[Corpora-List] PS:minimal changes in a paragraph (based on a corpus it appeared) ... (2nd attempt (after first one was deleted))

[John F. Sowa]

On 8/9/2011 8:20 PM, Albretch Mueller wrote:
> I think if you harness syntax really tight in a totally exaustive way
> (think of keeping all possible n-grams of all texts) "semantics"
> becomes some sort of illusion (how functional|irrelevant illusion it
> be, remains to be seen)

Semantics is *never* an illusion.  Communication doesn't happen by
a random accident with randomly generated words.  Somebody generated
those words to communicate some specific information for some purpose.

I'm sure that if you used N-grams with a very large N, every sentence
generated would *seem* to be meaningful.  But the apparent meaning in
the generated sentence would *not* convey the meaning that the speaker
had intended to communicate.

Semantics originates in the mappings of perception and action
to and from whatever is stored in the brain.  For example,

    Situation --> sensory icons --> interpretation by percepts -->
    concepts and conceptual structures --> words and syntax --> speech

I agree that statistics can be *helpful* in learning the mappings
and determining which mappings to prefer in any particular case.

But the meaning resides in the mappings that some agent chooses
for some communication to some listener for some purpose.
That meaning does not reside in a random selection of N-grams.

My question:  How do you use the speaker's intentions to control
the selection of N-grams (or other patterns) to convey the meaning?

John

_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora