Ryan Georgi
Ling 567

Lab 7 Writeup

----===== MATRIX YES/NO QUESTION ====---

Thankfully, from what I've read in both my primary grammar and other sources,
the question marking in Arabic is quite regular, with simply the particles /a/
or /hal/ prefixing the matrix clause. (The distinction between the two
is one of habitual aspect; for simplicity, I'm sticking with the more
general (non-habitual) particle, /a/).

However, as I will get to later, these complementizers are not acceptable
in embedded clauses, so I placed the constraint that it be [MC +] and the
constraint on my embedding-verb-lex that its complement be [MC -], which
successfully allows my embed-comp-lex (constrained to MC -) but not the
matrix-clause qpart-lex-item.

Examples:

a     t-ins.arifu   ?al-bint-u
QUES  3FSG-leaves   the-girl
'Does the girl leave?'

  ... and

y-istakh   ?al-rajul-u  ?in ?al-bint-u    t-ins.arifu
3MSG-thinks the-man-NOM  if  the-girl-NOM  3FSG-leaves
'The man asks if/whether the girl leaves'

... but not

y-aftikir  ?al-rajul-u   a     ?al-bint-u  t-ins.arifu
3MSG-thinks the-man-NOM  QUES  the-girl-NOM  3FSG-leaves
'* The man asks if/whether the girl leaves'

.. nor

?in   t-ins.arifu ?al-bint-u
QUES  3FSG-leave   the-girl-NOM
'* Did the girl leave?'

The qpart-lex-item also contributes [SF ques] to the matrix clause, which
shows up correctly in the MRS.


----===== EMBEDDED CLAUSES & COMPLEMENTIZERS ====----

It appears that the usual suspects for embedding clauses do much
the same in Arabic, such as believe, want, say. I'm not sure of the analysis
for "missing" complementizers in English ("I think -NULL- should..."), but
I didn't find any examples in my grammar, so I won't worry about optionality.

The complementizers are as follows:

?anna ?an (p. 278) -- 'that' (declaratives)

?in, ?iDa:, or law -- 'if/whether' (interrogatives)

?anna and ?in are followed by S(V)COMP, and ?an by VSCOMP (or VCOMPS)

?iDa:, law, and ?an all have, again, various different aspectual meanings
that I'm not prepared to cover, so I'll stick with ?anna and ?in.

The multiple forms of the above complementizers correspond with variations
in word order and different shades of semantics, along the lines of
what was known and when, as well as the speaker's evaluation of the 
belief states of the clause being embedded. The most common (and
generic) usage, however, is ?anna or ?in followed by a S(V)O clause.

To handle these forms, I created two types; interr-comp-lex and
declar-comp-lex, which constrained their SF to ques/prop, respectively.
These types both inherited from embed-comp-lex, which adds the constraint
[MC -], so that these embedding complementizers can't appear in the matrix
clause.

I also made seperate types for the verbs, depending on whether they take
interrogatives (ask) or declaratives (think), and gave them a supertype
embedding-verb-lex which constrains the SF of its COMPS item to be
the same as its own, as well as [MC -]. This ensured that the if/whether
and 'that' appeared with the right verbs.

Examples:

y-istakh    ?al-rajul-u  ?in ?al-bint-u   t-ins.arifu
3MSG-asks  the-MAN-NOM  whether  the-girl-NOM 3FSG-leaves
' The man asks if/whether the girl leaves'

... and

y-aftikir    ?al-rajul-u  ?anna ?al-bint-u   t-ins.arifu
3MSG-thinks  the-MAN-NOM  that  the-girl-NOM 3FSG-leaves
'The man thinks that the girl leaves'

...but not

yistakh     ?arrajulu    ?anna lbintu       tins.arifu
y-istakh    ?al-rajul-u  ?anna ?al-bint-u   t-ins.arifu
3MSG-asks  the-MAN-NOM  that  the-girl-NOM 3FSG-leaves
'* The man asks if/whether the girl leaves'

 ... or ...

yaftikir     ?arrajulu    ?in lbintu       tins.arifu
y-aftikir    ?al-rajul-u  ?in ?al-bint-u   t-ins.arifu
3MSG-thinks  the-MAN-NOM  whether  the-girl-NOM 3FSG-leaves
'* The man thinks that the girl leaves'

In addition to getting agreement for SF, there was also a little
legwork in getting the word order to change within the embedded clause.

My solution might not be ideal considering the deflected agreement I've
still to implement, as it currently blocks the alternate SVO word order
from occurring in the matrix clause.

What I did was to simply create an 'alternate' to the head-subj-phrase
in my types that, instead of being decl-head-subj-phrase & head-initial
is ...& head-final . I called this type embedded-phrase and constrained
it to be [ MC - ] .

Also, to keep the head-subj rule from applying
and allowing VSO in embedded clauses, I added the constraint to the
standard head-subj rule that it must be [ MC + ].

Ultimately, this kept a constraint of VSO in the matrix clause, but
made word order in embedded clauses a mandatory SVO.

Examples:

y-aftikir    ?al-rajul-u  ?anna ?al-bint-u   t-ins.arifu
3MSG-thinks  the-MAN-NOM  that  the-girl-NOM 3FSG-leaves
'the man thinks that the girl leaves'

...but not

y-aftikir    ?al-rajul-u  ?anna t-ins.arifu ?al-bint-u
3MSG-thinks  the-MAN-NOM  that  3FSG-leaves the-girl-NOM
'*The man thinks that the girl leaves'


------======== IMPERATIVES =======------------

Since verbs and verbal nouns are very close in the semitic
languages, there appear to be a number of ways to form constructions
with meanings similar to the second-person imperative using forms
like jussive verbal nouns, but I'm going to stick with the only
certainly imperative construction given by the book, which, similar
to English, is a special verb form without the subject.

In Arabic, this special verb form is the zero-marked stem of the
verb, so I made an imperative-lex-rule lrule of type 
const-add-only-ltol-rule that added the constraint [ FORM imp ]
to the verb.

By also adding a subtype imperative-phrase that inherited from
imp-head-opt-subj-phrase and constrained the HEAD-DTR's FORM to imp,
this took care of intransitive imperatives.

As I posted in Epost, however, there were issues in getting transitives
to work. I changed basic-head-opt-subj-phrase not to constrain its
COMPS list to be empty, and this seemed to fix it. I am now, however,
getting some strange things in generation when I check my imperatives
where I get suffixes standing alone before the object. I got imperatives
working a little late, so I haven't had time to check into this, but
I think I just need to constrain some irules a bit better.

I also have a problem where these irules don't appear to be applying
as mandatory, so zero-marked verbs seem to parse just fine everywhere.
Thus, I haven't been able to ensure that they are correctly staying
out of embedded clauses. Below are some examples that should be
a test case, however.

y-istakh    ?al-rajul-u  ?in ?al-bint-u       t-ins.arifu
3MSG-asks  the-MAN-NOM  whether  the-girl-NOM 3FSG-leaves
' The man asks if/whether the girl leaves'

... but not

y-istakh    ?al-rajul-u  ?in ?al-bint-u        ins.arifu
3MSG-asks  the-MAN-NOM  whether  the-girl-NOM  leaves
'* The man asks if/whether the girl leaves'

---=== WRAP UP =====-------

Though the numbers aren't quite comparable to other labs, since
I've been adding test data, my grammar now says its coverage is
31 examples, for an overall coverage of 40%.

I've still some overgeneration problems to solve, and really to
boost coverage I need to add in the copulas, but this is quite
an improvement!