Since the inception of voice assistants like Amazon Echo and Google Home, a lot more products have voice assistants built in now – be it smartphones, headphones, various devices around the house, and even vehicles. It is estimated that by 2023 there will be 8 billion voice assistants in use!
Voice Assistants are becoming a way of life for many people. With just a simple voice command, these assistants can perform simple to complex actions – be it telling what the weather is to ordering a pizza for you. However, for people with speech impairment, using a voice assistant is not easy because of its inability to understand the person’s speech, People with Down Syndrome, for example, have a unique speech pattern that makes it difficult for voice technologies to understand them,
Why do voice assistants have trouble understanding people with speech impairments to begin with? Because today’s voice assistants are trained on what’s known as “typical speech”. There is a ton of typical speech data available for training but not enough that’s not considered typical. Because of lack of training data, voice assistants have trouble understanding people with speech impairment.
In order to change this, Google has teamed up with Canadian Down Syndrome society to make speech technology accessible for those with disabilities. Project Understood is the joint effort of CDSS and Google’s Project Euphonia that aims to get thousands of voice recordings from people with Down Syndrome to help train and improve voice technology. The project invites people with Down Syndrome or speech impairments (because of which other people cannot understand them) to donate their voices by reading and recording simple phrases that will be used tro improve Google’s system. At the time of writing this post, the project had accumulated 300 voices with a goal of 500.
Anyone over the age of 18 that has difficulties being understood by others can apply. They will be expected to have access to a computer or phone with a microphone to record their voice. They are also expected to speak English fluently. If eligible, they will receive an email from Google within 3-5 days with instructions to record voice samples. The email will also have a link to the full list of 1700 samples that can be recorded. The user doesn’t have to finish recording all of them at once but are encouraged to finish recording them over 1-2 weeks. Once recorded, Google’s engineers will process the recordings and use it to improve voice technology. For participating, each person will receive a gift card as a thank you from Google.
Watch the video below and hit the source link to learn more about Project Understood.