Paul Jones wrote:
Since I was looking at relationships between words, I of course
stumbled upon Wordnet, and wanted to know if anyone had an idea about
what, if any, techniques they were using to "measure" how close words
were related, or is it all done manually. If so do I presume there is
no ranking when they work out that crimson and cerise are "related"
to the noun "red", i.e they clustered around the word "red" but all
equidistant from it.
From previous work:
http://www.google.com/search?q=lexical+chaining
The original work on lexical chaining was based on the index of Roget's
thesaurus, but since that wasn't available at the time this work started
(the early 90s at the University of Toronto), people started using WordNet.
<shameless-plug>
Don't miss a classic of the genre:
ftp://ftp.cs.toronto.edu/csri-technical-reports/366/uttr366.ps.gz
</shameless-plug>
Seriously, though, there's been a lot of work on this, and you should
get some good ideas from some of the work that was done in this century
:-) Man, I'm getting old.
Steve
--
Stephen Green // [email protected]
Principal Investigator \\ http://blogs.sun.com/searchguy
Advanced Search Technologies Group // Voice: +1 781-442-0926
Sun Microsystems Labs \\ Fax: +1 781-442-1692
