Michael Tepper
Ling 567: Grammar Engineering

Lab 2 Writeup

1. The basic word order of my language is SOV.  However, the written word
   order is flexible between SOV and OSV with the NP being emphasized
   moving closer to the verb.  I do not yet have a sense for how common
   the OSV word order is in Uzbek writing, so I do not know whether it makes
   sense to call Uzbek a verb final language with free word order of the
   verbal arguments or simply default to SOV, which is the default word 
   order. For now, I am calling the word order in Uzbek verb-final, which may
   overgenerate w/respect to written Uzbek.  Here are several examples that
   show the flexible ordering of main constituents. 
     Odam bogi kiryapti.
     man+NOM garden+ACC enters
     The man enters the garden.

     Bogi odam kiryapti.
     garden+ACC man+NOM enters
     The man enters the garden.
   
   The order of noun phrases in Uzbek is determiner-noun.  As far as I can 
   tell, this is not an oversimplification of Uzbek grammar.  An example is
   the phrase "bir bog" which means "one/a garden"

2.  This comment talks about the indefinite determiner: 
;;; Indefinite determiner: may want to call this something else in the future. 
;;;
;;; There is an analysis of 'bir' that classifies it as limited determiner or a
;;; weak indefinite article.  There is another analysis (in Bodrogligeti) that 
;;; calls it a member of a small class of adjectives that can modify indefinte nouns. 
;;; I may not be doing his analysis justice, but so far it seems incomplete.  He
;;; tries to use it to explain the nominative/accusative case alternation in object
;;; NPs (acc case is necessary for definite interpretation, nom case *can* imply an 
;;; indefinite interp, but only when it occurs with one of these indefinite-selecting 
;;; adjectives (including 'bir').
;;; 
;;; This does not seem to work, first and foremost because it doesn't capture the fact
;;; that 'bir' adds an indefinite quality to all phrases it modifies, regardless of
;;; its case or its status as object/subj/adjunct, etc.  It seems to be working like
;;; an actual determiner in this way.   
;;; For now I will treat 'bir' as an indefinite determiner and allow it to specify  
;;; all noun phrases.
To sum up, I ended up giving bir the pred value for indefinite determiners, and
it should be a good match for now.

For the demonstrative determiners, Uzbek has a three-way distinction that should
be well matched to the PRED values that are already specified in the Matrix for
the three way demonstrative distinction.

3. I discussed bare noun phrase in this comment: 
; Bare Noun Phrase
; If a bare NP subject (topic) or object (argument) has been previously referred to 
; in the discourse (or is somehow known to both speakers), then it is definite, and
; this definite character would likely be translated using english "the".  
; This also applies to other types of nouns/phrases, to be elaborated on later.  So 
; far this is the best analysis that I could find (for Turkish) and the evidence in 
; Bodrogligeti seems to suggest a similar pattern for Uzbek.
; 
; For now, I will call all bare NPs definite.  There may be a slight wrinkle in this 
; solution because there are some instances of bare NPs that act like they have 
; undergone verb incorporation: u  kitob o'qiyapti 
;                               he book  reads+NAR+PER
;                               he reads./he reads books.
; We might just want to treat these as bare nouns, and have a derivational rule that
; creates an incorporating-verb, which takes the first thing to its left as a bare
; NP..  I need help with this because I think this will not work with MRS, but
; I think that creating another class of bare noun phrase just for this will add
; lots of edges with only a very small gain in coverage..  Answer to this may come
; more easily once I add case. 

I will repeat the glossed examples that I gave in (1.):  
When I tested these examples with bare NPs and displayed their MRS matrices, 
they showed up as being quantified (better word for this?) by a definite determiner.
     Odam bogi kiryapti.
     man+NOM garden+ACC enters
     The man enters the garden.

     Bogi odam kiryapti.
     garden+ACC man+NOM enters
     The man enters the garden.

4.   Odam bogi kiryapti.
     man+NOM garden+ACC enters
     The man enters the garden.
 When I generate, this sentence spits out spurious strings for the case 
 alternation between nominative and accusative.  For instance, it will
 spit out man as accusative (man+acc odam-i).  This is wrong, for sure, 
 but it will be fixed once case agreement is added in to the grammar.  It
 also spits out plural (men+PL odamlar), and this will also be fixed when
 we add agreement (number agreement this time).  The LKB also
 generates sentences with the SOV OSV alternation, but this is ok, since
 right now we are treating Uzbek as verb final verb-argument free. 