Corpora: Call for assistance

Ken Litkowski (ken@clres.com)
Tue, 09 Feb 1999 14:58:33 -0500

As part of its support for the Dictionary Parsing Project
(http://www.clres.com/dpp.html), CL Research is currently articulating
defining patterns that can be reliably used to identify semantic
relations (semrels) in dictionary definitions. We have found that, like
semantic roles in sentences, there is no widespread agreement on a fixed
inventory (we have provided links to several inventories at the above
web page). Nonetheless, the inventories are substantial and the
defining patterns will be even more numerous. We would be grateful for
any assistance in their development. The goal of this effort is the
production of publicly available lexical resources similar to WordNet or
Microsoft's MindNet. We are currently using the publicly available
Webster's 2nd International Dictionary (120,000 headwords and 280,000
definitions) as the immediate resource, employing the publicly available
CL Research DIMAP software for parsing definitions. (We are also
parsing WordNet glosses, as well as using WordNet in parsing W2 to test
the validity of identified semantic relations that have WordNet
counterparts.)

In addition, we are expanding the parsing technology used in DIMAP for
the recently completed SENSEVAL word-sense disambiguation task to make
use of semrel information. We will be using this technology in the
upcoming Q & A Track of TREC-8, as well as anticipated future SENSEVAL
tasks. We would welcome any companies or individuals who would like to
join us in this participation.

Thank you for your consideration. My apologies for any cross-posting.

-- 
Ken Litkowski                     TEL.: 301-482-0237
CL Research                       EMAIL: ken@clres.com
9208 Gue Road
Damascus, MD 20872-1025 USA       Home Page: http://www.clres.com