(032) previous ~ index ~ next
To: tdt-distrib@ldc.upenn.edu
From: Christopher Cieri <ccieri@ldc.upenn.edu>
Subject: Annotation Exercise
Date: Thu, 30 Apr 1998 13:56:33 -0400
TDT-2 Participants,
George and Charles have asked LDC to set-up a topic annotation exercise
so that all participants will have a chance to experience this aspect of
the project first-hand in time to discuss it in Gaithesburg. They feel
that such an exercise will:
1) give the sites a visceral understanding of topics
2) provide data to help calibrate human performance
3) possibly generate feedback on the process of defining topics
We have created a environment for this test. You will need a Java
capable browser. Netscape 3 or later will work. If you point your
browser at: http://www.ldc.upenn.edu/tdt-exercise you will see a login
screen. To generate both login names and passwords for the group we
took, for each name in our tdt-distrib mailing list, everything before
the @. For example, my login and password are both "ccieri". If you have
any problems logging in, send me e-mail and I'll confirm your login and
password. Once you've logged in you will see a page with our newest
labelling instructions, a description of each topic and six file ids --
one for each source. From there, the instruction built into the
interface should be adequate. Remember to log-out when you have finished
a session.
Our goal was to get coverage of all sources but provide just enough
material to engage the average participant in 2-4 hours of work. A
skilled annotator could probably finish those six sources in 4 hours.
However, since the newswires are the most labor intensive, it may make
sense to maximize the overlap by having us all begin with the broadcast
sources.
Please e-mail us with any questions or problems. See you soon.
Chris
--
Christopher Cieri phone:215-573-5489
Executive Director fax:215-573-2175
Linguistic Data Consortium http://www.ldc.upenn.edu
(032) previous ~ index ~ next
Last updated Wed Sep 9 09:40:47 1998