Suddenly I’m talking to a lot more inanimate objects.
I recently pulled a muscle in my shoulder, and I’m not supposed to type Thus DragonDictate and I have become good friends. DragonDictate is almost magical, I speak in the words appear in front of me.
The problem of course, is that small recognition errors, such as the one in the previous sentence, are very hard to detect when editing your document. Computers do a good job finding spelling errors, but distinguishing when you mean in and when you mean and, appears to be beyond its ken. That’s fair enough because it’s almost beyond mine.
When reading over a document, it is particularly hard to find properly spelled and grammatically correct words that are not what you meant. But that is exactly the kind of error you need to look for when using dictation software.
The folks at Nuance who make DragonDictate claim that you can get “up to 99% accuracy out-of-the-box.”
Let’s assume that in fact I am getting 99% accuracy. That missing 1% represents about 3 to 4 words in every page of a Word document; enough, if you don’t find and correct them, to sound semiliterate. That said, I can dictate a whole lot faster than I can type, even though I type about 100 wpm, and even though I have to read through in order to fix those three or four errors.
I’m not just talking to my computer however. I’ve also been talking to my GPS. Some of the ways it wants you to talk are a bit cumbersome, like saying “drive to an address”and then waiting for the prompt and then saying the address, and then affirming that is what you meant. Still it’s better than fussing with buttons while you’re driving. And it has some really cool features, like you can say “drive to the nearest Starbucks.” Now no matter where I go I can always get a triple shot vente extra dry cappuccino.
Then there’s the new iPhone. I love my Windows phone, and on balance I honestly believe it is the best phone on the market, but the new iPhone Siri feature is astounding. Combining advanced speech recognition with the capabilities of a smart phone, gives you something to write home about.
For example you can say, “remind me to call Seth when I leave here” and it will. It uses its built-in GPS to detect when you leave your current location, and then it pops up the reminder.
You can ask it what your next appointment is, what all your appointments are on a given day, what Jim’s address is; or you can ask it questions that you would otherwise look up in an online encyclopedia. It certainly feels like the phone understands what you are saying.
We’re still a long way from true semantic recognition, let alone the HAL 9000, but the progress of the past few years has been nothing short of astounding.