[Corpora-List] Moving Lexical Semantics from Alchemy to Science

[Angus B. Grieve-Smith]

On 1/29/2011 2:26 PM, Yorick Wilks wrote:
> We could probably have guessed that if you live in the UK, rather than in the US where (lots!) more English speakers live.
     I'm not sure what your prediction was, but I do live in the US, and 
I've only been to the UK for a total of three long weekends.  I've seen 
plenty of walking stick bugs, and eaten a bunch of banquet chicken, for 
what it's worth.  I have no memory of hearing that sense of "rubber 
chicken," until earlier this week as part of the phrase "rubber chicken 
circuit."  Apparently the first attestation of the "rubber chicken 
circuit" was in the L.A. Times in 1937:

http://www.wordorigins.org/index.php/forums/viewthread/1718/

     Now, can anyone write a program that could disambiguate "rubber 
chicken circuit"?

     Actually, it strikes me that one thing that's missing from this 
discussion of compound noun formation is the role of analogy.  "Rubber 
chicken circuit" clearly comes from other circuits like the lecture 
circuit and the Keith circuit.  "Devil's food cake" makes a lot more 
sense if you contrast it with "angel food cake."

     Compound formation also interacts with clipping in unpredictable 
ways, so "cheesesteak" actually comes from "cheese steak sandwich," 
while "buffalo burger" and "veggie burger" both resulted from the 
clipping of "hamburger."  There are also borrowing and dialect issues, 
which can be seen from the great debate over "eggplant parmesan":

http://www.cliffordawright.com/caw/food/entries/display.php/topic_id/4/id/109/

     My guess is that you'd be able to figure out the meanings of some 
large fraction of compounds automatically, but there would be others 
that would be opaque.  And I think that'd be a fairly good model of the 
way that native speakers go about understanding their language.

-- 
				-Angus B. Grieve-Smith
				Saint John's University
				grvsmth at panix.com


_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora