Lab 4 Grading Rubric


Notes:

I think we discussed this at some point.  The way to get
the free order between the two complements of a ditransitive
is to instantiate basic-head-2nd-comp-phrase.

You didn't describe what you did for object optionality,
nor what the facts are in Finnish.

Your head-adj rules are probably overgenerating because
you haven't constrained your non-modifier parts of speech
to be [MOD <>].  Adding [MOD <>] to the defintion  (addendum
statement, actually) for noun, brought the number of parses
for "minA annan tomaatit miehelle" down to one.

Your write up does not describe agreement between adjectives
and the nouns they modify, but your grammar seems to be 
implementing it.

Please use full sentences in your write up to illustrate
your points.  I'd like to test the "big red house" example
and give you some feedback on the attachment sites, but
I can't without a full sentence (I don't feel like poking
around in the parse chart).

Your inflectional rules for adjectives need to specify
the HEAD value of the mother: since you're inheriting
from infl-head-change-only-ltol-rule, the type isn't copying
up the HEAD values.  If adj-lex doesn't say anything about
the CASE value inside MOD, you could actually use 
infl-add-only-nocont-ltol-rule instead.

The write up instructions requested that you describe how you
tested your analyses (preferably with examples I can play
with).  You should also include more detail in the write up
about the lexical types you added for modifiers.

The constraints on the SUBJ appear to be common to all your
verb types, and could be pulled up to verb-lex.

dtr-verb-lex seems to be underspecified: it's not requiring,
as far as I can tell, any of its arguments to be saturated
([SPR <>]) constituents.  I see that you're handling it this
way:

;; RGM 4-28-05 Fixed a head-comp-phrase overgeneration bug by restricting
;; the NON-HEAD-DTR to be a headed-phrase. This killed complements that were
;; only words. That is to say, complements can be single words, but they
;; better be phrasally licensed by a bare-np-phrase first.

... but you may want complements that are only words at some
point.  (Or, with coordinate NPs, complements that are phrasal
but not, at the top level, headed-phrases.)
Better, in my opinion, to constrain dtr-verb-lex
(parallel to what you have for tr-verb-lex and itr-verb-lex).

You know, any one of these overgeneration issues would have
been a perfectly fine thing to post to EPost about...

You're right that you need to add something to the adv-lex
definition to force VP attachment (and rule out S attachment).
Requiring a non-empty SUBJ value (parallel to your non-empty
SPR value) on the MOD should do it.