Original Solutions:

1) Create an app that records babies’ babbling and can detect if the infant is on track to have normal speech and comprehension abilities or whether the infant will develop SLI.

2) Use this same app to allow infants for which SLI has been detected to get extra practice responding to different speech-related stimuli.

Yes And’s:

  • “Yes, and I think you should do a combination of the both. This sounds like a great feature that could be integrated into your first solution.”
  • We think this is a good improvement to our idea as it will help both detect SLI and assist with its treatment. Even if a child does not end up having SLI, providing more dialogue opportunities for an infant will help their speaking skills in general.
  • “Yes, and can this model be used to recognize other speech disorders?”
  • For now, we are going to focus on the recognition and treatment of SLI. We imagine that there will be specific indicators of SLI that are not present in other speech disorders, so we need to have this machine learning model work specifically with SLI. If we can successfully create this SLI detection model, then we will later be able to create similar models for other speech disorders.
  • “Yes, and it can use the data on whether the baby actually did develop SLI or not to become an even more robust model.”
  • We’re glad this point was mentioned because we want to clarify how machine learning will be used. We would need to perform a longitudinal study of infants who are beginning the babbling phase and gather recordings of these infants babbling. A couple years later, we will find which infants did and did not develop SLI and input their babbling data into our machine learning model. As the model accumulates more babbling data of those with SLI versus those without SLI, it will be better able to differentiate between “SLI babbling” and “non-SLI babbling.” This is how indicators of the disorder will be discovered.

Yes But’s:

  • “Yes, but will you have to manually start the recording or is it happening all the time?”
  • This was a good point as we did not realize how exactly this would work. After discussing, we decided that the best approach would be to have a separate Alexa-like device that, when active, would be listening to any incoming sound and would distinguish and only record babbling inputs while ignoring regular speech inputs. The parents would also have the ability to turn the device off manually or even allow it to be active only certain times of the day.
  • Yes, but how long do you expect it to take for parents/doctors to get the results? If it is a long time, how will you ensure that people stick with it long enough to get results
  • The device will be structured in a way for convenience and accuracy.  Finding results and delivery should not take too long as to miss the window for diagnosing SLI in infants. To ensure this, there will initially be a “progress meter” on the app that shows, as a percentage, how much more data is needed to make a prediction about whether the infant will have SLI. Once the prediction is made, it will give a percentage likelihood of the infant developing SLI and will change as more babbling data is provided. The timing of detection will vary based on how often the device is used, but consistent use of the device will likely generate results within two to three months.
  • “Yes, but how would this app differ across languages?”
  • We were thinking of initially starting with English, and if it proves successful, we plan to expand the device’s use to different languages by creating additional machine learning models.

Revised Solution:

After the feedback we got we have decided to merge our two best solutions into one. Thus, the new solution would be:

Create a device that selectively records babies’ babbling and can detect if the infant is on track to have normal speech and comprehension abilities or whether the infant will develop SLI. There will also be an associated phone app for parents to monitor likeliness of their child developing SLI. Lastly, this app will be used to allow infants to get extra practice responding to different speech-related stimuli.