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

[John F. Sowa]

On 1/29/2011 9:26 PM, Mike Maxwell wrote:
> At one end of some kind of continuum, we have compounds that are
> probably one-offs; maybe 'rubberneck' (a verb).  At the other end, we
> have compounds that appear to be completely general, like 'X sensor'...

Yes.  The general principle is that a compound N1-N2 implies that
there is some relation R between the two components.

In the one-off case, R is a two-argument lambda abstraction from some
uncommon (but highly salient) sentence that includes N1 and N2.

In the highly productive cases, R is also a two-argument lambda
abstraction from a sentence, but it's one that is very commonly
associated with either N1 or N2 (or one of their hypernyms).

The continuum results from the commonality of the base sentence.

> So are the completely general compound patterns analogies, or rules?

Both.  There is a continuum between informal, case-based analogical
reasoning and formal rule-based deduction.  Every application of
a rule (If p, then q) depends on an analogy that matches (unifies)
p with some current case p' to derive a conclusion q'.

For the sensor example, the sentence is

    "A sensor measures something to derive a value."

The relation is

    R = (lambda x y)(A sensor x measures y to derive a value)

The rule is

    If (noun n) ('sensor' x), then sensor x measures n to derive a value.

When the compound occurs, the analogy matches the two words to the
antecedent to derive the consequent.

I discuss the relationship between formal reasoning and case-based
analogical reasoning in Section 4 of the following article:

    http://www.jfsowa.com/pubs/cogcat.htm

John

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