----------------------------------------------
Comments

 Hausa has demonstrative determiners, as well as demonstrative
 pronouns. I did not include demonstrative pronouns in my test suite,
 so I am mainly focusing on demonstrative determiners(wata, "this.fem"
 and wani "this.masc).  I hope to add demonstrative pronouns later on.
 Demonstrative pronouns are split by m/f/non-sg, and include this(by
 me), that(by you), that(there), and that (distant).

"there" is ambiguous (in English) between "by you" and "distant",
so I'm puzlzed what the "that (there)" category is meant to indicate.
If there really is a four-way distinction in Hausa, I'll need
to revise my hierarchy of demonstrative rels.

 I added a demonstrative-determiner-lex type, but I can't get it to
 work.  I've been having problems with unification between the type and
 the lexicon.  I'm trying to add the gender and number information in
 the entries in the lexicon (through SYNSEM.LOCAL.CONT.HOOK.INDEX.PNG),
 but it says that this doesn't unify with the type constraints.  I'm
 particularly confused about why the LKB doesn't complain when I put
 [GEN fem] into the demonstrative-determiner-lex itself, but I can't
 write a subtype that adds the GEN information (f-demonstrative-lex,
 see hausa.tdl file, directly under demonstrative-determiner-lex).  I
 left demonstratives inheriting from determiner-lex, as this gave them
 the correct behavior (but not the "exist_n_rel" or COG-ST
 information).

It's far better to post to EPost in these cases (esp. since I haven't
been getting to the grading until a few days after you turn the
labs in). 

At any rate, I tried making an instance of f-demonstrative-lex in your
lexicon.tdl, and it worked just fine.  That is, the LKB didn't complain,
and a sentence with the word I created parsed.  All that was missing
was the specific PRED value on the demonstrative relation.  You should
use the ALTKEY feature to get this information in on the lexical entries.
(This means that the demonstrative-determiner-lex type needs to 
identify the second RELS element with the LKEYS.ALTKEY value.

 Hausa has subject pro-drop, in that it requires that pronouns be
 dropped when they appear as subjects.  This means that the subject is
 understood by context.  To take care of this I added an instance of
 basic-head-opt-subj-phrase.  This allows subject-less sentences, such
 as "taa-zoo", (she-came).  In the last lab I added a PRO feature to
 distinguish pronouns from other words, and made all nouns PRO -.  I
 then required verbs to take a PRO + subject.  This took care of making
 sentences with pronoun subjects ungrammatical (such as "*ita taa-zoo",
 *she came).

As you are probably noticing in Lab 7, you actually want to use
decl-head-opt-subj-phrase, so that you get the right SF value.

 As far as I can tell, Hausa doesn't have optional objects.  There are
 a few exception, but these seem to be rare enough that I'm not going
 to cover them.  I can't find the page in my reference grammar, but the
 example given for these exceptions was something like "my wife, she is
 doing lunch", when it is understood that she is cooking lunch.

What about the verb 'eat'?  Can it really not drop its object?

 Adjectives in Hausa appear both before and after their head noun.
 When appearing before the noun, they have to take on a linker suffix
 -n/-r~/-n (m/f/pl).  To cover this, I made an adj-lex-rule which adds
 POSTHEAD - to an adjective. I also made three subtype rules for m/f/pl
 which add gender/number information.  made instances of these three
 rules in the irules.tdl file to add the correct suffix.  As I have it
 now, the rule fails to give POSTHEAD- and suffixes to prenominal
 adjectives.  I think

If your adjectives start off underspecified for POSTHEAD, though, then
the suffix-less ones should also show up prenominally.  (Unless you've
paired this rule with a zero rule that puts in the POSTHEAD + information.)

 f-adj := 
 %suffix (!c!c !c-r~) (!c !c-r~)
 f-adj-lex-rule.

 m-adj := 
 %suffix (!c!c !c-n) (!c !c-n)
 m-adj-lex-rule.

 ;;;Adjectives
  faraa := f-adjective-lex &
 [ STEM < "faraa" >,
 SYNSEM [ LKEYS.KEYREL.PRED "_white_j_rel" ]].

 farii := m-adjective-lex &
 [ STEM < "farii" >,
 SYNSEM [ LKEYS.KEYREL.PRED "_white_j_rel"]].

From this, it looks like adjectives also have gender suffixes
that appear before the linker suffix.  Have you considered making
a lexical rule for this?

Also, I couldn't get any examples to parse with your m-adj-lex-rule.
The problem seems to be that you're constraining the adjective's own
HOOK.INDEX (whose value is an event variable, and thus can't take the
feature PNG) instead of the HOOK.INDEX of the thin on its MOD list.


 To get the right behavior, I needed to constrain the adjectives to
 have the same gender and number information as their head noun.  I
 tried to make a co-reference on the PNG value of the first item of the
 adjective-lex's mod list, and adjective-lex itself, but I kept getting
 an error in "unification with ARG1-EV-RELATION at
 synsem.local.cont.rels.list.first".  To get around this, I made three
 subtypes m-adjective, f-adjective and pl-adjective which placed
 gender/number constraint on the MOD value.  I then had my adjectives
 inherit from the apropriate subtype.  I think it's simpler to do it
 the first way, but I just couldn't figure out that error, and I wanted
 to move on to adding a suffix rule for adjectives.

Right --- there is no need for the adjectives to have inherent
PNG.  The lexical rules which add the suffixes can constrain the
MOD..PNG, and that's it.  No generalization lost, since the lexical
rules will all need to say something about PNG anyway.

 For adverbs, I added an adverb-lex rule, and tested it with jiyà
 "yesterday".  Adverbs in Hausa generally show up at the end of the
 sentence, after the subject, verb and object (if any) information.
 Temporal verbs like jiyà can show up at the beginning of the sentence
 as well.  I added three examples to my test suite to show adverbs and
 stacking adverbs (taa-dafà miyàa sòosai jiyà, "she cooked soup
 perfectly yesterday").  One is ungrammatical, since it appears between
 the verb and its object.  My grammar overgenerates, since this
 sentence is parsed as grammatical.

It sounds like you should constrain the adverbs to modify only S
(COMPS <>, SUBJ <>, HEAD verb), and leave POSTHEAD underspecified
on the temporal ones.  Would that do it?