Actually, Mr. Red, you do train the software. The best way to use the program is to speak at your normal pace, normal volume, and normal pronunciation (don't try to be overly precise like separating syllables or you'll "confuse" the program—it may try to interpret them as separate words), so that it can get used to the way you speak.

Dragon Naturally Speaking has a whole training routine. Its training program consists of a list of documents (provided) that you read so that it can become familiar with your way of speaking. It also reads documents that you have written (it scans your "My Documents" and e-mail folders) so it can learn your vocabulary and writing style (now, how it learns your style, I really don't know). The more training sessions you do, the more accurate the program gets. Another very important part of training is to make use of the "correct that" feature as you're dictating, rather than simply going back and typing in corrections. That way, it learns your way of pronouncing those particular words. It can get used to mispronunciations as long as they are consistent, and it can also get used to various speech impediments.

In short, you train it. Otherwise, having separate files for each user would not be necessary. It is a two-way street, of course. You do have to speak clearly and use consistent pronunciations. I haven't tried it, but I imagine that if you suddenly start speaking with, say, a French accent, you'll thoroughly confuse the program and it will start making mistakes.

To say that the program "learns" may sound sort of anthropomorphic, but not so. Some programs are getting pretty sophisticated. Still as dumb as an earwig, but good at specific things.

More later.

Don Firth