Saturday, August 24, 2013

State of the Art: The Voice-Off: Android vs. Siri

That’s the kind of e-mail that brightened my day after I reviewed Google’s Moto X phone two weeks ago.

My correspondents seemed especially unhappy with one sentence in that review: “Android’s voice commands are still no match for Siri.”

Man, I really was stupid. Who’d be dumb enough to take sides in a religious war? I’d have been better off writing, “Conservatives are better-looking than liberals” or “Pro-life people are worse drivers than pro-choice.”

But the superiority of cellphone speech-recognition technology is not an idle question. Once touch screens became the future of phones, voice recognition became desperately important. Without physical keys or buttons, entering text and manipulating software controls are fussy, multistep procedures.

So I’ve just spent two weeks immersed in voice recognition. I carried an iPhone and a phone running Google’s Android operating system with me everywhere. I spoke to both phones simultaneously. I wanted to get to know the differences, the strengths, the weaknesses.

When people talk about speech recognition, they mean, and often confuse, three different functions. There’s dictation, where the phone converts speech to text; commands, where you operate the phone by talking; and Internet information searches. There are vast differences among the successes of the three.

Dictation, for example, is still fairly poor on both systems. Both Android phones and Siri, the iPhone’s speech feature, make many transcription errors. When you hear people bashing cellphone transcription, declaring, “I gave up on it,” they’re usually referring to dictation.

That’s forgivable, but come on. You’re asking your phone to understand varying accents at varying distances from its microphone, in rooms with varying background noise. It’s a wonder this feature works at all.

The latest Android version doesn’t require an Internet connection to do basic dictation. And in Android, the words appear on the screen as you utter them; Siri doesn’t transcribe until you stop talking.

On the other hand, Siri understands formatting controls like “capital,” “all caps” and “no space,” as well as all kinds of punctuation — “colon,” “dash,” “asterisk,” “ellipsis” and so on. Android understands only the basic symbols, like “period,” “comma” and “exclamation point.”

The second category, phone-control commands, is far more successful for far more people. This is when you say: “Call Mom,” “Text Emily,” “Wake me at 7:30,” “Play some Billy Joel,” “Remind me to feed the cat when I get home,” and so on.

Controlling your phone without touching it is important for safety, of course. If you must interact with your phone while driving, speaking to it certainly seems safer than looking at it.

But don’t forget the convenience factor. It’s much faster to say, “Open Angry Birds” than to flip through home screens full of icons. And “Set my alarm for 8 a.m.” is about 375 finger-taps quicker than using the clock app.

Here, Siri has the edge. As you’re driving along, for example, and you hear the incoming message sound, you can say, “Read my new messages,” and Siri reads them aloud. It even invites you to dictate a reply, without ever taking your eyes off the road. Android can’t do that.

Both systems can tap into some of the phone’s own apps. They recognize commands like “Make a meeting with Bob Barnett Thursday at noon” (a calendar interaction), “Make a note to pay back Harold” (notes), “Send an e-mail to Danny Cooper” (mail) and “What’s Steve Alper’s home address?” (contacts).

Android blows away iOS, though, in Web searches. Both kinds of phones do an amazing job fetching weather updates (“What will the weather in Detroit be this weekend?”), times (“What time is it in Belgium?”), stock prices, sports information (“When’s the next Cowboys game?”), conversions (“How many dollars in 32 euros?”), calculations (“How many days until Valentine’s Day?”) and every kind of Web-search query (“How many calories are in a Hershey bar?”, “When is the next solar eclipse?”, “How do you spell schadenfreude?”, “Show me pictures of a 1985 Corvette,” and so on).

No comments:

Post a Comment