Redstart Systems’ Voice Command Software Replaces the Keyboard and Mouse—and Not Just for Dictation

4/21/09Follow @wroush

If you want to dictate notes into your computer without typing, speech recognition software like Dragon Naturally Speaking, from Burlington, MA-based Nuance Communications (NASDAQ: NUAN), works surprisingly well these days. Even without training, dictation software can hit accuracy rates of 99 percent; once it learns your personal speech patterns, it’s nearly flawless.

But using speech commands to do almost anything else on your computer is far more difficult. Nuance says the latest version of Dragon Naturally Speaking can be used to control applications such as Microsoft Word, Outlook Express, and Internet Explorer, and the software even includes “voice shortcuts” that let users interact with search engines using natural-language utterances like “Search the Web for global warming articles.” But for complex, oft-repeated command-and-control operations, like opening or closing windows or moving blocks of text in a document, using natural language commands can be tedious. It also tends to be slower than using mouse and keyboard commands, since the software has to spend a good deal of time figuring out what you meant before it acts. For the large group of computer users who turn to speech recognition software because of repetitive strain injuries (RSI)—and who aren’t supposed to touch their computers at all, lest they aggravate their condition—that’s a dangerous situation.

That’s the problem—and the opportunity—that Boston-based Redstart Systems has set out to address. After nearly 15 years of behind-the-scenes software development, the tiny, self-funded startup today launched a program called Utter Command that vastly speeds up command-and-control operations for Windows computer users who already have Dragon Naturally Speaking Professional.

The secret to Utter Command isn’t speech recognition—it depends on Naturally Speaking as its speech engine—but rather its ability to parse “stacked,” shorthand commands. For example, instead of laboriously saying, “Move the cursor to the end of the sentence, select the last three words of the sentence, and delete them,” an Utter Command user would simply say “End 3 befores delete.” (In this example, “befores” is shorthand for “words before”—and is a good example of the way Utter Command clips things down.)

At $395, the new program isn’t cheap. But it may be a worthwhile investment for people who really can’t touch their computers. And if you view it as a powerful add-on that makes up for features missing in Naturally Speaking Pro, which retails for $899, the price tag seems even more reasonable.

Utter Command OverviewRedstart president and founder Kimberly Patch, a science writer who has worked at PC Week and Technology Research News, says she first conceived the software in the mid-1990s when she developed a repetitive strain injury from typing on her computer. “I started out using Dragon Dictate 1.0, but I got frustrated with it and starting writing macros to speed things up,” Patch says. “But I’d forget half the macros I wrote, and then I’d have to rewrite them. I realized it was easier to remember standardized commands. I was writing about things like cognition and linguistics, and it turned out that this made sense according to the cognitive studies; there are MRI studies that show that certain things are easier to say than others.”

What Patch was discovering (as she explains in a series of papers on the Redstart website) was that sticking to a small set of commands, and arranging them according to a precise grammar, might actually create a lower cognitive load on a user than trying to speak to a computer as if it were a person. It would probably ease the load on the computer, too, since the software wouldn’t have to anticipate all the different ways a person might phrase a command in natural language.

Patch started writing down the commands in her grammar to make sure she was using them consistently. A bit later on, she found a programmer to help her incorporate the commands and the grammar into an application. And about five years ago, she decided to turn the application into a product.

But getting it working the way she wanted and writing up the documentation “took a lot longer than I thought it would,” she says. On top of that, there was an ethical concern. “With RSI, you can type, but you don’t want to, and if you get frustrated with something, you will hurt yourself,” she says. “We wanted to make sure our system was not going to … Next Page »

Wade Roush is a contributing editor at Xconomy. Follow @wroush

Single Page Currently on Page: 1 2

By posting a comment, you agree to our terms and conditions.