---------------------------------------------------------
Notes

  After adding these rules and redefining the top level type, our VTA
  verbs were close to 100% correct parses in the sentences where they
  were isolated phenomena.

Yay --- I'm glad it works!

  Missing from our analysis is reflexive/reciprocal inflectional rules
  for our VTA verbs. For example, we currently do not handle AC.3SG >>
  GL.3SG. Valentine has several sections dedicated to discussing this
  phenomena.

Reflexive/reciprocal are outside the scope of the quarter, anyway.

  Nishnaabemwin does not have a part of speech corresponding to what
  we call adjectives in English.  All descriptive concepts are
  expressed by means of nominal derivational morphology (in which the
  descriptive property is introduced as a prefix such as gchi- 'big')
  or by intransitive verbs.  We have elected to treat these nominal
  prefixes as attributive adjectives.  There are two reasons for doing
  so.  The first is that, while we refer to them as prefixes, they
  behave phonologically as separate words, giving us motivation for
  dealing with them in our grammar.  Second, this is an efficient way
  to introduce the attributive predications involved without creating
  redundant lexical entries for nouns derived with various attributive
  prefixes.  We would miss significant generalization if we did not
  treat these prefixes as adjectival morphemes.  These morphemes
  always attach to the beginning of nouns:

Well that's certainly simpler.  Why are they called prefixes at all?

  Some adverbs do force the verbs they modify to be in conjunct order,
  although attested examples of this are adverbs glossed as English
  non-adverbs (if, when, while), so it is debatable whether these are
  truly adverbs in Nishnaabemwin. ("Adverb" and "particle" seem to be
  catch-all terms for grammars of Nishnaabemwin -- nouns and verbs are
  the only other parts of speech.)

If, when, and while clauses are adverbial in the sense that they
modify main clasuses.  So it's probably not terrible to call these
adverbs.  But they're outside the scope for the quarter :)

  While such complements are typically objects of the verb in English
  (e.g., I think *that you are nice*), in Nishnaabemwin the subjects
  of such sentences are typically promoted to the goal semantic role
  for the purposes of inflecting the verb.  Sentential complements can
  appear as complements to both transitive (X knows that Y) and
  intransitive (X is happy that Y) verbs.

I don't know waht "promoted to the goal semantic role" means here.
Is this a description of how agreement works in these verbs?
Also, X knows that Y and X is happy that Y are both equally transitive
in the sense of taking two arguments (X and Y).  Are they morphologically
distinguished in ojg?

  Demonstrative complementizers maanda (this) and iw (that) can be
  used to introduce sentential complements.

I've not come across the phrase "demonstrative complementizers" before.
Are the complementizers homophonous with the demonstratives?  
That's interesting.  What is the semantic difference between using
maanda and iw as complementizers?

  Focus is somewhat clearer in questions, in which the particle "na"
  questions the element it follows, and the particle must appear
  second in the sentence.  This ensures that the focused element in
  questions is in second position.

It sounds to me like the focused element is *first* in the sentence,
with "na" in second position.  Do you have examples of this in the test suite?
It sounds very similar to Russian "li", so I hope we can implement a solution...

Thank you for filling in the missing test suite material from
Lab 4.

  The adverb-lex type was automatically entered into our grammar by the questionnaire:

adverb-lex := basic-adverb-lex & intersective-mod-lex &
  [ SYNSEM [ LOCAL [ CAT [ HEAD.MOD < [ LOCAL.CAT.HEAD verb ]>,
			   VAL [ SPR < >,
				 SUBJ < >,
				 COMPS < >,
				 SPEC < > ]]]]].

I don't think you could have gotten this type from the questionnaire,
as the type for the negative adverbs is a scopal modifier, not
an intersective modifier.

I'm confused about what's going on with adjectives.  At the top
of your write up, you say that you're going to treat them as 
independent words (or so I thought).  Then you describe an unsuccessful
analysis of them as prefixes.  

I'm not clear on what exactly you were trying to do, and especially on
why you thought you needed to modify the class1-anim-sg-lex-rule.
The customization system should have given you a lexical rule slot for
the adjectives that fits in the right spot in the morphotactics, and all
you needed to do was modify that slot (to get rid of no-ccont) and
the subtypes (to put in the semantics).  That is, of course, if you're
treating them as prefixes.

#175   
Source: author
Vetted: f
Judgment: g
Phenomena: wo, agr, adv
Giimooj giiwwabmigoon Nenbozhoon nnahaangnikwem
Giimooj w-gii-waabmaad-igw-an Nenbozh-oon nnahaangnikwem
suddenly GL.3SG-PAST-see-INV.THM-AC.3.OBV Nenabush.AN-3.OBV my+daughter+in+law.AN.3SG
'Suddenly Nenabush saw my daughter-in-law'

You're getting a lot of parses for this one.  At least some of the
problem is some ambiguity regarding the lexical rules applied to the
verb.  Other sources are the possibility of coordinating the two nouns
(if you made the NP-coord top rule say [NUM pl] that would probably
block these), and the fact that the adverb can attach to V, VP or S from
that same spot (not sure what to do about that).    Also, the MRS has
"stealthily", while your gloss has "suddenly".  Which is it?

  Another confounding issue is that we created these demonstratives as
  determiners in the questionnaire. The result is that we had to make
  a choice about det-noun or noun-det without truly analyzing if these
  distinctions make sense for our demonstrative pronouns.

  ; proximal demonstrative pronouns

While demonstrative pronouns are often homophonous with demonstrative
determiners or adjectives, they are not the same morphosyntactically,
and we are not analyzing them this quarter.  

maaba := det-dem-prn-anim-3SGProx-determiner-lex &
  [ STEM < "maaba" >,
    SYNSEM.LKEYS.KEYREL.PRED "_this_det_rel",
    SYNSEM.LKEYS.ALTKEYREL.PRED proximal+dem_a_rel ].

This is not the desired semantics ... as given in the lab instructions,
the KEYREL.PRED should be "exist_q_rel".  Furthemore, this should
be given on the supertype, since it's the same for all the determiners.

Also, the repeated paths are bad tdl form.  Assuming you did want
both of those constraints, it should look like this:

maaba := det-dem-prn-anim-3SGProx-determiner-lex &
  [ STEM < "maaba" >,
    SYNSEM.LKEYS [ KEYREL.PRED "_this_det_rel",
                              ALTKEYREL.PRED proximal+dem_a_rel ]]

  the most common ordering is for a demonstrative to immediately
  precede the noun, however in cases where a participle must be in
  second position, the noun and demonstrative are separated:


  Giw dash Kweway wgii-gnawenmaawaan niw binoojiinyan.
  those.PR.3PL.Prox and.AV women.AN.3PL.Prox took+care+of.VTA.IND.3PL.Prox those.PR.3.OBV children.AN.3.OBV
.

"dash"/and.AV is a particle?  What is it coordinating?  What
does this sentence mean?  (Where's the free translation line?)

  We may have incorrectly intrepreted demonstrative determiners in our
  language. We implemented the requirements of the Lab on our
  demonstrative pronoun determiners in our grammar. The MRS results
  from our test sentences are fairly consistent for our demonstrative
  determiners, only a few of them do not assign the proper number of
  relations and indexing.

That's because you left the length of the RELS list on the demonstrative
determiners unbound, breaking the diff-list append.

determiner-dem-prn-lex := determiner-lex-supertype &
  [ SYNSEM [ LOCAL [ CAT.VAL.SPEC.FIRST.LOCAL.CONT.HOOK [ INDEX #ind & [COG-ST activ+fam],
                                LTOP #ltop ],
                              CONT.RELS.LIST.REST #altkeyrel ],
             LKEYS.ALTKEYREL #altkeyrel & [ ARG1 #ind, LBL #ltop ]]].


This should say:

CONT.RELS <! [], #altkeyrel !>

  We were unsure of how to deal with optional arguments in our
  language. Our unexpressed arguments on verbs are not really dropped
  subjects. The addition of indefinite pronouns and an analysis of the
  unspecified actor could greatly contribute to our understanding of
  the correct COG-ST values that are required. We did not implement
  any COG-ST values in our optional argument framework for this Lab.

I need more here.  Why aren't they dropped subjects?  

I see we discussed this some on GoPost:


    On 2/05/2010 5:23 PM klassp said:
    I think they are always interpreted as pronouns. We do have explicit intransitive and transitive verbs for eat for example. We have indefinite pronouns (for example waya -- anyone animate and gegoo -- something inanimate).


In that case, the transitive verbs can say [ OPT-CS in-foc ] on their complement.  There remains the interesting question of whether intransitive eat et al should have an ARG2 in the semantics, but let's assume not for now.

... did you not have time to incorporate this into your lab?

- we need to map the adverb for negation to the required negation suffix

Post to GoPost with the details of how this works, and I'll make suggestions.
In particular: are both parts of the negation always required?  Where does
the negation suffix attach in the pipeline with respect to affixes that
involve the object?  What are the constraints on the position of the negative
adverb?

- we need to analyse the descriptive intransitive verbs in our language

Do they need to be treated any different from ordinary intransitives?

- yes/no is still not parsing in our grammar. We will need to "wire up" the rules for yes/no in our grammar

Please post to GoPost with the details so I can advise.

- determine COG-ST for our unexpressed arguments

This was indeed on the menu for Lab 5.



