Newsgroups: comp.speech
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!cs.utexas.edu!uunet!portal.austin.ibm.com!bocanews.bocaraton.ibm.com!news
From: sumner@vnet.ibm.com (Charlie Sumner)
Subject: Re: Which speech recognition package?
Sender: news@bocanews.bocaraton.ibm.com (News Admin ID)
Message-ID: <D6H9wp.D2y@bocanews.bocaraton.ibm.com>
Date: Mon, 3 Apr 1995 21:19:37 GMT
Lines: 26
X-Nntp-Posting-Host: sumner.bocaraton.ibm.com
Reply-To: sumner@vnet.ibm.com (Charlie Sumner)
References: <3lfdv7$onp@spool.cs.wisc.edu>
Organization: IBM Boca Raton
X-Newsreader: IBM NewsReader/2 v1.02

In <3lfdv7$onp@spool.cs.wisc.edu>, jgoldst@tiger.cs.wisc.edu (Jonathan Goldstein) writes:

> I was wondering two things: 
>
>1) Are there any word processors that the IBM Speech Dictation System
>can dictate directly into? If not, when are they coming to market?

Not at this time... hopefully soon.  Officialy I can't comment on future products.
Most people have developed a simple voice macro to 'transfer-into-MS-word' or similar.
>
>2) Occasionally, my sister's speech becomes slightly slurred, how
>does the IBM Speech Dictation compare to Dragon Dictate in this
>situation.

It's hard to predict how any speech reco system will respond to slurred words.  I
will say that the VoiceType Dictation system improves with usage by remembering 
word contexts.  By this I mean if you are fairly consistent with your word usage 
patterns, the system will recognize the correct word as long as the accoustics are 
'close enough' ie. If I say 'A stitch in DIME, saves nine'.  The system will produce 
'A stitch in time, saves nine'... 'stitch in dime' makes no sense, and therefore since
dime and time sound pretty much the same, the statistical language model predicts
'time' to be the correct word.

I don't know how Dragon handles this particular scenario.

Charlie

