Ryan Mattson
5-1-05
Lab#4 Writeup
LING 567

The ditransitive verb sentences I added to my testsuite are designed to take
advantage of the relatively free word order in Finnish. The following
illustrate the general paradigm:

1) He   annavat tomaatit     miehelle
   they give    tomatoes-acc man-allat
   'They are giving the tomatoes to the man'

2) He   annavat miehelle  tomaatit
   they give    man-allat tomatoes-acc
   'They are giving tomatoes to the man'

3) He   annavat miehelle
   they give    man-allat
   'They are giving [something known in context] to the man'

In any case, the word order is free regarding the objects of the verb, but
in the case of a verb like 'antaa' (seen above) one object is not necessary
to the grammaticality of the sentence. The allative case object is needed,
though, to establish the target that is being given something. I have not
had the time yet to implement a freer word order, but I will get to that
soon. As of yet I haven't included duplicate types to allow for the switching
of the order of the objects, either, thus no parses are expected for test
items that are in a different order than the currently implemented dtr-verb-lex
type. (To note: an abundant amount of parses are showing up for sentences
using the complement order that isn't implemented. These are spurious parses
being licensed by a mishmash of other rules, which I will begin to track
down soon.)

In Finnish, only first and second person subjects to verbs are optional. The
third person subject is always required. In order to accommodate this
phenomenon I created two subtypes to 'person:' third and non-third, with two
sub-types to 'non-third:' first and second. This allowed me to constrain
my head-opt-subj-phrase to only allow for optional subjects that agreed
with the type non-third.

The grammar is behaving as expected with subject optionality. Third person
subjects are required (no parses show up with a third person conjugated
verb that is missing the third person personal pronoun as a subject) and
first and second parses can omit the subject.

However the grammar is overgenerating when it comes to complement optionality.
The complement rules are not constrained enough, it seems, to rule out
structures licensed by the recently implemented adjective rules, and this
rule is showing up in a lot of places to allow for some very bizarre trees.
The adjective rule, then, needs to be further constrained to only work on
things of type adjective.

Modification in Finnish seems to want adjectives to come in front of nouns, but
wants adverbs to come after verbs. I'm going to talk to my Finnish instructor
and ask about the exact argument structure when it comes to these modifiers,
but from my current experience it sounds ungrammatical to put adjectives after
their nouns and adverbs before their verbs. The constraints needed to be 
placed on these adjectives are [POSTHEAD -], but the adverbs need to be 
[POSTHEAD +]. One easy to deal with aspect of Finnish adjectives is that 
there is no set order for them, unlike in English:

4) a) The big red house
      *The red big house

   b) Punainen iso talo
      red      big house
      Iso punainen talo
      big red      house

Both 4a and 4b are perfectly valid to say in Finnish. The POSTHEAD constraints
seem to license the appropriate structures. On a whim I tried stacking
adjectives, and the grammar licenses appropriate structures, except it is
currently overgenerating. The adjectives are being stuck in at various levels,
and the grammar needs to be further constrained so that adjectives attach at
the correct part of the tree. I believe the same needs to be done for adverbs.

One aspect of adverb behavior that I'm not sure about is the licensing of
adverbs that are sisters to S and license a second root node above. This is
the one overgeneration of adverbs, and I'm guessing that to rule this out
the type that adverbs can modify needs to be something that excludes the root
node more than just requesting something of [HEAD verb] (which root nodes
do get).

A final note: there is an extreme amount of overgeneration happening with
some ditransitive verb sentences. I'm not exactly sure why all these structures
are being licensed, but it's going to take either a lot of work going through
and correcting each thing, or there's one or two rules that I'm not writing
correctly that are causing the overgeneration. I have yet to track down the
exact problem.