(188) previous ~ index ~ next
To: hub4@jaguar.ncsl.nist.gov, tdt-distrib@ldc.upenn.edu
From: john.garofolo@nist.gov
Subject: re: Use of TDT-2 for Hub-4 training is NOT allowed
Date: Fri, 2 Oct 1998 14:03:03 -0400
Folks,
The TDT-2 data is not to be used this year in training systems for
the Hub-4 evaluation. There are several reasons for this:
1. There is overlap in the Hub-4 evaluation material and the
TDT corpus.
2. We may wish to use a substantial portion of the TDT-2 corpus for
Hub-4 test material next year.
3. It is too late to reasonably expect everyone to be able to acquire
the TDT-2 corpus from the LDC and integrate it into their systems in
time for the Hub-4 evaluation since it includes about 800 hours of
BN on well over 150 CD-ROMs.
-John G.
--
National Institute of Standards and Technology
Information Technology Laboratory
Spoken Natural Language Processing Group
Building 225 (Technology), Room A-216
Gaithersburg, MD 20899
U.S.A.
Phone: (301) 975-3193
Email: john.garofolo@nist.gov
(188) previous ~ index ~ next
Last updated Fri Oct 2 19:04:21 1998