Google Adds Voice Recognition to iPhone App

News from Google

I have a very good feeling about Google’s new iPhone app that does voice recognition. I’ve been playing with this voice recognition application for several weeks and I have to say that I’m really impressed. First and foremost, the voice recognition works really well. Crazy long-tail specialized vocabulary is tricky (more on that later), but for queries with normal words in them, the voice recognition is really accurate and I think it will get even better. You can say “population of Troy, New York” and you’re pretty likely to get good search results:

I like the slick interface, because all you have to do is start the app. When you want to do a search, just hold the iPhone to your ear. The iPhone’s accelerometer senses the movement and makes a “baBUM” noise to let you know when to talk. Then just say a query like [daffodil pictures] or whatever. It’s much smoother to experience than it is to write down. The net effect is as if you had some kind of Star Trek communicator device, except powered by Google instead of Spock and the rest of the crew.

I’m really impressed with the team that worked on this. They pulled this together in a very short time and they’ve produced an extremely polished application. Every time I’ve emailed someone with feedback or a question, they replied quickly, but also thoughtfully. It’s clear to me that this application is a labor of love and they want it to be outstanding. I can’t wait to see what they do next; maybe the application can start to learn your specific voice and its inflections?

Do you remember Battelle and O’Reilly’s definition of Web 2.0? At the risk of mangling it, they define Web 2.0 as “applications that harness collective intelligence (either implicitly or explicitly) to get better as people use them.” I expect the voice recognition of Google’s search app to get better as people use it, much in the same way that the intent of GOOG-411 was partially to improve our text-to-speech models.

The last thing I like is subtle. This app has changed the way that I do queries. I’ll be the first to admit that I’m a huge search geek. I’m hyperaware of when my query habits change, and I notice myself much more likely to do off-the-cuff queries such as [what's the average price per square foot for carpet?] or [how many miles per gallon do Audis get?] or [what is a softshell jacket?] or [what does the 11-99 foundation stand for?]. Marissa Mayer once kept a diary of all the searches she wanted to do during one day, and mentioned about 20 queries that came to mind. I feel like I had almost that many queries just driving into work. This app lowers the bar to doing searches. For a few days I was like a five-year-old just doing queries as they popped into my head. The easier/faster it is to search, the more I searched.

Not only have I started to do more queries, but I also say longer, more natural-language queries. Why? Because the more contextual clues I can give to the voice recognition engine, the better it will do. So a query like [mount everest elevation feet] might work, but [how high is mount everest] is more likely to be recognized (in my limited experience). The way you formulate queries is different when you’re speaking compared to when you type. I’m still pondering the implications of that.

Is the app perfect? Of course not. Right now it keeps a history of my last six queries. Personally, I’d like to keep all my queries so I can go back and find previous searches. And the voice recognition, while very solid, will continue to improve over time. I’d also love a way to add my own personal vocabulary of terms such as “PageRank” or “webspam” (which currently comes up as webcam). So the app will improve, but I still feel like I have K.I.T.T. from Knight Rider in my pocket a lot of the time.

The first time you hold a phone up to your ear and just say “18 percent of 33 dollars” and get a Google calculator answer back that it’s 5.94 dollars, it’s an epiphany. Now I don’t need a tip calculator application; I just talk to my phone and it tells me to leave a six dollar tip:

You may have a similar epiphany when you say the phrase “miles per gallon” and the app knows to type “mpg” or you say “one hundred and seventy-six” and the app returns “176.”

That rocks, and it feels like just the tip of the iceberg in terms of potential. Imagine if you had voice “bookmarks” like “Go to my email” or “Bring up today’s calendar” and you’d automatically see which room your next meeting was in. I think this will be a fun application. If you have an iPhone, take it for a test drive yourself: go to the App Store and search for “Google Mobile App,” then try it out. If you have iTunes installed, you can also click on this link to install from your computer. Oh, and if you’re a non-U.S. user, the app turns off voice search by default. To enable voice recognition, just click on “Settings” and slide the “Voice Search” setting to ON.

Read other people’s impressions of the new application over on Techmeme, if you’re interested.

www.mattcutts.com

published @ November 18, 2008