Degenerative diseases like ALS are known for causing speech impairments. Even with how good the voice recognition technology has gotten, if the vocal input isn’t loud or clear enough, the A.I. system fails to recognize and compute tasks. Because of this, Google is working on Project Euphonia.
Voice recognition uses training from essential voice recordings that sound like a typical speech, without much deviation in accents or speech deficiencies. Therefore, researchers recorded many different people with different speech conditions to train its a.I. However, it was not the ideal solution since each person for ALS, for example, has a unique speech deficiency based on the progression of the disease. So google decided that to reduce errors, it used a baseline voice recognition model alongside different tweaks.
The A.I. has made quite a significant progress; however, it still struggles somewhat with phonemes. Because of this, the A.I. uses context to determine and correct the word in any given phrase. By analyzing the entire sentence, the program can correctly replace any wrong word with the correct one.
This is an excellent effort by Google to improve its voice recognition software further. Including more people who can use this will bring them into the Google ecosystem, and make it easier to use technology. Google isn’t the only that is working hard on voice A.I., in fact in June Apple announced that you would be able to control your entire Mac by just your voice, no need for a mouse or keyboard.