Lesson 2: Initial Training

Release 3, 4, 5 Hints

The goal of this lesson is to establish a "voice file." Each user of a system has a different voice file. (A single individual may choose to have multiple such files, such as when they use recorders for input.)

Start your system. Observe your environment -- is it your normal dictation environment? Do you normally have a radio on, background music, etc.? It is best to do this training in your normal environment.
Initiate NaturallySpeaking, either from the Start Menu or from your desktop icon. (1 min.)
Establish a new user - use your name (1 min.) (ALTERNATE: Your name may appear when you start NaturallySpeaking as it may have been partially set up, in which case the "Base" vocabulary has already been selected for you.)
Select the "Base" for your speech files. We suggest that you use the Speech Model that the computer marks as RECOMMENDED, unless you are using a laptop Pentium-III, in which case we suggest that if it recommends "Standard" that you change this to "BestMatch 3".

If you are using Medical or Legal Suite, and you have a new machine, you will usually want to select the last base vocabulary option -- Base Medical/Legal Professsional -- for best results. This is NOT the default option as of August 1998.

Whatever version you use, we suggest you open the "Vocabulary" drop-down list to verify you are choosing the most appropriate vocabulary for your needs and machine. In Release 5, you must select the "Advanced$quot; button to obtain this detailed information.

You should plan on retraining on Release 5.x using a "BestMatch" vocabulary, unless your machine has less than 128MB or is slower than 500MHz, in which case there may be a performance advantage in retaining a Standard voice model.
Put on the microphone.
Breath sounds are the "enemy" - but the microphone needs to be fairly close to the mouth. The key is to avoid placing the microphone element directly in front of your mouth. About ½" away from the corner of your mouth is good for most microphones.
Audio Setup Wizard (3 min.)
Follow the steps, choosing the options that apply to your system. Don't be overly concerned with the speech to noise number unless you feel you can easily improve it by making small changes to your system. We've seen microphones that score poorly work quite well for actual speech recognition. If you get "Unacceptable", it is likely the microphone is not properly connected or you have a poor sound system.
General Training
You will read about 5 screens, taking about a minute, then Dragon will do a short calibration. Then you will read for about 5 minutes or 25 minutes (depending on your computer speed), followed by a longer (10 minute) calibration. You can pause during the reading. Stop (Pause) and get a drink if your throat gets dry. Try to read in your DICTATION voice - which is often different than your reading voice. Don't force a monotone voice, and don't try to be overly expressing. A NATURAL voice will work best.

More General Training Hints

Note for Release 2.0 and 1.0: If there is a "hang" during the calibration, then do not panic. Minimize the top window and any other windows. You should then see a Retry box. Select Retry and continue.
NaturallySpeaking installation will walk you through and give you options to run Vocabulary Builder (skip for now) and to view training material. View the multimedia training material if you have speakers connected to your system.
Try using NaturallySpeaking by reading some of your material to it. Some will get good results immediately, others will get funny results at first. Read/dictate a paragraph or two - don't worry about formatting.

There will be Miss steaks. Try these voice commands, using a slight pause before and after the commands:

SCRATCH THAT - erases what was just said

SELECT <a word or phrase> -- this should highlight <a word or phrase>, where <a word or phrase> is something that Dragon has written on your monitor. DON'T pause between saying "SELECT" and <a word or phrase>. Then dictate something to replace the word or phrase you selected, i.e. replace the highlighted word(s). For example, say "select miss steaks". This would highlight the phrase "Miss steaks". Then say "mistakes" and normally the word will be changed.

CORRECT <a word or phrase> -- try this when you say something correctly but Dragon does not understand ewe correctly. You will get a menu of alternate words. For now, use the keyboard and type what you said into the correction window; then use the "Enter" key to cause Dragon to make the correction. For instance, say "correct ewe". You should get a "Correction" window. Type "you" then the Enter key. This is the strongest type of correction, and the method by which Dragon "learns" your voice.

You may either save or discard the document. If what you have dictated is useful, you may use Windows copy/paste to move the information elsewhere, or you may print it using normal Windows functions.

Later you will learn to dictate directly into your application, or to copy/paste by voice. But first, it is advisable to learn to dictate directly into NaturallySpeaking so that you become comfortable with the correction mechanisms and so that NaturallySpeaking starts to adapt to your voice.

When you exit the NaturallySpeaking program, you will be asked if you wish to save the changes to your speech files. You should answer "Yes" to this question.

DO NOT BE DISCOURAGED AT THIS POINT! WITH SOME SYSTEMS, WITH SOME PEOPLE, IT WILL APPEAR AS THOUGH THERE IS NO HOPE. AMAZINGLY, AFTER A FEW CORRECTIONS, THAT IS, A LITTLE MORE VOICE TRAINING, THESE PROBLEMS USUALLY VANISH. FAULTY/SUBSTANDARD EQUIPMENT OR IMPROPER SETUP CAN CAUSE SIMILAR RESULTS.

Lesson 3

top

Helping you make the most
of Dragon NaturallySpeaking

Release 3, 4, 5 Hints