Google rolls out a new App for people with speech impairments

A beta app developed by Google for persons with speech impairments has just been released, and it will be used as a voice assistant while also contributing to a multiyear research project to improve Google’s speech recognition. It is, therefore, seeking assistance creating the Android app in order to provide more communication alternatives for persons with speech problems. 

Users with neurological problems that impair their ability to speak will benefit from the enhancement of Google Assistant, as well as other speech-to-text and speech-to-speech functions in general. 

It is hoped that Project Relate, as the initiative and app are now known, will provide speech transcription and synthesis, making it easier for users to be understood. 

Volunteers can sign up for Project Relate at g.co/ProjectRelate. They must be at least 18 years old and “have difficulties being understood by others” to be eligible. A Google account and an Android phone running OS 8 or later are also required. English speakers in the United States, Canada, Australia, and New Zealand are the only ones who can access it for the time being. They’ll have to record 500 phrases, which should take between 30 and 90 minutes to complete. 

Due to the large amount of data required to train a speech recognition engine, the results are biased toward typical speech patterns. Those with accents are underrepresented in these data sets, and as a result, they are not as well understood— and people with speech impairments are even less frequently represented, making it virtually impossible for them to use typical voice-powered gadgets. 

However, capturing and analyzing the highly distinctive speech patterns of persons with impairments and disabilities requires extra effort. Every voice is unique, but unusual patterns like those caused by a stroke or injury might be difficult for machine learning to grasp. 

A improved voice transcription tool for persons with speech problems is Project Relate’s core. It can be pasted or read by others using the “Listen” option. After listening, “Repeat” repeats their words in a louder voice. When they ask Google Assistant to play music or tell them the weather, they can just say “Assistant”.

Before implementing these features, Google researchers say they generated a library of over a million volunteer speech samples. In order to train the speech-recognition AI, this was employed. Of course, more data is better in ML, and more particular data is better in ML.

Google is looking for a group of people who will use the app on a daily basis to test it out. So that the speech model can better match their speech patterns, they must first record a series of phrases. It’s not too late to sign up as a volunteer if this app has the potential to be useful in your day-to-day activities.