Katie Sickles
LING 567
lab 8

NEGATION:

	Sentential negation is created by using a negative adverb. This
	adverb usually appears directly to the right of the verb it's
	modifying, but in looking at corpora, it seems that it can appear 
	attach at the VP level as well. 
	
		eg borda ekki mat-inn
		I eat.1sg not food.MASC.ACC
		'I do not eat food'
		
	The lkb parses this sentence, but when generating I noticed two
	parses for:
		eg borda mat-inn ekki
	In inspecting the parses from this utterance, it seems that ekki
	is attaching allowed to attach at both the VP and S level, thus
	creating the two parses. This is caused by the constraints on
	neg-verb-lex which are HEAD.MOD < [ LOCAL.CAT.HEAD verb ] >, and 
	Ss are HEAD verb, same as VPs. I haven't been able to get this
	constrained properly yet. Despite this problem, the following
	sentenced parses correctly, just slightly overgenerates:
	
		tad meida-ar ekki mig
		it hurt.3sg no me
		'it doesn't hurt me'
	
	And the follow will have two parses as described above:
	
		tad meida-ar mig ekki
		it hurt.3sg me no
		'it doesn't hurt me'
		
CAN:

	The auxillary verb equivalent of the English "can", appears
	directly before the regular verb of the sentence and takes
	that VP as its complement.
	
		eg get elska hann
		I can.1sg love.1sg him
		'I can love him'
		
	To get the proper syntantic structure for aux verbs, I constrained
	the SUBJ of the aux to be coindexed with the first element on its
	ARG-ST, and also be coindexed witht the SUBJ of the VP that is
	its COMP.
	
	The aux verb has different conjugations than the regular verbs,
	so I had to create a new set of lexical rules to get the proper
	inflection. These lexical require DTR aux-verb-lex. This was
	allowed for correct parsing, but would get too many parses since
	it was still allowing the aux-verb-lex have the standard verb
	lexical rules to be applied. To fix this problem I had to create
	a new type of verb, reg-verb-lex, which inherits from verb-lex.
	The standard lexical verb rules require this type as its DTR, so
	aux-verb-lex can no longer be input for these rules. This no
	allowed only one correct parse for:
	
		eg get borda gler
		I can.1sg eat.1sg glass.NEUT.ACC
		'I can eat glass'
		
	However, now the question can't be formed since it was contrained
	to have a verb that had gone through the standard lexical rules
	as its DTR. This problem hasn't been fixed yet.
	
TO BE FIXED:

	Currently in my grammar I need to:
	
	Finish adding all the inflections so that the test suite will be 
	able to parse/not parse everthing rather than not find morphemes.
	
	Fix the problems with the negative adverb and can that I described
	above.
	
	Get ditransitive verbs and obj optional verbs working properly.
	This would entail cleaning up the transitive types to inherit
	and share constraints from a supertype.
	
	Get imperatives working properly. The lexical rule is applying, but
	still won't parse.
	
	Fix determiners. There is the free determiner which isn't really
	working, and I might not be able to get it working for this class.
	Everything is getting "_unspec_q_rel" for the semantics. I need
	to add the proper PRED relation in the determiner lexical rules
	and probably leave BARE-NP underspecified for this.
	
	Adjective problems. Adjectives have a weak and strong form, and
	this depends on the def/indefiniteness of the noun. This may be
	too complicated to complete for this class, but I will attempt
	it in the next lab.