Michael Tepper
Ling 567
Lab 7 Writeup

For sentential negation, Uzbek uses a verb suffix that appears directly after the main verb stem.  In the example below, the negation suffix, "-ma", appears after the verb stem "keltir-", which means "to cause," or "to bring about.":
	Men-ga zarar keltir-ma-s.
	I+DAT harm cause+NEG+AOR
	It doesn't hurt me.
This suffix is consistent across all verb forms, and it can basically be thought of as creating a negated stem.  In my grammar, I implemented negation with the neg_verb-lex-rule, which is the first rule that applies to verbs in my lexical rule hierarchy.  It looks different from my other rules because it relies on a change in sign type to prevent it from spinning during generation.  The rest of my rules use a kludgy feature I introduced called INFLECTION (which serves basically the same role as 'INFLECTED bool' from the Matrix), indicating the current state of inflection on both verbs and nouns.  Since, syntactically, the negation suffix derives a stem, I did not think it was elegant to use the INFLECTION feature.  Plus, using INFLECTION would have required two rules: one to simply pass on the bare stem, and one to derive the negated stem.  As it is, my trees are spindly enough (I have not even touched the magically productive processes of agglutination and cross POS class derivation in Uzbek).  
In order to get negation working, I did not have to add anything beyond what the assignment suggested.  I had to work out the above wrt my lexical rule and I had to connect up the semantics.  I noticed that the assignment suggested we use the MRS type "adv-relation," which does not exist in the matrix version that I have.  To get the right MRS representation, I just used arg1-ev-relation.  As far as actually deriving the correct MRS, I simply followed the instructions in the assignment.  The only real frustration that I had was when I noticed that I kept losing the handle to my verb in the qeq introduced by the negation suffix.  Then I went back and read the instructions for "can" as a bound morpheme, which kindly suggested to use the semantics (the LTOP) of the lexical rule's daughter.  Considering that I am inheriting from 'infl-cont-change-only-ltol-rule', I needed the information in the bound morpheme section that detailed what was necessary to hook-up semantically between the mother and the daughter.  I am not sure that my implementation is correct, but I seem to be getting the right semantics.
In Uzbek, "can" could be thought of as either a raising verb or a bound morpheme.  Orthographically, there is usually a space (but sometimes not) between the "can" and the main verb of the sentence, which is in one of the infinitival verb forms:
	Men-ga zarar keltir-a ol-ma-s.
	I+DAT harm cause+GER can+NEG+AOR+(3RD)
	It can't cause harm to me.  (It can't hurt me.)
The alternative representation for the above sentence would be:
	Men-ga zarar keltir-ol-ma-s,
where verb and auxiliary lean together.  However, in general this process is restricted, and I do not know details about which verbs this is actually the more likely representation.  Anyhow, I chose to represent can as a raising auxiliary verb.  I followed the instructions carefully, and did not have to implement anything beyond what the assignment suggested. 
The only thing that I am missing, with respect to the use of the abilitative auxiliary in Uzbek, is the fact that 'olmoq,' the auxiliary that I implemented, sometimes alternates with 'bilmoq,' which means something more like "to know how to do."  For some verbs, 'bilmoq' is actually preferred as the "can" auxiliary.   Considering I do not know how strict the constraint is, I am not sure what the most reasonable or elegant way might be to include this constraint.  Maybe this would be something to pass on to a semantic component, since the only solutions I can think of involve a bifurcation of my verb-lexeme types.
I checked the semantics for my can verbs and I appear to be getting the correct MRS.  Implementing this part was very easy, simply following the instructions from the lab assignment.  The most difficult part was simply trying to figure out what words to use. Uzbek did not have any attested sentences among the "I can eat..." websites, so I had to make up my sentences.  I hope my summary is not too brief.  
  