[Corpora-List] Tables and lexis [was Re: Bootcamp]

[Linas Vepstas]

Hi,

2008/8/21 Wolfgang Teubert <w.teubert at bham.ac.uk>:
>
> Dear Linas,

I suppose I should have replied in private, so as to avoid
cluttering the list.

> something about the meaning of 'table:'

I'm not sure what to make of your email. Perhaps it is a collection
of 'random' sentences, all of which use the word 'table' in them.
In this case, one can pose a task of corpus linguistics: given
these sentences, obtain something that might be called a
'lexis' for 'table'.

Not easy, this task, as your sentences are all out of context, and
many of them exhibit sophisticated grammatical constructions,
while also conveying, and depending on, complex concepts only
tangential to 'table'. (Is Picasso a person? Who knows? Perhaps
he is a house painter? A table painter? A painter of tableaux?)

> Picasso does not paint a table with the inscription that it is not a table; he paints a table which is not a table.

Magritte has a painting of a pipe, on which he has painted
'Ceci n'est pas un pipe'  You have to see it to understand,
and you have to study art history to understand why. Perhaps
this is why Picasso never painted 'Esto no es una tabla'.

However, with access to a larger context -- i.e. more sentences
around that sentence that uses the word 'table' in it, one might,
perhaps, deduce that tables are physical objects, and that they're
inanimate.  How could one do this? Perhaps the use of
prepositions might suggest physical location: "on the table",
"near the table", and so we deduce with some probability that
maybe tables should be understood to exist in physical space.

Of course, one can easily be mislead: after reading "the heart
is in the chest" and "guilt weighed heavily on his heart", one
could erroneously conclude that "guilt" is a physical object,
quite heavy, and presumably located in the chest, as well.
Alas. Without some grounding in human emotions, one would
likely conclude many such idiocies.

> Point out a hundred reasons for saying that a hippopotamus is not a table: you'll end up agreeing that neither is a table a table -- it only seems to be a ...

A currently popular representation for 'meaning' or 'lexis' is in
terms of 'intension' and 'extension': lists of examples, and lists
of properties. You see this in wordnet, for example (is-a, has-a).
I discovered yesterday, from Wikipedia, that Karl Popper has
some math formulas capturing this, see
http://en.wikipedia.org/wiki/Karl_Popper
from which I quote:

   He defines it as: Vs(a)=CT_v(a)-CT_f(a)    where Vs(a) is the
    verisimilitude of a, Ctv(a) is a measure of the content of truth
    of a, and CTf(a) is a measure of the content of the falsity of a.

Similar definitions, but more deeply articulated, can be found in
Pei Wang's NARS (Non-Axiomatic Reasoning System).  He
proposes a reasoning system (deduction, induction, abduction)
that works with fractional measures of intension and extension --
so that when you assert "a table is not a hippopotamous", then,
I should add it to my list of things of what a table is not.

My mentor Ben Geortzel will very happily explain how NARS fails:
he points out that its not based on probabilities, and so has
interpretational difficulties; there are also some reasonable
constructions that lead to absurd output. But all is not lost: only
some minor tweaks rescue the basic concepts.

These last two paragraphs just barely scratch the surface; so,
for example, it would be quite foolish to add "a table is not a
hippopotamous" to my list, as it is something that can surely
be deduced from other knowledge, and serves only to clutter.
But then, we'd need to talk about what it means to "deduce" or
to "induce", etc. and so the conversation wanders.

--linas

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