Please do not restate anything I say, because your statements come out totally different from anything I would ever say.
> In particular John is suggesting somewhat the reverse of
> my position, by insisting that we need a(n unknown) semantic
> representation, separate from syntax, to make up for the
> shortcomings of syntax.
I never said anything remotely like that. Following are the points I would say:
1. Syntax and semantics are fields of study -- subfields
of a larger field called linguistics. They are *not*
representations. Therefore they are not and cannot
be *separate* representations.
2. Syntax talks about how little pieces of a language are
related to bigger patterns.
3. Semantics talks about how those pieces refer to things in some
domain and whether what they say about them are true or false.
4. Pragmatics is another field of study, which addresses the
purpose or intention of the speaker and/or hearer of a
language.
5. Finally, since syntax, semantics, and pragmatics address different
aspects of language, none of them makes up for any shortcomings
of the other two.
> I don't think there is a lot of work which takes my position that
> we can model semantics by making ad-hoc syntactic generalizations
> over corpora.
I am happy nobody is doing that, because it is impossible. One might discover syntactic generalizations from a corpus, but it is not possible to learn anything about the semantics and pragmatics without getting further information about how people use those the sentences in social activities.
Children cannot learn a language just by listening to the radio, and they can't learn much just by watching it used on TV. But they can learn language very rapidly by actively using it with other people in social situations.
John