[Fwd: Re: [Corpora-List] automatic definition construction/retrieval]

Ken Litkowski ken at clres.com
Fri Sep 2 17:07:37 UTC 2005


While it may be the case that some single word new words arise and need 
definition, I would suggest that most of them will be multiword units. 
HpaB is probably an abbreviation for a longer MWU.  While some of the 
definition seeking behavior evolving in the TREC QA track is certainly 
relevant, that is not "defining" in the lexicographical sense.

An important useful first step these days is the use of the Google 
'define' operator to pick whatever formal definitions might exist. 
Next, one can Google "HpaB is" and get a whole range of answers.

But, creating new definitions is the more interesting part of the 
problem, particularly with MWUs.  For this, I would suggest the work of 
OntoLearn (Velardi and Navigli).  There was a recent article in 
Computational Linguistics, and they have several other papers which 
describe algorithms for inducing definitions.  Keep an eye out for 
further progress from them.

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



More information about the Corpora mailing list