Speech recognition
 One of the hottest areas in data entry technology is Speech Recognition.

There are two classes of speech recognition technology:

    1. Speaker dependent -- which requires that the user train the software to his voice
    2. Speaker independent--works for everyone, "right out of the box"
There are also two degrees of implementation of speech recognition: Keypad speech recognition has a limited number of commands, likened to a telephone keypad, with only a few predefined messages.  Recognition of and distinction among these commands is a less sophisticated problem than recognizing a broad vocabulary.  Since only a limited number of words need to be distinguished,  keypad applications are easily adapted for speaker-independent implementation.

Keyboard applications, on the other hand, allow users to input any variety of text into the computer, and present a much greater challenge.

Speech recognition engines are also divided into

    1. discrete word and
    2. continuous speech types.
Before 1997, major players Dragon Systems' DragonDictate and IBM's VoiceType simplySpeaking both required the user to pause briefly between words.  These discontinuous speech products were awkward to use and slow.  Click to download two demos of IBM's discontinuous-speech product, Voicetype.
USdemo1.exe
USdemo2.exe
after downloading, execute the files through the file manager.

there was a major breakthrough in 1997 with the introduction of large-vocabulary continuous-speech full dictation programs, IBM's ViaVoice and Dragon Systems' NaturallySpeaking.  Continuous speech engines use the context of a phrase to determine words.
NaturallySpeaking has more extensive hands-off editing features at this time.  Here are some NaturallySpeaking help files as AVI's which have to be viewed using the media player utility:
correct1.AVI
delete1.AVI
there is a better demo available, but it's a LOTUS screenCam97 that will only run under windows95
nsdemo.exe
this file is about 1.88 Mb.   there is also as a self-extracting zip file of this demo that will fit on a floppy disk( about 1.02 Mb):
nsdemzip.exe

These products are advertised as requiring minimally machines with 166 Mhz Pentium processors and 32 Meg of RAM,  but users report adequate performance requires machines greater than 200 MHz and 64 meg of RAM.  Thus, these products would not work well with the large installed base of slower machines.

 return to technology page Return to MiniMBA Technology Page
Return to Bud's Page  E-mail: rbanis@jinx.umsl.edu 
Campus office CCB 230 
On-campus phone X6136 
Off-campus 394-4950 
Go to UM-St. Louis Home Page 
 		 Go to UM-St. Louis School of Business Page 
 

last modified February 28, 1998 ~~ comments and suggestions to rbanis@jinx.umsl.edu