(156) previous ~ index ~ next
To: tdt-distrib@unagi.cis.upenn.edu
From: Thomas C Pierce <tp26+@andrew.cmu.edu>
Subject: Re: MISC Files in Tracking Output
Date: Sat, 5 Sep 1998 18:43:55 -0400 (EDT)
i can speak for the CMU folks doing detection and tracking.. yes, we
treat all boundaries (even when no Brecid/Erecid fields are present) as
delimiting a document. to deal with this, we've had to make sure that
our systems can deal with various degenerate conditions, such as "empty"
documents (ie there are no tokens).
we'd originally been doing the same thing as Bowden descibed (with the
same "result"). at the time, Jon basically told me the same thing he's
posted to the group about not using MISC/NEWS tags... so that's what
we've been doing.
-tom
Excerpts from mail: 5-Sep-98 Re: MISC Files in Tracking .. by G. Bowden
Wise@markab.cr
>
> Well, we took the specs to literally so when it says
> scoring will be done only on "news" stories, when boundaries
> are given we don't retrieve the non-news stories at all.
>
> Will the eval software fail to score a file if the MISC
> stories are not output?
>
> Are other sites retrieving all boundaries for a particular
> file (even MISC and UNTRANSCRIBED?) or just retrieving
> MISC?
(156) previous ~ index ~ next
Last updated Wed Sep 9 09:40:57 1998