|
Speech recognition is improving
constantly, and in particular, the evolution
of personal computers with faster processing
power has allowed developers to design better
speech recognition engines.
Speech recognition most invariably
implies digital dictation,
since, with the exception of some telephony
dictation solutions, the sound input has
to be digital.
There are two kinds of
deployments: standalone solutions
and network solutions.
Standalone solutions:
the speech recognition takes place on the
user’s PC, and in most cases in real-time.
It is fairly easy to deploy, and works well
for users who do not need support staff
and wish to generate their own documents
quickly.
Network solution:
the speech recognition takes place at a
central location on your network (speech
recognition server). The main asset of this
deployment type is that the “dictionary”
(Context, Vocabulary) used for speech recognition
is common to all users, and improves faster.
This leads to the definition of the two
types of speech recognition: batch
recognition and real time speech recognition.
In batch recognition, the author dictates
a sound file, but doesn’t see the
text appearing on the computer immediately;
the sound file is forwarded to a central
speech recognition server, converted to
text, and then sent back to a predefined
support staff (or the original author) for
proofreading. This situation is well suited
for authors who cannot afford to spare time
in document finalising; there is no real
change in work pattern on the author’s
side; however, the support staff is able
to process many more documents.
In real time speech recognition, the author
dictates and the recognised text appears
immediately on the screen; the document
can then be finalised immediately, or forwarded
to support staff for archiving, printing…
In addition to the digital
dictation benefits, additional
benefits lay in:
• the reduction of the
turnaround time for the average document
• the ability to generate
documents during off-hours, when the support
staff is unavailable
|
|
