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

  These two examples above are not being tested (see below the
  Testsuite Examples Not Covered in Starter Grammar)

I see no such section, and it's not clear to me why you wouldn't be
testing it now ... if it's in the testsuite, you're testing it, even if you
don't expect to currently get the right answer.  Or maybe those are
marked as ungrammatical, though there are some verbs that take
dative objects?

  Only imperfective is used in present tense. Cannot really construct
  an ungrammatical example here because if we plug in the Perfective
  aspect, it will be read as Future Tense Perfective form (because
  they are identical) or as Perfective Past (which would just give it
  a past reading), which will be grammatical, so we have only
  grammatical example here:

Still, you can model this by constraining the perfective to be
non-present.  (You can make a non-present type by defining
your own tense hierarchy, with past and future inheriting from
non-present.)

  To cater for this we added one main verb slot and four morphemes: -l
  (sg.masc), -la (sg.fem), -lo (sg.neut), -li (pl). The inputs were
  either verbs whose stem ends in a vowel (such as chita) or another
  verb slot which first added a vowel to a verb whose stem ends in a
  consonant (such as spitj, stem: sp, first verbslot: sp-i, past tense
  verb slot: spi-l for sg.masc).

These ones don't mark person, only number and gender?

  //We don't have ungrammatical examples FOR IMPERFECTIVE FUTURE
  because only verb "byt'" is inflected and it is covered in the
  Future Tense section.

Do you have examples somewhere of "byt" taking a complement with
the wrong inflection (e.g., agreement markers)?

  We noticed that the customization system only allows one negatin
  adverb. We were not sure whether nikogda (never) would be considered
  a negation adverb and if so, how we would specify that in the
  customization system.

This is complicated, since we don't want to get two negative predicates
in the sentence when nikogda appears with ne.  Is ne always immediately
adjacent to the verb?  That might help some, but either way we're dealing
with a negative polarity item, which is a tricky thing...

  Although several sources note this, they didn't describe the process
  in detail.  It seems though that the first noun phrase would be more
  definite.  I.e. in the sentence "Sobaku kupil Ivan" (Ivan bought the
  dog) the NP "dog" occupies the first position in the sentence and
  precedes the subject, thus being equivalent roughly to "It was the
  dog that Ivan bought" emphasizing that it was the dog (and not a
  cat).  We have covered the different word order variations in the
  Word Order section.

This sounds like it might be relevant for the information structure lab.
Keep an eye on it...

   ARGUMENT OPTIONALITY.  We have several examples in testsuite, in
   which the subject and object can be dropped, if the context is
   clear or if the verb describes a natural phenomena (then dropping
   the subject is actually required).  We are not testing all of them
   yet, though, because these are special cases.

Again, I'm not clear what you mean by "not testing"...

  It actually involves adding a specific morpheme to the stem for just
that case, but does not lend itself to an easy rule. For example verb
spitj (sleep), in 1st.sg.present is splju, but for kupitj is kupchu,
the rest of the conjugation follow some rules 9after we add a vowel
morpheme. This also affects perfective future for those same verbs.

I'd be inclined to just call this morphophonology and abstract away from
it...

  We got a lot more ungrammatical examples that parsed and we do not
  know yet why. For example, Ja chitaem (which is wrong due to the
  noun - verb agreement) is listed as an ungrammatical example which
  parsed, however when I parse that same sentence in lkb, it is
  correctly flagged with errors. I suspect maybe some test suite
  formatting issues, but have not seen anything yet.

In cases like this, please include the [incr tsdb()] item number so I can
easily check and see what's going on. ... In fact, [incr tsdb()] only shows
one ungrammatical example parsing.  Why do you think there are many?
Furthermore, the example that is incorrectly parsing is identical to
other examples you listed as grammatical.  Maybe you meant to edit
it and didn't?

   We also could not see how to add more than one negation verb
   (nikogda (never) in addition to ne (no)). But this may be possible
   through adding auxilliaries and changind complement features.  Did
   not have time to explore further.

If ne is always right next to the main verb, you could potentially
treat it as an auxiliary, or rather as two:  One contributes the negation relation,
and the other doesn't but does select for an adverb that it specifies to
scope over the verb.    This has to be done via tdl editing, however.

You have 13 examples that each get two readings.  Several of these seem
to be due to the fact that knigu is underspecified for animacy.
