Recognition Window



The Recognition window is your interface for refining Dragon Dictate’s voice model by telling it about recognition errors it has made. This is called training the voice model. When the Recognition window is showing, it floats over all applications.

When Dragon Dictate has typed what you said in Dictation mode, and has made a mistake, use the Recognition window to correct Dragon Dictate’s interpretation of what you said, as well as to correct what Dragon Dictate typed.

There are thus two ways to modify the existing dictated content of a document: you can use Dictation mode commands, or you can use the Recognition window to train the voice model. However, there is a big difference between them:

When you use a Dictation mode command to select and delete a phrase, or to select and dictate to replace a phrase, and so forth, you are implying that you made a mistake.

But when you use the Recognition window to train the voice model, you are implying that Dragon Dictate made a mistake; it misrecognized your speech, or didn’t know a word you used. Use training, in other words, to teach Dragon Dictate and to improve its accuracy in recognizing your speech and the kinds of things you say.

To summon the Recognition window:

  • Say Show Recognition Window.

  • In Dictation mode, say one of the following:

    • Train “Someword”

    • Train “Someword” Through “Otherword”

    • Train “Someword” to “Otherword”

    • Correct “Someword”

    • Correct “Someword” Through “Otherword”

    • Correct “Someword” to “Otherword”

  • Click the “Open Recognition Window” button in the Status window. (It is the rightmost button.)

  • Press the Recognition window keyboard shortcut.

  • If “Show Recognition Window when dictating” is checked in the Recognition Preferences pane, the Recognition window will open automatically when you are dictating in Dictation mode or Spelling mode as soon as the insertion point is within or after a phrase capable of being trained. Usually this will be as soon as you dictate anything.

The Recognition window (and voice model training in general) operates on dictated utterances. Dragon Dictate has its own ideas about what constitutes an utterance, based partly on where you paused while dictating. Therefore, the phrase that is selected when you say Train…, or the phrase that appears in the Recognition window (and is selected when you summon the Recognition window), might be longer or shorter than what you initially specified.

  • When you are done making a change, Dragon Dictate's smart formatting feature opens a new window that asks how you want to automatically handle this formatting change in the future. You will be given one or more options, such as "Never abbreviate titles," which would mean that dictating "Mister Smith" would always be rendered as "Mister Smith," not "Mr. Smith."

If you don’t want to use any of the smart formatting rule choices, you can click the X in the top left corner of the window, or simply resume dictating text, and the window will disappear. Smart formatting rules are turned on by default when you install Dragon Dictate, but you can turn them off in the Recognition Preferences pane, which is accessible in the Dictate menu.

To close the Recognition window:

  • Say Hide Recognition Window.

  • Say Cancel Training.

  • Say Cancel Edit.

  • Say Cancel Recognition.

  • Click the close button at the top left of the Recognition window.

  • Press the Recognition window keyboard shortcut.

  • If “Close Recognition Window after each choice” is checked in the Recognition Preferences pane, the Recognition window will close automatically as soon as you use it to replace a dictated phrase with its correct version.

Try different Recognition Preferences pane settings and different ways of summoning and closing the Recognition window as you train your dictation. You’ll soon settle on a favorite way of working (which might depend on external factors, such as the size of your screen).

You can adjust the transparency of the Recognition window. Say Increase Transparency or Decrease Transparency.

How to Train

The Recognition window presents alternative interpretations of the phrase containing the current selection or insertion point in the dictation document. The first alternative is the interpretation that Dragon Dictate entered in the document. The number of additional alternatives is limited by your setting for “Maximum number of alternatives” in the Recognition Preferences pane.

You can use the Recognition window to train what you just dictated, because the insertion point is still at the end of that phrase, so that is the phrase shown in the Recognition window. You can also select a dictated phrase (using Dictation mode navigation commands); if Dragon Dictate still has cached dictation information for this part of the document, the Recognition window will change to show alternatives for that phrase.

Thus, one way of working with the Recognition window is to train phrases as soon as you dictate them; another is to dictate a larger amount, such as a paragraph or two, and then go back and train.

In most cases, Dragon Dictate will have preserved the recording of your speech from when you dictated the phrase. To play the recording of the phrase selected in your text:

  • Say Press Play.

  • Click the Play button at the top right of the window.

This can be a useful precaution, because you may not have said what you think you said. You do not want to mistrain Dragon Dictate by using the Recognition window to enter a trained phrase that is not a correct interpretation of what you actually said! If you made a mistake, don’t use the Recognition window!

If one of the Recognition window’s alternatives is correct, enter it in the document in place of the alternative that’s already there. To do so:

  • Say Pick 2 or Choose 2 (or whatever the actual number at the left of the correct alternative is — but most of the time, it probably will be 2!).

  • Click the numbered button at the left of the correct alternative.

You have now performed phrase training. Internally, Dragon Dictate will modify its speech model, and will save this information into your profile the next time your profile is saved.

If none of the Recognition window’s alternatives is correct, you can edit one of the alternatives so that it is correct, and then enter the edited, correct alternative as in the previous step. It doesn’t matter which of the alternatives you edit; edit whichever is most convenient. To edit an alternative:

  • Say Edit 1 (or whatever the actual number at the left of the desired alternative is).

  • Click once to select the alternative, pause, and then click once inside the text of the alternative.

The alternative you are editing is now the “target” for dictation and commands. You can edit using Dictation mode or Spelling mode (and it is also fine to use your fingers to type in the Recognition window). The only standard command that doesn’t work is “Scratch that”; instead, use Scratch Word.

To exit editing mode without entering the corrected phrase into your document:

  • Say Cancel Edit or click elsewhere.