Have you talked to your computer lately?

Do you talk to your computer?

Microsoft's Cortana, Google Voice on Android, Amazon's Echo (and to a lesser extent) Apple's Siri are pretty good at recognizing human speech. Cortana even has a rudimentary sense of humor. For example ask Cortana to “open the pod bay doors”.

Two things actually happen to make voice recognition work.

• The first is getting the computer to recognize the phonemes. These are the individual sounds that make up words. While there are 26 letters in the English language there are between 40 and 44, (depending on who you ask) distinct phonemes or sound units. These phonemes are mixed together and co-articulated into the words we use. While young children pick this up rather easily getting the computer to correctly recognize speech is a little more difficult.
• The second part is converting the speech to text. The computer then filters out the so called “stop words” like and, or, the and so on then it tries to select the words that describe the meaning of what you said. Like if you were talking to your dog and said something like, “After I get done working we can go for a walk.” The dog hears a bunch of blah blah blah blah go walk. The dog doesn't care about all those other words just the “go walk” part. Like the dog your computer is searching for the important words.

Talk to you computer.

While this may seem simple to most humans and their pets it actually takes some rather advanced artificial intelligence for the computer to get this right.

In fact with the October 2018 update you can control your Windows 10 computer using only your voice. You can open and close apps, search for files on your computer or on the Internet, and dictate text anywhere you would normally type it like documents, searches, text input boxes and e-mails. This feature is available for computers that have English as the display language. Support for other languages is coming. This is good for seriously the handicapped people but also for people with lesser handicaps like just being a poor typist.

You can also control your Android, (4.1 Jelly Bean or newer) device entirely by voice commands using the Voice Access app. The setup process may require a little patience but one you have done that voice commands can allow you to unlock the phone with your “trusted voice”. This can actually be more secure than face recognition.

Other tips to improve voice recognition:

The main catch to all of this is you will need a good microphone. Most smart phones already have a good microphone, most desktop computers do not. A good microphone usually means a headset. Desktop microphones can be used but remember to speak to them. Don’t turn away to look at the clock while you are talking. Also noisy environments can interfere with the voice recognition.

Use the Voice Training feature in Windows 10. This will help your computer better understand how you speak. Faults like various accents, talking with your mouth full or dropping the trailing “g” as I often do can be overcome with a little training session to teach the computer what walkin, talkin and chewin gum actually mean.

Keep it simple. If your car breaks down don't ask Cortana to “find me a tow truck” or it will start searching through you device looking for documents or e-mails with the words tow truck because you used the word “find”. Just say “Tow Truck” and let Cortana do the rest using your location to find a tow truck near by.

You may have noticed that Apple/Mac was scarcely mentioned in this article. That is because to-date they have yet to get Siri to work right. At this time Microsoft and Google are way ahead of Apple in the Artificial Intelligence game.

Go back to the top