to support my summer research at HCRI. Over the course of two weeks, I was on-site during one of HCRI's two-week therapy sessions. I worked under the guidance of Dr. Webster's exceptionally dedicated team of experts at HCRI who graciously accepted and patiently taught me about their research and plans for advances in stuttering research. My research was grounded in the goal to develop quantitative measures of voice onsets made to four classes of speech sounds by stuttering clients before and after participation in HCRI therapy program. While conducting my research and work, I learned about the inner-workings of the institute and observed clients' experiences. I also had the opportunity to interact with clients to collect data for my research.
I worked with Kevin, the lead and highly experienced Information Technology Specialist at HCRI, to collect each client's consent and record them reading aloud 10 monosyllabic words. Each utterance was entered into FluencyNet, a proprietary fluency-detection software designed by Kevin and used by clients throughout the duration of their program. The program then awarded each utterance a numerical grade to differentiate and classify the onset of each sound. At the conclusion of the two weeks, this process was repeated with the same 10 sounds but shuffled. The premise of this project was to determine the potential of the program to help soften the clients' onsets.
Gentle onsets are an essential component of an HCRI graduate's toolbelt, given that gentle onsets are a common part of phoneme and sound formation for speech. But stutterers require explicit instruction to identify and adopt the strategies needed to create these critical onsets. Many stutterers and the clients are initially unable to perform gentle onsets, so an essential part of the intervention program is to teach them how to construct sounds with gentle onsets. Proficiency in gentle onsets is typically demonstrative of proficiency in speech and fluency.
A substantial portion of my research was spent working on other methods to determine onsets and fluency beyond the limits of only FluencyNet. Using Apple's XCode programming engine, I was able to access Create ML, a program that has sound classification training models that can be taught to recognize certain sounds, for instance, a drum or guitar. Using speech data from the 10-word recordings and past recordings of clients reading passages, I separated each utterance and fed them into the program and categorized each utterance into files of fluency and onset classification. Ultimately, the trained program could largely recognize soft or hard onsets when given a recording but failed when it came to detecting stuttering or when given a live recording. Ultimately it seems the program is not sophisticated enough yet to differentiate similar sounds like different classifications from the same voice, but it has an exciting future nonetheless.
I plan to continue to remotely collect client data with the intention to apply my findings to my senior year independent research. Again, I am so appreciative of Dr. Webster's talented team at HCRI for this exceptional opportunity. It was a privilege, especially as a high school student, to access HCRI clinicians' exceptional expertise and experience. I will be forever grateful to Dr. Webster for his leadership and guidance, Kevin for his patience and generosity, and Linda for her limitless encouragement and confidence in me.
Stan Craig
SSEAT Founder and Director
June 2022