--- "Jessica P. Hekman" <http://www.arborius.net/~jphekman> wrote: > > Date: Thu, 7 Nov 2002 10:43:14 -0500 (EST) > From: Jessica P. Hekman > To: Ricky Houghton > Subject: Re: Update > > On Thu, 7 Nov 2002, Ricky Houghton wrote: > > use -langwt 5.0 | 4.0 | 3.0 just to see what happens. As you language model > > matches better you input speech you can raise it, lower it when there is a > > strong mismatch. > > I will try this. I set "-langwt 3.0" and it's still pretty slow for me. (I tried "emergency broadcast system" and it sat there for 1 minute before I killed it. The memory footprint kept growing and growing during this minute.)