Trees, pheno, tectogrammar

[Carl Pollard]

Hi everyone,

Just a clarification of Shalom's comment about CCG. Steedman's framework
differs from type-logical grammar in the sense that two combinators (=
constants that are expressible as closed terms of "pure" lambda calculus,
i.e. terms which contain no nonlogical constants as subterms) which are
equivalent (in the sense that the corresponding terms are term-equivalent)
can have different prosodic interpretations. My proposal resembles TLG,
not CCG, in that respect. Also, like TLG, there is no way to make some
combinators go away, or only apply to arguments of certain categories.
Perhaps unsurprisingly, Glyn Morrill views it as a form of categorial
grammar and Mark Steedman isn't so sure.

I hope my previous message did not convey that order and prosody are
uninterpreted. They can correlate with semantic distinctions, but the
architecture requires that the correlation be mediated by syntax.
E.g. the English topicalization rule would be treated as a
(polymorphic) function Top_X that takes as arguments an X and an Sfin
missing an X (where the polymorphism schematizes over X). The
contrastive meaning would come from the semantic interpretation of
Top_X and the contrastive prosody (the pitch accents on the
phonological realizations of the arguments) from the phonological
interpretation of Top_X. So the pitch accents signal a contrastive
meaning, but Top_X, which lives in tectostructure, is the go-between.

Carl