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

*Please* submit your writeups as plain text files so I can cut and paste
into these notes more easily.

The tarball you turned in has lots of extraneous files in it.  I shouldn't
have to guess which one is your actual test suite text file.  I went with
testsuite.txt since it was the longest.

"disfluency" does not mean the same things as "ungrammatical example".

Why would unspecified actors be "4th person"?  It just sounds to me
like a kind of 3rd person indefinite.

   Accordingly, person is the primary distinction, with 2nd person
   taking precedence over 1st, and both taking precedence over 4th,
   which takes precedence over 4th.

I assume the first 4th here is supposed to be 3rd.  Is this hierarchy
part of treating the unspecified actors as a separate person? I think you
can still handle it without that, given the hierarchy defining stuff.

  If person cannot distinguish the verb’s arguments, then the animate
  noun takes precedence over the inanimate one.  Finally, if both
  gender and person are shared, an additional feature called
  “obviation” is used, with proximate (near) nouns taking precedence
  over obviative (distant) ones.  The noun which is higher in the
  topicality hierarchy is the subject, and the other is the object.
  The sentence below displays the effects of this hierarchy, with the
  direct meaning of the sentence given:

I'm not sure this cascaded description is right.   For the first step,
all 1st and 2nd person elements are probably animate anyway, so it's
only if you have two 3rd person animates.  Is it really the case that
a 3rd person inanimate outranks a "4th" person animate?  Likewise, going
to the next step, is it really the case that a 3rd person inanimate proximate
outranks a 3rd person animate obviative?

  As mentioned above, a new feature, “obviation” was required to
  complete the topicality hierarchy for Nishnaabemwin.  In fact,
  obviation is a feature marked on nouns and verbs which also must
  show agreement.  We have elected, however, to leave implementing
  this agreement to a later date.

Actually, I'm not sure you need to treat it as agreement.
Just add obviation to the scale you define on the direct-inverse page.
I'm going to forward your write-up to Scott Drellishak for his
input on how best to model this using that page.

 When a morphophonological analyzer is introduced, the type system
 will simplify considerably.  Verbs will be divided into intransitive,
 transitive, and ditransitive types, as well as being grouped
 according to the animacy of its potential subjects and objects.

It sounds like you're not clear on what a morphophonological analyzer
is: It's a bit of software that takes surface forms and gives you
regularized underlying representations.  For example, a
morphophonological analyzer for English would take "children" and give
"child+s".  We're not going to build or otherwise introduce these this
quarter.  Rather, we're going to assume one.  In other words, if you
have any morphophonological effects (sound changes at morpheme
boundaries, infixes, ablaut...), represent this in your test suite by
differentiating the standard orthography/transliteration line from the
morpheme-segmented line.  Then have your grammar target the
morpheme-segmented line (in the spelling of the stems and affixes.

What you need to do (and will start doing this week) is to define
lexical rules in the customization system.  These will handle the combinatory
explosion of morphologically complex languages, allowing you to input
just verb (and noun) stems, and definitions of the (regularized) affixes
that attach to them.  If any of this isn't clear, please ask!

  Verbs will be divided into intransitive, transitive, and
  ditransitive types, as well as being grouped according to the
  animacy of its potential subjects and objects.

We won't be doing ditransitives (and can't through the customization system).
Is it really a syntactic constraint that some verbs can't take animate (or
inaminate) arguments?  

Re: splattered NPs --- cool!  I have an analysis of this worked out
from Wambaya, and we can try to integrate it later in the quarter, if you
are interested.

Regarding Det-N and N-Det order, this at least can be easily fixed
by instantiating both the head-spec and spec-head rules.  Don't bother
with that until we're done with the customization system, though, since
you don't want to do any hand-editing of tdl until then.

  One grammatical sentence which we expected not to parse (and didn’t)
  was an instance of the phenomenon in Nishnaabemwin by which numbers
  and demonstrative pronouns can co-occur:

I think you have demonstrative adjectives in this language, rather
than demonstrative determiners.

  Finally, a significant but understandable omission from the
  customization system is information structure.  While this is a
  poorly developed area of the Grammar Matrix, it is an essential
  aspect of Nishnaabemwin because it is the only phenomenon which has
  a predictable influence on word order.  Improvements in this area
  may not change many grammaticality judgments given by the system,
  but they would be of great use in generating more realistic text.

And this is something we will work on later this quarter, so please take
notes along the way as your sources talk about the relationship of
information structure to word order (or other aspects of morphosyntax).

Your lab2/item file has only one line in it.  I was able to create a profile
from it anyway, though, suggesting that if you're having problems with
[incr tsdb()], it's not your skeletons/ directory.

Are the hyphens part of the standard orthography?

#1
Source: a:776
Vetted: s
Judgment: g
Phenomena: wo
Waagaakwad wii-giinboode
Waagaakwad wii-giinboode
axe.3.SG.INAN FUT-want.sharpening.X.SG.INAN
'The axe needs sharpening'

If not, they shouldn't be in that first line.

What is X?

This:

    verb1_stem1_pred=_walk_in_a_circle_v_rel

should be _walk+in+a+circle_v_rel

This:

    verb4_stem1_pred=_is_abundant_v_rel

can just be _abundant_v_rel

Pred values shouldn't indicate case:

    verb6_stem1_pred=_made_v_rel

should be _make_v_rel

