Lab 8 Grading Rubric

-------------------------------------------------------------
Comments:

 To test it, I created a new transitive verb whose complement is head
 adp and whose complements are fulfilled.  This prepositional phrase
 takes the same spot as a noun phrase on regular transitive verbs.  I
 choose the irregular verb 'essere' to test this.

 Questi libri sono per te.  (These books are for you)

I doubt you want to say there's a special 'essere' that takes
a PP[per] complement.  That looks more like a predicative use of
the PP after the copula.  Furthermore, the semantics is broken:
the pronoun's relations aren't showing up and the pronoun isn't
getting linked to the verb's relation.  You need your preposition
lexical type to inherit from raise-sem-lex-item and then the
semantics will be fixed :-)

The semantics for this:

 I amici danno un libro ad una ragazza.
 The friends give a book to a girl.

are similarly broken, but should also be fixed by the change
mentioned above.

 Since I had created the two types, adding this argument optionality
 didn't increase coverage, but actually decreased it.  I had originally
 analyzed the noun/verb inversion rule to only apply to verbs with no
 arguments, and so I created it as a lexical rule.  However, verbs with
 a null argument keep their argument until after it passes the
 head-opt-comp-phrase, disallowing the inversion rule from taking
 effect.  It worked fine with both the transitive and intransitive
 interpretation, but didn't reflect what was happening in the grammar.
 Sentences like 'Mangi-o io' (eat I) don't parse, and sentences like
 'Mangi-a Maria' (Eat Maria) only parse with the interpretation that
 something is eating Maria, rather than Maria is eating.  I have not
 taken the time to fix this yet.

I have an idea of how to do this, if I have the right generalization:

  *V Obj Subj
  V Subj  (inherently intransitive V or transitive V with DNI/INI)
  obj-V Subj (transitive V with object prefix)

This suggests that the inversion rule should be a phrase structure
rule that requires LIGHT + on its daughter (and then you'll need to
make verbs HC-LIGHT -).  That should make sure it can only apply
as part of a 'spindle'.  Alternatively (and this might be the better plan),
the rule can make the mother's COMPS list be a list of unexpressed
synsems. As we discussed today, build a type ulist (and ucons and unull)
parallel to olist, but without the [OPT +] constraint.

You probably want the accusative and dative pronoun rules to make
the corresponding arguments DEF +.

 The adjective could also be used with a copula.  In this case it must
 agree with the gender of the subject of the copula.  To do this, I
 created a verb copula that took an adjective as its complement and
 stipulated that the PNG value of the subject and the PNG value of the
 noun that the adjective is modifying be the same.

PNG isn't enough --- you'd like the whole index to be identified
so that in the MRS the adjective relation ends up predicated of
the subject.
