Possessives

Sean Crist kurisuto at unagi.cis.upenn.edu
Wed Oct 20 01:29:57 UTC 1999


On Sat, 16 Oct 1999, Patrick C. Ryan wrote:

> [SC]
>> Just to amplify on this point, consider "a counterfeit dollar".  The whole
>> point here is that the item in question is _not_ a dollar; it doesn't
>> belong to some subset of dollars in the way that a black dog belongs to a
>> subset of the dogs.  Likewise with "a fake moustache", "an imposter
>> policeman", etc.

> [PR]
> One of the most challenging things about interpretation of any language is
> that words, spelled and pronounced the same, do not always mean the same
> things.

> In the phrase, "a counterfeit dollar", the "dollar" cannot mean a 'real
> dollar', obviously that would be self-contradictory. It means, in this
> phrase, rather 'something that looks, feels, etc.  like a dollar'. In that
> category, there are the 'real' ones and, another circle for the
> 'counterfeit' ones.

> But, for the sake of discussion, let us pretend to assume that it means
> 'real, genuine dollar'. Then what part of speech is 'counterfeit', and how
> would you say it is used syntactically?

Based on its distribution, it's obviously an adjective.  Despite its
semantics, it has exactly the same syntactic distribution as, say,
'green'. The point was that you can't always model adjectives strictly as
things which form subsets over nouns, which is what I understood you to be
saying in an earlier post.

Some commentary.  In mathematical terms, a language is a set of finite
strings over some set of symbols (and, folks, I am determined to ignore
anyone who takes me to task for defining the word 'language' in this sense
in this context).  The task of the syntactician of natural languages is to
produce a model which correctly assigns the right strings to the 'accept'
and 'reject' bins.  For example, a model of English is incorrect if it
accepts "John Mary loves"; it's also incorrect if it rejects "John loves
Mary".

Virtually all syntactic analysis over the last 40 years has defined word
categories in terms of the distribution of words within sentences- what
you have called the 'slot' approach.  It's quite true that one can at
least imagine other approaches, such as one where word categories are
defined in terms of word semantics rather than word distribution. I might
say, tho, that we rejected that approach many decades ago for the simple
reason that it doesn't work as well, as Larry Trask has aptly illustrated.

If you feel otherwise, then show us a model of language such as English
which does a more accurate job of assigning strings of English words to
the 'accept' and 'reject' bins than the best distributional model to date.
If you can do so, we'll be glad to accept your model.

  \/ __ __    _\_     --Sean Crist  (kurisuto at unagi.cis.upenn.edu)
 ---  |  |    \ /     http://www.ling.upenn.edu/~kurisuto/
  _| ,| ,|   -----
  _| ,| ,|    [_]
   |  |  |    [_]



More information about the Indo-european mailing list