Share this story...
Latest News

Do some thinking, testing before you spend big on voice recognition software

(AP Photo/Gregory Bull)

Q: I’d like to use voice recognition software for dictation but don’t want to spend money unless I know it’s going to work. What would you suggest?

Since its humble beginnings in the 1950s, voice recognition technology has made great strides over the years, but there are still many challenges to making it work the way most people envision it should.

Managing your expectations about what it can and can’t do will have as much impact on your success as the technology itself.

If you’re looking for the kind of perfection portrayed in sci-fi movies, don’t bother looking at anything that’s commercially available just yet. Frankly, I’m not sure we’ll ever see an error free speech-to-text recognition system any time soon.

Understanding accuracy claims

You’ll likely see various claims being made about the accuracy rate of today’s technology, but keep in mind, a 90 percent accuracy rate means that every tenth word could be wrong. Even at 95 percent accuracy every 20th word could be wrong.

This means you’ll always have to spend time reviewing and correcting anything you generate, especially when it comes to things like homonyms and punctuation.

If you’re okay with that, then you’re ready for the next step.

Hardware is crucial

Everything starts with the microphone that generates the sound patterns that the software will attempt to recognize, so trying to use the built-in mic on a laptop or webcam isn’t going to cut it.

Ambient noise can make recognition even tougher than it already is, so you’ll need to invest in a decent headset mic so you’re providing the program with the cleanest audio possible.

Cadence is key

To get started with any voice recognition program, you always have to go through a training process so the software can get to know your voice and, more importantly, you train yourself on how to talk to the program.

Your cadence is the first thing you’ll need to change, because speaking to the program like you would to another human being is going to generate more errors.

This one area is where I’ve seen most people give up, because they aren’t willing to go through the learning/training curve in order to make the system provide a reasonable level of productivity.

Let’s face it: If you’re spending as much time cleaning up errors as it would have taken to type it out in the first place, it’s pointless.

Start with what you already have

You most likely already have voice recognition capabilities in your computer if the operating system is reasonably recent.

Mac users can follow these instructions to try using the Dictation tool that’s built in, while Windows 7, 8  and 10 users can go to the Control Panel and click on “Ease of Access” then “Speech Recognition” to turn it on.

None of these built-in technologies will compare with what most consider the industry leading software from Nuance, called Dragon Naturally Speaking, which can range from $75 to $500.

Nuance also offers a smartphone app called Dragon Anywhere that you can try out for free to see if using your mobile devices works better for you.

Related Links