Dan Matei wrote:
True, but we should use (programatically) the already hand-crafted data (for concepts). For
instance, in Nat Turner case, we could use the WordNet entry on "rebellion", at
http://wordnetweb.princeton.edu/perl/webwn?o2=&o0=1&o7=&o5=&o1=1&o6=&o4=&o3=&s=rebellion&i=4&h=01000
0#c
There we find:
Noun
•S: (n) rebellion (refusal to accept some authority or code or convention) "each generation must
have its own rebellion"; "his body was in rebellion against fatigue"
•S: (n) rebellion, insurrection, revolt, rising, uprising (organized opposition to authority; a
conflict in which one faction tries to wrest control from another)
so we get 5 useful synonyms. Then we could use some of its relationships:
-----------------------------------------
This might work, but there are other words as well: Nat Turner. I could see the machine changing "Turner" into "spatula" and perhaps "nat" as a variant of "gnat" and somehow coming up with "fly swatter." (Joking!)
While working with the bits and pieces of the individual words of a concept may work, that would require some analysis before deciding if it could work. What would an automatic translator do with something like "Persian Gulf War, 1991?" That could be really bizarre!
Sounds like a fun project for a graduate student!
I think data-mining titles from our databases would be more productive. Since LCSH is based on literary warrant, it might stand a chance. The conceptual links are there already, e.g. "650 Persian Gulf War, 1991" has a high coincidence of these other terms related to it from the 245 field.
Jim Weinheimer
Received on Thu May 07 2009 - 03:36:38 EDT