4/19/06: lecture in 567

Description of coordination strategies: "the coordinand"

If aux info is filled out, make aux even if not need for neg/q

globals.lsp and hyphens

instance identifiers can't contain . or :, but script
uses orthography to make them for lexical items, and the
orthography may contain those characters.

When I did fake_dutch I specified obligatory determiners and
got a bare np phrase anyway.  Are we just assuming that all
languages want it?  (Neither of the nouns can go through it
in this grammar, but the phrase is there.)

Change comment in irules.tdl to specify spelling *changing*
lexical rules, rather than spelling preserving.

same-nonlocal-lex-rule probably shouldn't inherit from no-ccont-lex-rule,
since cont-change-only-lex-rule inherits from it.

head-subj in matrix.tdl should not say POSTHEAD +.

fix lex rules so that COORD is specified  (cf. V-M type labels)

come up with a better analysis of bare-nps ... this year
the hack of constraining noun types by digging in the RELS list
was the first thing they hit.  This is probably related to
disentangling discourse status and quantifiers.

make a file/module that patches the semantics of head-filler, opt-subj
etc that are ready for x-classification with for languages
that don't do it that way

Why isn't MC a head feature?

Need to figure out what to do about messages on VP coordination (and S).
If the root condition requires a contentful message (forcing things to
go through a clausal 'spindle') the bottom-coord rule seems to lose
the 'MSG no-msg' information, leading to spurious extra parses.

Remove "LIGHT" machinery --- it's English specific.  Keep the features
there and come up with something to patch in.  Current problems: need
HC-LIGHT - on verb-lex anytime want to be able to pick out verbs.  Also,
adj-head phrases copy up LIGHT instead of being LIGHT -.  Wanted them
to be LIGHT - for Tok Pisin 'ol'.  Also, I'm a bit worried about whether
the class of phenomena that pick out 'LIGHT' things is really that
consistent.  That is, for any given language, is there one set of
things that count as LIGHT for all LIGHT-sensitive phenomena?  Distinguish
between LIGHT and lexical?


Floodeen:
 This was because head-adj left SLASH constrained to be the SLASH of
 its 2nd argument, which is the adjunct, not the head.  I handled this
 by making a new type inheriting from head-adj-int-phrase, which I then
 constrained to take its SLASH from the head daughter.

Dutch & ?Hebrew:
 You could consider making a head-opt-2nd-comp rule (on the model of
 basic-head-opt-comp-phrase and basic-head-2nd-comp-phrase from
 matrix.tdl). --- particularly useful if you don't want
 head-2nd-comp-phrase but have other reasons to constrain head-opt-comp
 to apply very low.

Floodeen:
 Most of the time went into the argument compositional 'can', which I
 ended up scrapping because I can't seem to get any MC restrictions to
 work with my subject extraction rule without breaking strange things
 that I don't really understand.  Clearly extracted-subj constrains its
 daughter to be 'MC na' for a good reason.

Bullock:
 I also noticed that negated sentences can be incorrectly interpreted
 as imperatives, even when the subject is not 2nd person. The grammar
 requires the subject of imperatives to be 2nd person, but this
 restriction doesn't seem to be working if the matrix verb is
 negated. For example, the following sentence is interpreted as both a
 declarative and an imperative:

This sounds like a matrix bug:  The scopal adverb is the semantic
head, so the HOOK of the adj-head phrase is coming from the adj daughter,
but it's not saying anything about the HOOK.XARG.  Thanks for pointing
this out.

Am I copying MSG everywhere I should be in the lex rule types?

Johanson

 It doesn't hurt me

 The best translation of this sentence is 'Non mi-f-a male' (It doesn't
 hurt me.)  However, when I tried to generate from the semantics I got
 'Non f-a male' (It doesn't do poorly.) as an alternative.  It got this
 because when I implemented the argument optionality, I said that all
 verbs could drop one of its arguments with a definite interpretation.
 This is not the case.  Dropping 'mi' changes the verb 'farsi' to the
 verb 'fare'.  Dropping an argument in a reflexive verb changes the
 verb.  The verbs are actually listed separetly in the dictionary.  A
 clearer example could be the dropping of the object in the verb
 'chiamarsi'.  "Mi chiamo Luigi' means 'My name is Luigi', but 'Chiamo
 Luigi' means 'I call Luigi'.  It doesn't matter how obvious it is that
 you are calling yourself, it cannot be interpretted with the definite
 interpretation.

 In order to fix this, I would have to add a different type of verb for
 reflexives.  These verbs would not allow a definite argument
 optionality and would require that its object would be a pronoun. For
 now, we will just have to pretend that 'farsi' is the same thing as
 'fare'.

You're right that the reflexive verbs need to be handled
separately.  I think the way to go is with the ones that are
*lexically* required to be reflexive, that affix is really just
an agreement marker with the subject.  

Still, that's not what's going on with mi-duole, and I'm surprised
that it generates duole, since duole shouldn't have 1sg on the ARG2.
