4.0379 Notes and Queries: Collocations; IBM Greek (2/36)

Elaine Brennan & Allen Renear (EDITORS@BROWNVM.BITNET)
Thu, 9 Aug 90 16:54:30 EDT

Humanist Discussion Group, Vol. 4, No. 0379. Thursday, 9 Aug 1990.


(1) Date: Thu, 9 Aug 90 13:57 EST (25 lines)
From: LEWIS@cs.umass.EDU
Subject: inference of semantic properties of collocations

(2) Date: Thu, 09 Aug 90 11:01 CDT (11 lines)
From: A10PRR1@NIU
Subject: Greek alphabet on an IBM

(1) --------------------------------------------------------------------
Date: Thu, 9 Aug 90 13:57 EST
From: LEWIS@cs.umass.EDU
Subject: inference of semantic properties of collocations from statistical ones

I recently encountered in papers by Church, etal and Smadja the idea that
collocations with different characteristics occur on different scales.
For instance, two words which are part of compound ("White House",
"bowling alley", "computer scientist") will have a mean separation
between word occurrences close to 1.0 and very low variance in this
separation. Collocations which are based on some semantic relationship
will have a much larger variance in the separation between occurences of
the word in a document.

Can anyone give me some pointers to other literature on this phenomenon?
I'm particularly interested in the possibility of statistically
inferring the extent to which a compound is compositional ("computer
programmer" > "bowling alley" > "White House") in its meaning.

Best, Dave

David D. Lewis ph. 413-545-0728
Computer and Information Science (COINS) Dept. fax. 413-545-1249
University of Massachusetts, Amherst NET: lewis@cs.umass.edu
Amherst, MA 01003
USA
(2) --------------------------------------------------------------14----
Date: Thu, 09 Aug 90 11:01 CDT
From: A10PRR1@NIU
Subject: Greek alphabet on an IBM

Norman Coombs,

Tell your colleague to look into Nota Bene (from Dragonfly Software); it
handles English and Greek easily.

Phil Rider
Northern Illinois University