Orð og tunga - 01.06.2006, Page 53

Mattheiv Whelpton: Argument Structure
51
linguistics. Such work however is also related to the other develop-
ment mentioned earlier - corpora. Levin's work would not have been
possible without the availability of computer corpora. Corpora make
it possible to investigate directly the patterns in which particular
words appear. A grammatically tagged corpus allows us to investigate
the full range of "altemate syntactic environments" in which verbs ap-
pear.
2.3 Corpora and lexical analysis
This brings us to another example of work where both lexicographers
and linguistics can share and interact - the WordNet project
(http://wordnet.princeton.edu/).
"WordNet® is an online lexical reference system whose design is
inspired by current psycholinguistic theories of human lexical mem-
ory. English nouns, verbs, adjectives and adverbs are organized into
synonym sets, each representing one underlying lexical concept. Dif-
ferent relations link the synonym sets."
o The verb cut has 41 senses (first 10 from tagged texts)
o 1. (58) cut — (separate with or as if with an instrument; "Cut the rope”)
o 2. (18) reducc, cut down, cut back, trim, trim down, trim back, cut, bring down — (cut
down on; make a reduction in; "reduce your daily fat intake"; "The employer wants to cut back
health benefits")
o 3. (3) swerve, shcer, curve, trcnd, veer, slue, slew, cut -- (tum sharply; change dírection
abruptly; "The car cut to the lcft at the interseetion"; "The motorbike veered to the right")
o 4. (2) cut - (make an incision or separation; "cut along the dotted line")
o ...
Figure 5 - WordNet entry for verb “cut”
Such work is obviously of direct interest to lexicographers but as the
description of WordNet shows, there is a strong linguistic motivation
for the project, again in terms of the conceptual structures that imder-
lie lexical organisation.
This development is taken much further by Charles Fillmore
(Berkley) in the FrameNet project (http://framenet.icsi.berkeley.edu). This
is an attempt to model word properties within a cognitive grammar
framework in which syntactic distributions and cognitive frames of
reference for word senses come together.
As in WordNet, a verb is listed with a number of senses, though
in this case, the senses are not defined in terms of synonym sets but