Voice Recognition on the iPhone

Posted by: Arik Hesseldahl on July 29, 2008

From our intern Jay Yarow:

Recently I spoke with a sports industry executive who purchased an iPhone 3G. While he loves the gadget, he isn’t abandoning his Blackberry because the iPhone’s touch screen hinders his ability to write email.

The touch screen complaint is all too common. When typing email you either like Apple’s on-screen QWERTY keyboard, or you don’t and prefer the Blackberry, the Treo or something else, no matter what great applications Apple puts on the App Store.

Eventually, Apple may build a model with a keyboard. Or not. It’s hard to say about a company whose CEO is said to despise buttons. But wouldn’t it be faster if you could speak your emails directly into the phone and have your words turned into text?

Meet Steve Chambers, president of Nuance Communications’ mobile division, the company behind Naturally Speaking, a voice recognition software package for the PC. “Voice is the answer to restrictive keyboards,” on mobile devices he says. Nuance has software on nearly 3 billion handsets worldwide. The firm developed owns the T9 predictive text software, which is the technology that tries to figure out the word you mean to use in a text message on many wireless phones. Of Nuance’s $900 million in total revenue, about $200 million comes from mobile software on mobile phones.

Why not speak what you mean to say into the phone rather than fumble with the tiny keys to type it? “Virtual keyboards are tough, we are actively working on active dictation for mobile devices,” says Chambers. In particular he sees the iPhone as a great opportunity.

A user can say, “Byte of The Apple” and a list of search results, culled from Google appears as a list. Click on the result and Safari opens up with the web page. Likewise, the search function works with maps. For instance, say, “1221 Avenue of the Americas, New York,” and you could click on an icon that shows you BusinessWeek’s New York offices on a map.

Nuance is still deciding what it’s going to do with this application. “The App Store is so new,” says Chambers, “how good a vehicle is it for us” as compared with selling to carriers? Or Nuance can work with software developers. “We can make the API available to people to integrate it,” Sejnoha says. “It could be the front end for someone like Powerset,” he says, referring the company that’s working on natural language search queries.

Chambers and Vlad Sejnoha, Nuance’s Chief Scientist demonstrated two prototype applications, one running on the iPhone, the other on a Treo, showing the potential, which you can see in a video (sorry, the close-ups are a little blurry) after the jump.

Chambers hopes to have dictation software for the iPhone, like these software featured above, ready by the beginning of 2009.

The great hurdle for Nuance (and others) is the relatively puny computing power of a mobile phone, when compared to a notebook or desktop. While the iPhone sports a beefy Samsung processor, its not nearly as powerful a the processor in a Mac or PC. To translate speech being dictated into the body of an email, you need some serious processing power. Nuance sends the speech to a server, which transforms it into text, then sends the text into the body of an email. The speed of the connection, determines the speed of the conversion.

There are already a few programs that do something similar on the iPhone. Jott turns short speech recordings into text that can then be sent by email. Unfortunately it takes 5-10 minutes to get the text translated and the message sent. Not ideal for when you’re in a hurry. There is also Reqall, which does a lot of the same things that Jott does. Both have their own dedicated iPhone App. Both hand off the voice message to a server in the cloud that does the heavy lifting required to turn the voice message into text.

One other issue: Bandwidth. Sending voice to a server in the cloud and getting a quick response takes some serious speed to be useful. You’ll need a solid 3G connection that isn’t always available or a Wi-Fi connection. Without it, you’re probably best taking your chance with the virtual keyboard.

TrackBack URL for this entry: http://blogs.businessweek.com/mt/mt-tb.cgi/