adjectives need more than SUBJs

[LFG List]

It is still too hot in Italy to be cool enough to disentangle all the
questions raised by Miriam Butt. However, Italian might be a good case to
discuss.
In the early '80s we at my University also worked on a Grammar for Italian
and the computational lexicon resulted with adjectives that not only needed
SUBJ but also selectional restrictions and syntactic position. There is
more than one reason for this:
1. Italian allows very free ordering of constituents: all adjectives may be
used in postnominal (or predicative) position - with a few exceptions - and
a certain number of them may be used also in prenominal position. Some of
those allowing both positions change their meaning accordingly.
    Un povero maestro vs un maestro povero
    A bad teacher  vs  a poor (not rich) teacher
In addition, in case the adjective allows the truncated form this is only
available in prenominal position, with a given meaning associated,
    Un buon *buono maestro vs un maestro buono *buon
    A good teacher  vs  a good (helpful) teacher
2. However, just looking at postnominal position, agreement might not be
sufficient to pick up the right governor,
    Gino ha incontrato Mario al cinema contento
    John met Mario at the cinema happy
where we do not want contento to modify cinema but Mario or Gino.
3. Anaphoric binding of morphologically unexpressed subjects in copulative
constructions: in case no selectional restrictions were available lots of
problems would arise whenever ambiguous case occur, such as
    Noto' che non era solida
    pro noticed_3rd_past that pro not was solid
    He noticed that it was not solid
where little pros have to be kept disjoint: the simplest solution is to
allow argument information to be made available as soon as possible, i.e.
after quantifier raising to serve anaphoric binding and before logical or
s-structure building. Of course there might be another solution in which
the discourse model should be made available whenever a MU pronoun is used
in order to fetch the antecedent and use that information to produce
disjointness for the two pros(!!).
4. I don't have a satisfactory solution for copula-less languages but there
is one thing to note, and that is: if some linguistic item such as a
predicate is omitted some other feature could be there to remind its
presence, like for instance a particular intonation - so I believe happens
in Russian, but I'm not aware of the other languages quoted.
In sum I am in favour of a lexical representation with fully expressed
functional and semantic information available at the same entry place, also
because I believe that parsing highly ambiguous languages - or structurally
underdetermined, as I define Italian - requires all linguistic information
to be available during the parse itself.
Cheers,
Rodolfo

**********************************************************
rodolfo delmonte Ph.D.
Associate Professor of Applied & Computational Linguistics
Section of Linguistic Studies
Ca' Garzoni-Moro, San Marco 3417
Universita' Ca' Foscari
30124 - VENEZIA (It)
tel.:39-41-2578464
lab.:39-41-2578452/19
fax.:39-41-5287683
**********************************************************