Michael Tepper
Ling 567
Lab 8 Writeup

Coordination in Uzbek is, on the surface, fairly straight forward.  There is a single lexical item that performs mono/polsyndetic coordination for nouns, verbs, and modifiers.  Here are several examples of its use:

	Odam va qiz bog'ni koryapti.
	man and girl garden+ACC see+PROG+3RD
	The man and the girl see the garden.

	Odam bog'ni ekyapti va sug'o'ryapti.
	man garden+ACC plant+PROG+3RD and irrigate+PROG+3RD
	The man is planting and irrigating the garden.

        Odam oq va yashil fermani koryapti.
	man white and green farm+ACC see+PROG+3RD
	The man sees the white and green farm.

Although none of my examples show it, it is also common in Uzbek to get monosyndetic coordination of more than just two coordinands (for each of these POS types).  

There are at least two other ways that Uzbek can coordinate nouns with a conjunctive meaning.  One of them is with n-polysyndetic use of "va".  So:

	Va odam va qiz bog'ni koryapti.
	and man and girl garden+ACC see+PROG+3RD
	(Both) the man and the girl see the garden.

I found that the above construction is in quite common use on several Uzbek online forums, but the structure was not included in my grammars (nor did I have time to consult a native speaker about it).  I assume it intensifies the inclusion of the first element ("va odam" in my example).  It would probably be wise to get a judgment on this case, but I just thought I would include it to see if I could get two versions of n-polysyndeton (as well as polysyndenton) working side by side for noun phrases.

There is another construction that uses n-polysyndeton to coordinate noun phrases.  The Uzbek word for "also" or "too" is "ham," and it appears postposed after the phrases that it modifies.  When it appears n-times after n noun phrases, it is interpreted as having the meaning "both X1 and X2..." or "this X1 as well as X2".  Here is an example:

	Odam ham qiz ham bog'ni koryapti.
	odam also girl also garden+ACC see+PROG+3RD
	Both the man and the girl see the garden.

For verbs, there is an interesting morphological coordinator that I decided not to cover.  It is a gerund-forming suffix, (-ib), and has multiple uses in Uzbek phrase structure.  When used to coordinate several VPs (that are required by the construction to share the same *referential* subject), this suffix (and only this suffix) appears on the first n-1 verbs, and then the last verb carries the regular inflection for the whole sentence.  An example would be:

	Odam bog'ni ekib sug'o'ryapti.
	man garden+ACC plant+GER irrigate+PROG+3RD
	The man is planting and (then) irrigating the garden.

This construction usually implies that the coordinated events occur in sequence, but this is not always true.  However, I did not implement this construction because I am not clear to what extent it carries the "and then" meaning in Uzbek.  Apparently there is at least one other way to conjoin verb phrases (or at least their meaning) with another gerund-type construction.  However, both -ip and other gerund constructions in Uzbek are required by light verbs (like olmaq - "can"), and I think it is more likely that verb structures coordinated in this way end up being interpreted as some type meaning incorporation with only a single event for both verbs.

Whew.. anyhow, on to how I handled these phenomena.

First I subtyped each rule in the monopoly coordination process, for general use with my lexical types.  The problem is that I have this clunky analysis of case right now and I am assigning case to all head types to capture a funky verb-participle construction in Uzbek.  So, I am enforcing that each of my coordination rules not only agree in case (this is too strong, more on that later) but also base up their case information to the the top node.  I also introduced two subtypes of the rules for n-polysyndeton that pass up case.

For noun coordination coordination, I simply subtyped the bottom, mid, and top rules above.  I did have to make one big change in order to capture the n-polysyndetic coordinator "ham."  Since it appears phrase-finally, I added a second binary bottom rule to the coordination module.  This rule is probably necessary for languages that have regularly postposed coordinators.  The rule I added is called "binary-bottom-coord-rule", and the only way it differs from "binary-bottom-coord-rule" is that its argument structure is reversed.

Since adjectives do not participate in agreement in Uzbek, adjective coordination is incredibly simple.  Since adjectives are connected with the mono/polysyndeton "va", I just subtyped a set of the three monopoly coordination rules for adjectives.  I did have to modify the bottom rule so that it would connect the handle of the right-most coordinated phrase to the coord-rel, because it was not done automatically.  

Verb coordination is still fairly messy.  So far, I have just connected up the three mono/poly coordination rules (i.e. I subtyped them for verbs).  Since I do not know how Uzbek handles V-coordination (I think that the gerund constructions are much much more common for this type of coordination), I decided to only implement VP coordination by making sure that all the arguments taken by my coordinators and coordinating rules are COMPS empty.  Next, since it seemed from E-post that S-coordination is not quite working yet (Jonathon said something about missing m-rels), I decided to wait on that as well.  Therefore, I made sure that each of my VP coordination rules only applied to things that were in need of a subject.  

With all of this hedging and limiting my task, I am still overgenerating in the extreme.  The problem is that number and person agreement is lost with my coordinated phrases.  So, I can parse a sentence like:
	*Odam va qiz bogni kiryapman.
	man and girl garden+ACC enter+TNS+1ST
where the subject completely disagrees with the verb.  I know how to go about fixing this (from discussions on E-post etc) but I do not really know how number and person agreement works in Uzbek.  I am secretly hoping that it is really easy, like maybe the verb just agrees with the last item in a coordinated subject.  However, I am probably going to have to implement a *matrix* of subtypes for my verb coordination rules, in the same way that Emily suggested for Ryan to do, that explicitly define the person and number on the verb based on what has occurred in the subject for all combinations of person and number.  

I hope this is enough for you and Scott to diagnose some problems with the module.  I think that the only significant problem I had was the ordering restriction on coordinators and coordinands, in the bottom rule.  