Katie Sickles
LING 567
Lab7

General Changes:

	I made the change to have my test suite now be the morpheme line
rather than the standard orthography line. The inflection in the irules
now reflect this as well (eg -um). I've decided that this better
represents the stems and suffixes of the language.

Fixes:
	
	The adverbs now work correctly. There was just a syntax error in the
lexical entry that prevented the grammar from being loaded. Sentences such
as:

	Kona borda-ar helst mat-inn
	woman.FEM.NOM eat.3sg gladly food.MASC.ACC-the.MASC.ACC
	'a woman gladly eats the food'
	
	The ditransives and transitives with optional object now load in the
grammar, but still no parses for them yet.

Matrix yes/no questions:
	
	In Icelandic the questions are former by inverting the subject and verb:
	
	Elska-ar kott-ur-inn hund-inn?
	love.3sg cat.MASC-NOM.MASC-the.NOM.MASC dog.ACC.MASC-the.ACC.MASC
	'does the cat love the dog?'
	
The customization script already provided the types and rules to account 
for this. This sentence parsed, but I found through generation that the
grammar would erronously allow the base form of the verb:

	*Elska kott-ur-inn hund-inn
	
This is was happening since the subject-inversion rule was applying to any
verb. I fixed this by requiring that it's daughter had to be of type
num+pers+lex+rule (the top verb lex rule). This constrained it to only
applying to verbs that had already been conjugated (gone through one of
the verb lex rules).

Embedded clauses:

	This embedded clauses appear with complementizers and following a few
set of verbs such as 'ask', 'say', or 'wish':

	eg heyra ad kott-ur-inn elska-ar hund-inn
	I hear.1sg CL cat-NOM.sg-the.NOM.sg love.3sg dog.ACC.SG-the.ACC.sg
	'I hear that the cat loves the dog'
	
The structure of these embedded clauses is the complementizer taking a
sentence as its complement. To implement this I created the clause-emb-verb-lex
type and had the verb "heyra" be of that type rather than a transitive
verb. This new type required it's comp be [ HEAD comp ], so that it
must take an complement with a complementizer as it's head.
	I had to create a new lex item for the complementizer. This type
has a ARG-ST of one item that is made of its comp. The comp list constrains
its element to be [ HEAD verb ]. This originally overgenerated, so I also
constrained the comp to be [ MC - ]. This fixed the overgeneration and
still parsed the grammatical phrases.

Imperatives:

	There are three imperative forms for the verbs. One for 2sg and 2pl
as well as one for 1pl:

	Hoppa
	jump.2sg.imp
	
	Hoppa-id
	jump.2pl.imp
	
	Hoppa-um
	jump.1pl.imp
	
Due to this phenomena, I didn't constrain the imp-head-opt-phrase to be
2nd person. The person distinction was taken care of in the imperative
lexical rules. This phenomena still doesn't parse. When I look at the 
parse chart, it shows that imperative lexical rule applied, but it
doesn't form the phrase.

Test Suite:

	The overall coverage was greatly improved with this lab. It went from
around 35% to 55%.