Michael Tepper
Ling 567: Grammar Engineering

Lab 3 Writeup

Summary

The work I did most closely resembles the first package, since my language 
has both case and agreement.  I decided to implement lexical rules for plural 
common nouns, noun case, verb person/number agreement, and one verb tense.  
I ended up writing many more than the required two lexical rules.

Some Details

The case system in Uzbek is fairly straightforward.  Subjects of both
transitive and nontransitive verbs take nominative case.  For now I model the
simplest verb paradigm, where objects take acc. or nom. case.  There are 
other paradigms where verb objects can take dative or ablative case 
(and maybe others).  Bare accusative objects are understood as definite,
bare nominative objects are understood as a "concept" undergoing verb 
incorporation.  Nominative objects that are specified by an overt 
indefinite specifier are (no surprise here) understood as indefinite. 

My sources disagree about what other explicit case marking (eg. 
acc, dat, abl,gen etc) of bare NP objects means for determiner semantics.
Right now, I am assuming that accusative-marked bare-NPs are the only NPs
that are interpretable as definite with syntactic information only.  

Therefore, acc-case marked NPs go through one bare-NP rule
and get marked for definiteness.  Other cases go through another bare-NP
rule and get marked as ambiguosly def/indef.  It is up to the pragmatic
/semantic component to decide whether these ambiguous cases become definite,
indefinite, or maybe go to some conceptual, verb incorporating space (for now).

I also implemented a piece of the Uzbek agreement system.  There is 
very little det-noun agreement in Uzbek (for instance, demonstrative 
determiners are always the singular form of the demonstrative pronoun). There
are plural determiners, but it is difficult for me to see how they are used.
Some plural determiners, like the one for "some," seem fine with either a
plural or a singular noun argument.  However, others seem more picky.  I will
have to implement this when I know a bit more.  I know how it works for 
Turkish, and if need be I can just assume the two systems are the same.

The chunk that I did implement was the verb agreement system.  In Uzbek, 
non-3rd person verbs agree in both person and number with their subjects.
I implemented this with lexical rules, taking each verb to an according
per/num by making sure that it would only accept an NP Subj with that
per/num.  

