While some consider speech recognition a gimmick, the smartphone industry continues to develop it.
When Motorola released its Moto X smartphone earlier this month, the voice recognition feature was considered one of the most noteworthy. The first device Motorola conceived and built since Google acquired the company, the Moto X used voice-recognition technology to access Google Now, a digital personal assistant of sorts that makes recommendations based on the user’s search history.
All a Moto X user has to do is say “OK Google Now” while in the vicinity of the phone, and the personal assistant would activate to execute whatever command followed. That means no clumsy typing on a touchscreen while the users’ hands are busy, like when they’re driving, for example. David Pogue at The New York Times called voice command feature the phone’s “most useful,” claiming that its no-touch capability makes it a safer and more convenient alternative to Apple’s Siri.
Clearly, voice recognition technology has made some progress in the past few years. So how come nobody seems to want to use it?
Adib Ghubril, a research director at Gartner, recently tried to use Siri while on vacation to get directions to Chateau Laurier, a historical hotel in Ottawa. After Siri failed to process the request several times, Ghubril found that it worked when he stopped saying the word in French and instead pronounced it the way an English speaker would interpret it.
“It was unbelievable how the tone is expected to be from English, with an English bias,” Ghubril says. “If you think about these quirks, and how far the technology still has to go – if I’m francophone, I’d be irritated.”
Although Ghubril says voice recognition technology in general has made some good strides in the past few years, he points to these inconvenient shortfalls as a barrier to mainstream adoption.
“If you’re using your phone as kind of a virtual assistant, you have to change the way you think a little bit,” Ghubril says. “You have to change the way you say ‘um,’ you have to change the way you talk, you have to think about sentences, because as you pronounce them, your virtual assistant is analyzing them. So certain personal behaviors change a little bit and have to change a little bit.”
Rather than adapt to the technology, users may just abandon voice command tools at the first sign of an issue.
“I think when you look at it from a user perspective – when it works well, it’s great. When it doesn’t, it’s a huge turnoff,” Carolina Milanesi, a research vice president at Gartner, says.
Milanesi saw Google Now voice recognition in action at the Google I/O conference in May, and although she thought the demonstration was great, she couldn’t help but wonder how much the performance owed to the particularly clear-speaking woman who gave the demonstration.
“I’ve never heard anyone speak so clearly on a stage. The reality is that we’re not like that,” she says. “Apart from accents and jargon, there are broken sentences. There’s ‘um,’ and pauses, and things like that. So how clever is the engine to allow me to find the things that I want? So it’s going to take a little bit more time, and also trust. You know, trust in what you get is going to be what you want.”
Until phone developers can establish that kind of trust in their voice-command tools, users will see the technology as just another gimmick, Milanesi says.
Ghubril, meanwhile, says speech recognition is becoming more useful every day. Indeed, Gartner has identified speech recognition as one of the emerging technologies that are beginning to become productive.
More specifically, Ghubril says the kind of speech recognition that can be used effectively in enterprise settings is already possible, and will be used on a day-to-day basis within the next two years.
“In other words, let’s say I’m a medical technician, and I work in a very specific environment and I deal with specific day-to-day things,” Ghubril says. “It’s possible to build an application for me where all the repertoire of the words that I use that are specific to this application can be covered and supported. So I can speak quasi-naturally into the thing, and it’ll know what I’m talking about.”
The more ambitious, natural-language speech recognition that smartphone developers would need to accommodate the idiosyncrasies of their users’ speech, to the point that a device can immediately understand users’ commands and put them in context, is more like five to 10 years away, Ghubril says.
However, that doesn’t mean speech won’t be successful with mainstream markets. Ghubril says that although “we had a good time making fun of Siri,” speech comes too naturally to humans for the technology to fail. People already yell at their phones or their PCs when they aren’t working, even when they know that won’t fix anything. Considering how error-prone and labor-intensive touchscreen navigation can be, particularly when it comes to type-based functions, Ghubril says the obvious next step is to incorporate speech. Just as the combination of a keyboard and mouse was needed to make a PC useful, touch-based devices will include touch to reach their full potential.
Colin Neagle covers emerging technologies and the startup scene for Network World. Follow him on Twitter and keep up with the Microsoft, Cisco and Open Source community blogs. Colin’s email address is cneagle@nww.com.
