Episode 50: Google’s Project Euphonia: a research initiative to make speech recognition possible for people with impaired speech with Julie Cattiau & Katie Seaver, MS, CCC-SLP

 

Automatic Voice Recognition has become a tool for all of us to access our technology and our environments. However, for people with speech impairments, this is not always easy.  Google has committed to a research project to determine how to make speech recognition accessible for all through Project Euphonia. You will learn how speech recognition models are currently built and the need for data (speech samples) from people with impaired speech to retrain current models.  Google’s Project Manager and a supporting SLP discuss the who, what, where, when, and why about Project Euphonia.

topics covered:

  • How Google is researching the utilization of A.I. (artificial intelligence) to improve speech recognition for people with impaired speech.

  • What Project Euphonia needs to work: speech samples.

  • Discussion on traits of potential candidates, how to submit speech samples, and how SLPs can support this endeavor. 

  • Brief discussion on the YouTube video, “Healed through A.I.” which chronicles the work of Project Euphonia and DeepMind.

  • Katie discusses her involvement as the SLP consultant on Project Euphonia.

  • How the finished product of Project Euphonia could benefit individuals with dysarthria.

  • Positive, future impacts of the finished product of Project Euphonia on SLPs and our practice.

 

Resources to dig deeper:

How AI can improve products for people with impaired speech” by Julie Cattiau

Project Euphonia Purpose, Scope, and Timeline: all your questions answered

Project Understood is the awareness campaign that was spearheaded by the Canadian Down Syndrome Society and this is their website for the campaign.

Watch the linked video to see Project Euphonia in action with The YouTube Originals, The Age of AI: Episode 2  (more context provided with this Google blog post).

If you'd like more insight into the technical workings on the voice synthesis project, here is the DeepMind blog post about it, which is the team Euphonia partnered with.

Interested in partnering with Google to assist the research? Check out: g.co/euphonia to take to the SLP survey and access the interest form for participants. 

Partnering with Project Euphonia:


Julie Cattiau is a Product Manager in Google AI. In her current role, she focuses on building technologies that can help users with neurologic conditions (e.g: ALS, stroke, down syndrome...) communicate more easily through improved speech recognition. She is also the Product Manager for the bioacoustics project, which goal is to use AI to help monitor endangered species thanks to sounds. Prior to this role, she was the product manager responsible for the Google Translate mobile apps.

Julie has a master of science in computer science from French engineering school Telecom ParisTech.

juliecattiau.jpg

KatieSeaver.jpg

Katie Seaver is a Speech Language Pathologist and AAC Specialist.  For over 16 years Katie has focused on adult neurorehabilitation.  Most recently, at the Leonard Florence Center in Chelsea MA, Katie works with residents living with ALS and MS who believe that until medicine proves otherwise, technology is the cure. Earlier this year, Katie joined the Project Euphonia team at Google, acting as an SLP in residence for the project. Katie believes that engaging the SLP community will be integral to making technology useful in the daily lives of people with communication disorders.


SUBSCRIBE & REVIEW

Have you subscribed to the podcast? Subscribe today on Apple Podcasts/iTunes and future episodes will magically arrive on your doorstep/iPhone. For Android users, check out the 10 options available to you to listen!

You’ll earn my undying gratitude by leaving a review on iTunes! Reviews help other SLPs find the podcast. Also, I love reading your feedback! Just click here to review, select “Ratings and Reviews” and “Write a Review,” and let me know what your favorite part of the podcast is. Malaho plenty!