What Siri has taught us about voice recognition

Norman Winarsky acknowledged the shortcomings of Apple’s Siri last week, saying the voice-activated personal assistant had been stretched “beyond its initial capabilities.” Siri was initially developed for travel and entertainment, he said, but consumers have tried to use the software for everything from scheduling meetings to pondering the meaning of life.

Winarsky knows what he’s talking about: He’s the vice president of SRI Ventures and was a co-founder and board member of Siri, the developer of the eponymous technology that Apple acquired for a reported $200 million in 2010. Winarsky went out of his way to spare Apple from any blame for those broader uses, but I don’t agree: I think Apple tremendously oversold Siri’s capabilities with big-budget commercials like this one, in which Samuel L. Jackson tells his phone to “Find me a store that sells organic mushrooms for my risotto.”

With Siri, less = more

Those ads give users the impression that Siri can perform a broad range of functions including search accurately and easily almost regardless of how users talk to it. As any Siri user knows, though, the technology simply isn’t good enough to talk to as if it were human. I’ve tried repeating Jackson’s risotto line to Siri at least a half-dozen times and have yet to receive accurate results. (It usually directs me to organic-food stores in Arizona.) The disconnect between Siri’s real-life performance and Apple’s commercials led former Apple executive Jean-Louise Gassee wrote several weeks ago to write that Apple should pull the ads: “Doubtless, (Siri) works for some people,” Gassee wrote, but how many?” Gassee wrote. “How many give up after a few tries?”

And the heightened expectations for Siri have resulted in some disappointed users. A survey of 482 iPhone 4S users by Parks Associates earlier this year found that while 87 percent of respondents actually used Siri each month, only 55 percent are happy with it; 9 percent remain unsatisfied. And it’s worth noting that many Siri users have learned not to expect too much from the technology: Roughly one-third of regular users tap Siri only for making calls, searching the web or sending text messages. Thirty-five percent say they would never use it to schedule meetings, 30 percent wouldn’t send e-mail with Siri and 32 percent wouldn’t even use it to play music. So it’s understandable that Nick Bilton of The New York Times would write a few months ago that users’ love for Siri has soured.

Voice recognition: Very powerful, but very flawed

The key for Apple to get as many people as possible to use Siri as much as possible, then, is to ensure the best possible user experience. And that can be done by encouraging the kind of use cases where Siri is most likely to succeed: Performing a web search using just two or three words rather than a complicated sentence, for instance. Finding an email address from an address book, or getting the score from last night’s Yankees game. Or seeing what the weather’s like in San Francisco.

And Apple isn’t the only player that should be encouraging consumers use voice-recognition technology only in the most appropriate cases. Google should be taking that same tack with Android’s voice search offering, and developers looking to leverage technologies such as Nuance’s new Nina should make sure their apps deliver the best possible experience – that they capitalize on the power of voice without asking too much of those technologies. Because if there’s anything Siri has taught us, it’s that voice recognition software is a very powerful – but very flawed – technology. And it has a long way to go.

Relevant Analyst
Colin Gibbs

Colin Gibbs

Mobile Curator Gigaom Network

Do you want to speak with Colin Gibbs about this topic?

Learn More
You must be logged in to post a comment.

No Comments Subscribers to comment

Explore Related Topics

Latest Research

Latest Webinars

Want to conduct your own Webinar?
Learn More

Learn about our services or Contact us: Email / 800-906-8098