Hi,

Reminders about submissions which are due at NIST by midnight
on Sunday, April 14 - earlier if possible... The basic information is 
on the guidelines webpage under "Submissions and Evaluation". Please 
take a look at that.

  <http://www-nlpir.nist.gov/projects/duc/2002.html> 

Your submission must conform to the submissions SGML DTD provided 
there. If it does not parse, then it will not run through the 
evaluation setup software.

  <http://www-nlpir.nist.gov/projects/duc/duc2002/duc2002submission.dtd>

Make up your own "sysid" - something likely to be unique, not tooooo 
cryptic, and preferably short (8 chars) so it can eventually be used 
conveniently on charts etc. 

The body of the abstracts should be just plain English text. 

The body of the extracts will be a sequence of <s>...</s> sentence 
elements (see the example off the web page).

  <http://www-nlpir.nist.gov/projects/duc/duc2002/example_duc2002submission.sgml>

The words associated with each sentence in an extract should be included 
if possible so the extract is readable by a human without reference to 
the documents, but the evaluation will be solely in terms of the docid-num 
pairs - not the words of the sentences. 

(Note that for abstracts, the document attribute is "docref"
while in extracts it is "docid". The unintended inconsistency
is my fault, but may serve to underscore the difference between
abstracts and extracts.)

Please double check that your submission is complete and correct
- correct docsets, docrefs/docids, summary types, sizes, etc. It's
a shame to do all the work and then not get it evaluated due to
a typo or scripting mistake.

Please email me directly and early if you have looked over the webpage 
info and still have questions.

Thanks,
Paul