Stuttered Speech Data for Stutterers, with Stutterers, by Stutterers: Insights from Stuttering Advocates

Automatic Speech Recognition (ASR) systems have long struggled to accurately recognize and process stuttered speech, largely due to their training on fluency-biased datasets. As a result, people who stutter (PWS) frequently encounter challenges when using voice-activated technologies, automated transcription services, and other speech-based AI tools. This lack of inclusivity and accessibility also reinforces social and technological barriers for people with speech differences. Funded by the Patrick J. McGovern Foundation, our project seeks to change that by creating a representative, community-driven dataset of stuttered speech.  

Centering the perspectives of the community from the very beginning, we conducted in-depth interviews with eight stuttering advocates, including researchers, nonprofit leaders, and community organizers to gather their insights on how to structure a community-centered data collection that is ethical, empowering, and inclusive. Six of our interviewees are PWS and two are long-term allies for the stuttering community.

Background of the Stuttering Advocates

• Maya Chupkov – PWS, CEO & founder of Proud Stutter, a podcast challenging societal perceptions of stuttering, and filmmaker working on a documentary about race and stuttering.
• Naomi Rodgers, Ph.D., CCC-SLP – PWS, assistant professor at the University of Iowa and leader of the Iowa Stuttering Lab, focusing on clinical intervention, counseling, and community support.
• Edmund Metzold – PWS, senior healthcare technology specialist and leader in the Passing Twice LGBTQ+ and stuttering support network.
• Jia Bin, M.A. – PWS, Ph.D. researcher at Michigan State University, specializing in stuttering research, and co-founder of StammerTalk (口吃说), an online community for Chinese-speaking PWS.
• Rong Gong, Ph.D. – PWS, principle research scientist at Microsoft and co-founder of StammerTalk (口吃说), a grassroots online community for Mandarin speaking people who stutter. With the support of StammerTalk community, Rong led the development of AS-70, the first and largest Mandarin stuttered speech dataset for ASR.
• Tracy Wang, Ph.D. – PWS, Principal Applied Scientist at Oracle and a core member of StammerTalk (口吃说), co-leading the development of AS-70.
• Nan Bernstein Ratner, Ed.D., F-, H-ASHA, F-AAAS, Board Certified Specialist in Stuttering, Cluttering, and Fluency Disorders – Professor at the University of Maryland, expert in fluency disorders, and co-director of FluencyBank, one of the most popular stuttering research corpus.
• Aidan Sank – Executive Director of SPACE, a nonprofit dedicated to increasing visibility and agency for PWS through storytelling and community engagement.

Community Perspectives on Empowering Data Collection Process

Our work is deeply valued by the stuttering community, as reflected in the voices of our interviewees. Nan Bernstein Ratner, professor in the Department of Hearing and Speech Sciences at the University of Maryland, College Park, commented on the potential impact of the project and shared her past data collection experience:

“I think it’s a really good idea. We definitely have done it before. I’m happy to consult with you on how to do it again.”

Jia Bin, a speech and language researcher and co-founder of StammerTalk (口吃说), emphasized,

“I love working with you because of the human-centered, stutter-centered approach. That’s why we do what we do. You are creating a space where people who stutter can contribute in a meaningful way, and I fully support that.”

Edmund Metzold, co-founder of Passing Twice, an LGBTQ+ and stuttering support organization, also noted:

“Just the fact that the dataset of people who stutter is created by people who stutter, I think, will naturally appeal to many people.”

This community-driven approach ensures authenticity and trust, as echoed by Aidan Sank, Executive Director of SPACE, a nonprofit organization working to create more space for stuttering and change the way the world listens, commenced on our data collection plan,

“The data collection plan sounds effective, and deeply needed. I think this is going to make a huge difference, and I love that you are engaging with such a community-centered process. The combination of unscripted conversation and voice command recitation will create a really robust dataset for learning and development.”

Rather than treating data collection as a transactional process, we focus on the personal benefits of this data collection process to people who stutter. Maya Chupkov, founder of Proud Stutter, emphasized the design of spontaneous conversation of our data collection process:

“Personal interaction—introducing yourself, sharing a bit of your background, and explaining why you’re doing this—could make participants feel more empowered and invested.”

Naomi Rodgers, a researcher leading the Iowa Stuttering Lab, commented on our plan of having people who stutter as data collectors,

“If you’re aiming for this to be community-led and to capture as much stuttering as possible, then having interviewers who stutter could be a good approach. It helps create a sense of safety within the interview.”

Summary and Next Steps

Our research highlighted the importance of recruiting both PWS and allies as data collectors, diversity in participants, and designing tasks that encourage natural speech patterns. Additionally, we are prioritizing transparent data governance, fair compensation, and ethical consent practices. We have already made significant progress:

• Finalizing legal agreements for data contributor and usage.
• Recruited a diverse group of participants and data collectors.
• Designed a balanced set of speech tasks that authentically represent stuttering patterns.
• Implementing improved annotation strategies to accurately capture stuttering events.

As we begin data collection in the coming weeks, we remain committed to an inclusive, ethical, and community-driven process. We are grateful to our stuttering advocates for their guidance and look forward to continued collaboration in making ASR technology more accessible for all. For a detailed breakdown of our findings, including insights from each interviewee, please continue reading the full report here.

Contributing Your Voice

If you are a person who stutters and are interested in contributing your stuttered speech, please complete this survey to sign up.