CBC Interviews Our Founder On Voice Dictation

CBC Interviews on Voice Recognition.jpg

“OK Google I’m listening”

“Flight to Portland is delayed by 30 minutes”

“Change my dinner reservation for tonight from 7:30-8:00”

In the commercials for Google Home like this one and the Amazon series of Alexa Devices, the people who talk to them are always clearly understood even when they speak quickly.

“Alexa ask Dominos to send me my last order”

“Okay, order placed”

You might think that fast talkers would be more of a challenge but as it turns out in some cases the opposite is true.

“Alexa…What is the Capital of…Canada?”

“I think I missed part of your question, try asking it again.”

“Alexa, what’s the Capital of Canada?”

"Canada's Capital City is Ottawa”

Talking naturally is tip number one when it comes to interacting with these devices, that is according to Shawn Wilkie. He's a software developer who specializes in voice-recognition. He lives in Canada but vacations in Nicaragua, which is where I reached him via Skype.

"Hey this sounds pretty good."

"Good stuff!"

While abroad, Wilkie says he uses voice-recognition on his phone all the time.

“So my Spanish is very broken so in Google Translate there is a new feature, you just flick a button and it listens for both languages at the same time. So if I say "how much" it says “Cuánto cuesta” and the cab driver will say a price “$5” and then I’ll say “okay” and if it says “Bueno” and it listens, then and speaks.

Just like real people, these latest technologies are better at drawing meaning from full sentences rather than individual words so don't slow down.

"It struggles with it because speech recognition nowadays is contextual, so it is looking at the words behind and in front of the words you are saying to make it make sense and it all of a sudden you stop in the middle of a sentence, the contextual part of the speech recognition just fails."

Wilkie says talking naturally is easier with Google Home and Amazon's Alexa because they don't have screens, there is nothing to look at, you have to say what you want and wait to see if you are understood. He said people should apply that same strategy when using a voice to text on their phones.

"Don't look at the screen when you are dictating, dictate freely like you are having a conversation like you and me are right now. Give yourself the ability to complete a thought and then take a look at the screen."

Otherwise Wilkie says people tend to watch to see if each word they are saying is being processed and that slows her speech which makes it harder for the device to understand you. Another tip, be specific;

"Alexa, who is Canada's Prime Minister?"

"The Prime Minister of Canada is the primary minister of the crown, chairman of the cabinet..."

"Alexa stop!"

"Alexa, who is Canada's current Prime Minister?"

The current Prime Minister is the Liberal party's Justin Trudeau who was appointed on November 4, 2015."

Sometimes it is just a single missing word that results in a miscommunication. Shawn Wilkie says being unintentionally vague can have real consequences. He told his phone to call a local pizza shop, but wasn't specific about which location. He ended up ordering from the wrong place. So talk normally and be specific and voice-recognition will deliver.

"Alexa repeat after me:
For CBC Radio, I'm Blair Sanderson"

"For C B C radio I'm Blair Sanderson."