Michael Tepper
Grammar Eng. Ling 567
Lab 6 Writeup

Before I forget, my test suite this week is lab5.items.  It contains
instances of the structures that we were supposed to capture for lab6.

Matrix interrogative clauses in Uzbek are made with a question
particle that is almost always tacked on to the very end of the verb.
To my knowledge there is only one tense (aorist) that violates this
principle.  Here are a few examples of this construction:

Sen bogni koryapsan+mi?
;you see+TNS+2ND+Q
;Are you seeing the garden?

This tense is narrative present tense, so I am not really sure of the exact 
meaning of this phrase (or any of the phrases I have been working with for the
past seven weeks :-))  

Embedded interrogatives are made in a similar way to
participial-clause embedded declaratives. One of the hardest parts
about this lab was detective work I had to do in order to find this
construction.  Initially, I was on the wrong track with my embedded
clauses because there are at least three ways to make them in Uzbek.
The first way is the +ki clause, which I do not think I explained to
well last time.  Basically it works alot like the English that-clause.
The matrix verb of the sentence comes first, then the sentential
complement:

Men bilyapman+ki bogni kiryapman.
; I know+TNS+PERS+COMP garden+ACC enter+TNS+PERS.
; I know that I am entering the garden.

However, there is an alternative construction, which was well attested
(and even recommended as grammatical Uzbek) by the Uzbek universe of
blogs and online forums.  This construction uses a participle that is
actually possessed by its subject, and then case-marked as the
complement of the matrix verb:

Men bogni senning kirganing+ni bilyapman.
;I garden+ACC you+GEN enter+NOM+2ND+ACC know+TNS+1ST.
;I know that you entered the garden. 

The +ni is emphasizing the fact that the participle is taking
accusative case.  Notice that the object of the participle,
garden+ACC, is also case marked. It turns out that, when used this
way, the participle's tense is ambiguous tense between present and
past.  It is possible to remove that ambiguity with time adverbs.

So, now that I have described the participle embedded clause
construction, I can talk about the whether clause.  Basically, whether
clausees are made by an interesting reduplicated fashion, where the
verb root is first stated in positive, which takes a conjunction
suffix, and then this bit is hyphened to the verb in the negative,
which is suffixed by the full participle aperatus.  For example, the
verb kormaq - to see, would become kor+CONJ-kor+NEG+NOM+PERS+CASE,
which in 3rd person singular would look like: korup-kormagani+ni,
where +ni is the accusative case marker.  An example sentence with
this construction is:

Men bogni senning kirip-kirmaganingni bilyapman.
Men bogni senning kirip-root-maganingni bilyapman.
;I garden+ACC you+GEN enter+CONJ-enter+NEG+NOM+2ND+ACC know+TNS+1ST.
;I know whether you entered the garden.

This sentence demonstrates my little trick for dealign with these,
morphosyntactically speaking.  Since I do not know of a way to copy
over the root so that I can inflect it twice, using LKBs morphological
component, I instead assume a preprossessing step where the second
root is written out as the english "root".

Syntactic Coverage, in General

As far as syntactic coverage goes, for matrix and embedded questions,
I think that I am covering the phenomena fairly broadly at this point.
Uzbek actually does have a floating question particle (kinda like
Finnish and Latin) to mark the emphasis of the polar question, and I
am not representing that yet.  I have heard that the places it floats
to are quite restricted.  I am pretty sure that the Q particle can
float to the subject position, but I do not think that it can float to
the object position, or to any other case-marked adjunct position.
Concerning embedded questions, the only thing that I *know* I am not
covering is the orthographical form of the question participle.  Of
course, I am concerned that I have not seen the full range of what
Uzbek can do with embedded questions.  Above, I described only two out
of the *three* ways that Uzbek can handle embedded declaratives.  I
would not be surprised if there was another way to make embedded
questions as well.

How I am handling the syntax.

For matrix interrogatives and matrix imperatives, I introduced several
features on the verb head to indicate whether or not the verbs had
been inflected for an interrogative or imperative sense.  Since my
matrix interrogatives and matrix imperatives have the same syntactic
structure as my matrix declaratives, all I had to do was make sure
that some marking of the structure underlying was visible at the top
of the tree.  Then, I introduced non-branching rules to apply the
message semantics to the new interrogative and imperative clauses.

It was somewhat difficult getting the semantics right for questions,
since this was really the first time that we have not had the
semantics done for us.  However, it was informative to play with the
HCONS and RELS difference lists, and to help build an intuition as to
how to build an MRS in general.  I think that I am getting the
semantics right for both types of clauses, at this point.

Now, embedded interrogatives were a *beast* to implement.  Basically,
my strategy was to first handle the participial matrix declaratives
and then the matrix interrogatives, which are done in the same way,
would be very little extra work on top of that.

In order to get the embedded participial clauses working, I made a few
false starts.  I introduced a few types into the verb semantics.  I am
most embarassed by the feature that I introduced under verb mood,
called 'nominalized' which may or may not be doing any actual work for
me at this point.  I have a general questiona bout what mood is used
for (or should be used for) in the matrix.  Should it ever be used
discriminatively, or should it just be used for semantic
interpretation later on?

Anyhow, the main *hack* that allowed me to get participial embedded
clauses was introducing case as a head feature on all words.  Then, I
introduced two main subtypes of case, "no-case" and "noun-case."  In
Uzbek, everything except nouns starts with "no-case."  However, in
verbs that become nominalized, like these participle constructions,
end up with the "noun-case" feature, which is filled in by the lexical
rule that derrives the participles from the root verb.  The same rule
also fills in the fact that the subject of these participle
constructions must appear in the genitive, as opposed to nominative
case with matrix verb subjects.

As I suspected, once I introduced declaratives using this
construction, getting interrogative embedded complements was fairly
trivial.  Both are just participles, after all.  However, I had to
make sure that I marked the interrogative participle QUES +, so that
it would go through my unary rule question clause rule, as opposed to
the unary declarative rule.

Finally, the only thing I really have not talked about much is the
imperatives: Here is an example of an Uzbek imperative sentence: Bogni
kir!  ;garden+ACC enter!  ;Enter the garden!  I did a halfway job on
these (I still need to implement the plural imperative, which takes a
morphological affix, as opposed to just the bare verb stem.)  However,
after implementing a unary rule to get the message semantics for
commands, I think that I am at least getting the semantics right for
these things.
