Complex Applications and Dictation
Dictation relies on "language models"
created from preprocessed text such as the Wall Street Journal. Dictation
works best on longer "natural" sentences, and generally requires a
requires good conditions:
- A quiet setting
- A good
microphone
- A trained and dedicated user
- All words must be known to the speech engine
It is possible for dictation to work well in
certain applications if the user is dedicated to being understood by the
voice-recognition software. However, when possible dictation should
be avoided and replaced where possible by command and control. For
example, a report that might otherwise be dictated could be replaced by a form
with many choices.
Multiple voice applications can be active
simultaneously. All applications could be listening at all times and the
speech recognition engine gives each recognition to the application for which it
is the best fit. Or, applications might be listening only for an
activation command, otherwise giving way to the application with focus. It
is possible to listen for voice commands during dictation; however, this
usage model is often confusing and error prone. The number of commands
available during dictation should be limited to formatting and activation
commands.
SpeechStudio FAQ
SpeechStudio
Products
|
