(099) previous ~ index ~ next
To: Rich Schwartz <schwartz@bbn.com>
From: George Doddington <doddington@nist.gov>
Subject: Re: Example index files for the first TDT3 dry run
Date: Fri, 14 May 1999 12:57:40 -0700
Rich Schwartz wrote:
>
> Jon,
>
> Is this the development set? (That is, the set we should tune our
> algorithms on and hope similar performance for the dry run?) Your
> statement about releasing the complete TDT2 corpus for the first dry run
> is confusing me. I guess I'm probably misunderstanding what is TDT3 vs
> TDT2. I'm expecting (at least) 3 sets:
>
> 1. The first one now for developing algorithms
> 2. The second one for a dry run test in mid June (or are we doing this
> on the development test like last September?)
> 3. The third one for the evaluation in the fall.
>
> --Rich
I think that Jon has left for the day, so I will respond on his behalf:
Please refer to the TDT3 web page (http://www.nist.gov/speech/tdt3/tdt3.htm)
The TDT2 corpus (with Mandarin) will serve as the training/development/test
corpus for both the June and the September dry runs. What I've provided in
the subject index files is part of the dry run test, namely the first half
of the TDT2 corpus along with the half of the 20 bilingual topics.
--
George Doddington in Orinda, CA: doddington@nist.gov or 925/631-6628
(099) previous ~ index ~ next
Last updated Mon Jun 21 11:03:30 1999