Speech recognition, the new PC power tool

Speech recognition, the new PC power tool

Sarah Jacobsson Purewal @geeklil

The Siri effect is spreading. Starting this fall, Nuance, a company that makes advanced speech recognition and transcription software, will bundle its Dragon Assistant, a Siri-like, vocal sidekick, with select Intel-powered Ultrabooks and All-in-One desktops.

Dragon Assistant could change how people use speech recognition in everyday computing. Even if you’re a good typist, speech recognition can enhance your productivity by letting you handle everything from basic commands to emails without touching the keyboard.

My husband was an early and unexpected convert. In 2010, he had surgery on his left shoulder and was unable to use his left hand for a couple of months. His company bought him a copy of Dragon Dictate. He liked it so much that he still uses it today, long after his shoulder healed. He says it makes him faster and more productive (and he already types, on average, 130 words per minute), and it’s also easier for him to work in non-traditional environments, such as in the car or outside by the pool.

Everyone worries about the training time—and there is some of that—but it varies from product to product. Choosing the right program will ensure that the benefits are worth the investment. These tips will help you get the most out of Nuance’s speech recognition programs, whether you’re using the traditional desktop program, the mobile app, or the new, Ultrabook-bundled Dragon Assistant.

Dealing with Dragon

PC users have four Dragon products available. Dragon Home ($75) is for basic users—people who are looking to write emails, school papers, and grocery lists, and use basic voice commands to control their computer. The $150 Dragon Premium adds support for working in spreadsheets and presentations, while the $600 Dragon Professional adds advanced custom commands and transcription tools. The $800 Dragon Legal includes everything in Dragon Professional, along with preloaded legal terms and the ability to automatically format legal citations.

Haswell tabletPCWORLD (US)

Intel’s reference design for the sort of Haswell-based Ultrabooks now shipping with Dragon Assistant built in, as shown at CES 2013.

Nuance’s newest offering, Dragon Assistant, works more like Nuance’s mobile apps, like Dragon Mobile Assistant for Android and Dragon Dictation for iOS—it’s a verbal gofer, rather than a dictation app. According to Sarah Gaeta, vice president and general manager of the Dragon Desktop department at Nuance, anything beyond very quick email replies is better done with Nuance’s more advanced products.

How to train your Dragon

According to Gaeta, there’s no substitute for spending some quality time training your Dragon program of choice. Unlike Siri, Dragon software is designed to “learn” your voice as you use it, so it can better respond to your commands and dictation. The more you use Dragon, the more accurately it recognizes your accent, how you pronounce different vowels, the cadence of your speech, and any vocal quirks. Dragon stores that information in your personal profile, and it can store profiles on multiple people so they can all use Dragon on one computer.

Set up your own profile in Dragon and use it for a few hours so the software can learn your unique speech patterns and vocal mannerisms.

Take a few hours to use it as much as possible: dictating short emails, practicing voice commands, or even just speaking to the program like you would to another human being. According to Gaeta, Dragon Assistant requires about 30 minutes, while initially training Dragon Dictate or Dragon NaturallySpeaking might take two to three hours.

It’s best to do this initial training in a quiet room, so Dragon can learn your voice and, more importantly, learn what isn’t your voice. Once you train your Dragon, you’ll be able to dictate to your PC in a noisy coffee shop, and it will pick up only the words you say, ignoring the loud, oversharing couple behind.

Speak slowly and clearly

If you want Dragon to understand you, speak slowly and enunciate clearly. “This sounds like common sense, but it’s not,” Gaeta says. “It’s crazy; we put someone in a room with a microphone and they feel like they’re performing. They get nervous and speak way too quickly.”

Dragon software will walk you through a few simple tutorials that will train you to speak slower and more clearly, which will improve your productivity when using Dragon products (as well as your real-life conversation skills.)

Learning to speak slowly and enunciate is more challenging than it sounds. There’s a reason that people take public speaking lessons, and while I wouldn’t recommend doing that just for the sake of improving your productivity, a few great, free speech tutorials on YouTube will help you spiff up your speech patterns. Speaking clearly also makes a difference with Dragon Assistant.

Speak in full sentences

Working with speech recognition software is ultimately faster than typing, but it won’t feel that way at first. When you’re typing, you’re probably not thinking through entire sentences before you type them. There’s no reason for you to have a properly formulated sentence in your head when you start writing—you have plenty of time to perfect it while you type.

If Dragon fails to accurately determine which word you meant to write, you can use the built-in voice commands to correct the error and Dragon will learn from the mistake.

You can’t just say one word at a time as you speak a sentence, though—whether to a real person or to Dragon software. Dragon uses context to determine which words you’re saying, so it needs full sentences to be able to distinguish whether you mean you’re or your, or they’re, their, or there.

Luckily, you have some help: In the desktop clients (Dragon NaturallySpeaking and Dragon Dictate for Mac), you can use the keyboard alongside the voice dictation. You do have to be aware of when you’re using the keyboard, though, because basic commands (such as “scratch that” to erase the words you just said) won’t work for typed text.

Use Dragon for the right tasks

The traditional Dragon programs are most useful for email, or email-like tasks, because spoken English is so different from written English—many things that can be said cannot be written, and vice versa. For example, it’s acceptable to say, “I got this from the restaurant we ate at last night,” but you’d have to write “I got this from the restaurant at which we ate last night.”  Likewise, writing “My brother, an athletic director, lives in Tokyo,” looks better than if you said conversationally, “My brother’s an athletic director who lives in Tokyo.”

SOFTPEDIA

Dragon Professional includes a streamlined transcription module that’s built for writing long-form articles and essays.

Because spoken and written English are so different, native speakers will find that Dragon is better suited to writing that mimics spoken language. So you’ll be able to fly through tasks such as note-taking and emails, but you’ll find it much more difficult to dictate formal or creative writing.  Both Dragon Professional and Dragon Legal include a transcription agent module that makes recording and revising such writing much easier.

Dragon Assistant is useful for more basic tasks, such as Web searches, social network status updates, and writing short, quick emails. Like Siri, Dragon Assistant accepts natural-language prompts.  For example, you can say “What’s the weather like today?” instead of “Weather in San Jose on September 10.”  Nuance’s other Dragon programs don’t yet accept natural-language prompts—you’ll have to learn their built-in commands—but Gaeta says it’s coming soon.

Hack your typing

Even the speediest typist can benefit from Dragon software by learning to use it in situations when accessing a full QWERTY keyboard is impractical. I type about 125 words per minute when I’m at my desk, but when my hands are otherwise occupied—eating, getting ready for work—or when I’m in a bumpy car ride or similar situation that makes typing difficult, it’s nice to know that I can still get some work done.

Just remember that Dragon is a tool, not a crutch; Nuance may be trying to build software that loves you, but it’s not going to do your homework for you (yet). Proofread and double-check all Dragon-dictated documents before sending them out.