(058) previous ~ index ~ next
To: Jon_Yamron@dragonsys.com
From: Rich Schwartz <schwartz@bbn.com>
Subject: Re: TDT3 dry run
Date: Mon, 29 Mar 1999 19:26:09 -0500 (EST)
Again, I'd like to agree with all of Jon's responses.
If it is more powerful to use a tightly coupled system, as Doug Oard says
is true in CLIR, then people should try that, but surely we would want to
compare their results with those of a factored system. I, for one, am not
sure which approach (tightly coupled or factored) would be better -- there
are arguments both ways and it depends on how good each of the sytems is--
but it's clearly one of the right questions to ask.
As to the judgements, I assumed those were being done in the native
language anyway. (I would be slightly -- though not strongly -- concerned
if the annotator judgements were done on automatically translated text.)
But the only thing that Jon is proposing is that there will be a system
(in the form of the output) that we are each free to use IF WE WANT TO as
part of our CL-TDT system.
--Rich
(058) previous ~ index ~ next
Last updated Thu May 13 09:28:21 1999